CN118321519A - Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium - Google Patents
Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium Download PDFInfo
- Publication number
- CN118321519A CN118321519A CN202410454500.6A CN202410454500A CN118321519A CN 118321519 A CN118321519 A CN 118321519A CN 202410454500 A CN202410454500 A CN 202410454500A CN 118321519 A CN118321519 A CN 118321519A
- Authority
- CN
- China
- Prior art keywords
- bonding
- thermocouple
- crystallizer
- time sequence
- column
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 115
- 239000010959 steel Substances 0.000 title claims abstract description 115
- 238000000034 method Methods 0.000 title claims abstract description 64
- 230000005540 biological transmission Effects 0.000 claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 47
- 230000009467 reduction Effects 0.000 claims description 53
- 239000011159 matrix material Substances 0.000 claims description 44
- 238000001514 detection method Methods 0.000 claims description 40
- 238000005266 casting Methods 0.000 claims description 26
- 238000009749 continuous casting Methods 0.000 claims description 26
- 238000009826 distribution Methods 0.000 claims description 24
- 230000002159 abnormal effect Effects 0.000 claims description 23
- 238000012360 testing method Methods 0.000 claims description 20
- 238000012546 transfer Methods 0.000 claims description 19
- 238000012795 verification Methods 0.000 claims description 17
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 claims description 12
- 229910052802 copper Inorganic materials 0.000 claims description 12
- 239000010949 copper Substances 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 abstract description 32
- 230000009286 beneficial effect Effects 0.000 abstract description 8
- 238000009434 installation Methods 0.000 abstract description 4
- 238000013277 forecasting method Methods 0.000 abstract description 2
- 238000004519 manufacturing process Methods 0.000 description 27
- 230000000630 rising effect Effects 0.000 description 22
- 230000008859 change Effects 0.000 description 13
- 230000008569 process Effects 0.000 description 12
- 230000006870 function Effects 0.000 description 11
- 230000015654 memory Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 238000009499 grossing Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 9
- 238000013461 design Methods 0.000 description 8
- 238000003912 environmental pollution Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 8
- 239000002699 waste material Substances 0.000 description 8
- 238000004891 communication Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 230000003449 preventive effect Effects 0.000 description 6
- 238000009529 body temperature measurement Methods 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000001816 cooling Methods 0.000 description 3
- 230000004907 flux Effects 0.000 description 3
- 239000007788 liquid Substances 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000011897 real-time detection Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000007599 discharging Methods 0.000 description 2
- 230000002706 hydrostatic effect Effects 0.000 description 2
- 238000005461 lubrication Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000009851 ferrous metallurgy Methods 0.000 description 1
- 239000003574 free electron Substances 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
- 230000005676 thermoelectric effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Landscapes
- Continuous Casting (AREA)
Abstract
The invention provides a crystallizer bonding steel leakage forecasting method, a device, electronic equipment and a medium, wherein the method comprises the following steps: acquiring an original thermocouple temperature time sequence and a pull-speed time sequence corresponding to each sample from a bonding sample and a non-bonding sample; model training is carried out according to the pull speed time sequence and the thermocouple temperature original time sequence, and a crystallizer bonding steel leakage prediction model is obtained; obtaining a single-column thermocouple bonding probability prediction value through a crystallizer bonding breakout prediction model, and providing a crystallizer bonding breakout prediction method by combining bonding point transverse transmission characteristics, wherein the method is suitable for a scene that a certain heat extraction thermocouple in a plurality of rows of thermocouples is disabled, and is beneficial to reducing the bonding phenomenon missing judgment rate and the thermocouple installation row number; compared with a logic threshold judgment method, the method can not only realize more accurate and timely identification of the bonding phenomenon in multiple application scenes, but also help to reduce human errors and time cost caused by continuously debugging threshold parameters according to site scenes.
Description
Technical Field
The application relates to the technical field of ferrous metallurgy continuous casting detection, in particular to a crystallizer bonding steel leakage prediction method, a device, electronic equipment and a medium.
Background
In the continuous casting process, molten steel enters a tundish from a ladle through a long nozzle and flows into a crystallizer through a submerged nozzle. The high-temperature molten steel starts to cool and solidify to form a blank shell with a certain thickness due to the heat transfer effect between the high-temperature molten steel and the water-cooled copper plate of the crystallizer. The blank shell is continuously pulled out into a secondary cooling zone under the vibration of a crystallizer and the blank pulling action. However, in actual production, under the influence of factors such as poor lubrication effect of continuous casting mold flux, overlarge fluctuation of the liquid level of a crystallizer or improper control of superheat degree of molten steel, a casting blank is easy to bond with a copper plate of the crystallizer, if related measures are not timely taken, a blank shell is extremely easy to be torn continuously and solidified for the second time under the actions of vibration and blank pulling of the crystallizer to form a new bonding point, and when the bonding point moves down to an outlet of the crystallizer, the blank shell is broken to form steel leakage due to insufficient strength of the bonding point to resist internal hydrostatic pressure and blank pulling resistance. The bonding breakout not only can lead to the shutdown of a continuous casting machine, thereby causing production interruption and reducing the production efficiency of steel, but also is accompanied with certain energy waste, safety risk and environmental pollution, so the construction of the crystallizer bonding breakout prediction system has important significance.
Patent document CN107096899a discloses a crystallizer steel leakage prediction system based on logic judgment, which mainly aims at all thermocouples on a crystallizer copper plate, and based on each logic judgment rule parameter, carries out temperature rise detection, temperature rise rate detection, temperature continuous drop detection, temperature inversion detection and temperature change delay detection on three thermocouples in an inverted triangle area of a temperature rise abnormal thermocouple, and judges whether bonding alarm is carried out according to the comprehensive detection result. However, temperature inversion is not desirable as a characteristic parameter for identifying sticking phenomenon, for two reasons. Firstly, after the bonding alarm time point corresponding to temperature inversion is located at the intersection of the two heat extraction couple temperature time sequence curves, certain alarm hysteresis exists, and the risk of steel leakage is high. Secondly, in actual production, due to the influence of factors such as fluctuation of molten steel liquid level, continuous casting mold flux performance, molten steel fluidity and the like, the phenomenon that the temperature measurement value of the lower heat extraction thermocouple is higher than that of the upper heat extraction thermocouple occurs sometimes, and meanwhile, if the temperature inversion detection rule designed by the patent is adopted, the missing report is very easy to occur. Meanwhile, only when the bonding sample and the non-bonding sample are linearly available, the crystallizer bonding steel leakage prediction method constructed based on the logic threshold judgment method can achieve theoretical zero leakage and zero false alarm, the design of the logic threshold judgment method is highly dependent on process conditions, equipment quality, physical parameters and the like, parameter thresholds are required to be adjusted continuously according to site conditions in time, the robustness of the crystallizer bonding steel leakage prediction method constructed based on the logic threshold judgment method is poor, and the method is reflected in the aspects of high later maintenance cost, extremely easy initiation of the new production mode, and the like. Finally, the invention patent takes all thermocouples into account during feature extraction, which is not conducive to cost reduction due to reduced thermocouple mounting rows.
Disclosure of Invention
In view of the above-mentioned drawbacks of the related art, the present application provides a method, an apparatus, an electronic device and a medium for predicting bonding steel leakage of a mold, so as to solve the above-mentioned problems.
The application provides a crystallizer bonding steel leakage prediction method, which comprises the following steps: acquiring a thermocouple temperature original time sequence and a pull rate time sequence corresponding to each original sample, wherein the original samples comprise bonding samples of casting blanks and a crystallizer and non-bonding samples of the casting blanks and the crystallizer; obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the feature matrix to obtain a crystallizer bonding steel leakage prediction model; inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and the pulling speed time sequence corresponding to the temperature time sequence of any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and generating a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period; and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion.
In an embodiment of the present application, a temperature sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in any two rows of the same distribution range extracted in advance or in real time and a pull rate sequence corresponding to the temperature sequence are input to the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability prediction values, and a corresponding single-row thermocouple bonding probability prediction value sequence is generated for the plurality of single-row thermocouple bonding probability prediction values based on continuous detection in a preset time period, including: acquiring any two rows of thermocouple temperature time sequences of part or all columns in the continuous casting machine within preset time, and taking the thermocouple temperature time sequences as any two rows of thermocouple temperature time sequences to be detected of each column within the preset distribution range; inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in the preset distribution range and the pulling speed time sequence corresponding to the temperature time sequence into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability prediction values; forming a single-column thermocouple bonding probability predicted value sequence by using the single-column thermocouple bonding probability predicted value corresponding to each moment in the preset time to obtain a plurality of single-column thermocouple bonding probability predicted value sequences; and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion.
In an embodiment of the present application, obtaining the feature matrix according to the pull-rate sequence and the thermocouple temperature original sequence includes: filtering abnormal values in the original thermocouple temperature time sequence to obtain a thermocouple temperature time sequence; performing smooth noise reduction treatment on the thermocouple temperature time sequence by utilizing a Bezier curve to obtain a thermocouple temperature smooth noise reduction time sequence; and obtaining a characteristic value according to the pull-speed time sequence and the thermocouple temperature smooth noise reduction time sequence, and constructing a characteristic matrix through the characteristic value.
In an embodiment of the present application, model training is performed through the feature matrix to obtain a mold bonding breakout prediction model, including: setting the label value of the bonding sample to a first preset value, and setting the label value of the non-bonding sample to a second preset value; combining the first preset value and the second preset value, and carrying out layered sampling on the feature matrix based on a preset data dividing proportion to obtain a training set, a verification set and a test set, wherein the training set and the verification set are used for model training, and the test set is used for testing a trained crystallizer bonding steel leakage prediction model; fitting the initial crystallizer bonding breakout prediction model with the training set, and adjusting the initial crystallizer bonding breakout prediction model through the verification set to obtain the crystallizer bonding breakout prediction model.
In one embodiment of the present application, determining whether a bonding phenomenon exists by combining a predicted value of a bonding probability of a single thermocouple in each sequence with a bonding point lateral transfer characteristic recognition criterion includes: if all the single-column thermocouple bonding probability predicted values in the at least one single-column thermocouple bonding probability predicted value sequence are larger than the first preset probability, marking the thermocouple column corresponding to the single-column thermocouple bonding probability predicted value sequence at the latest moment as a third preset value; if the thermocouple column corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple column corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon.
In one embodiment of the present application, determining whether a bonding phenomenon exists by combining a predicted value of a bonding probability of a single thermocouple in each sequence with a bonding point lateral transfer characteristic recognition criterion includes: obtaining a predicted value of the bonding probability of each row of thermocouples at continuous moments to obtain a predicted value of the bonding probability of the thermocouples at a first moment, a predicted value of the bonding probability of the thermocouples at a second moment and a predicted value of the bonding probability of the thermocouples at a third moment, which correspond to each row of thermocouples; if the predicted value of the bonding probability of the single-column thermocouple at the first moment and the predicted value of the bonding probability of the thermocouple at the second moment are both larger than the second preset probability, and the predicted value of the bonding probability of the single-column thermocouple at the third moment is larger than the third preset probability, marking the thermocouple column corresponding to the predicted value of the bonding probability of the single-column thermocouple at the third moment by the third preset value; if the thermocouple column corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple column corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon.
In an embodiment of the present application, if the marking value meets the transverse transmission characteristic of the bonding point, determining that the bonding phenomenon exists includes: extracting a thermocouple column corresponding to a third preset value, and obtaining the transverse propagation velocity of the bonding point based on the time difference of the thermocouple column corresponding to the third preset value and the corresponding pulling speed average value; obtaining detection time through the transverse propagation rate of the bonding point and the arrangement interval between thermocouple columns corresponding to the third preset value; if the number of the thermocouple columns corresponding to the third preset value in the detection time is larger than or equal to a fourth preset value, judging that the crystallizer has a bonding phenomenon, and if the number of the thermocouple columns corresponding to the third preset value in the detection time is smaller than the fourth preset value, judging that the crystallizer has no bonding phenomenon.
The application provides a crystallizer bonding steel leakage prediction device, which comprises: the time sequence acquisition module is used for acquiring a thermocouple temperature original time sequence and a pull-speed time sequence corresponding to each original sample; the original sample comprises a bonding sample of a casting blank and a crystallizer copper plate, and a non-bonding sample of the casting blank and the crystallizer copper plate; the model training module is used for obtaining a characteristic matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the characteristic matrix to obtain a crystallizer bonding steel leakage prediction model; the time sequence detection module is used for inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time and the pulling speed time sequence corresponding to the temperature time sequence into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period; and the bonding judgment module is used for judging whether the crystallizer has bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transmission characteristic recognition criterion.
The beneficial effects are that: according to the crystallizer bonding steel leakage prediction method, device, electronic equipment and storage medium, the original thermocouple temperature time sequence and the pull speed time sequence corresponding to each sample are obtained from bonding samples and non-bonding samples of casting blanks and the crystallizer; obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the feature matrix to obtain a crystallizer bonding steel leakage prediction model; inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and the pulling speed time sequence corresponding to the temperature time sequence of any two rows of thermocouples to be detected in a preset distribution range into a crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a plurality of thermocouple bonding probability predicted value sequences based on the plurality of single-row thermocouple bonding probability predicted values; and judging whether a bonding phenomenon exists or not by combining the single-column thermocouple bonding probability predicted value in each sequence with a bonding point transverse transmission characteristic recognition criterion. Obtaining single-column thermocouple bonding probability predicted values through a neural network model, comprehensively considering the time sequence of the single-column thermocouple bonding probability predicted values and the transverse transmission characteristics of bonding points, being applicable to a scene that a certain heat removal thermocouple in a plurality of rows of thermocouples is disabled, being beneficial to reducing the missed judgment rate of bonding phenomenon, reducing the number of thermocouple installation rows and saving the cost; compared with the traditional logic threshold judgment method, the method can not only realize more accurate and timely identification of adhesion phenomena such as multi-type temperature change characteristics, multi-production site and the like under multi-application scenes, avoid shutdown of a continuous casting machine due to bonding steel leakage, improve the production efficiency of steel, avoid energy waste, safety risk and environmental pollution caused by bonding steel leakage, and help to reduce human errors and time cost without continuously debugging threshold parameters according to site live. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application as claimed.
Drawings
FIG. 1 is a flow chart illustrating a method of crystallizer bonding bleed-out prediction in accordance with an exemplary embodiment of the present application;
FIG. 2 is a timing diagram of bond sample single column thermocouple temperature versus pull rate illustrating an exemplary embodiment of the present application;
FIG. 3 is a timing diagram of bonded sample single column thermocouple temperature versus pull rate illustrating an exemplary embodiment of the present application;
FIG. 4 is a timing diagram of bonded sample single column thermocouple temperature versus pull rate illustrating an exemplary embodiment of the present application;
FIG. 5 is an exemplary diagram of a set of mold bond leak prediction model features shown in an exemplary embodiment of the application;
FIG. 6 is a flow chart illustrating a mold bond bleed-out forecast according to an exemplary embodiment of the present application;
FIG. 7 is a graph comparing marked time points of the sticking phenomenon shown in an exemplary embodiment of the present application;
FIG. 8 is a block diagram of a mold sticking leak prediction apparatus according to an exemplary embodiment of the present application;
fig. 9 is a schematic structural view of an electronic device according to an exemplary embodiment of the present application.
Detailed Description
Other advantages and effects of the present application will become apparent to those skilled in the art from the following disclosure, which describes the embodiments of the present application with reference to specific examples. The application may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present application. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict.
It should be noted that the illustrations provided in the following embodiments merely illustrate the basic concept of the present application by way of illustration, and only the components related to the present application are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
It should be noted that, in the present application, "first", "second", and the like are merely distinguishing between similar objects, and are not limited to the order or precedence of similar objects. The description of variations such as "comprising," "having," etc., means that the subject of the word is not exclusive, except for the examples shown by the word.
It should be understood that the various numbers and steps described in this disclosure are for convenience of description and are not to be construed as limiting the scope of the application. The magnitude of the present application reference numerals does not mean the order of execution, and the order of execution of the processes should be determined by their functions and inherent logic.
In the following description, numerous details are set forth in order to provide a more thorough explanation of embodiments of the present application, it will be apparent, however, to one skilled in the art that embodiments of the present application may be practiced without these specific details, in other embodiments, well-known structures and devices are shown in block diagram form, rather than in detail, in order to avoid obscuring the embodiments of the present application.
Embodiments of the present application respectively propose a mold bonding breakout prediction method, a mold bonding breakout prediction apparatus, an electronic device, and a storage medium, and these embodiments will be described in detail below.
Firstly, the continuous casting process mainly comprises the steps of casting refined molten steel into steel billets through a continuous casting machine, wherein the main process comprises the following steps of: and (3) conveying the ladle filled with refined molten steel to a rotary table, after the rotary table rotates to a pouring position, pouring the molten steel into a tundish, and distributing the molten steel into each crystallizer through a water gap by the tundish. The mold is one of the core devices of the continuous casting machine, the mold can form castings and solidify and crystallize rapidly, molten steel is injected from above the mold, and the molten steel begins to solidify inside the mold to form a blank shell through the cooling effect of the mold. As the billet moves down, the shell gradually thickens until the desired thickness is reached. The casting blank is pulled out of the crystallizer by the withdrawal and straightening machine and enters a secondary cooling zone to be further cooled and solidified, and the design and operation of the crystallizer have important influences on the quality of the casting blank and the production efficiency of the continuous casting machine. The mold should have good heat conductivity, wear resistance and structural rigidity to ensure uniform solidification of the cast slab and to prevent the occurrence of problems such as steel leakage. However, in actual production, under the influence of factors such as poor lubrication effect of continuous casting mold flux, overlarge fluctuation of the liquid level of a crystallizer or improper control of superheat degree of molten steel, a casting blank is easy to bond with a copper plate of the crystallizer, if related measures are not timely taken, a blank shell is extremely easy to be torn continuously and solidified for the second time under the actions of vibration and blank pulling of the crystallizer to form a new bonding point, and when the bonding point moves down to an outlet of the crystallizer, the blank shell is broken to form steel leakage due to insufficient strength of the bonding point to resist internal hydrostatic pressure and blank pulling resistance. The phenomenon is bonding breakout, is a main breakout form in the continuous casting process, and the bonding breakout not only can lead to the shutdown of a continuous casting machine, thereby causing production interruption, reducing the production efficiency of steel, but also is accompanied by certain energy waste, safety risk and environmental pollution.
Referring to fig. 1, fig. 1 is a flowchart illustrating a method for predicting mold sticking leak according to an exemplary embodiment of the present application. As shown in fig. 1, in an exemplary embodiment, the method for predicting the mold adhesion leak includes at least steps S110 to S140, which are described in detail as follows:
Step S110, obtaining a thermocouple temperature original time sequence and a pull speed time sequence corresponding to each original sample, wherein the original samples comprise bonding samples of casting blanks and a crystallizer, and non-bonding samples of the casting blanks and the crystallizer.
In one embodiment of the application, the drawing speed is the speed of a continuous casting machine drawing roller, the non-bonding sample of the casting blank and the crystallizer comprises a sample with greatly-fluctuated thermocouple temperature time sequence, a sample with irregular and small-fluctuated temperature time sequence and a sample normally poured in a period of time before the bonding example, in the embodiment, each sample comprises a first heat rejection couple temperature original time sequence and a second heat rejection couple temperature original time sequence, each sample further comprises a drawing speed time sequence, the obtained first heat rejection couple temperature original time sequence is P= { P 0,p1,…,pn }, the obtained second heat rejection couple temperature original time sequence is Q= { Q 0,q1,…,qn }, and the drawing speed time sequence is S= { S 0,s1,…,sn }. Specifically, a total of 744 original samples are obtained, a pull-rate time sequence corresponding to each original sample is collected, a pull-rate time sequence S= { S 0,s1,…,s59 } is obtained, a first heat extraction thermocouple temperature original time sequence { p 0,p1,…,p59 } and a second heat extraction thermocouple temperature original time sequence { q 0,q1,…,q59 } corresponding to each original sample are collected, the data collection time is recorded as { t 0,t1,…,t59 }, and the time interval is 1 second unequal. As an example, as shown in fig. 2, fig. 2 is a timing chart of temperature inversion and typical bonding characteristics of a single-column thermocouple of a bonding sample, which is shown in the present application, in this embodiment, by extracting the single-column thermocouple temperature and the pull-rate timing in the bonding sample. As an example, as shown in fig. 3, fig. 3 is a timing chart of temperature inversion and M-type temperature change of a single thermocouple of a bonding sample, which is shown in the present application, in this embodiment, by extracting the single thermocouple temperature and the pull time in the bonding sample, the timing chart of temperature inversion and M-type temperature change characteristics of the single thermocouple of the bonding sample is obtained.
As an example, as shown in fig. 4, fig. 4 is a timing chart of temperature and pull speed of a single thermocouple of a bonding sample, in this embodiment, by extracting the temperature and pull speed of the single thermocouple in the bonding sample, a timing chart of temperature measurement of the single thermocouple of the bonding sample showing a larger temperature difference between an upper row and a middle-lower row and a characteristic of approximately parallel descent of the upper-middle row is obtained.
And step S120, obtaining a characteristic matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the characteristic matrix to obtain a crystallizer bonding steel leakage prediction model.
In one embodiment of the application, the model training is performed through the feature matrix, so that the most relevant and representative data can be extracted from the original data, the feature matrix is convenient for machine algorithm batch processing data by converting the original data into a structured format, and the model can be more easily debugged and optimized through the feature matrix. When the model performance is poor, the problem can be identified by checking the feature matrix and the model parameters, and corresponding adjustment is performed.
Step S130, inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in the preset distribution range extracted in advance or in real time and the pulling speed time sequence corresponding to the temperature time sequence into a crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and generating a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period.
In one embodiment of the present application, a thermocouple is an important temperature measuring device, typically used to measure the temperature of molten steel. In the continuous casting process, the temperature of molten steel is a key technological parameter, and the temperature of molten steel can directly influence the quality of casting blanks and the production efficiency of continuous casting machines. Therefore, the temperature of the molten steel needs to be monitored and controlled in real time. The working principle of the thermocouple is based on thermoelectric effect, namely free electrons in a conductor or a semiconductor can be influenced by heat energy to generate potential difference when the temperature changes, the temperature value can be determined by measuring the potential difference, and the temperature change of the thermocouple is the most visual and representative typical characteristic when the bonding phenomenon occurs because the temperature data of the thermocouple is relatively easy to monitor and accurately measure, so the application detects the bonding phenomenon of the crystallizer through the real-time temperature measurement data of the thermocouple.
And step S140, judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence and a bonding point transverse transfer characteristic recognition criterion.
In one embodiment of the application, compared with the traditional method for detecting the bonding phenomenon based on the threshold value, the method predicts the bonding probability of the crystallizer through a deep learning model, is more accurate and reliable, because the classification of bonding samples and non-bonding samples is not necessarily a simple linear separable problem, and the prediction mode based on the deep learning is beneficial to mining hidden features which are contained in the thermocouple temperature time sequence and are beneficial to the identification of the bonding phenomenon, meanwhile, the artificial error and the time cost which are introduced by continuously debugging the threshold value parameter according to the site live are reduced, and the change trend of the bonding risk can be better known by generating a plurality of thermocouple bonding probability prediction value sequences. The method is favorable for making a more reasonable maintenance plan, reducing unnecessary maintenance activities, and simultaneously ensuring that proper measures are taken when the bonding risk is high, thereby reducing the times and time of production interruption, reducing the equipment failure rate, further improving the stability and reliability of the production line and further improving the production efficiency.
In the technical scheme shown in fig. 1, the original thermocouple temperature time sequence and the pull-up time sequence corresponding to each sample are obtained from bonding samples and non-bonding samples of a casting blank and a crystallizer; obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, performing model training through the feature matrix to obtain a crystallizer bonding breakout prediction model, inputting any two rows of thermocouple temperature time sequences to be detected in each row and any two rows of thermocouple temperature time sequences corresponding to the pull speed time sequences in a preset distribution range into the crystallizer bonding breakout prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a plurality of thermocouple bonding probability predicted value sequences based on the plurality of single-row thermocouple bonding probability predicted values; and judging whether a bonding phenomenon exists or not by combining the single-column thermocouple bonding probability predicted value in each sequence with a bonding point transverse transmission characteristic recognition criterion. Obtaining single-column thermocouple bonding probability predicted values through a neural network model, comprehensively considering the time sequence of the single-column thermocouple bonding probability predicted values and the transverse transmission characteristics of bonding points, being applicable to a scene that a certain heat removal thermocouple in a plurality of rows of thermocouples is disabled, being beneficial to reducing the missed judgment rate of bonding phenomenon, reducing the number of thermocouple installation rows and saving the cost; compared with the traditional logic threshold judgment method, the method can not only realize more accurate and timely identification of adhesion phenomena such as multi-type temperature change characteristics, multi-production site and the like under multi-application scenes, avoid shutdown of a continuous casting machine due to bonding steel leakage, improve the production efficiency of steel, avoid energy waste, safety risk and environmental pollution caused by bonding steel leakage, and help to reduce human errors and time cost without continuously debugging threshold parameters according to site live. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
In one embodiment of the present application, obtaining the feature matrix according to the pull rate sequence and the thermocouple temperature original sequence includes: filtering abnormal values in the original thermocouple temperature time sequence to obtain a thermocouple temperature time sequence; performing smooth noise reduction treatment on the thermocouple temperature time sequence by utilizing a Bezier curve to obtain a thermocouple temperature smooth noise reduction time sequence; and obtaining a characteristic value according to the pull speed time sequence and the thermocouple temperature smooth noise reduction time sequence, and constructing a characteristic matrix through the characteristic value. By eliminating abnormal values in the original thermocouple temperature time sequence, the accuracy of acquired data can be ensured, the Bezier curve is utilized to carry out smooth noise reduction treatment on the thermocouple temperature time sequence with abnormal values eliminated, more available temperature characteristics are provided while the influence of noise on the acquired data is reduced, the accurate calculation of characteristic values is facilitated, the miss judgment rate and the misjudgment rate of a model on a bonding phenomenon are reduced, and meanwhile the problem of filling of the missing values is solved.
In one embodiment of the application, the thermocouple temperature raw timing includes an up heat rejection thermocouple temperature raw timing and a down heat rejection thermocouple temperature raw timing. The thermocouple temperature sequence includes an up heat rejection thermocouple temperature sequence and a down heat rejection thermocouple temperature sequence. The thermocouple temperature smooth noise reduction time sequence comprises an upper heat extraction thermocouple temperature smooth noise reduction time sequence and a lower heat extraction thermocouple temperature smooth noise reduction time sequence. The original time sequence of the upper heat extraction couple temperature is P= { P 0,p1,…,pn }, the original time sequence of the lower heat extraction couple temperature is Q= { Q 0,q1,…,qn }, and the pull-up time sequence is S= { S 0,s1,…,sn }. Filtering abnormal values in the original time sequence P of the upper heat extraction thermocouple temperature and the original time sequence Q of the lower heat extraction thermocouple temperature to obtain an upper heat extraction thermocouple temperature time sequence X and a lower heat extraction thermocouple temperature time sequence Y, and performing smooth noise reduction treatment on the upper heat extraction thermocouple temperature time sequence X and the lower heat extraction thermocouple temperature time sequence Y through a Bezier curve to obtain an upper heat extraction thermocouple temperature smooth noise reduction time sequence U and a lower heat extraction thermocouple temperature smooth noise reduction time sequence V. In the process of collecting thermocouple temperature, abnormal values are likely to occur due to equipment delay and other reasons, the accuracy of collected data can be ensured by eliminating the abnormal values in the original time sequence of the thermocouple temperature, and in the process of collecting thermocouple temperature, the abnormal values are also eliminated by interference of various noises such as electromagnetic interference, mechanical vibration and the like, so that the Bezier curve is utilized to carry out smooth noise reduction treatment on the time sequence of the thermocouple temperature after eliminating the abnormal values, more available temperature characteristics are provided while the influence of the noise on the collected data is reduced, the accurate calculation of the characteristic values is facilitated, the missing judgment rate and the misjudgment rate of a bonding phenomenon of a model are reduced, and meanwhile the problem of filling the missing values is solved.
In one embodiment of the present application, the upper heat rejection couple temperature raw timing is p= { P 0,p1,…,p59 }, the lower heat rejection couple temperature raw timing is q= { Q 0,q1,…,q59 }, and the pull rate timing is s= { S 0,s1,…,s59 }. The outlier P 47 in the original timing P of the upper thermocouple temperature is filtered out, and the resulting timing is not limited to x= { X 0,x1,…,x58 }. The outlier Q 36 in the original timing Q of the heat rejection thermocouple temperature is filtered out, and the resulting timing is not limited to y= { Y 0,y1,…,y58 }. On the premise of not causing ambiguity, the data acquisition time corresponding to the thermocouple temperature time sequence after abnormal values are removed is denoted by { w 0,w1,…,w58 }. And carrying out smooth noise reduction on the upper heat extraction thermocouple temperature time sequence X and the lower heat extraction thermocouple temperature time sequence Y through a Bezier curve to obtain an upper heat extraction thermocouple temperature smooth noise reduction time sequence U= { U 0,u1,…,u59 } and a lower heat extraction thermocouple temperature smooth noise reduction time sequence V= { V 0,v1,…,v59 }. In the process of collecting thermocouple temperature, abnormal values are likely to occur due to equipment delay and other reasons, the accuracy of collected data can be ensured by eliminating the abnormal values in the original time sequence of the thermocouple temperature, and in the process of collecting thermocouple temperature, the abnormal values are also eliminated by interference of various noises such as electromagnetic interference, mechanical vibration and the like, so that the Bezier curve is utilized to carry out smooth noise reduction treatment on the time sequence of the thermocouple temperature after eliminating the abnormal values, more available temperature characteristics are provided while the influence of the noise on the collected data is reduced, the accurate calculation of the characteristic values is facilitated, the missing judgment rate and the misjudgment rate of a bonding phenomenon of a model are reduced, and meanwhile the problem of filling the missing values is solved.
In one embodiment of the present application, the smoothing noise reduction processing for the upper heat rejection couple temperature time sequence X by using the bezier curve specifically includes:
wherein B x (k) is the smoothing noise reduction processing result of the upper heat extraction thermocouple temperature time sequence X, For the combination number, x is the temperature time sequence of the upper heat extraction thermocouple, i is the acquisition time of the vertical axis, and the value of the parameter k in the formula is continuous, so the value number and the specific value of k can be flexibly set according to different requirements. In this embodiment, the smoothing noise reduction processing for the heat rejection couple temperature timing Y by the bezier curve specifically includes:
wherein B y (k) is the smoothing noise reduction processing result of the heat discharging couple temperature time sequence Y, For the combination number, y is the temperature time sequence of the heat discharging couple, i is the acquisition time of the vertical axis, and the value of the parameter k in the formula is continuous, so the value number and the specific value of k can be flexibly set according to different requirements. In this embodiment, the number of values of the parameter k is set to 60, and specific assignment is performed according to the following criteria, that is, the temperature smooth noise reduction time sequence and the original time sequence are ensured to be consistent in the time dimension.
As one example, in the pull-up timing S, the upper heat extraction thermocouple temperature smoothing noise reduction timing U, and the lower heat extraction thermocouple temperature smoothing noise reduction timing V, characteristic values related to the thermocouple temperature and the pull-up timing are extracted. The characteristic values specifically include: the upper row smooth noise reduction temperature value, the upper row temperature slope, the upper row temperature directional curvature, the lower row temperature slope, the lower row temperature directional curvature, the difference between the upper row and lower row smooth noise reduction temperature values, the pull rate, the upper row temperature rising duration, the upper row temperature rising amplitude, the ratio of the upper row temperature rising amplitude to the upper rising value, the upper row temperature falling duration, the upper row temperature falling amplitude, the lower row temperature rising duration, the lower row temperature rising amplitude, the ratio of the lower row temperature rising amplitude to the upper rising value, the lower row temperature falling duration, the lower row temperature falling amplitude, the longitudinal time lag, the ratio of the bonding longitudinal propagation rate to the average pull rate.
Specifically, characteristic values related to thermocouple temperature and pull rate are extracted by, first, obtaining an upper-row temperature gradient by the following formula:
wherein, B' X (k) is the upper row temperature slope, For the combination number, x is the temperature time sequence of the upper heat extraction thermocouple, i is the acquisition time of the vertical axis, the value number of the parameter k is set to 60, and the specific assignment is carried out according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
the lower row temperature slope is obtained by the following formula:
Wherein, B' y (k) is the gradient of the lower row temperature, For the number of combinations, y is the temperature time sequence of the heat removal couple, i is the acquisition time of the vertical axis, the value number of the parameter k is set to 60, and the specific assignment is carried out according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
The upper row temperature directional curvature is obtained by the following formula,
Wherein K x (K) is the directional curvature of the upper row temperature, w is the data acquisition time sequence corresponding to the thermocouple temperature time sequence after abnormal values are removed, B ' x (K) is the first-order derivative function of the upper row temperature slope B ' x (K) on the parameter K, B ' w (K) is the first-order derivative function of the Bezier curve obtained based on the control point w on the parameter K,Is the square term of B' w (k),For the square term of B 'x (k), B' w (k) is a second derivative function of the Bezier curve obtained based on the control point w with respect to the parameter k, the value number of the parameter k is set to be 60, and the specific assignment is performed according to the following criteria, namely, the temperature time sequence and the original time sequence are ensured to be consistent in the value in the time dimension;
the downward temperature directional curvature is obtained by the following formula,
Wherein K y (K) is the directional curvature of the lower row temperature, w is the data acquisition time sequence corresponding to the thermocouple temperature time sequence after abnormal values are removed, B ' y (K) is the first-order derivative function of the lower row temperature slope B ' y (K) on the parameter K, B ' w (K) is the first-order derivative function of the Bezier curve obtained based on the control point w on the parameter K,Is the square term of B' w (k),For the square term of B 'x (k), B' w (k) is a second derivative function of the Bezier curve obtained based on the control point w with respect to the parameter k, the value number of the parameter k is set to be 60, and the specific assignment is performed according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
Specifically, B' w (k) is calculated by the following formula,
Wherein B' w (k) is a first order derivative function of the Bezier curve obtained based on the control point w with respect to the parameter k, i is the acquisition moment of the vertical axis,For the combination number, w is a data acquisition time sequence corresponding to the thermocouple temperature time sequence after the abnormal value, the value number of the parameter k is set to be 60, and the specific assignment is carried out according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
b "w (k) is calculated by the following formula,
Wherein B' w (k) is a second derivative function of the Bezier curve obtained based on the control point w with respect to the parameter k,For the combination number, i is the acquisition time of a vertical axis, w is the data acquisition time sequence corresponding to the thermocouple temperature time sequence after the abnormal value, the value number of the parameter k is set to be 60, and the specific assignment is carried out according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
B "x (k) is calculated by the following formula,
Wherein B 'x (k) is a first order derivative of the upper row temperature slope B' x (k) with respect to the parameter k,For the combination number, i is the acquisition time of a vertical axis, w is the data acquisition time sequence corresponding to the thermocouple temperature time sequence after the abnormal value, the value number of the parameter k is set to be 60, and the specific assignment is carried out according to the following criteria, namely, the temperature smooth noise reduction time sequence is ensured to be consistent with the value of the original time sequence in the time dimension;
calculating the difference between the upper and lower row smooth noise reduction temperature values by the following formula:
U-V＝ui-vi,i＝0,1,…,59,
Wherein U is the upper heat rejection couple temperature smoothing noise reduction time sequence, V is the lower heat rejection couple temperature smoothing noise reduction time sequence, U i is the temperature value at the ith moment in the upper heat rejection couple temperature smoothing noise reduction time sequence, and V i is the temperature value at the ith moment in the lower heat rejection couple temperature smoothing noise reduction time sequence;
Determining the difference between the corresponding time of the upper row temperature peak value and the starting time of the upper row temperature rising as the upper row temperature rising duration time; determining the difference between the corresponding time of the lower row temperature peak value and the starting time of the rising of the lower row temperature as the rising duration time of the lower row temperature;
Determining a difference between the upper row temperature peak value and an upper row temperature rising value as an upper row temperature rising amplitude; determining a difference between a lower row temperature peak value and a lower row temperature rising value as a lower row temperature rising amplitude;
Dividing the upper row temperature rise amplitude by an upper row temperature rise value to obtain a ratio of the upper row temperature rise amplitude to the upper row temperature rise value; dividing the lower row temperature rise amplitude by the lower row temperature rise value to obtain the ratio of the lower row temperature rise amplitude to the upper rise value;
Determining the difference between the corresponding time of the peak value of the upper temperature at t 59 and the peak value of the upper temperature as the duration of the falling of the upper temperature; determining the difference between the corresponding time of the peak value of the temperature of the lower row and t 59 as the falling duration time of the temperature of the lower row;
Obtaining the temperature drop amplitude of the upper exhaust line through the difference value between the temperature peak value of the upper exhaust line and the temperature measurement value of the upper exhaust heat couple t 59; obtaining the descending amplitude of the lower exhaust temperature according to the difference between the peak value of the lower exhaust temperature and the temperature measurement value of the lower exhaust thermocouple t 59;
The longitudinal time lag is obtained through the difference value between the rising starting time of the lower row temperature and the rising starting time of the upper row temperature;
The ratio of the bonding longitudinal propagation rate to the average pull rate time sequence is obtained through longitudinal time lag, the pull rate average value from the rising starting time of the upper row temperature to the rising starting time of the lower row temperature and the interval between the upper heat extraction couple and the lower heat extraction couple, specifically, the ratio of the interval between the upper heat extraction couple and the lower heat extraction couple to the product of the longitudinal time lag and the pull rate average value from the rising starting time of the upper row temperature to the rising starting time of the lower row temperature is determined as the ratio of the bonding longitudinal propagation rate to the average pull rate; in this embodiment, as shown in fig. 5, fig. 5 is a diagram showing a characteristic set of a mold sticking and breakout prediction model, and in the pull-rate sequence S, the upper heat rejection couple temperature smooth noise reduction sequence U, and the lower heat rejection couple temperature smooth noise reduction sequence V, characteristic values related to thermocouple temperatures and pull-rate sequences are obtained by the calculation methods listed above, and a diagram showing a characteristic set of a mold sticking and breakout prediction model is obtained. As an example, all the calculated eigenvalues are built into a matrix of 19×60 to obtain an eigenvalue matrix, which is used as an input in a subsequent initial mold bonding breakout prediction model to train the initial mold bonding breakout prediction model to obtain a mold bonding breakout prediction model. By organizing the characteristic values into the characteristic matrix form, a structured input is provided for the model, so that the model can learn the influence of more characteristics on the accurate identification of the crystallizer bonding phenomenon, and the accuracy of model prediction is improved.
In one embodiment of the application, model training is performed through a feature matrix to obtain a crystallizer bonding breakout prediction model, which comprises the following steps: setting a label value of the bonding sample to a first preset value, and setting a label value of the non-bonding sample to a second preset value; combining the first preset value and the second preset value, and carrying out layered sampling on the feature matrix based on a preset data dividing proportion to obtain a training set, a verification set and a test set, wherein the training set and the verification set are used for model training, and the test set is used for testing a trained crystallizer bonding steel leakage prediction model; fitting the initial crystallizer bonding breakout prediction model with a training set, and adjusting the initial crystallizer bonding breakout prediction model through a verification set to obtain the crystallizer bonding breakout prediction model. Through clear label definition, data layering sampling and reasonable data set division, the accuracy, stability and generalization capability of the crystallizer bonding steel leakage prediction model are improved. The bonding probability prediction value is obtained through the crystallizer bonding breakout prediction model, whether the bonding phenomenon exists or not is judged according to the single-column thermocouple bonding probability prediction value in each sequence, the bonding phenomenon can be accurately and timely identified, the continuous casting machine is prevented from being stopped due to bonding breakout, the production efficiency of steel is improved, the energy waste, the safety risk and the environmental pollution caused by bonding breakout are avoided, the threshold value parameters are not required to be continuously debugged according to site conditions, and the reduction of human errors and the time cost is facilitated.
As an example, the first preset value is 1, the second preset value is 0, the label value of each sample is determined to be a first preset value "1" according to whether bonding occurs, and the label value of the non-bonded sample is set to be a second preset value "0". The feature matrix extracted in the embodiment is layered sampled according to the ratio of 7:2:1 based on the label value of each sample to obtain a training set, a verification set and a test set, wherein the training set and the verification set are used for model training, and the test set is used for testing a trained crystallizer bonding steel leakage prediction model; fitting the initial crystallizer bonding breakout prediction model with a training set, and adjusting the initial crystallizer bonding breakout prediction model through a verification set to obtain the crystallizer bonding breakout prediction model. Through clear label definition, data layering sampling and reasonable data set division, the accuracy, stability and generalization capability of the crystallizer bonding steel leakage prediction model are improved. The bonding probability prediction value is obtained through the crystallizer bonding breakout prediction model, whether the bonding phenomenon exists or not is judged according to the single-column thermocouple bonding probability prediction value in each sequence, the bonding phenomenon can be accurately and timely identified, the continuous casting machine is prevented from being stopped due to bonding breakout, the production efficiency of steel is improved, the energy waste, the safety risk and the environmental pollution caused by bonding breakout are avoided, the threshold value parameters are not required to be continuously debugged according to site conditions, and the reduction of human errors and the time cost is facilitated.
In one embodiment of the present application, the initial crystallizer bonding breakout prediction model is a long short-term memory neural network model (LSTM), and the crystallizer bonding breakout prediction model is an LSTM crystallizer bonding breakout prediction model. The loss function adopts binary cross entropy, and an Adam algorithm is selected as an optimizer for model training. Inputting the training set and the verification set into the LSTM model, and optimizing the super parameters contained in the model based on GPGO. Finally, the super-parameter value result of the bonding breakout prediction model of the LSTM crystallizer is trained as follows: the number of memory cells is set to 512; simultaneously regularization of L1 and L2 is used, and weight coefficients of the regularization are set to be 0.01; the learning rate was set to 0.001; setting the iteration number epoch to 200; the batch size was set to 128. The trained LSTM crystallizer bonding breakout prediction model was used to predict the bonding probability of the test set, with time periods of 5 to 35 milliseconds for each predicted sample varying, centered around 15 milliseconds. For the bonding samples, the predicted bonding probability values are all greater than 0.98; for non-bonded samples, the bonding probability predictions were all less than 0.6. It can be seen that for this test set, the LSTM crystallizer bonding bleed-out prediction model constructed in this example can completely distinguish between bonding and non-bonding samples.
In one embodiment of the application, any two rows of thermocouple temperature time sequences to be detected in each row in a preset distribution range extracted in advance or in real time and corresponding pulling speed time sequences are input into a crystallizer bonding breakout prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and corresponding single-row thermocouple bonding probability predicted value sequences are generated for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period, wherein the method comprises the steps of acquiring any two rows of thermocouple temperature time sequences of part or all rows in a continuous casting machine in the preset time as any two rows of thermocouple temperature time sequences to be detected in each row in the preset distribution range; inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in a preset distribution range and the pulling speed time sequence corresponding to the temperature time sequence into a crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability prediction values; forming a single-column thermocouple bonding probability predicted value sequence by using single-column thermocouple bonding probability predicted values corresponding to each moment in preset time to obtain a plurality of single-column thermocouple bonding probability predicted value sequences; and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion. Through monitoring and analysis of the single-column thermocouple bonding probability predicted value time sequence, the risk of bonding steel leakage can be found in advance, and the characteristics of bonding phenomena can be reflected more comprehensively by comprehensively considering the plurality of single-column thermocouple bonding probability predicted value time sequences and the bonding point transverse transmission characteristics, so that the occurrence of the bonding phenomena can be judged more accurately.
In one embodiment of the application, combining the single column thermocouple junction probability predictions in each sequence with the criteria for cross-transfer characteristic identification with the junction points to determine whether a junction phenomenon exists comprises: if all the single-column thermocouple bonding probability predicted values in the at least one single-column thermocouple bonding probability predicted value sequence are larger than the first preset probability, marking the thermocouple column corresponding to the single-column thermocouple bonding probability predicted value sequence at the latest moment as a third preset value; if the thermocouple row corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple row corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon. By comprehensively considering the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, so that the occurrence of the bonding phenomenon can be more accurately judged.
In one embodiment of the application, the LSTM crystallizer bonding bleed-out prediction model trained based on historical production data of steel mill a is included for real-time detection of steel mill B bonding phenomena. 107 production records of steel mill B were collected, each covering a time frame of 5 minutes. For each record, the LSTM crystallizer bonding steel leakage prediction model obtained by training is adopted to detect bonding phenomena in real time on the combination of the upper row, the middle row, the lower row and the middle row of thermocouples in each row, the detection of bonding phenomena in different combinations is mutually independent, wherein the upper, the middle and the lower heat removal thermocouples respectively correspond to the first, the second and the third heat removal thermocouples. As an example, the thermocouple temperature sequence to be detected is an upper heat rejection thermocouple temperature sequence and a lower heat rejection thermocouple temperature sequence, the upper heat rejection thermocouple temperature sequence and the lower heat rejection thermocouple temperature sequence are detected in real time through an LSTM crystallizer bonding leak steel prediction model, for each row of thermocouples, a single row of bonding probability prediction values are output, in this embodiment, based on the thermocouple bonding probability prediction values with continuous values in three time dimensions, each row of thermocouple bonding probability prediction values output at the first moment are marked as P1, each row of thermocouple bonding probability prediction values output at the second moment are marked as P2, each row of thermocouple bonding probability prediction values output at the third moment are marked as P3, in this embodiment, the third preset value is 1, the first preset probability is 0.965, if the thermocouple bonding probability prediction value sequences with continuous values in the three time dimensions are all larger than the first preset probability "0.965", the bonding probability prediction value sequence P3 at the latest moment is marked as a third preset value "1", if the thermocouple string corresponding to the third preset value "1" corresponds to the third preset value "and the third preset value" corresponds to the third preset value "and the bonding probability" corresponds to the third preset value, and the cross-phase junction "the bonding phenomenon is not determined to exist, and the cross-domain bonding phenomenon is determined if the thermocouple bonding phenomenon is not determined. By combining the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the accuracy of judging the bonding phenomenon can be improved. The time sequence data can reveal the time change trend of the bonding phenomenon, and the transverse transmission characteristic of the bonding point reflects the spatial expansion condition of the bonding phenomenon. By comprehensively considering the time sequence data of the predicted values of the bonding probability of the thermocouples in a plurality of single columns and the transverse transmission characteristics of the bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, and therefore the occurrence of the bonding phenomenon can be more accurately judged. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
In one embodiment of the present application, determining whether a sticking phenomenon exists by combining a single-column thermocouple sticking probability prediction value in each sequence with a sticking point transverse transfer characteristic identification criterion comprises: obtaining a predicted value of the bonding probability of each row of thermocouples at continuous moments to obtain a predicted value of the bonding probability of a single row of thermocouples at a first moment, a predicted value of the bonding probability of a single row of thermocouples at a second moment and a predicted value of the bonding probability of a single row of thermocouples at a third moment, which correspond to each row of thermocouples; if the predicted value of the bonding probability of the single-column thermocouple at the first moment and the predicted value of the bonding probability of the single-column thermocouple at the second moment are both larger than the second preset probability, and the predicted value of the bonding probability of the single-column thermocouple at the third moment is larger than the third preset probability, marking the thermocouple column corresponding to the predicted value of the bonding probability of the single-column thermocouple at the third moment by the third preset value; if the thermocouple row corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple row corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon. By combining the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the accuracy of judging the bonding phenomenon can be improved. The time sequence data can reveal the time change trend of the bonding phenomenon, and the transverse transmission characteristic of the bonding point reflects the spatial expansion condition of the bonding phenomenon. By comprehensively considering the time sequence data of the predicted values of the bonding probability of the thermocouples in a plurality of single columns and the transverse transmission characteristics of the bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, and therefore the occurrence of the bonding phenomenon can be more accurately judged.
In one embodiment of the application, an LSTM crystallizer bonding bleed-out prediction model trained based on historical production data of steel mill a is included for real-time detection of steel mill B bonding phenomena. 107 production records of steel mill B were collected, each covering a time frame of 5 minutes. For each record, the LSTM crystallizer bonding steel leakage prediction model obtained by training is adopted to detect bonding phenomena in real time on the combination of the upper row, the middle row, the lower row and the middle row of thermocouples in each row, the detection of bonding phenomena in different combinations is mutually independent, wherein the upper, the middle and the lower heat removal thermocouples respectively correspond to the first, the second and the third heat removal thermocouples. As an example, the thermocouple temperature timing to be detected is an upper heat rejection thermocouple temperature timing and a lower heat rejection thermocouple temperature timing, the thermocouple temperature timing to be detected, "upper heat rejection thermocouple temperature timing and lower heat rejection thermocouple temperature timing" is detected in real time by an LSTM crystallizer bonding leak prediction model, a single column bonding probability prediction value is output for each column of thermocouples, in this embodiment, based on thermocouple bonding probability prediction values that are continuous in three time dimensions, each column of thermocouple bonding probability prediction values output at the first time is denoted as P1, each column of thermocouple bonding probability prediction values output at the second time is denoted as P2, each column of thermocouple bonding probability prediction values output at the third time is denoted as P3, in this embodiment, the third preset value is 1, the second preset probability is 0.95, the third preset probability is 0.98, if the thermocouple bonding probability predicted value sequence P1 is greater than the second preset probability of 0.95, the thermocouple bonding probability predicted value sequence P2 is greater than the second preset probability of 0.95, the thermocouple bonding probability predicted value sequence P3 is greater than the third preset probability of 0.98, the thermocouple bonding probability predicted value sequence P3 at the latest moment is marked as a third preset value of 1, if the thermocouple column corresponding to the third preset value of 1 accords with the bonding point transverse transmission characteristic, the existence of bonding phenomenon of the crystallizer is determined, alarm information is sent out, and if the thermocouple column corresponding to the third preset value of 1 does not accord with the bonding point transverse transmission characteristic, the existence of bonding phenomenon of the crystallizer is determined. By combining the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the accuracy of judging the bonding phenomenon can be improved. The time sequence data can reveal the time change trend of the bonding phenomenon, and the transverse transmission characteristic of the bonding point reflects the spatial expansion condition of the bonding phenomenon. By comprehensively considering the time sequence data of the predicted values of the bonding probability of the thermocouples in a plurality of single columns and the transverse transmission characteristics of the bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, and therefore the occurrence of the bonding phenomenon can be more accurately judged. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
In one embodiment of the application, determining that a bonding phenomenon exists if the marking value meets the bonding point transverse transmission characteristic comprises: extracting a thermocouple column corresponding to a third preset value, and obtaining the transverse propagation velocity of the bonding point based on the time difference of the thermocouple column corresponding to the third preset value and the corresponding pull speed average value; obtaining detection time through the transverse propagation rate of the bonding point and the arrangement interval between thermocouple columns corresponding to a third preset value; if the number of the thermocouple columns corresponding to the third preset value in the detection time is larger than or equal to the fourth preset value, judging that the crystallizer has a bonding phenomenon, and if the number of the thermocouple columns corresponding to the third preset value in the detection time is smaller than the fourth preset value, judging that the crystallizer has no bonding phenomenon. By combining the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the accuracy of judging the bonding phenomenon can be improved. The time sequence data can reveal the time change trend of the bonding phenomenon, and the transverse transmission characteristic of the bonding point reflects the spatial expansion condition of the bonding phenomenon. By comprehensively considering the time sequence data of the predicted values of the bonding probability of the thermocouples in a plurality of single columns and the transverse transmission characteristics of the bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, and therefore the occurrence of the bonding phenomenon can be more accurately judged.
In one embodiment of the application, the lateral transfer characteristics of the bond point refer to the expansion and transfer characteristics of this bond phenomenon in the width direction of the mold (i.e. laterally) when bonding occurs between the molten steel and the copper wall of the mold within the mold. Such transfer characteristics may be affected by various factors such as the flow of molten steel, the temperature distribution of the copper walls of the mold, mold vibration, etc. The bond point cross transfer characteristics can be generally analyzed and judged by observing and monitoring predicted values of thermocouple bond probabilities at different locations within the mold. When a thermocouple at a certain position detects that the bonding probability exceeds a certain threshold value, the bonding phenomenon may gradually expand along the width direction of the crystallizer, namely, the bonding point is transferred transversely.
In one embodiment of the application, the fourth preset value is 2. In this embodiment, thermocouple columns with a mark value of "1" are extracted, two thermocouple columns with a mark value of "1" are extracted, a pull speed average value Q corresponding to a time difference of the thermocouple columns with the mark value of "1" is calculated, a bonding transverse transmission rate v=0.9×q, a detection time is obtained through bonding transverse transmission rate and an arrangement interval between the thermocouple columns corresponding to the third preset value of "1", if the number of the thermocouple columns marked as the third preset value of "1" is greater than or equal to a fourth preset value of "2" in the three adjacent thermocouples detected in the detection time, that is, the mark value of at least two thermocouple columns is "1", bonding exists in the crystallizer is determined, and if the number of the thermocouple columns marked as the third preset value of "1" is smaller than the fourth preset value of "2", bonding does not exist in the crystallizer. By combining the time sequence of the predicted values of the bonding probability of a plurality of single-column thermocouples and the transverse transmission characteristics of bonding points, the accuracy of judging the bonding phenomenon can be improved. The time sequence data can reveal the time change trend of the bonding phenomenon, and the transverse transmission characteristic of the bonding point reflects the spatial expansion condition of the bonding phenomenon. By comprehensively considering the time sequence data of the predicted values of the bonding probability of the thermocouples in a plurality of single columns and the transverse transmission characteristics of the bonding points, the characteristics of the bonding phenomenon can be more comprehensively reflected, and therefore the occurrence of the bonding phenomenon can be more accurately judged. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
Referring to fig. 6, fig. 6 is a flowchart illustrating a method for predicting mold sticking leak according to an exemplary embodiment of the present application. As shown in fig. 6, in an exemplary embodiment, the method for predicting the mold adhesion leak includes at least steps S601 to S610, which are described in detail as follows:
step S601, obtaining a thermocouple temperature original time sequence and a pull rate time sequence corresponding to each original sample, wherein the original samples comprise a bonding sample of a casting blank and a crystallizer and a non-bonding sample of the casting blank and the crystallizer, and then executing step S602.
Step S602, filtering abnormal values in the original thermocouple temperature time sequence to obtain a thermocouple temperature time sequence, performing smooth noise reduction processing on the thermocouple temperature time sequence by utilizing a Bezier curve to obtain a thermocouple temperature smooth noise reduction time sequence, and then executing step S603.
Step S603, obtaining characteristic values according to the pull-speed time sequence and the thermocouple temperature smooth noise reduction time sequence, constructing a characteristic matrix through the characteristic values, and then executing step S603.
Step S603, the label value of the bonded sample is set to a first preset value, the label value of the non-bonded sample is set to a second preset value, and then step S604 is performed.
Step S604, combining the first preset value and the second preset value, sampling the feature matrix layer by layer based on the preset data dividing proportion to obtain a training set, a verification set and a test set, performing model training to obtain a crystallizer bonding steel leakage prediction model, and then executing step S605.
Step S605, any two-row thermocouple temperature time sequence of part or all columns in the continuous casting machine is obtained and used as any two-row thermocouple temperature time sequence to be detected of each column in a preset distribution range, any two-row thermocouple temperature time sequence to be detected of each column in the preset distribution range and a pulling speed time sequence corresponding to the thermocouple temperature time sequence are input into a crystallizer bonding steel leakage prediction model to obtain a plurality of single-column thermocouple bonding probability prediction values, and then step S606 is executed.
Step S606, obtaining the predicted value of the bonding probability of each row of thermocouples at continuous moments, obtaining the predicted value of the bonding probability of the thermocouples corresponding to each row of thermocouples at a first moment, the predicted value of the bonding probability of the thermocouples at a second moment and the predicted value of the bonding probability of the thermocouples at a third moment, and then executing step S607.
In step S607, if the predicted value of the bonding probability of the thermocouple in the first time single row and the predicted value of the bonding probability of the thermocouple in the second time single row are both greater than the second preset probability, and the predicted value of the bonding probability of the thermocouple in the single row in the third time single row is greater than the third preset probability, the thermocouple row corresponding to the predicted value of the bonding probability of the thermocouple in the single row in the third time single row is marked by the third preset value, and then step S608 is executed.
Step S608, determining whether the thermocouple row corresponding to the third preset value accords with the bonding point transverse transmission feature, if the thermocouple row corresponding to the third preset value accords with the bonding point transverse transmission feature, executing step S609, and if the thermocouple row corresponding to the third preset value does not accord with the bonding point transverse transmission feature, executing step S610.
Step S609, determining that the crystallizer has a bonding phenomenon, sending out alarm information, and ending.
In step S610, it is determined that the mold is free from the sticking phenomenon, and then the process is ended.
In the technical scheme shown in fig. 6, the original thermocouple temperature time sequence and the pull-up time sequence corresponding to each sample are obtained from bonding samples and non-bonding samples of a casting blank and a crystallizer; obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature smooth noise reduction time sequence, performing model training through the feature matrix to obtain a crystallizer bonding breakout prediction model, inputting any two rows of thermocouple temperature time sequences to be detected in each row and any two rows of thermocouple temperature time sequences corresponding to the thermocouple temperature time sequences to be detected in a preset distribution range into the crystallizer bonding breakout prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming corresponding single-row thermocouple bonding probability predicted value sequences for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period; and judging whether a bonding phenomenon exists or not by combining the single-column thermocouple bonding probability predicted value in each sequence with a bonding point transverse transmission characteristic recognition criterion. Obtaining single-column thermocouple bonding probability predicted values through a neural network model, comprehensively considering the time sequence of the single-column thermocouple bonding probability predicted values and the transverse transmission characteristics of bonding points, being applicable to a scene that a certain heat removal thermocouple in a plurality of rows of thermocouples is disabled, being beneficial to reducing the missed judgment rate of bonding phenomenon, reducing the number of thermocouple installation rows and saving the cost; compared with the traditional logic threshold judgment method, the method can not only realize more accurate and timely identification of adhesion phenomena such as multi-type temperature change characteristics, multi-production site and the like under multi-application scenes, avoid shutdown of a continuous casting machine due to bonding steel leakage, improve the production efficiency of steel, avoid energy waste, safety risk and environmental pollution caused by bonding steel leakage, and help to reduce human errors and time cost without continuously debugging threshold parameters according to site live. On the other hand, by monitoring and analyzing the time sequence of the predicted value of the bonding probability of the single-column thermocouple, the risk of bonding steel leakage can be found in advance. By combining a plurality of single-column thermocouple bonding probability predicted value sequences and bonding point transverse transmission characteristic recognition criteria, the reasonable design of alarm signals can be realized, and the triggering of the alarm signals can prompt operators to take preventive measures in time or directly drop the pulling speed to stop so as to avoid bonding steel leakage.
In one embodiment of the present application, fig. 7 is a graph showing a comparison of marked time points of the bonding phenomenon according to an exemplary embodiment of the present application, and the real-time detection of the bonding phenomenon of the collected test sample of the steel mill B by using the LSTM crystallizer bonding breakout prediction model trained based on the historical production data of the steel mill a results in zero missing judgment and zero misjudgment. Further, for three very representative bonding samples shown in fig. 3-5, the marking time of the bonding phenomenon by the LSTM crystallizer bonding breakout prediction model and the logic threshold judgment rule used in steel mill B is shown in fig. 7. The following information is available from the analysis of fig. 7: on one hand, compared with the traditional logic threshold judgment method, the LSTM crystallizer bonding steel leakage prediction model is more timely in identifying bonding phenomenon; on the other hand, although the LSTM crystallizer bonding steel leakage prediction model is trained by the temperature information of the first heat rejection couple and the second heat rejection couple, the temperature time sequence of the two heat rejection couples is arbitrarily taken, and the bonding phenomenon can be accurately and timely judged. The test result of the LSTM crystallizer bonding steel leakage prediction model in the steel mill B is comprehensively compared with the recognition comparison of the logic threshold judgment rule to the bonding phenomenon, and compared with the traditional logic threshold judgment rule, the crystallizer bonding steel leakage prediction method based on deep learning can recognize the bonding phenomenon more accurately and efficiently.
Referring to fig. 8, fig. 8 is a block diagram illustrating a mold sticking leak prediction apparatus according to an exemplary embodiment of the present application.
As shown in fig. 8, the exemplary mold sticking leak prediction apparatus includes: a timing acquisition module 810, a model training module 820, a timing detection module 830, and a bond determination module 840. The time sequence acquisition module is used for acquiring the original time sequence of the thermocouple temperature and the pull-speed time sequence corresponding to each original sample; the original sample comprises a bonding sample of a casting blank and a crystallizer copper plate, and a non-bonding sample of the casting blank and the crystallizer copper plate; the model training module is used for obtaining a characteristic matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and carrying out model training through the characteristic matrix to obtain a crystallizer bonding steel leakage prediction model; the time sequence detection module is used for inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time and the pulling speed time sequence corresponding to the temperature time sequence into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period; and the bonding judgment module is used for judging whether the crystallizer has bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transmission characteristic recognition criterion. Acquiring thermocouple temperature original noise time sequence and pull rate time sequence corresponding to each sample from bonding samples and non-bonding samples of casting blanks and a crystallizer; obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, performing model training through the feature matrix to obtain a crystallizer bonding breakout prediction model, inputting any two rows of thermocouple temperature time sequences to be detected in each row and any two rows of thermocouple temperature time sequences corresponding to the thermocouple temperature time sequences to be detected in a preset distribution range into the crystallizer bonding breakout prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection of a preset time period; and judging whether a bonding phenomenon exists or not by combining the single-column thermocouple bonding probability predicted value in each sequence with a bonding point transverse transmission characteristic recognition criterion. The predicted value of the bonding probability of the single-row thermocouple is obtained through the neural network model, whether the bonding phenomenon exists or not is judged by combining the predicted value of the bonding probability of the single-row thermocouple in each sequence with the transverse transmission characteristic recognition criterion of the bonding point, the bonding phenomenon can be accurately and timely recognized, the continuous casting machine is prevented from being stopped due to bonding steel leakage, the production efficiency of steel is improved, the energy waste, the safety risk and the environmental pollution caused by bonding steel leakage are avoided, the threshold parameters are not required to be continuously debugged according to site situations, and the method is beneficial to reducing human errors and time cost.
In an exemplary embodiment of the present invention, the timing detection module is configured to obtain, in a preset time, any two rows of thermocouple temperature timings of part or all columns in the continuous casting machine, as any two rows of thermocouple temperature timings to be detected in each column in a preset distribution range; inputting the temperature time sequence of any two rows of thermocouples to be detected in each row within a preset distribution range into a crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability prediction values; forming a single-column thermocouple bonding probability predicted value sequence by using single-column thermocouple bonding probability predicted values corresponding to each moment in preset time to obtain a plurality of single-column thermocouple bonding probability predicted value sequences; and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion.
In an exemplary embodiment of the present invention, the model training module is configured to filter abnormal values in the original thermocouple temperature time sequence to obtain a thermocouple temperature time sequence; performing smooth noise reduction treatment on the thermocouple temperature time sequence by utilizing a Bezier curve to obtain a thermocouple temperature smooth noise reduction time sequence; and obtaining a characteristic value according to the pull speed time sequence and the thermocouple temperature smooth noise reduction time sequence, and constructing a characteristic matrix through the characteristic value.
In an exemplary embodiment of the invention, the model training module sets the label value of the bonded sample to a first preset value and the label value of the non-bonded sample to a second preset value; combining the first preset value and the second preset value, and carrying out layered sampling on the feature matrix based on a preset data dividing proportion to obtain a training set, a verification set and a test set, wherein the training set and the verification set are used for model training, and the test set is used for testing a trained crystallizer bonding steel leakage prediction model; fitting the initial crystallizer bonding breakout prediction model with a training set, and adjusting the initial crystallizer bonding breakout prediction model through a verification set to obtain the crystallizer bonding breakout prediction model.
In an exemplary embodiment of the present invention, the adhesion determination module is configured to mark, as a third preset value, a thermocouple column corresponding to the single-column thermocouple adhesion probability prediction value sequence at the latest moment if all the single-column thermocouple adhesion probability predictions in the at least one single-column thermocouple adhesion probability prediction value sequence are greater than the first preset probability; if the thermocouple row corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple row corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon.
In an exemplary embodiment of the present invention, the adhesion determination module is further configured to extract a thermocouple column corresponding to a third preset value, and obtain a lateral propagation rate of the adhesion point based on a time difference of the thermocouple column corresponding to the third preset value and a pull speed average value corresponding to the time difference; obtaining detection time through the transverse propagation rate of the bonding point and the arrangement interval between thermocouple columns corresponding to a third preset value; if the number of the thermocouple columns corresponding to the third preset value in the detection time is larger than or equal to the fourth preset value, judging that the crystallizer has a bonding phenomenon, and if the number of the thermocouple columns corresponding to the third preset value in the detection time is smaller than the fourth preset value, judging that the crystallizer has no bonding phenomenon.
The embodiment also provides an electronic device, including: one or more processors; and the storage device is used for storing one or more programs, and when the one or more programs are executed by the one or more processors, the electronic equipment realizes the crystallizer bonding steel leakage forecasting method provided in each embodiment.
Referring to fig. 9, fig. 9 is a schematic structural diagram of an electronic device according to an exemplary embodiment of the present application. It should be noted that, the electronic device 900 shown in fig. 9 is only an example, and should not impose any limitation on the functions and the application scope of the embodiments of the present application.
As shown in fig. 9, the electronic device 900 includes a processor 901, a memory 902, and a communication bus 903; a communication bus 903 is used to connect the processor 901 and the memory connection 902; the processor 901 is configured to execute computer programs stored in the memory 902 to implement the methods of one or more of the embodiments described above.
The present embodiment also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor of a computer, causes the computer to perform the crystallizer bonding breakout prediction method as before. The computer-readable storage medium may be included in the electronic device described in the above embodiment or may exist alone without being incorporated in the electronic device. The present embodiments also provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer apparatus reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer apparatus performs the crystallizer bonding breakout prediction method provided in the above-described respective embodiments.
The electronic device provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface, where the memory and the communication interface are connected to the processor and the transceiver and perform communication therebetween, the memory is used to store a computer program, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer program, so that the electronic device performs each step of the above method.
In this embodiment, the memory may include a random access memory (Random Access Memory, abbreviated as RAM), and may further include a non-volatile memory (non-volatile memory), such as at least one disk memory.
The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, abbreviated as CPU), a network processor (Network Processor, abbreviated as NP), etc.; but may also be a digital signal processor (DIGITAL SIGNAL Processing, DSP), application Specific Integrated Circuit (ASIC), field-Programmable gate array (FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components.
The above embodiments are merely illustrative of the principles of the present application and its effectiveness, and are not intended to limit the application. Modifications and variations may be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the application. It is therefore intended that all equivalent modifications and changes made by those skilled in the art without departing from the spirit and technical spirit of the present application shall be covered by the appended claims.
Claims (10)
1. A method for predicting mold sticking steel leakage, the method comprising:
Acquiring a thermocouple temperature original time sequence and a pull rate time sequence corresponding to each original sample, wherein the original samples comprise bonding samples of casting blanks and a crystallizer and non-bonding samples of the casting blanks and the crystallizer;
Obtaining a feature matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the feature matrix to obtain a crystallizer bonding steel leakage prediction model;
Inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and the pulling speed time sequence corresponding to the temperature time sequence of any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and generating a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period;
and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion.
2. The crystallizer bonding breakout prediction method according to claim 1, wherein the temperature sequence of any two rows of thermocouples to be detected in each column and the pulling speed sequence corresponding to the temperature sequence of any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time are input into the crystallizer bonding breakout prediction model to obtain a plurality of single-column thermocouple bonding probability predicted values, and the single-column thermocouple bonding probability predicted value sequences corresponding to the single-column thermocouple bonding probability predicted values are generated based on continuous detection of a preset time period, and the method comprises the steps of:
Acquiring any two rows of thermocouple temperature time sequences of part or all columns in the continuous casting machine within preset time, and taking the thermocouple temperature time sequences as any two rows of thermocouple temperature time sequences to be detected of each column within the preset distribution range;
Inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in the preset distribution range and the pulling speed time sequence corresponding to the temperature time sequence into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability prediction values;
Forming a single-column thermocouple bonding probability predicted value sequence by using the single-column thermocouple bonding probability predicted value corresponding to each moment in the preset time to obtain a plurality of single-column thermocouple bonding probability predicted value sequences;
and judging whether the crystallizer has a bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transfer characteristic identification criterion.
3. The method for predicting bonding steel leakage of a crystallizer according to claim 1, wherein obtaining a feature matrix according to the pull rate time sequence and the thermocouple temperature original time sequence comprises:
filtering abnormal values in the original thermocouple temperature time sequence to obtain a thermocouple temperature time sequence;
Performing smooth noise reduction treatment on the thermocouple temperature time sequence by utilizing a Bezier curve to obtain a thermocouple temperature smooth noise reduction time sequence;
and obtaining a characteristic value according to the pull-speed time sequence and the thermocouple temperature smooth noise reduction time sequence, and constructing a characteristic matrix through the characteristic value.
4. The method for predicting the bonding steel leakage of the crystallizer according to claim 1, wherein model training is performed through the feature matrix to obtain a model for predicting the bonding steel leakage of the crystallizer, and the method comprises the following steps:
setting the label value of the bonding sample to a first preset value, and setting the label value of the non-bonding sample to a second preset value;
Combining the first preset value and the second preset value, and carrying out layered sampling on the feature matrix based on a preset data dividing proportion to obtain a training set, a verification set and a test set, wherein the training set and the verification set are used for model training, and the test set is used for testing a trained crystallizer bonding steel leakage prediction model;
Fitting the initial crystallizer bonding breakout prediction model with the training set, and adjusting the initial crystallizer bonding breakout prediction model through the verification set to obtain the crystallizer bonding breakout prediction model.
5. The crystallizer bonding breakout prediction method according to claim 1, wherein determining whether a bonding phenomenon exists by combining a single-column thermocouple bonding probability prediction value in each sequence with a bonding point lateral transfer characteristic recognition criterion comprises:
If all the single-column thermocouple bonding probability predicted values in the at least one single-column thermocouple bonding probability predicted value sequence are larger than the first preset probability, marking the thermocouple column corresponding to the single-column thermocouple bonding probability predicted value sequence at the latest moment as a third preset value;
If the thermocouple column corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple column corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon.
6. The crystallizer bonding breakout prediction method according to claim 1, wherein determining whether a bonding phenomenon exists by combining a single-column thermocouple bonding probability prediction value in each sequence with a bonding point lateral transfer characteristic recognition criterion comprises:
Obtaining a predicted value of the bonding probability of each row of thermocouples at continuous moments to obtain a predicted value of the bonding probability of a single row of thermocouples at a first moment, a predicted value of the bonding probability of a single row of thermocouples at a second moment and a predicted value of the bonding probability of a single row of thermocouples at a third moment, which correspond to each row of thermocouples;
If the predicted value of the bonding probability of the single-column thermocouple at the first moment and the predicted value of the bonding probability of the single-column thermocouple at the second moment are both larger than the second preset probability, and the predicted value of the bonding probability of the single-column thermocouple at the third moment is larger than the third preset probability, marking the thermocouple column corresponding to the predicted value of the bonding probability of the single-column thermocouple at the third moment by the third preset value;
If the thermocouple column corresponding to the third preset value accords with the transverse transmission characteristic of the bonding point, determining that the crystallizer has bonding phenomenon, and sending out alarm information, and if the thermocouple column corresponding to the third preset value does not accord with the transverse transmission characteristic of the bonding point, determining that the crystallizer does not have bonding phenomenon.
7. The method for predicting sticking leak of a mold according to claim 5 or 6, wherein determining that a sticking phenomenon exists if the marking value meets a sticking point lateral transfer characteristic comprises:
extracting a thermocouple column corresponding to a third preset value, and obtaining the transverse propagation velocity of the bonding point based on the time difference of the thermocouple column corresponding to the third preset value and the corresponding pulling speed average value;
obtaining detection time through the transverse propagation rate of the bonding point and the arrangement interval between thermocouple columns corresponding to the third preset value;
If the number of the thermocouple columns corresponding to the third preset value in the detection time is larger than or equal to a fourth preset value, judging that the crystallizer has a bonding phenomenon, and if the number of the thermocouple columns corresponding to the third preset value in the detection time is smaller than the fourth preset value, judging that the crystallizer has no bonding phenomenon.
8. A crystallizer bonding breakout prediction apparatus, the apparatus comprising:
The time sequence acquisition module is used for acquiring a thermocouple temperature original time sequence and a pull-speed time sequence corresponding to each original sample; the original sample comprises a bonding sample of a casting blank and a crystallizer copper plate, and a non-bonding sample of the casting blank and the crystallizer copper plate;
the model training module is used for obtaining a characteristic matrix according to the pull speed time sequence and the thermocouple temperature original time sequence, and performing model training through the characteristic matrix to obtain a crystallizer bonding steel leakage prediction model;
the time sequence detection module is used for inputting the temperature time sequence of any two rows of thermocouples to be detected in each row and any two rows of thermocouples to be detected in a preset distribution range extracted in advance or in real time and the pulling speed time sequence corresponding to the temperature time sequence into the crystallizer bonding steel leakage prediction model to obtain a plurality of single-row thermocouple bonding probability predicted values, and forming a corresponding single-row thermocouple bonding probability predicted value sequence for the plurality of single-row thermocouple bonding probability predicted values based on continuous detection in a preset time period;
And the bonding judgment module is used for judging whether the crystallizer has bonding phenomenon or not by combining the single-column thermocouple bonding probability predicted value in each single-column thermocouple bonding probability predicted value sequence with a bonding point transverse transmission characteristic recognition criterion.
9. An electronic device, the electronic device comprising:
One or more processors;
Storage means for storing one or more programs that, when executed by the one or more processors, cause the electronic device to implement the crystallizer bonding breakout prediction method of any one of claims 1 to 7.
10. A computer readable medium, having stored thereon a computer program which, when executed by a processor of a computer, causes the computer to perform the crystallizer bonding breakout prediction method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410454500.6A CN118321519A (en) | 2024-04-16 | 2024-04-16 | Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410454500.6A CN118321519A (en) | 2024-04-16 | 2024-04-16 | Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118321519A true CN118321519A (en) | 2024-07-12 |
Family
ID=91779097
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410454500.6A Pending CN118321519A (en) | 2024-04-16 | 2024-04-16 | Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118321519A (en) |
-
2024
- 2024-04-16 CN CN202410454500.6A patent/CN118321519A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11105758B2 (en) | Prediction method for mold breakout based on feature vectors and hierarchical clustering | |
| CN105328155B (en) | A kind of bleed-out visual feature forecasting procedure based on improvement neutral net | |
| CN101332499B (en) | Slab continuous-casting bleedout forecast control method | |
| CN102151814B (en) | Bonding alarm method and system during continuous casting production | |
| CN102886504B (en) | Crystallizer bleed-out visual forecasting method based on machine vision | |
| CN109365769A (en) | A kind of crystallizer bleedout prediction electric thermo method based on mixed model judgement | |
| US4556099A (en) | Abnormality detection and type discrimination in continuous casting operations | |
| JP6358215B2 (en) | Method and apparatus for determining surface defects of continuous cast slab, and method for manufacturing steel slab using the surface defect determination method | |
| CN108705058B (en) | A method of forecast Crystallizer bleed-out is clustered based on K-Means | |
| CN102343427A (en) | Breakout prediction method for slab continuous casting mold based on withdrawal resistance | |
| CN103639385A (en) | Bleed-out forecasting method and system based on least squares | |
| Tirian et al. | Adaptive control system for continuous steel casting based on neural networks and fuzzy logic | |
| US9709515B2 (en) | Device and method for diagnosing cracks in a solidified shell in a mold | |
| Duan et al. | Integrated approach to density-based spatial clustering of applications with noise and dynamic time warping for breakout prediction in slab continuous casting | |
| Kumar et al. | Development of intelligent mould for online detection of defects in steel billets | |
| CN108469313B (en) | Copper plate of crystallizer temperature anomaly method for detecting area based on cellular automata | |
| CN118321519A (en) | Crystallizer bonding steel leakage prediction method and device, electronic equipment and medium | |
| CN111761040B (en) | Method and device for online identifying reverse sequence of thermocouple installation of crystallizer breakout prediction system | |
| CN115446276B (en) | Continuous casting steel leakage early warning method based on convolutional neural network recognition crystallizer copper plate V-shaped bonding characteristics | |
| CN111421119A (en) | Online prediction method for longitudinal cracks on surface of continuous casting slab | |
| JP2015160239A (en) | Method and apparatus for surface defect determination for continuously cast slab | |
| CN114653914A (en) | Crystallizer bleed-out early warning method based on morphological reconstruction and electronic device | |
| CN113935416A (en) | Continuous casting breakout prediction method for multi-state crystallizer | |
| JPH09248661A (en) | Method for predicting breakout caused by foreign matter biten in continuous casting | |
| JP6358199B2 (en) | Method and apparatus for determining surface defects of continuous cast slab, and method for producing steel slab using the surface defect determination method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |