CN117191394B - Fault detection method and device for air compressor equipment - Google Patents
Fault detection method and device for air compressor equipment Download PDFInfo
- Publication number
- CN117191394B CN117191394B CN202311460443.4A CN202311460443A CN117191394B CN 117191394 B CN117191394 B CN 117191394B CN 202311460443 A CN202311460443 A CN 202311460443A CN 117191394 B CN117191394 B CN 117191394B
- Authority
- CN
- China
- Prior art keywords
- bearing
- displacement signal
- time
- data
- period
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 19
- 238000006073 displacement reaction Methods 0.000 claims abstract description 74
- 238000000034 method Methods 0.000 claims abstract description 54
- 230000005484 gravity Effects 0.000 claims abstract description 35
- 230000036541 health Effects 0.000 claims abstract description 13
- 238000012423 maintenance Methods 0.000 claims abstract description 8
- 238000004364 calculation method Methods 0.000 claims description 29
- 239000013598 vector Substances 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 10
- 238000009434 installation Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 230000001427 coherent effect Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 17
- 230000008859 change Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a fault detection method and device for air compressor equipment, wherein the method comprises the following steps: acquiring a displacement signal of a bearing of the air compressor to obtain the current running state of the bearing; selecting a corresponding parameter equation according to the fault type of the bearing, and obtaining an axis track by utilizing the displacement signal of the bearing and the selected parameter equation; calculating the average Euclidean distance between the axis locus in each real data period and the center of gravity of each axis locus; if the average Euclidean distance is larger than the threshold value, stopping maintenance, otherwise predicting the futuremAverage Euclidean distance between the center of gravity of the axis track and the center of gravity of each axis track in each real data period; obtaining the expected failure time based on the position of the predicted average Euclidean distance greater than the threshold value, otherwise the air compressor equipment is restarted under the working conditionmThe real data cycle time status is normal. The invention can quantitatively analyze the faults of the air compressor and give out the specific health condition of the equipment in a future period of time.
Description
Technical Field
The invention belongs to the technical field of air compressors, and particularly relates to a fault detection method and device for air compressor equipment.
Background
The rotor in the air compressor is an important component for ensuring the effective operation of the air compressor, and the axial track diagram is widely studied as a way for intuitively showing the motion mode of the air compressor. The axial locus of the rotor consists of a group of mutually perpendicular displacement signals in the same section, and an axial locus diagram at a certain moment is obtained by plotting a plurality of pairs of groups of signals at the moment. The image can only reflect the running condition of the rotor at a certain moment, but cannot reflect the continuous change condition of the movement of the rotor in a certain time period.
The mode of early acquisition of the axis locus is mainly a one-dimensional signal processing method, namely, displacement signals in the horizontal direction x and the vertical direction y at a certain moment are acquired, then the signals in the two directions are combined, and the axis locus of the rotor is drawn on a plane according to the combined signals. In an actual scene, the rotation of the rotor is dynamically changed, the axis track is also changed, and the change condition of the axis track of the rotor along with time is often not well shown only by displaying the axis track on a two-dimensional plane. Therefore, the method is improved later, the time dimension is increased in the axis track diagram, the two-dimensional image is converted into three-dimensional image, the change process of the axis track can be dynamically displayed, and the method has important significance for analysis and research of the working state of the rotor.
It is noted that most of the existing patents for researching the axis track are to research faults corresponding to the real-time axis track, and the faults possibly occurring in the equipment are not researched. For example, the publication number is CN201510629547, and the patent name is a method and device for detecting vibration faults of a centrifugal compressor, and the patent can combine a three-dimensional axis track and fault judgment, but the calculation mode is relatively complex when the type of the axis track is identified, the calculation amount is too large, and the calculation occupies a large amount of resources, so that the method and device are unfavorable for large-scale use. And the degree of equipment faults and the development trend of the faults can be quantitatively analyzed, so that a overhaul time point with reference significance is provided.
Disclosure of Invention
The invention aims to provide a fault detection method and device for air compressor equipment, which solve the technical problems that the prior art scheme cannot quantitatively analyze the degree of faults and the development trend of the faults and cannot give overhaul time points.
The technical scheme adopted by the invention is as follows:
a fault detection method of an air compressor device, comprising the steps of:
collecting air compressor bearing continuityHorizontal displacement signal->And a vertical displacement signalForming a bearing displacement signal data set +.>;
From bearing displacement signal datasetsObtaining the current running state of the bearing;
if the current running state of the bearing is a fault, selecting a corresponding parameter equation according to the fault type of the bearing, and obtaining an axle center track of the bearing by utilizing a displacement signal of the bearing and the selected parameter equation;
calculating the average Euclidean distance between the axis locus in each real data period and the center of gravity of each axis locusThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is->The average Euclidean distance between the axial track in the real data period and the center of gravity of the axial track;
calculation ofJudging->Whether or not it is greater than threshold +.>If->If the machine is stopped for maintenance, if->Then predict the futuremAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>;
If it isAccording to greater than threshold ∈>The location of (2) gets the expected time of failure if +.>The air compressor installation is then re-run under such conditionsmThe real data cycle time status is normal.
Further, if the current running state of the bearing is normal, the bearing displacement signal is continuously collected at intervals, and the current running state of the bearing is continuously judged.
Further, two displacement sensors which are arranged at the bearing position of the air compressor and have the same cross section and are in mutually perpendicular directions are used for simultaneously collecting horizontal displacement signals of the bearing within a certain time period at the same sampling frequencyAnd a vertical displacement signal>。
Further, the displacement sensor is an eddy current sensor, and the acquisition time period is more than 30 seconds.
Further, the operating conditions of the bearing include normal operation, mass imbalance, misalignment and oil film whirl, wherein mass imbalance, misalignment and oil film whirl are bearing failure conditions.
Further, the current running state of the bearing is obtained by using the running state identification model; the running state identification model is an SVM model, the input of the running state identification model is a bearing displacement signal, and the output of the running state identification model is the running state of the bearing, and the running state identification model comprises normal running, unbalanced mass, non-centering and oil film whirl.
Further, when the failure category of the bearing is misalignment or oil film whirl, the corresponding parameter equationThe method comprises the following steps:
when the failure category of the bearing is mass unbalance, the corresponding parameter equationThe method comprises the following steps:
in the method, in the process of the invention,for horizontal displacement signal>Is a vertical displacement signal>、/>、/>、/>、/>、/>、/>、/>、/>And->For parameters->For angular frequency +.>Time is;
the method for solving the axis locus of the bearing comprises the following steps:
determining displacement signal frequency;
Calculating a real data periodTThe real data period T is the average rotation period of the rotating shaft in the period of time;
calculation of initiationValue of->;
With real data periodTLength of (2) will bearing displacement signal data setDividing to obtain->Segment data->,/>Data continuity, ->The length is->;
Solving corresponding parameter equation by applying each segment of data;
The method for calculating the center of gravity of the axis track of each real data period comprises the following steps:
in the method, in the process of the invention,and->For barycentric coordinates>Is->And->Density at the location.
Further, the real data periodThe calculation method of (1) comprises the following steps:
the horizontal displacement signal value in a period of timeArranged in order from small to large in a sequence selected to be arranged at [ -degree ] in the sequence>]Data on location->The data value is recorded as threshold +.>;
Sequentially recording the valuesIs greater than threshold->The time of (2) to obtain a length of +.>Is to be treated in the time series->I.e. +.>The total number of data of (2) is->A plurality of;
for a pair ofPerforming first order difference to obtain a length of +.>Differential sequence of->According to this sequence a threshold value is set>;
Recording differential sequencesIs greater than threshold->Is->The position of the data, based on which the time sequence is +.>Is divided into->Groups, calculating the mean value of the time in each group, combining them into a time series +.>;
For the obtained time seriesAfter differential, a differential sequence is obtained>Taking the average value as;
Real data period。
Further, predicting future using real-time health modelmAverage Euclidean distance between axis track and center of gravity of each axis track in real data periodComprising:
before utilizationnAverage Euclidean distance between axis track and center of gravity of each axis track in real data periodPredicting the future by means of a Conformer algorithmmAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>The method comprises the steps of carrying out a first treatment on the surface of the The Conformer algorithm body is consistent with the transducer algorithm, and the difference is that:
1) Position coding is carried out on an input sequence, then position coding is added, namely, the fault state of the air compressor is carried out, one-hot coding is carried out on h working conditions, one corresponding state information is selected as the state coding at each moment, and the original input data, the position coding and the state coding are combined and then input into a model;
2) In the decoding module, the length isnIs a true data vector of (2)And a length ofmIs +.>Combined into a new vector of length m+nVector->Sum vector->Is coherent in time;
combining into new vectorscAfter that, a mask module mask is introduced when calculating the self-attention mechanism, and thentMasking the data at a later time, in particular a new vectorcThe t-th element of the list only carries out self-attention calculation on all data at the moment before t;
after the self-attention mechanism is calculated, the calculation mode of the subsequent part is consistent with that of a transducer model;
conformer model will outputmThe feature vectors are input into the full connection layer to obtain the futuremPredicted values for each time instant.
A fault detection device for an air compressor apparatus, comprising:
the data acquisition module is used for acquiring the bearing continuity of the air compressorHorizontal displacement signal->And a vertical displacement signal>Forming a bearing displacement signal data set +.>;
An operation state module for determining a bearing displacement signal data setObtaining the current running state of the bearing;
the axis track module is used for selecting a corresponding parameter equation according to the fault type of the bearing if the current running state of the bearing is a fault, and obtaining an axis track of the bearing by utilizing the displacement signal of the bearing and the selected parameter equation;
the Euclidean distance module is used for calculating the average Euclidean distance between the axis track in each real data period and the center of gravity of each axis trackThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is->The average Euclidean distance between the axial track in the real data period and the center of gravity of the axial track;
a judgment prediction module for calculatingJudging->Whether or not it is greater than threshold +.>If->If the machine is stopped for maintenance, if->Then predict the futuremAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>The method comprises the steps of carrying out a first treatment on the surface of the If->According to greater than threshold ∈>The location of (2) gets the expected time of failure if +.>The air compressor installation is then re-run under such conditionsmThe real data cycle time status is normal.
Compared with the prior art, the invention has the following advantages:
(1) The time dimension is increased, so that the change condition of the axis track in the time dimension can be better analyzed;
(2) The parameter equation is introduced, so that the calculation difficulty of the axis track is greatly reduced;
(3) Optimizing a method for solving a parameter equation, and reducing the calculated amount of the solved parameter equation;
(4) And constructing a real-time health degree model, quantitatively analyzing the faults of the air compressor, and giving out the specific health condition of the equipment in a future period of time.
Drawings
FIG. 1 is a flow chart of a fault detection method of an air compressor apparatus of the present invention;
FIG. 2 (a) is a graph of the threshold h 1 Dividing a data schematic;
FIG. 2 (b) shows the acquisition threshold h 2 A schematic diagram;
FIG. 2 (c) shows a calculated time sequenceA schematic diagram;
FIG. 2 (d) shows the calculation of the differential sequenceA schematic diagram;
FIG. 3 (a) shows the present inventionT 0 A time two-dimensional axis track condition diagram;
FIG. 3 (b) shows the present inventionT 1 A time two-dimensional axis track condition diagram;
FIG. 3 (c) shows the present inventionT 2 A time two-dimensional axis track condition diagram;
FIG. 3 (d) shows the present inventionT 3 A time two-dimensional axis track condition diagram;
FIG. 3 (e) shows the present inventionT 4 A time two-dimensional axis track condition diagram;
FIG. 3 (f) is the present inventionT 5 A time two-dimensional axis track condition diagram;
FIG. 3 (g) is the present inventionT 6 Time two-dimensional axisA trace situation map;
FIG. 4 is a schematic view of the present inventionTo->A three-dimensional axis trajectory graph in time;
FIG. 5 is a flow chart of a Conformer algorithm of the present invention;
FIG. 6 is a schematic diagram of data combination according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. In addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not collide with each other.
The invention provides a fault detection method and device for air compressor equipment. The method is simplified aiming at the complexity of the original method operation, and can efficiently identify the fault condition of the air compressor equipment; meanwhile, the method also realizes the dynamic detection of the air compressor equipment faults, and clear suggestions can be given to the overhauling time of the air compressor faults by constructing a health degree model. The invention aims to improve the fault recognition efficiency, optimize the overhaul mode and avoid production loss caused by unreasonable shutdown.
The fault detection method of the air compressor equipment of the invention, as shown in fig. 1, comprises the following steps:
step 1, at intervals, simultaneously acquiring horizontal displacement signals of a bearing in a certain time period at the same sampling frequency by two displacement sensors which are arranged at the same section of the bearing of the air compressor and are in mutually perpendicular directionsAnd a vertical displacement signal>Forming a bearing displacement signal data set +.>;
Step 2, transmitting the bearing displacement signal to an operation state identification model to obtain the current operation state of the bearing;
step 3, turning to step 1 when the state is normal, selecting a proper parameter equation according to the fault type only when the bearing state is fault, and obtaining the axis track of the bearing by using the displacement signal of the bearing and the selected parameter equation;
step 4, calculating Euclidean distance between the axis track and the center of gravity of the axis track in each real data period;
Step 5, calculatingJudging->Whether or not it is greater than threshold +.>If (if)Immediately stopping the machine for maintenance, if +.>Then give future according to the real-time health degree modelmTThe health of the machine over time. Wherein,Tis the real data period.
As an alternative embodiment, step 1, two eddy current sensors installed at the bearing position of the air compressor and in the same cross section and in mutually perpendicular directions synchronously collect the mutually perpendicular eddy current sensorsSuccessive bitsShift signal->Andthe time period of acquisition should not be too short, the recommended time is more than 30 seconds, the acquired signals are correspondingly formed into a data set +.>. There is no upper limit on the acquisition time; if the acquisition time is long, a period of time can be taken from the acquisition time.
As an alternative embodiment, the operation state recognition model mentioned in step 2 is obtained as follows:
(1) And two displacement sensors which are arranged at the same position of the bearing of the air compressor and in the mutually perpendicular directions of the same section are used for simultaneously acquiring a plurality of groups of horizontal displacement signals and vertical displacement signals of the bearing under different running states in a fixed time length at the same sampling frequency to form a typical running state data set. Typical operating conditions include normal operation, mass imbalance, misalignment, and oil film whirl.
(2) And transmitting the displacement signal into an SVM model for training to obtain an operation state identification model.
The signal characteristics are not required to be extracted, so that the operation steps and the difficulty are greatly reduced. The SVM is adopted as the classifier, so that the complexity of the model can be simplified, and the robustness and accuracy of the model are improved.
The data acquired in the step 1 are processedAnd (5) inputting the trained model, and outputting the result to be the current running state of the air compressor.
It should be noted that other classification models may be used herein.
In some embodiments of the present application, the following operation is performed in step 3 only when the operation state of the air compressor output in step 2 is a fault.
When the fault is not centering or oil film whirling, the selected parameter equation is:
when the fault is unbalanced, the axis track is always in an elliptical shape, so the standard parameter equation of the axis track can be simplified as follows:
in the method, in the process of the invention,for horizontal displacement signal>Is a vertical displacement signal>、/>、/>、/>、/>、/>、/>、/>、/>And->For parameters->For angular frequency +.>Is time.
As an optional implementation manner, the axis track is solved according to a parameter equation corresponding to the displacement signal and the fault, and the specific optimization solving method is as follows:
(1) Determining the frequency of the displacement signal;
(2) Calculating a real data periodTReal data periodTThe average rotation period of the rotating shaft in the period of time is the average rotation period of the rotating shaft in the period of time;
(2) Calculation of initiationValue of->;
(4) In cycles ofTLength of (2) will shift signal data setDividing to obtain->Segment data->,/>Data continuity, ->The length is->;
(5) Solving corresponding parameter equation by applying each segment of data。
As an alternative embodiment, the real data periodThe calculation method of (2) is as follows:
(1) The value of the horizontal displacement signal is selected for a longer period of time (30 seconds or more, and can be selected according to practical situations)Arranged in order from small to large in a sequence selected to be arranged at [ -degree ] in the sequence>]Position data->The data value is recorded as threshold +.>;
(2) Sequentially recording the valuesIs greater than threshold->The time of (2) to obtain a length of +.>Is to be treated in the time series->I.e. +.>The total number of data of (2) is->A plurality of;
(3) For a pair ofPerforming first order difference to obtain a length of +.>Differential sequence of->According to this sequence a threshold value is set>;
(4) Recording differential sequencesIs greater than threshold->Is->The position of the data, based on which the time sequence is +.>Is divided into->Groups, calculating the mean value of the time in each group, combining them into a time series +.>;
(5) Differentiating the time sequence obtained in step (4) to obtain a differential sequenceTaking the mean value of->;
(6) Will be the real period。
FIGS. 2 (a) to 2 (d) are sequential calculations of the real data periodThe abscissa is time and the ordinate is sensor signal value. Wherein FIG. 2 (a) is a graph of the acquisition and based on a threshold h 1 The segmentation data is schematically shown in FIG. 2 (b) for the acquisition threshold h 2 FIG. 2 (c) is a diagram showing the calculation of time series +.>FIG. 2 (d) is a diagram showing the calculation of the differential sequence +.>Schematic diagram.
It should be noted that, here, a vertical displacement signal may be used as well; when the coordinate system is established, if the horizontal direction is taken as the y axis, the horizontal displacement signal is the displacement signal in the vertical direction. In addition, the reference positions may be divided into a front group or a rear group at the time of grouping. For example, assuming that there are 10 pieces of data in total from 1 to 10, where the 5 th position is the reference position, when dividing into the front group, from 1 to 5 is one group, and from 6 to 10 is another group.
As an alternative implementation manner, the specific operation flow for obtaining the axle center track is as follows:
firstly drawing a three-dimensional axis trajectory graph, dividing the three-dimensional axis trajectory graph according to a real data period T to obtain a plurality of single-period three-dimensional axis trajectory graphs, projecting the three-dimensional axis trajectory graph of each single period to a two-dimensional plane, solving a parameter equation of each two-dimensional plane axis trajectory, solving the gravity center of each two-dimensional axis trajectory graph by using the parameter equation, and calculating the gravity centers and corresponding axis trajectory dataAverage Euclidean distance>。
FIGS. 3 (a) to 3 (g) are in orderTo->In the case of two-dimensional axis locus in time, FIG. 4 is +.>To->Three-dimensional axial trace map (post-treatment integration case) in time.
As an alternative embodiment, step 4 applies the result of step 3Individual axis trajectory parameter equation->Calculate its corresponding center of gravity +.>. The specific calculation method is as follows:
if curve is formedThe parametric equation for (2) is:
the centroid is:
the density is set in this stepAnd if the center of gravity of the axis track is constant, obtaining the center of gravity coordinate of the axis track as follows:
re-calculating the data of each center of gravity and the corresponding axis trackAverage Euclidean distance>。
As an alternative embodiment, in step 5, the calculation is performedJudging->Whether or not it is greater than threshold +.>If->Immediately stopping the machine for maintenance, if +.>Then give future according to the real-time health degree modelmTThe health of the machine over time.
Because the axle center track can display the type of the fault and the equipment always develops from health to the fault along with the increase of time, the time is one of the key factors for judging the fault, so that the time factor is considered when the method for constructing the real-time health degree model of the equipment by utilizing the axle center track is adopted, and the specific construction method is as follows:
(1)taking the flat obtained in the step 4All European distance->;
(2) Based on the average Euclidean distance D from T to nT, as shown in FIG. 5, the future is predicted by Conformer (condition transformer) algorithmmTEuclidean distance of center of gravity corresponding to axis track of each period in time。
The Conformer algorithm body is consistent with the transducer, except that:
(1) The input sequence is subjected to position coding and then added with position coding, namely, the fault state of the air compressor (the identification result from the axis track), one-hot coding is carried out on h working conditions, one corresponding state information is selected as the state coding at each moment, and the original input data, the position coding and the state coding are combined and then input into a model, as shown in fig. 6. The improvement aims to add fault information into the model, so that extra working condition information which cannot be obtained from the quantity information is expanded, and the accuracy of the model for future prediction is improved;
(2) In the decoding module, the length isnIs a true data vector of (2)And a length ofmIs +.>(/>Initialization value of vector 0) into a new vector of length m+nVector->Sum vector->Is coherent in time. This is done in order to make the model output at one timemFuture predicted values, thereby speeding up the calculation; secondly, vector +.>Inputting the information as prior information to the decoder helps to improve the accuracy of the prediction of the unknown information.
Combined into vectorscThe mask module mask needs to be introduced later in the calculation of the self-attention mechanism. Because the model solves the time series problem and simultaneously predicts a plurality of values, the influence of the value at the moment t is only influenced by the data before the moment t, and therefore, the data at the moment t needs to be masked, and the specific operation is vectorcThe t-th element of (2) makes self-attention calculations only for all data at a time before t. The operation aims to avoid the problem of model overfitting caused by data leakage, and greatly improves the generalization capability of the model.
After the self-attention mechanism is calculated, the subsequent part is consistent with the calculation mode of the transducer model.
Conformer model will outputmThe feature vectors are input into the full connection layer to obtain the futuremPredicted values for each time instant.
Finally, ifThen can be according to the output->The position of the maximum value in (1) gets the expected failure time if +.>The air compressor installation is then re-run under such conditionsmTThe time status is healthy.
It should be noted that the number of the substrates,positions above the threshold value are the expected time of failure, and the position probability of the maximum valueMaximum, also more accurate.
The invention also provides a fault detection device for the air compressor equipment for realizing the method, which comprises the following steps:
the data acquisition module is used for acquiring the bearing continuity of the air compressorHorizontal displacement signal->And a vertical displacement signal>Forming a bearing displacement signal data set +.>;
An operation state module for determining a bearing displacement signal data setObtaining the current running state of the bearing;
the axis track module is used for selecting a corresponding parameter equation according to the fault type of the bearing if the current running state of the bearing is a fault, and obtaining an axis track of the bearing by utilizing the displacement signal of the bearing and the selected parameter equation;
the Euclidean distance module is used for calculating the average Euclidean distance between the axis track in each real data period and the center of gravity of each axis trackThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is->The average Euclidean distance between the axial track in the real data period and the center of gravity of the axial track;
a judgment prediction module for calculatingJudging->Whether or not it is greater than threshold +.>If->If the machine is stopped for maintenance, if->Then predict the futuremAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>The method comprises the steps of carrying out a first treatment on the surface of the If->According to greater than threshold ∈>The location of (2) gets the expected time of failure if +.>The air compressor installation is then re-run under such conditionsmThe real data cycle time status is normal.
In summary, the axial locus diagram is taken as the entry point, the axial locus is optimized and constructed, the axial locus can be obtained through a small amount of calculation, meanwhile, the axial locus data is fully utilized, a quantitative model of fault identification and equipment health degree is constructed, the defects of the existing method are overcome, and the utility of the axial locus data is improved.
It should be noted that, the sequence number of each step in the above embodiment does not mean the sequence of execution sequence, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.
It should be noted that each step/component described in the present application may be split into more steps/components, or two or more steps/components or part of the operations of the steps/components may be combined into new steps/components, as needed for implementation, to achieve the object of the present invention.
It will be readily appreciated by those skilled in the art that the foregoing is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (7)
1. A fault detection method for an air compressor apparatus, comprising the steps of:
collecting air compressor bearing continuityHorizontal displacement signal->And a vertical displacement signal>Forming a bearing displacement signal data set +.>;
From bearing displacement signal datasetsObtaining the current running state of the bearing; the running states of the bearing comprise normal running, unbalanced mass, non-centering and oil film whirl, wherein the unbalanced mass, the non-centering and the oil film whirl are bearing fault states;
if the current running state of the bearing is a fault, selecting a corresponding parameter equation according to the fault type of the bearing, and obtaining an axle center track of the bearing by utilizing a displacement signal of the bearing and the selected parameter equation; comprising the following steps:
when the failure category of the bearing is misalignment or oil film whirl, the corresponding parameter equationThe method comprises the following steps:
when the failure category of the bearing is mass unbalance, the corresponding parameter equationThe method comprises the following steps:
in the method, in the process of the invention,for horizontal displacement signal>Is a vertical displacement signal>、/>、/>、/>、/>、/>、/>、/>、/>And->For parameters->For angular frequency +.>Time is;
the method for solving the axis locus of the bearing comprises the following steps:
determining displacement signal frequency;
Calculating a real data periodTThe real data period T is the rotation axist 1 To the point oft k Average rotation period over time;
calculation of initiationValue of->;
With real data periodTLength of (2) will bearing displacement signal data setDividing to obtain->Segment data,/>Data continuity, ->The length is->;
Solving corresponding parameter equation by applying each segment of data;
The method for calculating the center of gravity of the axis track of each real data period comprises the following steps:
in the method, in the process of the invention,and->For barycentric coordinates>Is->Density at;
wherein the real data periodThe calculation method of (1) comprises the following steps:
the horizontal displacement signal value in a period of timeArranged in order from small to large in a sequence selected to be arranged at [ -degree ] in the sequence>]Data on location->The data value is recorded as threshold +.>;
Sequentially recording the valuesIs greater than threshold->The time of (2) to obtain a length of +.>Is to be processed in the time series of (a) is to be processedI.e. +.>The total number of data of (2) is->A plurality of;
for a pair ofPerforming first order difference to obtain a length of +.>Differential sequence of->Setting a threshold according to the sequence;
Recording differential sequencesIs greater than threshold->Is->The position of each data, based on the position, time seriesIs divided into->Groups, calculating the mean value of the time in each group, combining them into a time series +.>;
For the obtained time seriesAfter differential, a differential sequence is obtained>Taking the mean value of->;
Real data period;
Calculating the average Euclidean distance between the axis locus in each real data period and the center of gravity of each axis locusThe method comprises the steps of carrying out a first treatment on the surface of the Wherein,is->The average Euclidean distance between the axial track in the real data period and the center of gravity of the axial track;
calculation ofJudging->Whether or not it is greater than threshold +.>If->If the machine is stopped for maintenance, if->Then predict the futuremAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>;
If it isAccording to greater than threshold ∈>The location of (2) gets the expected time of failure if +.>The air compressor installation is then re-run under such conditionsmThe real data cycle time status is normal.
2. The fault detection method of an air compressor device according to claim 1, wherein if the current operation state of the bearing is normal, the bearing displacement signal is continuously collected at intervals, and the current operation state of the bearing is continuously judged.
3. The fault detection method of air compressor equipment according to claim 1, wherein the horizontal displacement signals of the bearings are simultaneously collected for a certain period of time at the same sampling frequency by two displacement sensors installed at the same cross section of the bearings of the air compressor in mutually perpendicular directionsAnd a vertical displacement signal>。
4. A fault detection method for an air compressor device according to claim 3, wherein the displacement sensor is an eddy current sensor and the period of acquisition is more than 30 seconds.
5. The fault detection method of an air compressor device according to claim 1, wherein a current operation state of the bearing is obtained using an operation state recognition model; the running state identification model is an SVM model, the input of the running state identification model is a bearing displacement signal, and the output of the running state identification model is the running state of the bearing, and the running state identification model comprises normal running, unbalanced mass, non-centering and oil film whirl.
6. The fault detection method of an air compressor device according to claim 1, characterized by usingReal-time health model prediction futuremAverage Euclidean distance between axis track and center of gravity of each axis track in real data periodComprising:
before utilizationnAverage Euclidean distance between axis track and center of gravity of each axis track in real data periodPredicting the future by means of a Conformer algorithmmAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>The method comprises the steps of carrying out a first treatment on the surface of the The Conformer algorithm body is consistent with the transducer algorithm, and the difference is that:
1) The method comprises the steps of performing position coding on an input sequence, adding state coding, namely, performing one-hot coding on h working conditions, selecting corresponding state information as the state coding at each moment, combining original input data, the position coding and the state coding, and inputting the combined state coding into a model;
2) In the decoding module, the length isnIs a true data vector of (2)And a length ofmIs +.>New vectors combined to length m+n +.>Vector->Sum vector->Is coherent in time;
combining into new vectorscAfter that, a mask module mask is introduced when calculating the self-attention mechanism, and thentMasking the data at a later time, in particular a new vectorcThe t-th element of the list only carries out self-attention calculation on all data at the moment before t;
after the self-attention mechanism is calculated, the calculation mode of the subsequent part is consistent with that of a transducer model;
conformer model will outputmThe feature vectors are input into the full connection layer to obtain the futuremPredicted values for each time instant.
7. A fault detection device for an air compressor apparatus, comprising:
the data acquisition module is used for acquiring the bearing continuity of the air compressorHorizontal displacement signal->And a vertical displacement signal>Forming a bearing displacement signal data set +.>;
An operation state module for determining a bearing displacement signal data setObtaining the current running state of the bearing; the running states of the bearing comprise normal running, unbalanced mass, non-centering and oil film whirl, wherein the unbalanced mass, the non-centering and the oil film whirl are bearing fault states;
the axis track module is used for selecting a corresponding parameter equation according to the fault type of the bearing if the current running state of the bearing is a fault, and obtaining an axis track of the bearing by utilizing the displacement signal of the bearing and the selected parameter equation; comprising the following steps:
when the failure category of the bearing is misalignment or oil film whirl, the corresponding parameter equationThe method comprises the following steps:
when the failure category of the bearing is mass unbalance, the corresponding parameter equationThe method comprises the following steps:
in the method, in the process of the invention,for horizontal displacement signal>Is a vertical displacement signal>、/>、/>、/>、/>、/>、/>、/>、/>And->For parameters->For angular frequency +.>Time is;
the method for solving the axis locus of the bearing comprises the following steps:
determining displacement signal frequency;
Calculating a real data periodTThe real data period T is the rotation axist 1 To the point oft k Average rotation period over time;
calculation of initiationValue of->;
With real data periodTLength of (2) will bearing displacement signal data setDividing to obtain->Segment data,/>Data continuity, ->The length is->;
Solving corresponding parameter equation by applying each segment of data;
The method for calculating the center of gravity of the axis track of each real data period comprises the following steps:
in the method, in the process of the invention,and->For barycentric coordinates>Is->Density at;
wherein the method comprises the steps ofReal data periodThe calculation method of (1) comprises the following steps:
the horizontal displacement signal value in a period of timeArranged in order from small to large in a sequence selected to be arranged at [ -degree ] in the sequence>]Data on location->The data value is recorded as threshold +.>;
Sequentially recording the valuesIs greater than threshold->The time of (2) to obtain a length of +.>Is to be processed in the time series of (a) is to be processedI.e. +.>The total number of data of (2) is->A plurality of;
for a pair ofPerforming first order difference to obtain a length of +.>Differential sequence of->Setting a threshold according to the sequence;
Recording differential sequencesIs greater than threshold->Is->The position of each data, based on the position, time seriesIs divided into->Groups, calculating the mean value of the time in each group, combining them into a time series +.>;
For the obtained time seriesAfter differential, a differential sequence is obtained>Taking the mean value of->;
Real data period;
The Euclidean distance module is used for calculating the average Euclidean distance between the axis track in each real data period and the center of gravity of each axis trackThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is->The average Euclidean distance between the axial track in the real data period and the center of gravity of the axial track;
a judgment prediction module for calculatingJudging->Whether or not it is greater than threshold +.>If (if)If the machine is stopped for maintenance, if->Then predict the futuremAverage Euclidean distance of center of gravity of axis track and respective axis track in each real data period +.>The method comprises the steps of carrying out a first treatment on the surface of the If->According to greater than threshold ∈>The location of (2) gets the expected time of failure if +.>The air compressor installation is then re-run under such conditionsmThe real data cycle time status is normal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311460443.4A CN117191394B (en) | 2023-11-06 | 2023-11-06 | Fault detection method and device for air compressor equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311460443.4A CN117191394B (en) | 2023-11-06 | 2023-11-06 | Fault detection method and device for air compressor equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117191394A CN117191394A (en) | 2023-12-08 |
CN117191394B true CN117191394B (en) | 2024-01-19 |
Family
ID=89003795
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311460443.4A Active CN117191394B (en) | 2023-11-06 | 2023-11-06 | Fault detection method and device for air compressor equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117191394B (en) |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR19990025593A (en) * | 1997-09-13 | 1999-04-06 | 이종규 | Partial discharge on-line diagnostic system |
JP2013164748A (en) * | 2012-02-10 | 2013-08-22 | Aisin Seiki Co Ltd | State estimation device, state estimation method and program |
CN107728178A (en) * | 2017-09-29 | 2018-02-23 | 上海斐讯数据通信技术有限公司 | A kind of processing method and processing device of movement locus |
CN108760268A (en) * | 2018-06-08 | 2018-11-06 | 济南大学 | A kind of Vertical Mill operation data both phase step fault diagnostic method based on comentropy |
CN110276416A (en) * | 2019-07-02 | 2019-09-24 | 广东省智能机器人研究院 | A kind of rolling bearing fault prediction technique |
CN111649921A (en) * | 2020-04-13 | 2020-09-11 | 南通大学 | Rotating machinery fault diagnosis test system and working method |
CN111950200A (en) * | 2020-08-11 | 2020-11-17 | 哈电发电设备国家工程研究中心有限公司 | Power generation equipment fault diagnosis method based on axle center track fusion |
CN112362238A (en) * | 2020-11-13 | 2021-02-12 | 中国第一汽车股份有限公司 | Gravity center measuring method |
WO2022139193A1 (en) * | 2020-12-21 | 2022-06-30 | 효성중공업 주식회사 | Method and device for diagnosing malfunction of rotating machine |
CN115355162A (en) * | 2022-08-08 | 2022-11-18 | 西安交通大学 | Diaphragm compressor diaphragm fault diagnosis method based on oil pressure |
CN115660198A (en) * | 2022-11-07 | 2023-01-31 | 中国科学院空间应用工程与技术中心 | Method for predicting residual service life of rolling bearing |
CN116226646A (en) * | 2023-05-05 | 2023-06-06 | 国家石油天然气管网集团有限公司 | Method, system, equipment and medium for predicting health state and residual life of bearing |
CN116465623A (en) * | 2023-05-10 | 2023-07-21 | 安徽大学 | Gearbox service life prediction method based on sparse converter |
CN116563711A (en) * | 2023-05-17 | 2023-08-08 | 大连民族大学 | Hyperspectral target detection method of binary encoder network based on momentum update |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10139311B2 (en) * | 2014-09-26 | 2018-11-27 | Palo Alto Research Center Incorporated | Computer-implemented method and system for machine tool damage assessment, prediction, and planning in manufacturing shop floor |
WO2017027855A1 (en) * | 2015-08-12 | 2017-02-16 | Massachusetts Institute Of Technology | Systems and methods for predicting adverse events and assessing level of sedation during medical procedures |
US10830747B2 (en) * | 2016-10-26 | 2020-11-10 | Northwestern University | System and method for predicting fatigue strength of alloys |
-
2023
- 2023-11-06 CN CN202311460443.4A patent/CN117191394B/en active Active
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR19990025593A (en) * | 1997-09-13 | 1999-04-06 | 이종규 | Partial discharge on-line diagnostic system |
JP2013164748A (en) * | 2012-02-10 | 2013-08-22 | Aisin Seiki Co Ltd | State estimation device, state estimation method and program |
CN107728178A (en) * | 2017-09-29 | 2018-02-23 | 上海斐讯数据通信技术有限公司 | A kind of processing method and processing device of movement locus |
CN108760268A (en) * | 2018-06-08 | 2018-11-06 | 济南大学 | A kind of Vertical Mill operation data both phase step fault diagnostic method based on comentropy |
CN110276416A (en) * | 2019-07-02 | 2019-09-24 | 广东省智能机器人研究院 | A kind of rolling bearing fault prediction technique |
CN111649921A (en) * | 2020-04-13 | 2020-09-11 | 南通大学 | Rotating machinery fault diagnosis test system and working method |
CN111950200A (en) * | 2020-08-11 | 2020-11-17 | 哈电发电设备国家工程研究中心有限公司 | Power generation equipment fault diagnosis method based on axle center track fusion |
CN112362238A (en) * | 2020-11-13 | 2021-02-12 | 中国第一汽车股份有限公司 | Gravity center measuring method |
WO2022139193A1 (en) * | 2020-12-21 | 2022-06-30 | 효성중공업 주식회사 | Method and device for diagnosing malfunction of rotating machine |
CN115355162A (en) * | 2022-08-08 | 2022-11-18 | 西安交通大学 | Diaphragm compressor diaphragm fault diagnosis method based on oil pressure |
CN115660198A (en) * | 2022-11-07 | 2023-01-31 | 中国科学院空间应用工程与技术中心 | Method for predicting residual service life of rolling bearing |
CN116226646A (en) * | 2023-05-05 | 2023-06-06 | 国家石油天然气管网集团有限公司 | Method, system, equipment and medium for predicting health state and residual life of bearing |
CN116465623A (en) * | 2023-05-10 | 2023-07-21 | 安徽大学 | Gearbox service life prediction method based on sparse converter |
CN116563711A (en) * | 2023-05-17 | 2023-08-08 | 大连民族大学 | Hyperspectral target detection method of binary encoder network based on momentum update |
Non-Patent Citations (1)
Title |
---|
基于改进SVD及参数优化VMD的轴承故障诊断;张莹 等;《噪声与振动控制》;51-58 * |
Also Published As
Publication number | Publication date |
---|---|
CN117191394A (en) | 2023-12-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jiao et al. | Deep coupled dense convolutional network with complementary data for intelligent fault diagnosis | |
CN110361176B (en) | Intelligent fault diagnosis method based on multitask feature sharing neural network | |
Wang et al. | An enhanced intelligent diagnosis method based on multi-sensor image fusion via improved deep learning network | |
Qiao et al. | An adaptive weighted multiscale convolutional neural network for rotating machinery fault diagnosis under variable operating conditions | |
Zhao et al. | Semisupervised deep sparse auto-encoder with local and nonlocal information for intelligent fault diagnosis of rotating machinery | |
CN112763214B (en) | Rolling bearing fault diagnosis method based on multi-label zero-sample learning | |
Zheng et al. | Sucker rod pumping diagnosis using valve working position and parameter optimal continuous hidden Markov model | |
CN110297479B (en) | Hydroelectric generating set fault diagnosis method based on convolutional neural network information fusion | |
Li et al. | Joint attention feature transfer network for gearbox fault diagnosis with imbalanced data | |
Rao et al. | A deep bi-directional long short-term memory model for automatic rotating speed extraction from raw vibration signals | |
CN114638784A (en) | Method and device for detecting surface defects of copper pipe based on FE-YOLO | |
Enshaei et al. | Automated detection of textured-surface defects using UNet-based semantic segmentation network | |
CN116448419A (en) | Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization | |
Li et al. | A novel method of bearing fault diagnosis in time-frequency graphs using InceptionResnet and deformable convolution networks | |
Miao et al. | Sparse representation convolutional autoencoder for feature learning of vibration signals and its applications in machinery fault diagnosis | |
CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
Ai et al. | Two-stream deep feature-based froth flotation monitoring using visual attention clues | |
DeVol et al. | Inception based deep convolutional neural network for remaining useful life estimation of turbofan engines | |
CN117191394B (en) | Fault detection method and device for air compressor equipment | |
Liu et al. | Incipient fault detection of planetary gearbox under steady and varying condition | |
CN117541717A (en) | Modeling method of digital twin body of ball mill | |
CN110779722B (en) | Rolling bearing fault diagnosis method based on encoder signal local weighting | |
CN116070126A (en) | Aviation plunger pump oil distribution disc abrasion detection method and system based on countermeasure self-supervision | |
CN110261108A (en) | Bearing fault method of identification when specified operating based on CNN color property figure | |
Liu et al. | Fault diagnosis of rolling bearing under limited samples using joint learning network based on local-global feature perception |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |