CN116561641A - Industrial equipment fault diagnosis method and system based on multi-view generation algorithm - Google Patents
Industrial equipment fault diagnosis method and system based on multi-view generation algorithm Download PDFInfo
- Publication number
- CN116561641A CN116561641A CN202310597693.6A CN202310597693A CN116561641A CN 116561641 A CN116561641 A CN 116561641A CN 202310597693 A CN202310597693 A CN 202310597693A CN 116561641 A CN116561641 A CN 116561641A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- fault diagnosis
- time
- industrial equipment
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses an industrial equipment fault diagnosis method and system based on a multi-view generation algorithm, relates to the technical field of machine learning, and aims to solve the problem that the fault diagnosis accuracy is low due to inaccurate feature extraction in the existing fault diagnosis method. The technical key points of the invention include: collecting one or more running state quantities of industrial equipment, and sorting the same running state quantities according to time to form an original unitary time sequence set; performing multi-view generation on a plurality of time sequence samples, and combining the time sequence transformation with feature vector splicing to generate a feature vector set of a plurality of views; inputting the feature vector of each view into a machine learning-based fault diagnosis classifier for training; performing multi-view generation on the time sequence of the running state sample to be detected to obtain a corresponding feature vector; and inputting the feature vector into a trained classifier to obtain a fault prediction result. The invention obviously improves the precision and performance of time sequence classification and fault diagnosis.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to an industrial equipment fault diagnosis method and system based on a multi-view generation algorithm.
Background
The industry is the actual embodiment of the country, and the industrial level is relevant to the national life. In industrial production, a large amount of industrial equipment participates, and the safety and reliability of the industrial equipment are the most critical. Once industrial equipment fails, the production efficiency is reduced and serious life and property losses are caused. In the modern industry, a large number of sensors are installed on industrial equipment to monitor various state quantities of the equipment operation, such as temperature, rotation speed, pressure, etc. The state quantities acquired by the sensors form time sequences according to time sequences, and fault diagnosis can be realized by processing and analyzing the time sequences.
In the existing industrial equipment fault diagnosis, the fault diagnosis based on a machine learning algorithm is adopted, and the time sequence classification is carried out on the unitary time sequence acquired by each sensor, so that the fault diagnosis is realized. The method has the advantages of strong interpretation, high speed and high accuracy, and has wide application.
However, in the traditional machine learning-based unitary time series classification method for fault diagnosis, the extracted features are too single, and the multi-domain fault mode cannot be completely characterized and extracted; one data description of an object represents one view, a plurality of descriptions represent a plurality of views, and a method of learning based on the plurality of view data to improve classification and diagnosis performance is called multi-view learning. Multiple view learning can extract features from multiple domains, resulting in comprehensive and diverse data descriptions. The descriptions between each view are not exactly the same and there is a diversity. The fault diagnosis performance of the industrial equipment can be enhanced by improving the accuracy of each view and the diversity among views and improving the classification accuracy.
Disclosure of Invention
Therefore, the invention provides an industrial equipment fault diagnosis method and system based on a multi-view generation algorithm, which are used for solving the problem that the fault diagnosis accuracy is low due to inaccurate feature extraction in the existing fault diagnosis method based on machine learning.
According to an aspect of the present invention, there is provided an industrial equipment fault diagnosis method based on a multi-view generation algorithm, the method comprising the steps of:
step one, collecting one or more running state quantities of industrial equipment, and sorting the same running state quantities according to time to form an original unitary time sequence set;
step two, preprocessing the original unitary time sequence set;
step three, carrying out multi-view generation on a plurality of time sequence samples in the preprocessed original unitary time sequence set, and generating a feature vector set of a plurality of views by combining time sequence transformation and feature vector splicing;
inputting the feature vector set of each view into a fault diagnosis classifier based on machine learning for training, and obtaining a trained fault diagnosis classifier;
step five, carrying out multi-view generation on the time sequence of the pre-processed running state sample to be detected, and obtaining a corresponding feature vector;
and step six, inputting the feature vector corresponding to the running state sample to be tested into a trained fault diagnosis classifier to obtain a fault prediction result of the running state sample to be tested.
Further, in the first step, the operation state quantity includes temperature, rotation speed and pressure; the industrial equipment comprises engineering machinery, vehicles, machine tools, environment-friendly machinery, electrical equipment and electronic equipment.
Further, the preprocessing in the second step includes: filling the time sequence with the missing value by adopting an interpolation method; processing the unequal long-time sequences into equal-time sequences by adopting a sliding window method; carrying out standardization processing on each long-time sequence; labeling each time series sample after normalization.
Further, the specific process of the third step comprises the following steps:
k quick sequence transformations are carried out on the plurality of time sequence samples, and K transformed time sequences are obtained; the sequence transformation comprises Hilbert transformation, differential transformation of each order, wavelet transformation or Fourier transformation;
the number of kernels of each of the plurality of time-series samples and the transformed time-series is C 1 ,C 2 Generating corresponding feature vectors, and generating K+1 feature vectors for each time sequence sample; the method specifically comprises the following steps: each convolution check carries out sliding dot product operation on each time sequence to generate a feature map, and P pooling features are extracted from each feature map, wherein the pooling features comprise positive value proportion, negative value proportion, zero value proportion, maximum value, minimum value and average value;
performing row stitching on the feature vectors generated by the corresponding time sequence samples and the feature vectors generated by the corresponding transformed time sequence samples to obtain feature vectors of K views, wherein each feature vector comprises (C 1 +C 2 ) P features.
Further, in the random convolution kernel transformation in the third step, the time-series samples s=(s) 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
where x is the dot product operation, d is the expansion coefficient, l represents the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
Further, the specific process in the step six comprises the following steps: respectively inputting the feature vectors of the K views into the corresponding trained fault diagnosis classifier, and outputting K prediction labels; integrating the K prediction tags through integrated voting to obtain a final prediction tag; wherein the method adopted by the integrated voting comprises hard voting, soft voting and weighted voting.
According to another aspect of the present invention, there is provided an industrial equipment fault diagnosis system based on a multi-view generation algorithm, the system comprising:
the data acquisition module is configured to acquire one or more operation state quantities of the industrial equipment, and sort the same operation state quantities according to time to form an original unitary time sequence set;
a preprocessing module configured to preprocess the original unitary time series set;
the feature set generation module is configured to generate multiple views of the preprocessed multiple time sequence samples in the original unitary time sequence set, and generate feature vector sets of the multiple views by combining time sequence transformation and feature vector splicing;
the classifier training module is configured to input the feature vector set of each view into a machine learning-based fault diagnosis classifier for training, and obtain a trained fault diagnosis classifier;
the fault prediction module is configured to perform multi-view generation on the time sequence of the pre-processed running state sample to be detected, and obtain a corresponding feature vector; inputting the feature vector corresponding to the running state sample to be tested into a trained fault diagnosis classifier to obtain a fault prediction result of the running state sample to be tested; the specific process comprises the following steps: respectively inputting the feature vectors of the K views into the corresponding trained fault diagnosis classifier, and outputting K prediction labels; and integrating the K prediction labels through integrated voting to obtain a final prediction label.
Further, the running state quantity in the data acquisition module comprises temperature, rotating speed and pressure; the industrial equipment comprises engineering machinery, vehicles, machine tools, environment-friendly machinery, electrical equipment and electronic equipment; the preprocessing process in the preprocessing module comprises the following steps: filling the time sequence with the missing value by adopting an interpolation method; processing the unequal long-time sequences into equal-time sequences by adopting a sliding window method; carrying out standardization processing on each long-time sequence; labeling each time series sample after normalization.
Further, the specific process of generating the feature vector sets of the multiple views in the feature set generation module includes:
k quick sequence transformations are carried out on the plurality of time sequence samples, and K transformed time sequences are obtained; the sequence transformation comprises Hilbert transformation, differential transformation of each order, wavelet transformation or Fourier transformation;
the number of kernels of each of the plurality of time-series samples and the transformed time-series is C 1 ,C 2 Generating corresponding feature vectors, and generating K+1 feature vectors for each time sequence sample; the method specifically comprises the following steps: each convolution check carries out sliding dot product operation on each time sequence to generate a feature map, and P pooling features are extracted from each feature map, wherein the pooling features comprise positive value proportion, negative value proportion, zero value proportion, maximum value, minimum value and average value;
performing row stitching on the feature vectors generated by the corresponding time sequence samples and the feature vectors generated by the corresponding transformed time sequence samples to obtain feature vectors of K views, wherein each feature vector comprises (C 1 +C 2 ) P features.
Further, the time series samples s=(s) in the random convolution kernel transformation in the feature set generation module 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
where x is the dot product operation, d is the expansion coefficient, l represents the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
The beneficial technical effects of the invention are as follows:
according to the invention, a time sequence sample is acquired through a sensor arranged on industrial equipment, a fault diagnosis function is realized through analyzing a unitary time sequence sample set acquired and judging whether the equipment has faults or not and the type of the faults; the invention realizes the fault diagnosis of the equipment through time sequence classification, extracts a large number of features based on machine learning and time sequence, and generates feature vectors of a plurality of views; the accuracy and the diversity of the multiple views are combined for integrated learning, so that the accuracy and the performance of time sequence classification and fault diagnosis are remarkably improved.
The invention extracts a large number of convolution features from different domains by introducing a variety of fast sequence transforms in combination with random convolution kernel transforms. The time sequence is subjected to various sequence transformations, whereby the original sequence is mapped to different domain spaces. Then extracting a large number of convolution features from different sequence spaces through random convolution kernel transformation, so as to realize the extraction of various features from different domains; feature vectors of a plurality of views are generated by adopting feature vector splicing, so that the accuracy and diversity of each view are ensured, and the classification performance is further improved by integrating and learning to synthesize the prediction results of each view; the accuracy of the various features extracted from different domains is different, and the performance in some domains is poor, so that the feature vectors in each domain and the feature vectors corresponding to the original sequence are spliced to generate each view through feature vector splicing, the accuracy of each view is ensured, and the classification performance is further improved through ensemble learning.
Drawings
The invention may be better understood by reference to the following description taken in conjunction with the accompanying drawings, which are included to provide a further illustration of the preferred embodiments of the invention and to explain the principles and advantages of the invention, together with the detailed description below.
Fig. 1 is an overall flowchart of an industrial equipment fault diagnosis method based on a multi-view generation algorithm according to an embodiment of the present invention.
Fig. 2 is a flow chart of the data acquisition step in an embodiment of the present invention.
Fig. 3 is a flowchart of a multi-view generation step in an embodiment of the present invention.
Fig. 4 is a flowchart of sample label prediction to be tested in an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, exemplary embodiments or examples of the present invention will be described below with reference to the accompanying drawings. It is apparent that the described embodiments or examples are only implementations or examples of a part of the invention, not all. All other embodiments or examples, which may be made by one of ordinary skill in the art without undue burden, are intended to be within the scope of the present invention based on the embodiments or examples herein.
The invention provides an industrial equipment fault diagnosis method and system based on a multi-view generation algorithm, wherein the multi-view generation algorithm is used for generating multi-view feature vectors by extracting multi-domain features through combination of various quick time sequence transformations and convolution features; the characteristics of a single domain are difficult to comprehensively represent the fault mode, and the characteristics of multiple domains are obtained through multiple sequence transformation, so that the fault characteristics can be extracted more comprehensively; the prediction results of all the views are integrated through ensemble learning, and high-performance time sequence classification and fault diagnosis are realized by utilizing the accuracy and diversity of all the views.
The embodiment of the invention provides an industrial equipment fault diagnosis method based on a multi-view generation algorithm, which is shown in fig. 1 and comprises the following steps: firstly, through data acquisition, state quantity acquired by a certain sensor of industrial equipment is stored as a unitary time sequence according to time sequence, and a large number of same time sequences jointly form an original unitary time sequence set; then, processing the time sequence in the acquired unitary time sequence set into a unitary time sequence which is equal in length, equal in interval, free of missing value and standardized through data preprocessing, and marking the label of each sample as the running state of industrial equipment, wherein the label comprises normal and fault types; these tagged unitary time series samples form a training set; then, through multi-view generation, feature diversity is increased based on various rapid time sequence transformations and feature vector splicing, K kinds of sequence transformations are carried out on the time sequence to generate a corresponding transformed sequence, and convolution features are extracted through random convolution kernel transformation to generate a corresponding feature vector; then, the feature vectors corresponding to the original sequences are respectively spliced with the feature vectors corresponding to the K transformation sequences, so that K feature vectors, namely K views, are obtained; training the K classifiers on the K views respectively in a training set with labels through classifier training; finally, in a sample label prediction stage to be detected, K view feature vectors obtained by generating the sample to be detected through multiple views are respectively input into a trained classifier to obtain K prediction labels; and finally, obtaining a predictive label of the sample to be tested, namely the running state of the industrial equipment through voting integration. The following describes embodiments of the present invention in detail.
Step one: and acquiring and storing data from sensors of industrial equipment to obtain an original unitary time sequence set.
According to an embodiment of the invention, the sensor measures a certain operating state quantity of an industrial device, such as temperature, rotation speed, pressure, etc., wherein the industrial device comprises, but is not limited to, engineering machinery, vehicles, machine tools, environmental protection machinery, electrical devices, electronic devices, etc. The time sequences in the original unitary time sequence set are all acquired by the same type sensor installed at the same position of the same type industrial equipment. And sequencing and storing the state quantities acquired by a large number of sensors according to the time sequence to obtain an original unitary time sequence set.
Step two: preprocessing data, namely preprocessing an acquired original unitary time sequence set to obtain a training set with labels in a unified format.
According to an embodiment of the present invention, the flow of preprocessing is shown in fig. 2, and includes the following steps:
(1) Filling the missing values;
because some time sequences in the original unitary time sequence set have missing values, each missing value is filled by adopting an interpolation method. Methods of interpolation include, but are not limited to, linear interpolation, lagrangian interpolation, spline interpolation, and the like. With missing value padding, each time sequence in the original unitary time sequence set is free of missing values.
(2) Equal-length division;
because the time sequences in the interpolated unitary time sequence set are not equal in length, in order to facilitate processing of the unified sample format, a sliding window method is used to process the longer unequal-length time sequences into shorter identical lengths. By equal length division, each time series in the interpolated unitary time series set is equal in length.
(3) Standardization treatment;
because the time sequence units acquired by the same type of sensor are the same, each time sequence of the unitary time sequence set after equal-length division is subjected to standardization processing. For a time series s=(s) of length n 1 ,s 2 ,…,s n ) The normalized formula of (2) is as in formula (1).
Wherein mean(s) is the mean of the time series s and std(s) is the standard deviation of the time series s.
Through the normalization processing, each time series in the unitary time series set is processed into a standard time series having a mean of 0 and a standard deviation of 1.
(4) Marking a label;
the time series in the unitary time series set processed through the first three steps are all unified in format, and each time series sample is marked with a label in this step. The tag values are discrete, including normal, various fault types. And marking the corresponding label with the time sequence sample according to the actual running state of the equipment. By tagging the labels, the original unitary time series set initially entered becomes an equal length, equal interval, no missing value, normalized and tagged time series training set. The time series training set contains m time series samples, and each time series has a length of n.
Step three: and generating multiple views, namely combining each time series sample with feature vector splicing through time series transformation, and generating feature vectors of the multiple views.
The process of multi-view generation according to an embodiment of the present invention is shown in fig. 3.
A time-series sample symbol of length n is defined as s=(s) 1 ,s 2 ,…,s n ) Wherein s is i Is the ith sequence value of s. Firstly, K quick sequence transformations are carried out on the time sequence, and K transformed time sequences are obtained. Including but not limited to hilbert transforms, differential transforms of various orders, wavelet transforms, fourier transforms, and the like.
Then, the kernel number of each of the original sequence and the transformed sequence is C 1 ,C 2 Generates a corresponding eigenvector, and generates k+1 eigenvectors for each time-series sample. The convolution operation in the random convolution kernel transform is shown as equation (2), where the time series samples s=(s) 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
where x is the dot product operation, d is the expansion coefficient, l is the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
Each convolution check time sequence carries out sliding dot product operation to generate characteristicsA drawing. P kinds of pooled features are extracted from each feature map. Wherein features extracted from the feature map include, but are not limited to, positive scale, negative scale, zero scale, maximum, minimum, mean, etc. Whereby each time series is converted into K+1 eigenvectors, the eigenvectors corresponding to the original and transformed sequences respectively contain V 1 ·P,C 2 P features. The corresponding feature vectors of the original sequence are respectively spliced with the feature vectors of the K transformation sequences to obtain the feature vectors of the K views, and each feature vector comprises (C 1 +C 2 ) P features.
Step four: and (5) training a classifier.
According to the embodiment of the invention, each time series sample of the training set generates K views with dimension of m× (C 1 +C 2 ) A P-dimensional feature vector training set. Training a classification algorithm on the training set of each view to obtain classifiers, and obtaining K classifiers in total. The classification algorithms employed on the various views may be the same or different. Classification algorithms include, but are not limited to, decision trees, support vector machines, logistic regression, ridge regression, k-nearest neighbors, random forests, rotational forests, bayes, neural networks, and the like.
Step five: and predicting the label of the sample to be detected.
According to the embodiment of the invention, a time series sample of a sample to be detected is input after data preprocessing, a label of the sample is output after prediction, and the prediction process is shown in fig. 4. Firstly, generating multiple views to obtain feature vectors of K views. And then, respectively inputting the feature vectors of the K views into the trained classifiers, and outputting K prediction labels. And finally integrating the K prediction tag sets through integrated voting to obtain a final prediction tag. Methods employed for the integrated voting include, but are not limited to, hard voting, soft voting, weighted voting, and the like.
Another embodiment of the present invention proposes an industrial equipment fault diagnosis system based on a multi-view generation algorithm, the system comprising:
the data acquisition module is configured to acquire one or more operation state quantities of the industrial equipment, and sort the same operation state quantities according to time to form an original unitary time sequence set;
a preprocessing module configured to preprocess the original unitary time series set;
the feature set generation module is configured to generate multiple views of the preprocessed multiple time sequence samples in the original unitary time sequence set, and generate feature vector sets of the multiple views by combining time sequence transformation and feature vector splicing;
the classifier training module is configured to input the feature vector set of each view into a machine learning-based fault diagnosis classifier to obtain a trained fault diagnosis classifier;
the fault prediction module is configured to perform multi-view generation on the time sequence of the pre-processed running state sample to be detected, and obtain a corresponding feature vector; inputting the feature vector corresponding to the running state sample to be tested into a trained fault diagnosis classifier to obtain a fault prediction result of the running state sample to be tested; the specific process comprises the following steps: respectively inputting the feature vectors of the K views into the corresponding trained fault diagnosis classifier, and outputting K prediction labels; and integrating the K prediction labels through integrated voting to obtain a final prediction label.
Further, the running state quantity in the data acquisition module comprises temperature, rotating speed and pressure; the industrial equipment comprises engineering machinery, vehicles, machine tools, environment-friendly machinery, electrical equipment and electronic equipment; the preprocessing process in the preprocessing module comprises the following steps: filling the time sequence with the missing value by adopting an interpolation method; processing the unequal long-time sequences into equal-time sequences by adopting a sliding window method; carrying out standardization processing on each long-time sequence; labeling each time series sample after normalization.
Further, the specific process of generating the feature vector sets of the multiple views in the feature set generation module includes:
k quick sequence transformations are carried out on the plurality of time sequence samples, and K transformed time sequences are obtained; the sequence transformation comprises Hilbert transformation, differential transformation of each order, wavelet transformation or Fourier transformation;
the number of kernels of each of the plurality of time-series samples and the transformed time-series is C 1 ,C 2 Generates a corresponding eigenvector, and generates k+1 eigenvectors for each time-series sample. The method comprises the steps of carrying out a first treatment on the surface of the The method specifically comprises the following steps: each convolution check carries out sliding dot product operation on each time sequence to generate a feature map, and P pooling features are extracted from each feature map, wherein the pooling features comprise positive value proportion, negative value proportion, zero value proportion, maximum value, minimum value and average value;
performing row stitching on the feature vectors generated by the corresponding time sequence samples and the feature vectors generated by the corresponding transformed time sequence samples to obtain feature vectors of K views, wherein each feature vector comprises (C 1 +C 2 ) P features
Further, the time series samples s=(s) in the random convolution kernel transformation in the feature set generation module 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
where x is the dot product operation, d is the expansion coefficient, l represents the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
The function of an industrial equipment fault diagnosis system based on the multi-view generation algorithm according to the embodiment of the present invention may be described by the foregoing industrial equipment fault diagnosis method based on the multi-view generation algorithm, so that the system embodiment is not described in detail, and reference may be made to the above method embodiment, which is not described herein.
While the invention has been described with respect to a limited number of embodiments, those skilled in the art, having benefit of the above description, will appreciate that other embodiments are contemplated within the scope of the invention as described herein. The disclosure of the present invention is intended to be illustrative, but not limiting, of the scope of the invention, which is defined by the appended claims.
Claims (10)
1. An industrial equipment fault diagnosis method based on a multi-view generation algorithm is characterized by comprising the following steps of:
step one, collecting one or more running state quantities of industrial equipment, and sorting the same running state quantities according to time to form an original unitary time sequence set;
step two, preprocessing the original unitary time sequence set;
step three, carrying out multi-view generation on a plurality of time sequence samples in the preprocessed original unitary time sequence set, and generating a feature vector set of a plurality of views by combining time sequence transformation and feature vector splicing;
inputting the feature vector set of each view into a fault diagnosis classifier based on machine learning for training, and obtaining a trained fault diagnosis classifier;
step five, carrying out multi-view generation on the time sequence of the pre-processed running state sample to be detected, and obtaining a corresponding feature vector;
and step six, inputting the feature vector corresponding to the running state sample to be tested into a trained fault diagnosis classifier to obtain a fault prediction result of the running state sample to be tested.
2. The industrial equipment fault diagnosis method based on the multi-view generation algorithm according to claim 1, wherein in the first step, the operation state quantity includes temperature, rotation speed, and pressure; the industrial equipment comprises engineering machinery, vehicles, machine tools, environment-friendly machinery, electrical equipment and electronic equipment.
3. The industrial equipment fault diagnosis method based on the multi-view generation algorithm according to claim 1, wherein the preprocessing in the second step comprises: filling the time sequence with the missing value by adopting an interpolation method; processing the unequal long-time sequences into equal-time sequences by adopting a sliding window method; carrying out standardization processing on each long-time sequence; labeling each time series sample after normalization.
4. The industrial equipment fault diagnosis method based on the multi-view generation algorithm according to claim 1, wherein the specific process of the third step comprises:
k quick sequence transformations are carried out on the plurality of time sequence samples, and K transformed time sequences are obtained; the sequence transformation comprises Hilbert transformation, differential transformation of each order, wavelet transformation or Fourier transformation;
the number of kernels of each of the plurality of time-series samples and the transformed time-series is C 1 ,C 2 Generating corresponding feature vectors, and generating K+1 feature vectors for each time sequence sample; the method specifically comprises the following steps: each convolution check carries out sliding dot product operation on each time sequence to generate a feature map, and P pooling features are extracted from each feature map, wherein the pooling features comprise positive value proportion, negative value proportion, zero value proportion, maximum value, minimum value and average value;
performing row stitching on the feature vectors generated by the corresponding time sequence samples and the feature vectors generated by the corresponding transformed time sequence samples to obtain feature vectors of K views, wherein each feature vector comprises (C 1 +C 2 ) P features.
5. The industrial equipment fault diagnosis method based on the multi-view generation algorithm according to claim 4, wherein in the third step, the time series samples s=(s) in the random convolution kernel transformation 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
where x is the dot product operation, d is the expansion coefficient, l represents the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
6. The industrial equipment fault diagnosis method based on the multi-view generation algorithm according to claim 1, wherein the specific process in the step six comprises: respectively inputting the feature vectors of the K views into the corresponding trained fault diagnosis classifier, and outputting K prediction labels; integrating the K prediction tags through integrated voting to obtain a final prediction tag; wherein the method adopted by the integrated voting comprises hard voting, soft voting and weighted voting.
7. An industrial equipment fault diagnosis system based on a multi-view generation algorithm, comprising:
the data acquisition module is configured to acquire one or more operation state quantities of the industrial equipment, and sort the same operation state quantities according to time to form an original unitary time sequence set;
a preprocessing module configured to preprocess the original unitary time series set;
the feature set generation module is configured to generate multiple views of the preprocessed multiple time sequence samples in the original unitary time sequence set, and generate feature vector sets of the multiple views by combining time sequence transformation and feature vector splicing;
the classifier training module is configured to input the feature vector set of each view into a machine learning-based fault diagnosis classifier for training, and obtain a trained fault diagnosis classifier;
the fault prediction module is configured to perform multi-view generation on the time sequence of the pre-processed running state sample to be detected, and obtain a corresponding feature vector; inputting the feature vector corresponding to the running state sample to be tested into a trained fault diagnosis classifier to obtain a fault prediction result of the running state sample to be tested; the specific process comprises the following steps: respectively inputting the feature vectors of the K views into the corresponding trained fault diagnosis classifier, and outputting K prediction labels; and integrating the K prediction labels through integrated voting to obtain a final prediction label.
8. The industrial equipment fault diagnosis system based on the multi-view generation algorithm according to claim 7, wherein the operation state quantity in the data acquisition module comprises temperature, rotation speed and pressure; the industrial equipment comprises engineering machinery, vehicles, machine tools, environment-friendly machinery, electrical equipment and electronic equipment; the preprocessing process in the preprocessing module comprises the following steps: filling the time sequence with the missing value by adopting an interpolation method; processing the unequal long-time sequences into equal-time sequences by adopting a sliding window method; carrying out standardization processing on each long-time sequence; labeling each time series sample after normalization.
9. The industrial equipment fault diagnosis system based on the multi-view generation algorithm according to claim 7, wherein the specific process of generating the feature vector sets of the multiple views in the feature set generation module comprises:
k quick sequence transformations are carried out on the plurality of time sequence samples, and K transformed time sequences are obtained; the sequence transformation comprises Hilbert transformation, differential transformation of each order, wavelet transformation or Fourier transformation;
the number of kernels of each of the plurality of time-series samples and the transformed time-series is C 1 ,C 2 Generating corresponding feature vectors, and generating K+1 feature vectors for each time sequence sample; the method specifically comprises the following steps: each convolution check carries out sliding dot product operation on each time sequence to generate a feature map, and P pooling features are extracted from each feature map, wherein the pooling features comprise positive value proportion, negative value proportion, zero value proportion, maximum value, minimum value and average value;
performing row splicing on the feature vectors generated by the corresponding time sequence samples and the feature vectors generated by the corresponding transformed time sequence samples to obtain feature vectors of K views, wherein each view is provided with a plurality of time sequence samplesThe feature vector contains (C) 1 +C 2 ) P features.
10. The industrial equipment fault diagnosis system based on the multi-view generation algorithm according to claim 9, wherein the time series samples s=(s) in the random convolution kernel transformation in the feature set generation module 1 ,s 2 ,…,s n ) Ith value s i The dot product operation formula with the convolution kernel omega is as follows:
wherein, is the dot product operation, d is the expansion coefficient, I is the length of the convolution kernel ω, ω j The j-th value of the convolution kernel ω is represented, and b represents the coefficient of deviation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310597693.6A CN116561641A (en) | 2023-05-25 | 2023-05-25 | Industrial equipment fault diagnosis method and system based on multi-view generation algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310597693.6A CN116561641A (en) | 2023-05-25 | 2023-05-25 | Industrial equipment fault diagnosis method and system based on multi-view generation algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116561641A true CN116561641A (en) | 2023-08-08 |
Family
ID=87487861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310597693.6A Pending CN116561641A (en) | 2023-05-25 | 2023-05-25 | Industrial equipment fault diagnosis method and system based on multi-view generation algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116561641A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056734A (en) * | 2023-10-12 | 2023-11-14 | 山东能源数智云科技有限公司 | Method and device for constructing equipment fault diagnosis model based on data driving |
-
2023
- 2023-05-25 CN CN202310597693.6A patent/CN116561641A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056734A (en) * | 2023-10-12 | 2023-11-14 | 山东能源数智云科技有限公司 | Method and device for constructing equipment fault diagnosis model based on data driving |
| CN117056734B (en) * | 2023-10-12 | 2024-02-06 | 山东能源数智云科技有限公司 | Method and device for constructing equipment fault diagnosis model based on data driving |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Metric-based meta-learning model for few-shot fault diagnosis under multiple limited data conditions | |
| CN111721535B (en) | Bearing fault detection method based on convolution multi-head self-attention mechanism | |
| CN112036301B (en) | Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| CN113343855B (en) | Rolling bearing fault diagnosis system and method based on guide type sub-field self-adaption | |
| CN112729835B (en) | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing | |
| JP6991412B2 (en) | Search device, search method, search program and learning model search system | |
| CN112257862B (en) | Semi-supervised identification method based on relation network marking sample expansion | |
| CN116561641A (en) | Industrial equipment fault diagnosis method and system based on multi-view generation algorithm | |
| CN112819059A (en) | Rolling bearing fault diagnosis method based on popular retention transfer learning | |
| CN115577248A (en) | Fault diagnosis system and method for wind generating set | |
| CN115358259A (en) | Self-learning-based unsupervised cross-working-condition bearing fault diagnosis method | |
| CN114564987A (en) | Rotary machine fault diagnosis method and system based on graph data | |
| CN111860447A (en) | Method for accurately identifying telemetry time sequence data mode of random phase shift satellite | |
| CN110686897A (en) | Variable working condition rolling bearing fault diagnosis method based on subspace alignment | |
| Lu et al. | Rotating Machinery Fault Diagnosis Under Multiple Working Conditions via A Time Series Transformer Enhanced by Convolutional Neural Network | |
| CN116894215B (en) | Gear box fault diagnosis method based on semi-supervised dynamic graph attention | |
| CN116034321A (en) | Interpreting machine learning output in industrial applications | |
| Sipola et al. | Gear classification and fault detection using a diffusion map framework | |
| CN114926702B (en) | Small sample image classification method based on depth attention measurement | |
| EP3690751A1 (en) | A method for building a deep latent feature extractor for industrial sensor data | |
| Lu et al. | An unsupervised intelligent fault diagnosis system based on feature transfer | |
| CN116756483B (en) | Mechanical fault diagnosis method, device and equipment under condition that target working condition data are unavailable | |
| Lyu et al. | A Novel Fault Diagnosis Method Based on Feature Fusion and Model Agnostic Meta-Learning | |
| Zhang et al. | TSViT: A Time Series Vision Transformer for Fault Diagnosis |
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 |