Mechanical composite fault diagnosis method and system based on multi-model collaborative decision, and medium

Abstract

The invention belongs to the related technical field of mechanical fault diagnosis, and particularly discloses a mechanical composite fault diagnosis method and system based on multi-model collaborative decision, and a medium, and the method comprises the steps: expanding a to-be-diagnosed equipment time sequence signal into Y samples through sliding window sampling; respectively inputting the Y samples into X pre-trained fault detection models to obtain fault detection results; correspondingly converting the fault detection result into a fault label; model fault tags are determined based on the sample number corresponding to the fault tags, and two decision tags are selected from the model fault tags; if the difference between the number of the fault detection models corresponding to the two decision labels is not smaller than a preset threshold value, taking the decision label corresponding to a larger number of the fault detection models as a final fault label; and otherwise, taking the decision tag corresponding to more samples as a final fault tag. According to the invention, multi-model collaborative mechanical composite fault diagnosis and detection are realized, end-to-end diagnosis can be carried out, and the accuracy of fault detection is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field related to mechanical fault diagnosis, and more specifically, relates to a mechanical composite fault diagnosis method, system and medium based on multi-model collaborative decision-making. Background Art

[0002] In modern industrial production, mechanical equipment plays an important role in the fields of machinery manufacturing, aerospace, transportation, etc., and its stable operation is directly related to production safety and efficiency. In the actual operation of modern intelligent equipment, due to factors such as material fatigue, improper assembly, and lubrication failure, various faults such as bearing damage, gear wear, and rotor imbalance are prone to occur. These problems will not only reduce equipment performance and bring economic losses, but may even cause major safety accidents.

[0003] In actual industrial production environments, although fault mechanism-based diagnosis provides an in-depth understanding of the root causes of faults in theory, it faces the challenges of high model complexity and difficulty in modeling. Diagnostic methods based on traditional signal processing rely on the experience of engineers in terms of feature extraction and selection, threshold setting, etc., lack universally applicable criteria, and it is difficult to predict the trend of fault development.

[0004] With the further development of sensor technology and big data technology, the number of data samples extracted in industry has gradually increased, and data-driven fault diagnosis methods have a good application prospect. Deep learning has been widely used in mechanical fault diagnosis due to its ability of automatic feature learning and extraction, strong expression ability, and adaptive ability. Deep learning uses multi-layer nonlinear transformation of neural networks to enable it to have a high degree of expression ability and can process complex fault signals and patterns, thereby improving diagnostic accuracy. At the same time, the deep learning model is an end-to-end learning model, that is, it realizes the output from raw input data to fault diagnosis results, which greatly simplifies the complexity of multiple independent processing steps in traditional diagnostic systems and improves the efficiency of the diagnostic process.

[0005] In the existing technology of using deep learning for fault diagnosis, the accuracy of fault diagnosis is low in some cases due to the randomness of the decision-making of a single deep learning model. In addition, due to the large consumption of computing resources required for training deep learning models, it is difficult to promote them in actual industrial diagnosis tasks. Summary of the invention

[0006] In view of the above defects or improvement needs of the prior art, the present invention provides a mechanical composite fault diagnosis method, system and medium based on multi-model collaborative decision-making, the purpose of which is to improve the accuracy of equipment fault diagnosis.

[0007] To achieve the above object, according to a first aspect of the present invention, a mechanical compound fault diagnosis method based on multi-model collaborative decision-making is proposed, comprising the following steps:

[0008] Expand the timing signal of the device to be diagnosed to Y samples through sliding window sampling;

[0009] Input Y samples into X pre-trained fault detection models respectively. Each fault detection model performs fault detection on each sample, and obtains X×Y fault detection results in total.

[0010] The fault detection results are converted into fault labels. The fault detection results reflect the probability that the samples belong to various types of faults, and the fault labels reflect the fault categories.

[0011] For any fault detection model: Y input samples correspond to Y fault labels. Based on the number of samples corresponding to the fault labels, the model fault labels are determined; thus, the model fault labels of all fault detection models are obtained;

[0012] Select the two model fault labels with the largest number of occurrences from all model fault labels as decision labels, and determine the number of fault detection models and the number of samples corresponding to the two decision labels respectively;

[0013] If the difference in the number of fault detection models corresponding to two decision labels is not less than the preset threshold, the decision label with the larger number of corresponding fault detection models will be used as the final fault label; otherwise, the decision label with the larger number of corresponding samples will be used as the final fault label; the fault category reflected by the final fault label is the equipment fault category.

[0014] As a further preferred method, converting the fault detection result into a fault label includes:

[0015] For any fault detection result, the probability that the sample reflected by the fault detection result belongs to each type of fault is greater than 50%, and the corresponding fault category is retained to form a fault label, which is a single fault or a compound fault.

[0016] As a further preferred method, determining the model fault label based on the number of samples corresponding to the fault label includes:

[0017] The fault categories corresponding to the Y fault labels are split into single faults, and the corresponding sample numbers are counted; all single faults with a sample number greater than Y / 2 are screened and retained. If there is no single fault sample number greater than Y / 2, the single fault with the largest sample number is retained; the retained single faults are combined to construct the model fault label.

[0018] As a further preferred embodiment, after obtaining all model fault labels, if no more than half of the model fault labels are of the same compound fault, the detected compound fault is regarded as several single faults, and the samples corresponding to the compound fault are used as samples of the corresponding single fault, and subsequent decision labels are determined.

[0019] As a further preference, there are 7 fault detection models, which are respectively constructed based on TFSCL, MSRCN, CFSPT, CFSPT_SD, HFPT, MSMCNN_SD, and Props2.

[0020] As a further preferred embodiment, pre-training X fault detection models includes: acquiring a time series signal of each fault category, and obtaining corresponding samples through sliding window sampling, thereby constructing a training set, and training each fault detection model through the training set.

[0021] As a further preferred embodiment, when determining the final fault label, if the number of fault detection models and the number of samples corresponding to two decision labels are the same, or the number of samples corresponding to two decision labels is less than Y / 2, the final fault label is decided manually.

[0022] As further preferred, the timing signal is a vibration, sound or rotation speed signal of the equipment.

[0023] According to a second aspect of the present invention, a mechanical complex fault diagnosis system based on multi-model collaborative decision-making is provided, comprising a processor, wherein the processor is used to execute the above-mentioned mechanical complex fault diagnosis method based on multi-model collaborative decision-making.

[0024] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the mechanical complex fault diagnosis method based on multi-model collaborative decision-making is implemented.

[0025] In general, the above technical solution conceived by the present invention has the following technical advantages compared with the prior art:

[0026] 1. The present invention designs a multi-sample and multi-model collaborative fault diagnosis; by converting equipment timing signals into multi-sample batch decisions and integration, the comprehensiveness and complementarity of information perception are effectively improved, and the generalization performance of the model is significantly improved; at the same time, unlike the traditional single model strategy, the present invention adopts multiple detection models to collaboratively fuse decisions, which significantly reduces the risk of overfitting of a single model to specific data and improves the accuracy of equipment fault diagnosis.

[0027] 2. The present invention uses several lightweight detection models, and the training process does not involve any target domain information. After the model is trained using historical data, it can directly perform diagnosis without any domain adaptation and transfer learning process to perform secondary processing on the test data, which has high real-time performance. At the same time, it can be very conveniently embedded in a distributed training framework, and multiple training devices can be used to speed up the overall training time and reduce computing resource consumption.

[0028] 3. The fault diagnosis technology of the present invention is based on data-driven deep learning, which can achieve end-to-end diagnosis without complex expert knowledge, effectively solving problems such as high mobility of experts and difficulty in accumulating deep business knowledge in vertical industry domains. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the structure of a mechanical compound fault diagnosis method based on multi-model collaborative decision-making according to an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of sliding window sampling according to an embodiment of the present invention;

[0031] Figure 3 A schematic diagram of a fault detection result and a fault label according to an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of a multi-model collaborative decision-making process according to an embodiment of the present invention;

[0033] Figure 5 This is a flowchart of multi-model voting rules according to an embodiment of the present invention;

[0034] Figure 6 (a) and (b) are schematic diagrams of the fault decision process of an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0036] The embodiment of the present invention provides a mechanical composite fault diagnosis method based on multi-model collaborative decision-making, such as Figure 1 and Figure 4 As shown, the following steps are included:

[0037] S1. Pre-train the fault detection model before diagnosis.

[0038] The time series signals of each fault category are collected through industrial sensors, and the data is stored as a csv file according to the sensor channel; the time series signal data in the known fault type file is sampled through a sliding window to obtain the corresponding samples, thereby constructing a training set; X fault detection models are trained respectively through the training set. Specifically, the time series signal is the vibration, sound or speed signal of the equipment.

[0039] Furthermore, there are a total of 7 fault detection models in this embodiment, including 4 CNN models and 3 Transformer models, specifically TFSCL, MSRCN, CFSPT, CFSPT_SD, HFPT, MSMCNN_SD, and Props2. Compared with mainstream models, these 7 models have the characteristics of smaller size, fast calculation speed, high efficiency and high accuracy.

[0040] S2. Obtain the timing signal of the device to be diagnosed, and expand the timing signal of the device to be diagnosed into Y samples through sliding window sampling; input the Y samples into the pre-trained X fault detection models respectively, and each fault detection model performs fault detection on each sample, and a total of X×Y fault detection results are obtained.

[0041] Specifically, the sliding window sampling in S1 and S2 is as follows Figure 2 As shown, the length of each sample in the sample unit, the step length is the length of the interval between the previous sample and the next sample, and the overlapping part represents the tail of the previous sample and the head of the next sample. This method can make use of the limited data set as much as possible to generate more samples.

[0042] S3. Based on the X×Y fault detection results obtained by the X fault detection models, the fault type of the equipment is determined according to the established voting rules.

[0043] Specifically, the voting rules of S3 mainly include four parts: preliminary processing of decision data, failure mode statistics, preliminary failure decision and final failure decision:

[0044] S31. Preliminary processing of decision data: converting the fault detection results into fault labels. The fault detection results reflect the probability that the samples belong to various types of faults, and the fault labels reflect the fault categories.

[0045] The specific rules for converting fault detection results into fault labels are as follows: for any fault detection result, the probability that the sample reflected by the fault detection result belongs to each type of fault is determined. If the probability exceeds 50%, the corresponding fault category will be retained to form a fault label; the fault label is a single fault or a compound fault.

[0046] The above process can be performed using a composite fault decoupler based on the Sigmoid function, such as Figure 3As shown in the figure, the deep learning model represents the fault detection model. The different shapes of the output icons represent different faults. The following P represents the probability of the corresponding fault. The output of the composite fault decoupling classifier is the fault category judged by the model. Figure 3 The output is a composite fault consisting of two faults.

[0047] S32, Fault mode statistics: For any fault detection model, Y input samples correspond to Y fault labels, and the model fault labels are determined based on the number of samples corresponding to the fault labels; accordingly, the model fault labels corresponding to the X fault detection models are determined respectively.

[0048] Furthermore, based on the number of samples corresponding to the fault labels, the model fault labels are determined, including: splitting the fault categories corresponding to the Y fault labels into single faults and counting the corresponding number of samples; screening and retaining all single faults with a sample number greater than Y / 2, if there is no single fault sample number greater than Y / 2, retaining the single fault with the largest sample number; and combining the retained single faults to construct the model fault label.

[0049] S33. Preliminary fault decision: Select the two model fault labels with the largest number of occurrences from all model fault labels, namely, decision labels, and determine the number of fault detection models and the number of samples corresponding to the two decision labels respectively.

[0050] Furthermore, after obtaining all the model fault labels, if no more than half of the model fault labels are of the same compound fault, the detected compound fault will be regarded as the corresponding several single faults, and the samples corresponding to the compound fault will be used as samples of several single faults to participate in the sample statistics of related single faults. After adding them up, the preliminary fault decision will be executed to determine the decision label, so as to avoid the situation where the single fault decision is abnormal due to the compound fault statistics.

[0051] S34, final fault decision: If the difference between the number of fault detection models corresponding to the two model fault labels is not less than the preset threshold, the model fault label with a larger number of corresponding fault detection models is used as the final fault label; otherwise, the model fault label with a larger number of corresponding samples is used as the final fault label. The fault category reflected by the final fault label is the equipment fault category, and the final decision result is recorded and converted into the required csv format.

[0052] In order to make the present invention more clear, specific embodiments are described below.

[0053] First, the timing signal of the device to be diagnosed is expanded to 200 samples through sliding window sampling, and the 200 samples are input into 7 models respectively. Each model performs 200 fault diagnosis, and a total of 1400 fault detection results are obtained. Figure 5As shown in the figure, the fault type of the device is determined according to the voting rules, including:

[0054] S31. Perform data statistics on the 200 fault detection results of each model according to a threshold of 0.5, and convert the probability into a fault label. For example, the continuous numerical probability output by the model is [0.98, 0, 0.86], which is converted into a label (1, 3), indicating that it is a compound fault of fault 1 and fault 3; 7 models produce 7 csv files.

[0055] S32. For any fault detection model, among the 200 fault labels, the compound faults are split into single faults, and their corresponding sample numbers are also corresponding to the single faults. Then, sample statistics are performed together with other single faults to screen out all single faults with sample numbers greater than 100 (if there are none, take the single fault with the largest sample number) and combine them into model fault labels.

[0056] S33, select the two model fault labels with the most occurrences from the seven model fault labels as decision labels, and determine the number of fault detection models and the number of samples corresponding to the decision labels. For compound fault labels, the number of samples corresponding to each single fault may be inconsistent, and the average value can be taken for subsequent statistics.

[0057] After the above process, the Figure 6 In the document shown, each line represents the diagnosis result of a timing signal, that is, Figure 6 In fact, it shows the results of the above diagnostic process multiple times. The blue part represents the model fault label and the corresponding number of decisions (i.e. the corresponding number of samples) obtained from the 200 decision statistics of each model; the green part combines the two decision labels of the 7 model results; the pink part represents the number of corresponding models predicted as the decision label category, with a maximum value of 7; the purple part represents the number of times the decision label is predicted in all models, with a maximum value of 1400.

[0058] S34. Combine the final votes of different models and the number of all samples to make a collaborative decision and determine the final fault label:

[0059] (1) For two decision labels, if the model voting gap is not less than 2, the number of models counted shall prevail (i.e., Max = 7 voting rounds).

[0060] For example: Figure 6 In the first row, 6 models identify it as [8,9], and 1 model identifies it as [8,9,13]. Since 6-1 is greater than 2, the final fault label is judged to be [8,9].

[0061] (2) For two decision labels, if the model voting gap is less than 2, the number of valid decision samples shall prevail (i.e., voting rounds with Max = 1400).

[0062] For example: If 3 models identify the decision label as A and 3 models identify the decision label as B, then compare the total number of samples of A and B. If 3A(332)>3B(256), the decision is A. If 2 models identify the decision label as A and 3 models identify the decision label as B, then compare the total number of samples of A and B. If 2A(320)>3B(300), the decision is A.

[0063] (3) When extreme situations arise that cannot be resolved by the above mechanisms, human decision-making can be used.

[0064] For example, if the number of votes and the number of samples of the model are the same, or all decisions are less than 100, they are considered invalid decisions. For example, 1A(80)=1B(80), which may mean that this type is an unknown fault category that has never appeared.

[0065] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A mechanical composite fault diagnosis method based on multi-model collaborative decision-making, characterized in that: The steps include: Expand the timing signal of the device to be diagnosed to Y samples through sliding window sampling; Input Y samples into X pre-trained fault detection models respectively. Each fault detection model performs fault detection on each sample, and obtains X×Y fault detection results in total. The fault detection results are converted into fault labels. The fault detection results reflect the probability that the samples belong to various types of faults, and the fault labels reflect the fault categories. For any fault detection model: Y input samples correspond to Y fault labels. Based on the number of samples corresponding to the fault labels, the model fault labels are determined; thus, the model fault labels of all fault detection models are obtained; Select the two model fault labels with the largest number of occurrences from all model fault labels as decision labels, and determine the number of fault detection models and the number of samples corresponding to the two decision labels respectively; If the difference between the number of fault detection models corresponding to two decision labels is not less than a preset threshold, the decision label with a larger number of corresponding fault detection models is used as the final fault label; Otherwise, the decision label with a larger number of corresponding samples is used as the final fault label; the fault category reflected by the final fault label is the equipment fault category.

2. The mechanical composite fault diagnosis method based on multi-model collaborative decision-making according to claim 1 is characterized in that: The fault detection results are converted into fault labels, including: For any fault detection result, the probability that the sample reflected by the fault detection result belongs to each type of fault is greater than 50%, and the corresponding fault category is retained to form a fault label, which is a single fault or a compound fault.

3. The mechanical composite fault diagnosis method based on multi-model collaborative decision-making according to claim 2 is characterized in that: Based on the number of samples corresponding to the fault label, determine the model fault label, including: The fault categories corresponding to the Y fault labels are split into single faults, and the corresponding sample numbers are counted; all single faults with a sample number greater than Y / 2 are screened and retained. If there is no single fault sample number greater than Y / 2, the single fault with the largest sample number is retained; the retained single faults are combined to construct the model fault label.

4. The mechanical composite fault diagnosis method based on multi-model collaborative decision-making according to claim 2 is characterized in that: After all model fault labels are obtained, if no more than half of the model fault labels are of the same compound fault, the detected compound fault will be regarded as several single faults, and the samples corresponding to the compound fault will be used as the samples of the corresponding single fault for subsequent decision label determination.

5. The mechanical compound fault diagnosis method based on multi-model collaborative decision-making according to claim 1 is characterized in that: There are a total of 7 fault detection models, which are built based on TFSCL, MSRCN, CFSPT, CFSPT_SD, HFPT, MSMCNN_SD, and Props2 respectively.

6. The mechanical compound fault diagnosis method based on multi-model collaborative decision-making according to claim 1 is characterized in that: Pre-training X fault detection models includes: obtaining time series signals of each fault category, and obtaining corresponding samples through sliding window sampling, thereby constructing a training set, and training each fault detection model through the training set.

7. The mechanical compound fault diagnosis method based on multi-model collaborative decision-making according to claim 1 is characterized in that: When determining the final fault label, if the number of fault detection models and the number of samples corresponding to the two decision labels are the same, or the number of samples corresponding to the two decision labels is less than Y / 2, the final fault label is decided manually.

8. The mechanical compound fault diagnosis method based on multi-model collaborative decision-making according to any one of claims 1 to 7, characterized in that: The timing signal is a vibration, sound or rotation speed signal of the device.

9. A mechanical composite fault diagnosis system based on multi-model collaborative decision-making, characterized in that: It includes a processor, which is used to execute the mechanical complex fault diagnosis method based on multi-model collaborative decision-making as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the mechanical complex fault diagnosis method based on multi-model collaborative decision-making as described in any one of claims 1 to 8 is implemented.

Images