Equipment fault diagnosis method for adaptively updating model weight

By constructing an auxiliary classification model to determine the accuracy of the fault diagnosis model and updating the model weights only when necessary, the performance degradation problem of the model when the equipment operating conditions change in the existing technology is solved, and efficient and accurate fault diagnosis is achieved.

CN118916743AActive Publication Date: 2024-11-08ZHEJIANG UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411028507.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-08
Estimated Expiration
2044-07-30

AI Technical Summary

Technical Problem

Existing fault diagnosis methods suffer from performance degradation when equipment operating conditions change, leading to unnecessary or untimely model updates, wasting computational resources, and potentially producing erroneous diagnostic results.

Method used

By constructing an auxiliary classification model, the accuracy of the fault diagnosis model is judged by the classification accuracy and classification energy. The model weights are updated only when necessary, and the domain adaptation method is used for adaptive updating.

Benefits of technology

Reduce unnecessary model updates, improve the accuracy and efficiency of fault diagnosis, and ensure timely and accurate identification of faults when equipment status changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916743A_ABST
    Figure CN118916743A_ABST
Patent Text Reader

Abstract

The invention discloses an equipment fault diagnosis method for adaptively updating a model weight. The method comprises the following steps: firstly, collecting historical data of equipment under a plurality of working conditions, performing data preprocessing to obtain a plurality of auxiliary data sets and test data sets, and constructing an auxiliary classification model by using classification accuracy and classification energy of the auxiliary data sets; the prediction classification accuracy of the fault diagnosis model is determined based on the classification energy of the test data set, whether the model weight is updated or not is determined according to the predicted classification accuracy, self-adaptive model weight updating is achieved, and therefore high-accuracy diagnosis of equipment faults is guaranteed. According to the method, unnecessary model updating can be reduced, the domain adaptation method is operated only when necessary, the fault diagnosis process can be accelerated, and therefore equipment in a fault state can be checked and overhauled timely and effectively.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of rotating equipment fault diagnosis, and relates to an equipment fault diagnosis method in the fields of machine learning, deep learning and time series classification, and specifically relates to an equipment fault diagnosis method for adaptively updating model weights. Background Art

[0002] In today's industry, rotating equipment is widely used, and its complexity and precision are constantly increasing, such as aircraft engines and gas turbines. However, when these rotating equipment have faults such as bearing wear or blade breakage, serious accidents may occur, resulting in huge economic losses. Therefore, timely and accurate identification of equipment operating status and taking measures when signs of failure appear are crucial to improving production efficiency and reducing catastrophic losses.

[0003] Currently, a large number of research and patents on fault diagnosis have emerged, among which the most popular ones are data-driven methods. These methods reveal the relationship between vibration signals and equipment status through the analysis of large-scale data. For example, patent CN113720605B uses a one-dimensional convolutional neural network and a fully connected neural network to extract vibration features, and uses ensemble learning technology to obtain diagnostic results from multiple classifiers. Patent CN112199888B uses a deep residual network to extract fault features from vibration signals.

[0004] Although existing methods have achieved a fairly high level of diagnostic accuracy, they require that the characteristics of the data used to train the model and the characteristics of the data encountered when the model is actually working follow the same distribution, which is different from the actual situation, because when the equipment conditions change, the data characteristics will also change. When the data characteristics are inconsistent, the performance of the model will be seriously degraded. Therefore, it is necessary to frequently test the accuracy of the model's classification of new data to ensure that the model's diagnostic results are still in an acceptable state. This includes humans labeling new data and comparing it with the model's classification results.

[0005] Existing fault diagnosis methods mainly adopt regular update methods, such as running domain adaptation methods (such as domain adversarial neural networks, DANN) every few days to update the weights of the diagnosis model to adapt it to the latest data distribution. Due to the irregularity of data feature distribution changes, the regular update method has unnecessary or untimely model updates, which wastes computing resources and may result in incorrect model diagnosis results. Summary of the invention

[0006] In order to solve the problems and needs existing in the background technology, the present invention proposes an equipment fault diagnosis method for adaptively updating model weights.

[0007] The technical solution of the present invention is as follows:

[0008] 1. A method for equipment fault diagnosis with adaptive updating of model weights

[0009] S1: After collecting historical data of the equipment under multiple working conditions and performing data preprocessing, several auxiliary data sets and test data sets are obtained; among which, the auxiliary data set is a collection of multi-label and multi-working condition data.

[0010] S2: input all auxiliary data sets into the fault diagnosis model respectively, so as to determine the classification accuracy corresponding to each auxiliary data set, and calculate the classification energy corresponding to all auxiliary data sets;

[0011] S3: Use the classification accuracy and classification energy of all auxiliary datasets to build an auxiliary classification model;

[0012] S4: Calculate the classification energy corresponding to all test data sets, and then use the auxiliary classification model to calculate the prediction classification accuracy corresponding to all test data sets;

[0013] S5: Determine whether the accuracy of the current fault diagnosis model meets the equipment accuracy based on the predicted classification accuracy of the test data set. If not, update the weight of the fault diagnosis model and repeat S2-S4 until the accuracy of the updated fault diagnosis model meets the equipment accuracy to obtain an updated fault diagnosis model; otherwise, keep the model weight unchanged.

[0014] S6: Repeat S1-S5 at a preset interval to adaptively update the weight of the fault diagnosis model, thereby ensuring high-accuracy diagnosis of equipment faults.

[0015] In S2, the calculation formula of the classification energy e of each auxiliary data set is as follows:

[0016]

[0017] Where N represents the amount of data in each auxiliary dataset, x i 、x j Respectively represent the i-th and j-th data in the current auxiliary data set, f k (x i ) represents the model's response to data x i The output belongs to the kth label, T is the energy parameter, and K represents the number of label types in the current auxiliary dataset.

[0018] The S3 is specifically:

[0019] The classification accuracy acc and classification energy e of each auxiliary data set are regarded as a point, and then a linear function is obtained by fitting the points of all auxiliary data sets and recorded as the auxiliary classification model.

[0020] In S4, the classification energy of each test data set is calculated in the same manner as the classification energy of the auxiliary data set.

[0021] In S5, if there are multiple test data sets whose predicted classification accuracy is less than the preset accuracy, the accuracy of the current fault diagnosis model does not meet the equipment accuracy; if the predicted classification accuracy of all test data sets is greater than the preset accuracy, the accuracy of the current fault diagnosis model meets the equipment accuracy.

[0022] 2. A computer device

[0023] The computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.

[0024] 3. A computer-readable storage medium

[0025] The storage medium stores a computer program, which implements the steps of the method when executed by a processor.

[0026] IV. A COMPUTER PROGRAM PRODUCT

[0027] The product comprises a computer program / instructions which, when executed by a processor, implement the steps of the method.

[0028] The beneficial effects of the present invention are:

[0029] The present invention determines the classification accuracy of the fault diagnosis model through an auxiliary classification model. If the accuracy is less than a preset threshold, the domain adaptation fault diagnosis method is used to adjust the model parameters. Otherwise, the model weights are kept unchanged. The present invention can reduce unnecessary model updates and only run the domain adaptation method when necessary, which can speed up the fault diagnosis process, so that equipment in a faulty state can be checked and repaired in a timely and effective manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic flow chart of the specific steps of the present invention. DETAILED DESCRIPTION

[0031] The present invention is further described below using the data of the rolling bearing test bench (DIRG) of the Polytechnic University of Turin as a specific embodiment. The data set can be obtained through the FTP protocol.

[0032] DIRG is a high-speed aviation bearing dataset. It provides information about the degree of bearing damage and its acceleration signals at different speeds and radial loads. The bearing is divided into three states: normal, inner race fault, and outer race fault. Inner race fault and outer race fault can be further subdivided into three levels: 150, 250, and 450 μm, resulting in a total of 7 states for the bearing. Sensors are placed at two locations on the bearing to collect acceleration signals on the x, y, and z axes, resulting in a total of 6 channels of data. The dataset covers 17 operating conditions, corresponding to 17 fields. In each field, data collection is performed 7 times, corresponding to the 7 states of the bearing. This results in a total of 119 sub-datasets. Data labels are assigned according to the state of the bearing. Each sub-dataset is obtained by sampling 10 seconds of data from the sensor at a frequency of 51200 Hz. In this embodiment, data from 7 operating conditions in the dataset are used. Each of these 7 operating conditions has a unique speed or load and is assigned an index as an identifier, as shown in Table 1.

[0033] Table 1 shows the 7 working conditions in the data set.

[0034]

[0035]

[0036] In this embodiment, Resnet is used as the fault diagnosis model, denoted as f, and data set No. 1 (denoted as D1) is selected as the training set training model. The following data processing process is performed on D1: Use a sliding window to cut each data into multiple non-overlapping parts, all with a length of 1024 and the same label. Use min-max scaling to standardize these data. In actual use, there is no restriction that the model must use Resnet.

[0037] like Figure 1 As shown, the equipment fault diagnosis method for adaptively updating model weights proposed by the present invention comprises the following steps:

[0038] S1: Datasets 2, 3, 4, 5, and 6 are considered as auxiliary data sets, and are recorded as D2, D3, D4, D5, and D6 respectively. The same data preprocessing method as the model training set (D1) is used: each data is divided into multiple non-overlapping parts with a length of 1024 and the same label using a sliding window. These data are standardized using min-max scaling.

[0039] S2: Input all auxiliary data sets into the fault diagnosis model to obtain the classification results of the data, which are recorded as c2, c3, c4, c5, c6, and compare the classification results with the actual labels. Count the correct classification results on each auxiliary data set and divide by the total data volume of the data set to obtain the classification accuracy, which are recorded as acc2, acc3, acc4, acc5, acc6. Data sets 7, 8, and 9 are regarded as test data sets, recorded as D7, D8, and D9. D7, D8, and D9 are input into the fault diagnosis model f to obtain the classification results of the data set, which are recorded as c7, c8, and c9.

[0040] In this embodiment, the classification accuracy rates acc of the model on the auxiliary data set are:

[0041] acc2=97.32, acc3=95.41, acc4=92.66, acc5=90.35, acc6=91.20

[0042] The calculation formula of the classification energy e of each data set is as follows:

[0043]

[0044] Where N represents the amount of data in each auxiliary data set, and N is 50. i 、x j Respectively represent the i-th and j-th data in the current auxiliary data set, f k (x i ) is the model for data x i The output of the kth label, T is the energy parameter, a hyperparameter, and T is 1. K represents the number of label types in the current auxiliary data set. In this embodiment, there are 7 types of data, so K is 7. The classification energies of the 5 auxiliary data sets are recorded as e2, e3, e4, e5, and e6, and the classification energies of the 3 test data sets are recorded as e7, e8, and e9. The specific values ​​are as follows:

[0045] e2=4.87, e3=5.56, e4=6.57, e5=7.41, e6=7.10,

[0046] e7=7.55, e8=7.11, e9=7.78

[0047] S3: Use the classification accuracy and classification energy of all auxiliary datasets to build an auxiliary classification model;

[0048] S3 is specifically:

[0049] The classification accuracy acc and classification energy e of each auxiliary data set are regarded as a point, and then a linear function is obtained by fitting the points of all auxiliary data sets and recorded as the auxiliary classification model. The formula is as follows:

[0050] acc = -2.739229422066556e+110.64828371278466

[0051] S4: Substitute the classification energy of the three test data sets into the auxiliary classification model to obtain the predicted classification accuracy. Substitute e7=7.45 into the above function to obtain the classification accuracy acc7=89.97 predicted by the model on data set D7, substitute e8=7.11 into the above function to obtain the classification accuracy acc8=91.17 predicted by the model on data set D8, and substitute e9=7.7 into the above function to obtain the classification accuracy acc9=89.56 predicted by the model on data set D9.

[0052] S5: In the above results, acc7 and acc9 are less than the set threshold of 90, and the fault diagnosis model weights are updated using the running domain adversarial neural network method. After the updated model is completed, steps S2 to S4 are re-executed, and the new classification accuracy rates are acc7 = 94.7, acc8 = 90.41, and acc9 = 93.56, which are all greater than the set threshold, and the update is completed.

[0053] S6: Repeat S1-S5 at preset intervals to adaptively update the weight of the fault diagnosis model. That is, the weight update time of the fault diagnosis model can be flexibly determined according to different working conditions or equipment status, and the fault diagnosis model can be managed more conveniently, thereby ensuring high-accuracy diagnosis of equipment faults. At the same time, equipment in a faulty state can be checked and repaired in a timely and effective manner.

[0054] The present invention determines the classification accuracy of the fault diagnosis model through the auxiliary classification model. If the accuracy is less than the preset threshold, the domain adaptation fault diagnosis method is used to adjust the model parameters. Otherwise, the model weights are kept unchanged. Therefore, the present invention can reduce unnecessary model updates and only run the domain adaptation method when necessary, which can speed up the fault diagnosis process.

[0055] The above-mentioned embodiment is an implementation method of the present invention on the rolling bearing test bench data set of Ling Polytechnic University, but the specific implementation of the method of the present invention is not limited to bearings. Any similar scheme that collects equipment operation data through sensors and performs model evaluation in accordance with the principles and ideas of the present invention should be regarded as the protection scope of the patent of the present invention.

Claims

1. A method for equipment fault diagnosis by adaptively updating model weights, characterized in that: The following steps are involved: S1: After collecting historical data of the equipment under multiple working conditions and performing data preprocessing, several auxiliary data sets and test data sets are obtained; S2: input all auxiliary data sets into the fault diagnosis model respectively, so as to determine the classification accuracy corresponding to each auxiliary data set, and calculate the classification energy corresponding to all auxiliary data sets; S3: Use the classification accuracy and classification energy of all auxiliary datasets to build an auxiliary classification model; S4: Calculate the classification energy corresponding to all test data sets, and then use the auxiliary classification model to calculate the prediction classification accuracy corresponding to all test data sets; S5: judging whether the accuracy of the current fault diagnosis model meets the equipment accuracy according to the prediction classification accuracy of the test data set; if not, updating the weight of the fault diagnosis model to obtain an updated fault diagnosis model; S6: Repeat S1-S5 at a preset interval to adaptively update the weight of the fault diagnosis model, thereby ensuring high-accuracy diagnosis of equipment faults.

2. The equipment fault diagnosis method for adaptively updating model weights according to claim 1, characterized in that: In S2, the calculation formula of the classification energy e of each auxiliary data set is as follows: Where N represents the amount of data in each auxiliary dataset, x i 、x j Respectively represent the i-th and j-th data in the current auxiliary data set, f k (x i ) represents the model's response to data x i The output belongs to the kth label, T is the energy parameter, and K represents the number of label types in the current auxiliary dataset.

3. The equipment fault diagnosis method for adaptively updating model weights according to claim 1, characterized in that: The S3 is specifically: The classification accuracy acc and classification energy e of each auxiliary data set are regarded as a point, and then a linear function is obtained by fitting the points of all auxiliary data sets and recorded as the auxiliary classification model.

4. The equipment fault diagnosis method for adaptively updating model weights according to claim 1, characterized in that: In S4, the classification energy of each test data set is calculated in the same manner as the classification energy of the auxiliary data set.

5. The equipment fault diagnosis method for adaptively updating model weights according to claim 1, characterized in that: In S5, if there are multiple test data sets whose predicted classification accuracy is less than the preset accuracy, the accuracy of the current fault diagnosis model does not meet the equipment accuracy; if the predicted classification accuracy of all test data sets is greater than the preset accuracy, the accuracy of the current fault diagnosis model meets the equipment accuracy.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Citation Information

Patent Citations

  • A method, system, and readable storage medium for fault diagnosis of rotating equipment based on deep residual networks.

    CN112199888B

  • A Machine Learning-Based Fault Diagnosis Method for Rotary Cement Production Equipment

    CN113720605B

  • A lung nodule auxiliary detection system and method based on an improved convex hull method

    CN109300113A

  • Deep adversarial diagnosis method for fan bearing fault under non-equilibrium small sample scene

    CN110567720A

  • Fault diagnosis method for deep convolution sparse automatic encoder of rolling bearing

    CN113255437A