CN116861798B - Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm - Google Patents
Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm Download PDFInfo
- Publication number
- CN116861798B CN116861798B CN202311116724.8A CN202311116724A CN116861798B CN 116861798 B CN116861798 B CN 116861798B CN 202311116724 A CN202311116724 A CN 202311116724A CN 116861798 B CN116861798 B CN 116861798B
- Authority
- CN
- China
- Prior art keywords
- downtime
- data set
- vacuum dry
- data
- dry pump
- 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
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 11
- 230000009467 reduction Effects 0.000 claims abstract description 7
- 230000011218 segmentation Effects 0.000 claims abstract description 3
- 230000008569 process Effects 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 11
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000003111 delayed effect Effects 0.000 claims description 5
- 229910052757 nitrogen Inorganic materials 0.000 claims description 5
- 230000000295 complement effect Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 150000001576 beta-amino acids Chemical class 0.000 claims 1
- 230000036541 health Effects 0.000 abstract description 5
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 6
- 230000004044 response Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Compressors, Vaccum Pumps And Other Relevant Systems (AREA)
Abstract
The invention provides an online real-time residual life prediction method for a vacuum dry pump based on an XGBoost algorithm, and relates to the technical field of vacuum dry pump health monitoring. It comprises: s1, acquiring characteristic data of a vacuum dry pump. S2, inputting the characteristic data into a pre-trained prediction model based on the XGBoost algorithm, and obtaining the early warning level. Training of the predictive model includes: a1, acquiring historical data sets of a plurality of downtime vacuum dry pumps, extracting data of each downtime operation period, and acquiring a first downtime period data set. A2, setting an early warning level label according to the time length of each time distance downtime time in the data set. A3, acquiring a plurality of state information with the maximum pearson correlation coefficient value according to the first downtime period data set and the early warning level label. And A4, extracting data of a plurality of state information, and acquiring a second downtime period data set. And A5, performing segmentation and dimension reduction processing according to the second downtime period data set to obtain a sample data set. A6, constructing a prediction model according to the sample data set.
Description
Technical Field
The invention relates to the technical field of vacuum dry pump health monitoring, in particular to an online real-time residual life prediction method of a vacuum dry pump based on an XGBoost algorithm.
Background
Vacuum dry pump apparatus are found in large numbers in the industrial field. The potential hidden trouble of the vacuum dry pump equipment is diagnosed in time, the safe operation of the vacuum dry pump equipment is ensured, the casualties and the economic loss caused by unexpected shutdown are avoided, and the method has great significance for society and enterprises. In the prior art, most enterprises use a mode of periodic maintenance and manual diagnosis to carry out equipment management on the vacuum dry pump.
With the development of neural networks, methods for health detection of mechanical devices by big data have emerged. It can be divided into three types: model-based methods, data-based methods, and a hybrid of the former two. Model-driven methods require sufficient prior knowledge and expert experience to build accurate mathematical models or physical models, making accurate mathematical models of mechanical devices or critical components difficult to build. The data driving-based method cannot adapt to high-dimensional characteristics, small data sets and other scenes when the residual life of the mechanical equipment is predicted.
In view of this, the applicant has studied the prior art and has made the present application.
Disclosure of Invention
The invention provides an online real-time residual life prediction method of a vacuum dry pump based on an XGBoost algorithm, which aims to improve at least one of the technical problems.
The embodiment of the invention provides an online real-time residual life prediction method of a vacuum dry pump based on an XGBoost algorithm, which comprises steps S1 to S2.
S1, acquiring characteristic data of a vacuum dry pump during operation;
s2, inputting the characteristic data into a pre-trained prediction model based on an XGBoost algorithm, and obtaining an early warning level of the vacuum dry pump; the early warning level is used for indicating the time length from the downtime moment;
the prediction model based on the XGBoost algorithm is obtained through training in the steps A1 to A6:
a1, acquiring historical data sets of a plurality of downtime vacuum dry pumps, and preprocessing to extract data of each downtime operation period and acquire a first downtime period data set; wherein the historical dataset contains a plurality of status information of the vacuum dry pump;
a2, calculating the time length of each moment from the downtime moment in each data set according to the first downtime period data set, and setting an early warning level label according to the time length;
a3, according to the first downtime period data set and the early warning level label, acquiring a plurality of state information with the maximum correlation coefficient value through pearson correlation coefficient analysis;
a4, extracting data of a plurality of state information with the maximum correlation coefficient values from the first downtime period data set, and obtaining a second downtime period data set;
a5, dividing according to the second downtime period data set and preset duration; performing dimension reduction processing on each segment of segmented data to obtain a sample data set;
a6, constructing a prediction model of the residual life based on the XGBoost algorithm according to the sample data set; wherein the optimal weight of XGBoostAnd objective function->The method comprises the following steps:
in the method, in the process of the invention,is the first order reciprocal>Is the reciprocal of the second order->Sequence number of leaf, ++>Limiting leaf node score for regularization parameters, +.>For leaf node number, & gt>The number of leaf nodes is limited for the learning rate.
By adopting the technical scheme, the invention can obtain the following technical effects:
the online real-time residual life prediction method for the vacuum dry pump, disclosed by the embodiment of the invention, not only fully considers the historical data of the vacuum dry pump, but also considers the up-down time correlation, and improves the accuracy and applicability of the residual life prediction model of the vacuum dry pump.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for predicting the online real-time residual life of a vacuum dry pump.
FIG. 2 is a training flow diagram of a predictive model based on the XGBoost algorithm.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a first embodiment of the present invention provides an XGBoost algorithm-based online real-time residual life prediction method for a vacuum dry pump, so as to improve at least one of the above technical problems.
The embodiment of the invention provides an online real-time residual life prediction method of a vacuum dry pump based on an XGBoost algorithm, which can be executed by online real-time residual life prediction equipment (hereinafter referred to as life prediction equipment) of the vacuum dry pump. In particular, by one or more processors in the lifetime prediction device to implement steps S1 to S2.
S1, acquiring characteristic data of the vacuum dry pump during operation.
Specifically, the feature data is a plurality of state information which are screened out and have highest residual life correlation when a prediction model based on the XGBoost algorithm is trained. Including correlation coefficient valuesThe maximum plurality of status information includes a device model numberRun time->Status of operation->Pump down power->Pump power->Lower pump temperature->Upper pump temperature->Nitrogen flow->Exhaust pressure->。
It is understood that the lifetime prediction device may be an electronic device with computing capabilities, such as a portable notebook computer, a desktop computer, a server, a smart phone, or a tablet computer.
S2, inputting the characteristic data into a pre-trained prediction model based on an XGBoost algorithm, and obtaining the early warning level of the vacuum dry pump. The early warning level is used for indicating the time length from the downtime moment.
Specifically, the prediction model based on the XGBoost algorithm greatly improves the adaptability of predicting the residual life of the mechanical equipment, and effectively improves the accuracy of predicting the vacuum dry pump faults.
According to the method for predicting the online real-time residual life of the vacuum dry pump, provided by the embodiment of the invention, the health state of the vacuum dry pump can be predicted in advance according to the change characteristics of the historical characteristic data of the vacuum dry pump, decision support is provided for an operation enterprise, the operation enterprise can make decisions according to the predicted health state of the vacuum dry pump, and personnel injury and property loss of the enterprise caused by downtime of the vacuum dry pump are avoided. Compared with other existing methods, the solution method provided by the invention has the advantages of high solution speed, high precision and strong stability.
The prediction model based on the XGBoost algorithm is obtained through training in the steps A1 to A6:
a1, acquiring historical data sets of a plurality of downtime vacuum dry pumps, and preprocessing to extract data of each downtime operation period and acquire a first downtime period data set. Wherein the historical dataset contains a plurality of status information of the vacuum dry pump.
Preferably, the history data set contains state information of each moment when the vacuum dry pump is operated. Specifically, status information +_of the vacuum dry pump during operation is obtained through each characteristic sensor in each vacuum dry pump>。Wherein the status information comprises a device model +.>Run time->Status of operation->Lower pump current->Upper pump current->Pump down powerPump power->Lower pump speed->Upper pump speed->Lower pump temperature->Upper pump temperature->Nitrogen flow->Exhaust pressure->。
On the basis of the above embodiment, in an alternative embodiment of the present invention, the step A1 specifically includes steps a11 to a14.
A11, acquiring historical data sets of a plurality of downtime vacuum dry pumps, and complementing the sampling frequency among different state information to be consistent through first frame inserting processing. The first frame inserting process is to complement the data of the empty time with the data of the time before the empty time of the data.
In particular, the acquisition frequencies of different sensors on the same vacuum dry pump may be different for different vacuum dry pumps. For example, there are two acquisitions once a second and one acquisition once a second. In this embodiment, according to the historical dataset of each vacuum dry pump, a dataset with a sampling frequency of two seconds is searched, and the data at the moment of the vacancy is complemented by the data of one second at the moment of the vacancy, namely. So that the sampling frequency of all data is unified to 1 second once.
A12, completing blank frames in the historical data set after the first frame inserting process through the second frame inserting process. The second frame inserting process is to complement the data of the empty time with the data of the time before the empty time of the data.
Specifically, when a part of the sensors sample, no information may be acquired, so that the data at a part of time in the data is blank. According to the data of each moment of each vacuum dry pump, the data of each vacuum dry pump is collected and searched for the time data of the frame loss, and the data of the frame loss is filled by the data of the moment of the last moment of the frame loss moment of the vacuum dry pump, namely. Thereby complementing the lost frame data.
A13, judging whether the vacuum dry pump is in downtime for a plurality of times according to the historical data set after the second frame inserting process.
And A14, when judging that the vacuum dry pump is in downtime for a plurality of times, dividing the historical data set into independent data sets according to downtime time to extract data of each downtime operation period, thereby acquiring a first downtime period data set.
Specifically, step A1 obtains a record that the historical dataset is a downtime device. However, the record is a full period during which there may be multiple downtime. Corresponding to the recording of multiple run time periods, it is therefore necessary to split each run time period into separate data sets. In this embodiment, according to a historical equipment downtime record data set, equipment which is downtime for a plurality of times and operation time periods thereof are acquired, and according to different equipment operation time periods, the historical equipment data set is split into different data sets.
A2, according to the first downtime period data set, calculating the time length of each moment from the downtime moment in each data set, and setting an early warning level label according to the time length. Preferably, step A2 specifically includes steps a21 to a22.
A21, according to the downtime period data set, calculating the time length RUL of each moment from the downtime moment in each data set. Wherein,
wherein:for downtime, add>For the current moment of data acquisition.
Specifically, according to the downtime of the historical equipment, calculating the residual life of each vacuum dry pump at each moment to obtain historical residual life characteristic data (RUL) of each vacuum dry pump, namely: and the time length RUL of each time in the data from the downtime time.
A22, setting early warning level labels for all moments of the data set according to the time length RUL of the downtime moment of each moment distance.
Specifically, each vacuum dry pump calculates that the early warning level of each vacuum dry pump at each moment is determined by the residual life of the moment, and when RUL < = 7, the early warning level label is set to 1, which indicates that the vacuum dry pump is down within seven days. When RUL >7, the early warning level flag is set to 0, indicating that the machine will not downtime for 7 days.
A3, according to the first downtime period data set and the early warning level label, acquiring a plurality of state information with the maximum correlation coefficient value through pearson correlation coefficient analysis.
Specifically, according to the historical data set of each vacuum dry pump processed in the step A1 and the residual life calculated in the step A2, pearson correlation coefficient analysis is carried out to obtain the correlation coefficient value of each vacuum dry pump on the residual life, so that a plurality of characteristic values with larger correlation coefficient values are screened out.
Preferably, step A3 specifically includes steps a31 to a32.
A31, calculating coefficient values of pearson correlation coefficients of each state information to the residual life of the vacuum dry pump according to the first downtime period data set and the early warning level label. Wherein, the calculation formula of the pearson correlation coefficient is:
wherein:is the pearson correlation coefficient, +.>Representing overall covariance,/->Representing a single feature->Is the target feature->Representation->Standard deviation of>Representation->Standard deviation of>Indicate->The value of the individual feature->Is->Is>Indicate->Target value corresponding to individual characteristic, +.>Is->Is>Representing the total number of features.
Wherein the overall covariance:
overall mean:
) Is the standard deviation of X
Preferably, when-0.1.ltoreq.No correlation is found when less than or equal to +0.1; when-0.3 +.>< -0.1 or +0.1 </l>When less than or equal to +0.3, weak correlation is achieved; when-0.5 +.>With-0.3 or +0.3 </l>When less than or equal to +0.5, the correlation is medium; when-1 is less than or equal to%>With-0.5 or +0.5 </l>And when the number is less than or equal to +1, the correlation is strong.
A32, selecting a plurality of state information of medium correlation and strong correlation according to the coefficient value of the Pearson correlation coefficient.
Specifically, according to the calculated correlation coefficient of each characteristic value for the early warning level, a plurality of characteristic values with larger pearson correlation coefficients are selected. Preferably, the plurality of state information having the largest pearson correlation coefficient value includes a device model numberRun time->Status of operation->Pump down power->Pump power->Lower pump temperature->Upper pump temperature->Nitrogen flow->Exhaust pressure->。
A4, extracting data of a plurality of state information with the maximum correlation coefficient values from the first downtime period data set, and obtaining a second downtime period data set.
On the basis of the above embodiment, in an alternative embodiment of the present invention, after step A4, before step A5, the method further includes:
and respectively carrying out normalization processing on the data of each operation period in the second downtime period data set. The formula of normalization processing is as follows:
in the method, in the process of the invention,for normalized data, ++>For data before normalization, ++>Minimum, & ->Is the maximum value.
And A5, dividing according to the second downtime period data set and the preset duration. And performing dimension reduction processing on each segment of segmented data to obtain a sample data set. Preferably, step A5 specifically includes steps a51 to a52.
A51, dividing the data of each hour of each operation period according to the second downtime period data set.
In this embodiment, one hour of history data is directly observed. Therefore, the data of each vacuum dry pump in each hour are divided according to the historical data set of each vacuum dry pump processed in the step A1.
A52, reducing the data of each hour after segmentation into one dimension through a flat function, and obtaining a sample data set.
The embodiment of the invention carries out dimension reduction treatment on the divided data of each hour, and reduces the data of each hour of each vacuum dry pump into one dimension. Specifically, each 3600 rows of data are flattened into one row by using a flat function, and one hour of data are all in one row, so that the calculated amount is reduced.
A6, constructing a prediction model of the residual life based on the XGBoost algorithm according to the sample data set. Preferably, step A6 specifically includes steps a61 to a63.
And A61, dividing the sample data set into a training set and a testing set. Wherein 70% of the sample data is used as a training set and 30% of the sample data is used as a test set.
Specifically, the historical data set of each vacuum dry pump processed in the step A5 is proportionally divided into a training sample and a test sample. Wherein the training samples form a training set and the test samples form a test set. Preferably, 70% of the pump data set is used as the training set and 30% of the pump data set is used as the test set.
A62, constructing a prediction model to be tested based on the residual life of the XGBoost algorithm through a training set.
Specifically, parameters of an XGBoost algorithm are set, a residual life prediction model based on the XGBoost algorithm is constructed by utilizing samples of a training set, and a nonlinear mapping relation of input features and early warning levels is established.
Wherein the objective function of XGBoost is
Wherein:is the number of samples; />Measuring and predicting value and target value as error function>Errors between;the regularization term is used for controlling the complexity of the model and avoiding the model from being over fitted; />The residual value of the last prediction;is a predicted value.
Wherein:penalty coefficients for leaf nodes; />Is a regularized term coefficient; t and->The number and the weight of the leaves of the kth tree are respectively; />Is->The weight coefficient of each leaf; />Representing an objective function of XGboostA training loss part and a regularization part; />Representing the predicted value of the t-1 step; />A score representing a leaf node; />Representing the regularized portion.
The objective function is approximately
In the method, in the process of the invention,is a second order bias; />Is a first order bias guide; />Score vectors on leaf nodes; />Representing that the storage is mapped to +.>Data of individual leaf nodes->Is a set of indices of (a).
From this, the optimal weight of XGBoostAnd objective function->The method comprises the following steps:
in the method, in the process of the invention,is the first order reciprocal>Is the reciprocal of the second order->Sequence number of leaf, ++>Limiting leaf node score for regularization parameters, +.>For leaf node number, & gt>The number of leaf nodes is limited for the learning rate.
A63, inputting the test set into a prediction model to be tested for prediction, and calculating the false alarm rate, the delayed alarm rate, the missing alarm rate and the confidence coefficient according to the prediction result. When the confidence is greater than the preset value, the to-be-tested is presetThe test model serves as a final prediction model. Otherwise, the historical dataset is added for retraining.
Specifically, the test set is predicted, that is, each test sample in the test set is classified and predicted by using the to-be-tested prediction model obtained in the step A62, and the false alarm rate, the delayed alarm rate, the missing alarm rate and the confidence coefficient are calculated according to the result of the classification and prediction performed by the test set. And judging whether the prediction model to be tested meets the requirements or not based on the confidence. The confidence coefficient reaches 70% and can meet the requirement, if the confidence coefficient does not reach the requirement, the training set is continuously supplemented, and the model is trained.
The judging rules of the false alarm, the delayed alarm and the missing alarm rate are as follows:
the time for recording the real early warning level mark as 1 isThe time for recording the predicted early warning level of 1 is。
Of vacuum pumpsIs not empty, i.e. the vacuum dry pump is down and +.>Is not empty, i.e. the early warning level of the vacuum dry pump is marked 1, when +.>Early->And if so, the prediction result is a later report.
Of vacuum pumpsIs not empty, i.e. the vacuum dry pump is down and +.>Is not empty, i.e. the early warning level of the vacuum dry pump is marked as1, when->Later than->And if so, the prediction result is false report.
Of vacuum pumpsIs empty and->And if the vacuum dry pump is not empty, namely the vacuum dry pump is not really down, but the early warning level is marked as 1, the prediction result is false.
Several vacuum pumpsIs not empty and is->If the vacuum dry pump is empty, namely the real data of the vacuum dry pump is recorded in a downtime mode, but the early warning level is not marked as 1, the prediction result is a missing report.
The calculation formula of the missing report rate is as follows
The calculation formula of the false alarm rate is as follows
The calculation formula of the delay report rate is as follows
The confidence coefficient is calculated according to the following formula
In the method, the data of all the vacuum dry pumps which are missed are counted asThe data of all the delayed vacuum dry pumps are counted as +.>Data of vacuum dry pump counting all false alarms is +.>The total number of all vacuum dry pumps in the test set is。
The online real-time residual life prediction method of the vacuum dry pump provided by the embodiment of the invention not only pays attention to the characteristic change of the running state, but also integrates the time dependency factor, the relevance of the vacuum dry pump data changing along with time is mined, the context of the upper period is captured, the higher-level sequence data representation capability is generated, and the model prediction precision is improved.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus and method embodiments described above are merely illustrative, for example, flow diagrams and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present invention may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be understood that the term "and/or" as used herein is merely one relationship describing the association of the associated objects, meaning that there may be three relationships, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.
Depending on the context, the word "if" as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to detection". Similarly, the phrase "if determined" or "if detected (stated condition or event)" may be interpreted as "when determined" or "in response to determination" or "when detected (stated condition or event)" or "in response to detection (stated condition or event), depending on the context.
References to "first\second" in the embodiments are merely to distinguish similar objects and do not represent a particular ordering for the objects, it being understood that "first\second" may interchange a particular order or precedence where allowed. It is to be understood that the "first\second" distinguishing aspects may be interchanged where appropriate, such that the embodiments described herein may be implemented in sequences other than those illustrated or described herein.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. The online real-time residual life prediction method for the vacuum dry pump based on the XGBoost algorithm is characterized by comprising the following steps of:
acquiring characteristic data of the vacuum dry pump during operation;
inputting the characteristic data into a pre-trained prediction model based on an XGBoost algorithm to obtain an early warning level of the vacuum dry pump; the early warning level is used for indicating the time length from the downtime moment;
the prediction model based on the XGBoost algorithm is obtained through training the following steps:
acquiring historical data sets of a plurality of downtime vacuum dry pumps, and preprocessing to extract data of each downtime operation period and acquire a first downtime period data set; wherein the historical dataset contains a plurality of status information of the vacuum dry pump;
calculating the time length of each moment from the downtime moment in each data set according to the first downtime period data set, and setting an early warning level label according to the time length;
according to the first downtime period data set and the early warning level label, a plurality of state information with the maximum correlation coefficient value are obtained through pearson correlation coefficient analysis;
extracting data of a plurality of state information with the maximum correlation coefficient values from the first downtime period data set, and obtaining a second downtime period data set;
dividing according to the second downtime period data set and preset time length, and performing dimension reduction processing on each piece of divided data to obtain a sample data set;
constructing a prediction model of the residual life based on the XGBoost algorithm according to the sample data set; wherein the optimal weight of XGBoostAnd objective function->The method comprises the following steps:
in the method, in the process of the invention,is the first order reciprocal>Is the reciprocal of the second order->Sequence number of leaf, ++>Limiting leaf node score for regularization parameters, +.>For leaf node number, & gt>Limiting the number of leaf nodes for the learning rate;
according to the first downtime period data set, calculating the time length of each moment from downtime moment in each data set, and setting an early warning level label according to the time length, wherein the method specifically comprises the following steps:
according to the downtime period data set, calculating the time length RUL of each moment from the downtime moment in each data set; wherein,wherein: />For downtime, add>The current moment for collecting data;
setting early warning level labels for all moments of the data set according to the time length RUL of each moment from the downtime moment; the early warning level of the early warning level label is determined by the current residual life, and when RUL < = 7, the early warning level label is set to be 1, which indicates that the machine is down in seven days; when RUL >7, the early warning level label is set to 0, which means that the machine cannot be down within 7 days;
dividing according to preset time length, and carrying out dimension reduction treatment on each piece of divided data, wherein the method specifically comprises the following steps:
dividing the data of each hour of each operation period according to the second downtime period data set;
and reducing the data of each hour after segmentation into one dimension through a flat function, and obtaining a sample data set.
2. The XGBoost algorithm-based vacuum dry pump online real-time remaining life prediction method according to claim 1, wherein the history data set contains state information of each moment in time when the vacuum dry pump is operatedThe method comprises the steps of carrying out a first treatment on the surface of the Wherein the status information includes a device model +.>Run time->Status of operation->Lower pump current->Upper pump current->Pump down power->Pump power->Lower pump speed->Upper pump speed->Lower pump temperature->Upper pump temperature->Nitrogen flow->Exhaust pressure->。
3. The XGBoost algorithm-based online real-time residual life prediction method for a vacuum dry pump of claim 2, wherein the method is characterized by obtaining a historical data set of a plurality of downtime vacuum dry pumps, and preprocessing the historical data set to extract data of each downtime operation period, and obtaining a first downtime period data set, and specifically comprises the following steps:
acquiring historical data sets of a plurality of down vacuum dry pumps, and complementing sampling frequencies among different state information to be consistent through first frame inserting processing; the first frame inserting process is to complement the data at the vacant moment with the data at the moment before the vacant moment of the data;
the blank frames in the historical data set after the first frame inserting process are completed through the second frame inserting process; the second frame inserting process is to complement the data of the vacant moment with the data of the moment before the vacant moment of the data;
judging whether the vacuum dry pump is downtime for a plurality of times according to the historical data set after the second frame inserting process;
when judging that the vacuum dry pump is in downtime for a plurality of times, dividing the historical data set into independent data sets according to downtime time to extract data of each downtime operation period, thereby acquiring a first downtime period data set.
4. The XGBoost algorithm-based vacuum dry pump online real-time remaining life prediction method according to claim 1, wherein the method is characterized in that according to the first downtime period data set and the early warning level label, a plurality of state information with the maximum correlation coefficient value is obtained through pearson correlation coefficient analysis, and specifically comprises the following steps:
according to the first downtime period data set and the early warning level label, calculating coefficient values of pearson correlation coefficients of each state information on the residual life of the vacuum dry pump respectively; wherein, the calculation formula of the pearson correlation coefficient is as followsWherein: />Is the pearson correlation coefficient,Representing overall covariance,/->Representing a single feature->Is the target feature->Representation->Standard deviation of>Representation->Standard deviation of>Indicate->The value of the individual feature->Is->Is>Indicate->Target values corresponding to the characteristics,Is->Is>Representing the total number of features;
selecting a plurality of state information of medium correlation and strong correlation according to the coefficient value of the pearson correlation coefficient; wherein, when the content of the beta-amino acid is less than or equal to-0.1%No correlation is found when less than or equal to +0.1; when-0.3 +.>< -0.1 or +0.1 </l>When less than or equal to +0.3, weak correlation is achieved; when-0.5 +.>With-0.3 or +0.3 </l>When less than or equal to +0.5, the correlation is medium; when-1 is less than or equal to%>With-0.5 or +0.5 </l>And when the number is less than or equal to +1, the correlation is strong.
5. The XGBoost algorithm-based vacuum dry pump online real-time remaining life prediction method of claim 4, wherein the plurality of state information with the largest correlation coefficient value comprises a device model numberRun time->Status of operation->Pump down power->Pump power->Lower pump temperature->Upper pump temperature->Nitrogen flow->Exhaust pressure->。
6. The XGBoost algorithm-based vacuum dry pump online real-time remaining life prediction method according to claim 1, wherein the method is characterized in that the method comprises the steps of dividing according to a preset time length according to the second downtime period data set; and performing dimension reduction processing on each segment of segmented data, and before obtaining a sample data set, further comprising:
respectively carrying out normalization processing on the data of each operation period in the second downtime period data set; the formula of normalization processing is as follows:wherein->For normalized data, ++>For data before normalization, ++>Minimum, & ->Is the maximum value.
7. The XGBoost algorithm-based vacuum dry pump online real-time remaining life prediction method according to any one of claims 1 to 6, wherein constructing a XGBoost algorithm-based remaining life prediction model according to the sample data set specifically comprises:
dividing the sample data set into a training set and a testing set; wherein, 70% of sample data is used as a training set, and 30% of sample data is used as a test set;
constructing a prediction model to be tested based on the residual life of the XGBoost algorithm through a training set;
inputting the test set into a prediction model to be tested for prediction, and calculating the false alarm rate, the delayed alarm rate, the missing alarm rate and the confidence coefficient according to the prediction resultThe method comprises the steps of carrying out a first treatment on the surface of the When the confidence coefficient is larger than a preset value, taking the prediction model to be tested as a final prediction model; otherwise, adding a historical data set for retraining; wherein, the time for recording the real early warning level mark as 1 is +.>The time for recording the predicted early warning level of 1 is +.>The method comprises the steps of carrying out a first treatment on the surface of the The ∈of the vacuum pump>Is not empty, i.e. the vacuum dry pump is down and +.>Is not empty, i.e. the early warning level of the vacuum dry pump is marked 1, when +.>Earlier thanIf yes, the prediction result is a later report; the ∈of the vacuum pump>Is not empty, i.e. the vacuum dry pump is down and +.>Is not empty, i.e. the early warning level of the vacuum dry pump is marked 1, when +.>Later thanWhen the predicted result is false alarm; the ∈of the vacuum pump>Is empty and->If the vacuum dry pump is not empty, namely the vacuum dry pump is really not down, but the early warning level is marked as 1, the prediction result is false; vacuum pump->Is not empty and is->If the vacuum dry pump is empty, namely the real data of the vacuum dry pump is recorded in a downtime mode, but the early warning level is not marked as 1, the prediction result is a missing report; the confidence formula is:wherein->For rate of missing report>Is the false alarm rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311116724.8A CN116861798B (en) | 2023-09-01 | 2023-09-01 | Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311116724.8A CN116861798B (en) | 2023-09-01 | 2023-09-01 | Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116861798A CN116861798A (en) | 2023-10-10 |
| CN116861798B true CN116861798B (en) | 2023-12-26 |
Family
ID=88230751
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311116724.8A Active CN116861798B (en) | 2023-09-01 | 2023-09-01 | Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116861798B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013443B (en) * | 2024-04-08 | 2024-07-23 | 华侨大学 | Online real-time vacuum dry pump abnormality detection method based on generation model algorithm |
| CN118094126B (en) * | 2024-04-17 | 2024-07-16 | 华侨大学 | Vacuum dry pump residual life prediction method and equipment based on clustering and transducer algorithm |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114416320A (en) * | 2022-01-20 | 2022-04-29 | 北京字节跳动网络技术有限公司 | Task processing method, device, equipment and storage medium |
| CN115168173A (en) * | 2022-07-25 | 2022-10-11 | 阿里巴巴（中国）有限公司 | Fault prediction model training method, equipment fault determination method, device and equipment |
| WO2023044770A1 (en) * | 2021-09-24 | 2023-03-30 | 京东方科技集团股份有限公司 | Dry pump downtime early warning method and apparatus, electronic device, storage medium, and program |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260720A1 (en) * | 2017-03-13 | 2018-09-13 | General Electric Company | Fatigue Crack Growth Prediction |
-
2023
- 2023-09-01 CN CN202311116724.8A patent/CN116861798B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023044770A1 (en) * | 2021-09-24 | 2023-03-30 | 京东方科技集团股份有限公司 | Dry pump downtime early warning method and apparatus, electronic device, storage medium, and program |
| CN116235148A (en) * | 2021-09-24 | 2023-06-06 | 京东方科技集团股份有限公司 | Early warning method and device for downtime of dry pump, electronic equipment, storage medium and program |
| CN114416320A (en) * | 2022-01-20 | 2022-04-29 | 北京字节跳动网络技术有限公司 | Task processing method, device, equipment and storage medium |
| CN115168173A (en) * | 2022-07-25 | 2022-10-11 | 阿里巴巴（中国）有限公司 | Fault prediction model training method, equipment fault determination method, device and equipment |
Non-Patent Citations (1)
| Title |
|---|
| 真空干泵用屏蔽电机无速度传感器带速重投控制***;安跃军;张志恒;张振厚;王光玉;孔祥玲;;电工技术学报(12);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116861798A (en) | 2023-10-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116861798B (en) | Online real-time residual life prediction method for vacuum dry pump based on XGBoost algorithm | |
| Wang et al. | Online class imbalance learning and its applications in fault detection | |
| DK3073450T3 (en) | SYSTEM AND PROCEDURE FOR MONITORING A DRIVER'S DRIVING CONDUCT | |
| CN112148955A (en) | Method and system for detecting abnormal time sequence data of Internet of things | |
| CN110032490A (en) | Method and device thereof for detection system exception | |
| CN111538311B (en) | Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining | |
| CN111782484B (en) | Anomaly detection method and device | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN113518011A (en) | Abnormality detection method and apparatus, electronic device, and computer-readable storage medium | |
| EP1958034B1 (en) | Use of sequential clustering for instance selection in machine condition monitoring | |
| US12007745B2 (en) | Apparatus for predicting equipment damage | |
| CN110858072B (en) | Method and device for determining running state of equipment | |
| CN116028870B (en) | Data detection method and device, electronic equipment and storage medium | |
| CN116930759A (en) | Power battery voltage time sequence abnormality detection method, electronic device and storage medium | |
| CN116257663A (en) | Abnormality detection and association analysis method and related equipment for unmanned ground vehicle | |
| CN110956331A (en) | Method, system and device for predicting operation state of digital factory | |
| CN118013443B (en) | Online real-time vacuum dry pump abnormality detection method based on generation model algorithm | |
| CN115438885A (en) | Method, device, equipment and storage medium for detecting abnormal fluctuation of service | |
| CN113535458A (en) | Abnormal false alarm processing method and device, storage medium and terminal | |
| CN118094126B (en) | Vacuum dry pump residual life prediction method and equipment based on clustering and transducer algorithm | |
| CN110705597A (en) | Network early event detection method and system based on event cause and effect extraction | |
| CN112149833A (en) | Prediction method, device, equipment and storage medium based on machine learning | |
| CN116302848B (en) | Detection method and device for bias of evaluation value, electronic equipment and medium | |
| CN118350279A (en) | Vacuum dry pump residual life prediction method, device and medium based on incremental training | |
| CN115511106B (en) | Method, device and readable storage medium for generating training data based on time sequence data |
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 |