Gearbox fault detection method, device and equipment and storage medium

Abstract

The embodiment of the invention provides a gearbox fault detection method and device, equipment and a storage medium, and the method comprises the steps: obtaining the target operation time sequence data of a gearbox, and dividing the target operation time sequence data into a plurality of target data segments; inputting each target data fragment into a multi-scale feature extraction model for mapping to obtain a multi-scale initial feature corresponding to each target data fragment, and performing multi-scale weighted fusion on each multi-scale initial feature to obtain a target operation feature corresponding to each target data fragment; inputting each target operation feature into a time sequence model for mapping to obtain a target time domain feature corresponding to each target data fragment; and performing attention processing on each target time domain feature, and inputting the processed target time domain features into a classification model for classification processing to obtain a target fault detection result of the gearbox. According to the embodiment of the invention, the accuracy of the target fault detection result can be improved.

Description

Technical Field

[0001] The present application relates to, but is not limited to, the field of fault detection technology, and in particular to a gearbox fault detection method, device, equipment and storage medium. Background Art

[0002] The gearbox is one of the core components of rotating mechanical equipment. The gearbox often works under harsh working conditions of high speed and high load. Under the influence of alternating strain for a long time, key components such as gears and bearings in the gearbox are prone to structural damage such as wear, fracture, pitting, cracks, etc., which poses a great safety hazard. Therefore, it is necessary to detect whether the gearbox is faulty.

[0003] At present, the traditional gearbox fault detection method mainly relies on convolutional neural networks, which extract spatial features from the gearbox operation data for classification prediction to determine the fault category. However, the spatial features that convolutional neural networks focus on cannot effectively capture the complex information in the gearbox operation data, resulting in low accuracy of fault detection results. Summary of the invention

[0004] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0005] The embodiments of the present application provide a gearbox fault detection method, device, equipment and storage medium, which can improve the accuracy of target fault detection results.

[0006] To achieve the above-mentioned purpose, the first aspect of an embodiment of the present application proposes a gearbox fault detection method, which is characterized in that it includes: obtaining target operation time sequence data of the gearbox, and dividing the target operation time sequence data into multiple target data segments; inputting each of the target data segments into a multi-scale feature extraction model for mapping, and obtaining multi-scale initial features corresponding to each of the target data segments; performing multi-scale weighted fusion on each of the multi-scale initial features, and obtaining target operation features corresponding to each of the target data segments; inputting each of the target operation features into a timing model for mapping, and obtaining target time domain features corresponding to each of the target data segments; performing attention processing on each of the target time domain features, and then inputting them into a classification model for classification processing, and obtaining the target fault detection result of the gearbox.

[0007] In some embodiments, the multi-scale feature extraction model includes a first feature extraction network and a second feature extraction network, the first feature extraction network includes N first convolutional layers connected in sequence, the second feature extraction network includes M second convolutional layers connected in sequence, the N is less than the M, the N and the M are both positive integers, and the multi-scale initial features include first-scale initial features and second-scale initial features; the inputting each of the target data segments into the multi-scale feature extraction model for mapping to obtain the multi-scale initial features corresponding to each of the target data segments, and performing multi-scale weighted fusion on each of the multi-scale initial features to obtain the target operation features corresponding to each of the target data segments, includes: for each of the target data segments, inputting the target data segment into the first feature extraction network for mapping to obtain the first-scale initial features corresponding to the target data segment, inputting the target data segment into the second feature extraction network for mapping to obtain the second-scale initial features corresponding to the target data segment; for each of the target data segments, performing multi-scale weighted fusion on the first-scale initial features and the second-scale initial features corresponding to the target data segment to obtain the target operation features corresponding to the target data segment.

[0008] In some embodiments, the first feature extraction network includes a first convolutional layer, a second convolutional layer and a first pooling layer connected in sequence, the second feature extraction network includes a third convolutional layer, a fourth convolutional layer, a second pooling layer, a fifth convolutional layer, a sixth convolutional layer and a third pooling layer connected in sequence, the convolution kernel size of the first convolutional layer is greater than or equal to the convolution kernel size of the second convolutional layer, the convolution kernel size of the first convolutional layer is greater than the convolution kernel size of the third convolutional layer, the convolution kernel size of the third convolutional layer is greater than or equal to the convolution kernel size of the fourth convolutional layer, the convolution kernel size of the third convolutional layer is greater than the convolution kernel size of the fifth convolutional layer, and the convolution kernel size of the fifth convolutional layer is greater than or equal to the convolution kernel size of the sixth convolutional layer; the target data segment is input into the first feature extraction network for mapping to obtain the first scale initial feature corresponding to the target data segment, and the The target data segment is input into the second feature extraction network for mapping to obtain the second-scale initial feature corresponding to the target data segment, including: calling the first convolution layer and the second convolution layer to perform convolution processing on the target data segment to obtain a first intermediate feature, calling the first pooling layer to perform a pooling operation on the first intermediate feature to obtain the first-scale initial feature corresponding to the target data segment; calling the third convolution layer and the fourth convolution layer to perform convolution processing on the target data segment to obtain a second intermediate feature, calling the second pooling layer to perform a pooling operation on the second intermediate feature to obtain a third intermediate feature; calling the fifth convolution layer and the sixth convolution layer to perform convolution processing on the third intermediate feature to obtain a fourth intermediate feature, calling the third pooling layer to perform a pooling operation on the fourth intermediate feature to obtain the second-scale initial feature corresponding to the target data segment.

[0009] In some embodiments, the inputting of each of the target operation features into a timing model for mapping to obtain the target time domain features corresponding to each of the target data segments includes: inputting each of the target operation features into a timing model, and determining the forward hidden layer state and the reverse hidden layer state corresponding to each of the target operation features based on the timing model; for each of the target operation features, performing weighted summation based on the forward hidden layer state and the reverse hidden layer state corresponding to the target operation feature to determine the target time domain features corresponding to the target data segment.

[0010] In some embodiments, the target time domain features are subjected to attention processing and then input into a classification model for classification processing to obtain a target fault detection result of the gearbox, including: performing attention processing on each of the target time domain features to obtain an attention feature; obtaining a working type identifier of the gearbox, splicing the working type identifier and the attention feature and inputting them into a classification model for classification processing to obtain a target fault detection result of the gearbox.

[0011] In some embodiments, the work type identifier and the attention feature are concatenated and input into a classification model for classification processing to obtain a target fault detection result of the gearbox, including: obtaining a historical failure probability of the gearbox; inputting the attention feature and the historical failure probability into a probability prediction model for regression processing to obtain a current failure probability of the gearbox; concatenating the work type identifier, the current failure probability and the attention feature and inputting them into a classification model for classification processing to obtain a target fault detection result of the gearbox.

[0012] In some embodiments, before performing multi-scale feature extraction on each of the target data segments, the gearbox fault detection method further includes: obtaining sample operation timing data of the gearbox and sample detection labels corresponding to the sample operation timing data, dividing the sample operation timing data into multiple sample data segments, wherein the sample detection labels are used to indicate that the sample operation timing data belongs to a fault category or a normal type; inputting each of the sample data segments into a multi-scale feature extraction model for mapping to obtain multi-scale sample features corresponding to each of the sample data segments, performing multi-scale weighted fusion on each of the multi-scale sample features to obtain sample operation features corresponding to each of the target data segments; inputting each of the sample operation features into a timing model for mapping to obtain sample time domain features corresponding to each of the sample data segments; performing attention processing on each of the sample time domain features and then inputting them into a classification model for classification processing to obtain a sample fault detection result of the gearbox; determining a model loss based on the difference between the sample fault detection result and the sample detection label, and jointly training the multi-scale feature extraction model, the timing model and the classification model based on the model loss.

[0013] To achieve the above-mentioned purpose, the second aspect of an embodiment of the present application proposes a gearbox fault detection device, including: an acquisition module, used to acquire the operating time data of the gearbox; a multi-scale feature extraction module, used to divide the operating time data into multiple data segments, perform multi-scale feature extraction on each of the data segments, obtain multi-scale initial features corresponding to each of the data segments, perform multi-scale weighted fusion on each of the multi-scale initial features, and obtain target operating features corresponding to each of the data segments; a time domain feature extraction module, used to input each of the target operating features into a time series model for mapping, and obtain time domain features corresponding to each of the data segments; a classification module, which performs attention processing on each of the time domain features and then inputs them into a classification model for classification processing to obtain the fault detection result of the gearbox.

[0014] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the gearbox fault detection method described in the first aspect when executing the computer program.

[0015] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the gearbox fault detection method described in the first aspect is implemented.

[0016] The embodiments of the present application include at least the following beneficial effects: by dividing the target operation timing data of the gearbox into multiple target data segments, and then inputting the multiple target data segments into a multi-scale feature extraction model for mapping, the multi-scale initial features corresponding to the multiple target data segments are obtained; since each target data segment belongs to each subsequence of the target operation timing data, and each target data segment contains the multi-scale feature information of the corresponding subsequence, it is possible to fully capture the local spatial features of the target operation timing data, and then multi-scale weighted fusion is performed on each multi-scale initial feature to obtain the target operation features corresponding to each target data segment, which can retain the features related to the fault in the multi-scale initial features and remove the invalid features, and then input each target operation feature into the timing model for mapping. The target time domain features corresponding to each target data segment are obtained by projection, and each target time domain feature is then subjected to attention processing, which is equivalent to focusing on different parts of the input sequence composed of each target time domain feature through the feature capture ability of attention processing, enhancing the feature extraction ability in the time domain, and realizing full capture of the global time domain features of the target operation sequence data. Finally, the global time domain features are input into the classification model for classification processing to obtain the target fault detection result of the gearbox. Therefore, when processing the target operation sequence data of the gearbox, it can fully capture the spatial characteristics of the data, as well as the dynamic changes and time dependence of the data, realizing the effective capture of the complex information in the target operation sequence data of the gearbox, and can improve the reliability of the analysis of the target operation sequence data, thereby improving the accuracy of the target fault detection result.

[0017] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0019] Figure 1 A schematic diagram of an optional flow chart of a gearbox fault detection method provided in an embodiment of the present application;

[0020] Figure 2 An optional flow chart of feature extraction and weighted fusion provided in an embodiment of the present application;

[0021] Figure 3 A schematic diagram of an optional process for multi-scale feature extraction provided in an embodiment of the present application;

[0022] Figure 4 An optional specific flow chart of time domain feature extraction provided in an embodiment of the present application;

[0023] Figure 5 An optional flowchart of attention processing and classification processing provided in an embodiment of the present application;

[0024] Figure 6 An optional specific flow chart of the classification process provided in the embodiment of the present application;

[0025] Figure 7 An optional flowchart of a joint training model provided in an embodiment of the present application;

[0026] Figure 8 An optional structural diagram of a multi-scale feature extraction model and a weighted fusion model provided in an embodiment of the present application;

[0027] Fig. 9 An optional flowchart of the reasoning process provided in the embodiment of the present application;

[0028] Fig.10 An optional flowchart of a training process provided in an embodiment of the present application;

[0029] Fig.11 A schematic diagram of an optional structure of a gearbox fault detection device provided in an embodiment of the present application;

[0030] Fig.12 A schematic diagram of an optional hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.

[0032] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the characteristics of the target object such as target object attribute information or attribute information set, the permission or consent of the target object will be obtained first, and the collection, use and processing of these data will comply with relevant laws, regulations and standards. Among them, the target object can be a user. In addition, when the embodiment of the present application needs to obtain the target object attribute information, the target object's separate permission or separate consent will be obtained by means of a pop-up window or jumping to a confirmation page, etc., and after the separate permission or separate consent of the target object is clearly obtained, the necessary target object-related data for enabling the normal operation of the embodiment of the present application will be obtained.

[0033] In the description of this application, “several” means one or more, “more” means more than two, “greater than”, “less than”, “exceed”, etc. are understood to exclude the number, and “above”, “below”, “within”, etc. are understood to include the number.

[0034] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims or the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0035] At present, the traditional gearbox fault detection method mainly relies on convolutional neural networks, which extract spatial features from the gearbox operation data for classification prediction to determine the fault category. However, the spatial features that convolutional neural networks focus on cannot effectively capture the complex information in the gearbox operation data, resulting in low accuracy of fault detection results.

[0036] In response to the problem of low accuracy of fault detection results, the present application provides a gearbox fault detection method, device, equipment and storage medium, the method comprising: obtaining target operation timing data of the gearbox, dividing the target operation timing data into multiple target data segments; inputting each of the target data segments into a multi-scale feature extraction model for mapping, obtaining multi-scale initial features corresponding to each of the target data segments, performing multi-scale weighted fusion on each of the multi-scale initial features, obtaining target operation features corresponding to each of the target data segments; inputting each of the target operation features into a timing model for mapping, obtaining target time domain features corresponding to each of the target data segments; performing attention processing on each of the target time domain features and then inputting them into a classification model for classification processing, to obtain the target fault detection result of the gearbox.

[0037] According to the solution provided by the embodiment of the present application, the target operation timing data of the gearbox is divided into multiple target data segments, and then the multiple target data segments are respectively input into the multi-scale feature extraction model for mapping, so as to obtain the multi-scale initial features corresponding to the multiple target data segments. Since each target data segment belongs to each subsequence of the target operation timing data, and each target data segment contains the multi-scale feature information of the corresponding subsequence, it is possible to fully capture the local spatial features of the target operation timing data, and then each multi-scale initial feature is respectively subjected to multi-scale weighted fusion to obtain the target operation features corresponding to each target data segment, so that the features related to the fault can be retained in the multi-scale initial features, and the invalid features can be removed, and then each target operation feature is input into the timing model for mapping. , the target time domain features corresponding to each target data segment are obtained, and then each target time domain feature is subjected to attention processing, which is equivalent to focusing on different parts of the input sequence composed of each target time domain feature through the feature capture ability of attention processing, enhancing the feature extraction ability in the time domain, and realizing the full capture of the global time domain features of the target operation sequence data. Finally, it is input into the classification model for classification processing to obtain the target fault detection result of the gearbox. Therefore, when processing the target operation sequence data of the gearbox, it can fully capture the spatial characteristics of the data, and can fully capture the dynamic changes and time dependence of the data, realizing the effective capture of the complex information in the target operation sequence data of the gearbox, and can improve the reliability of the target operation sequence data analysis, thereby improving the accuracy of the target fault detection result.

[0038] The gearbox fault detection method, device, equipment and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the gearbox fault detection method in the embodiments of the present application is described.

[0039] The gearbox fault detection method provided in the embodiment of the present application relates to the field of computer technology. The gearbox fault detection method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the gearbox fault detection method, etc., but is not limited to the above forms.

[0040] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0041] The embodiments of the present application are further described below in conjunction with the accompanying drawings.

[0042] like Figure 1 As shown, Figure 1 An optional flow chart of a gearbox fault detection method provided in an embodiment of the present application is provided. The gearbox fault detection method can be executed by a server, or by a terminal, or by a server in cooperation with a terminal. The gearbox fault detection method includes but is not limited to the following steps S110 to S140:

[0043] A gearbox fault detection method, characterized by comprising:

[0044] Step S110, obtaining target operating time data of the gearbox, and dividing the target operating time data into a plurality of target data segments.

[0045] Step S120, input each target data segment into the multi-scale feature extraction model for mapping, obtain the multi-scale initial features corresponding to each target data segment, perform multi-scale weighted fusion on each multi-scale initial feature, and obtain the target operation features corresponding to each target data segment.

[0046] Step S130 , input each target operation feature into the time series model for mapping, and obtain the target time domain feature corresponding to each target data segment.

[0047] Step S140, after attention processing, each target time domain feature is input into a classification model for classification processing to obtain a target fault detection result of the gearbox.

[0048] Among them, the vibration signal can be collected through the gearbox experimental platform to obtain the target operating timing data. The vibration signal or the target operating timing data reflects the change of the vibration acceleration of the gearbox over time.

[0049] The time period corresponding to the target runtime series data is divided into a plurality of time steps, and the target runtime series data corresponding to each time step is then divided into a plurality of target data segments.

[0050] Among them, the target fault detection result is used to indicate that the target operating timing data belongs to a fault category or a normal type. For example, the fault category may include gear tooth breakage, gear defect, tooth surface wear, tooth root crack, bearing ball fault, inner and outer ring combined fault, outer ring fault and inner ring fault, etc.

[0051] Based on this, the target operation timing data of the gearbox is divided into multiple target data segments, and then the multiple target data segments are respectively input into the multi-scale feature extraction model for mapping, so as to obtain the multi-scale initial features corresponding to the multiple target data segments. Since each target data segment belongs to each subsequence of the target operation timing data, and each target data segment contains the multi-scale feature information of the corresponding subsequence, it is possible to fully capture the local spatial features of the target operation timing data, and then each multi-scale initial feature is subjected to multi-scale weighted fusion to obtain the target operation features corresponding to each target data segment. It is possible to retain the features related to the fault in the multi-scale initial features and remove the invalid features, and then input each target operation feature into the timing model for mapping to obtain each The target time domain features corresponding to the target data fragments are then processed with attention, which is equivalent to paying attention to different parts of the input sequence composed of each target time domain feature through the feature capture ability of attention processing, enhancing the feature extraction ability in the time domain, and realizing the full capture of the global time domain features of the target operation sequence data. Finally, it is input into the classification model for classification processing to obtain the target fault detection result of the gearbox. Therefore, when processing the target operation sequence data of the gearbox, it can fully capture the spatial characteristics of the data, and can fully capture the dynamic changes and time dependence of the data, realizing the effective capture of the complex information in the target operation sequence data of the gearbox, and can improve the reliability of the analysis of the target operation sequence data, thereby improving the accuracy of the target fault detection results.

[0052] In addition, refer to Figure 2In one embodiment, the multi-scale feature extraction model includes a first feature extraction network and a second feature extraction network, the first feature extraction network includes N first convolutional layers connected in sequence, the second feature extraction network includes M second convolutional layers connected in sequence, N is less than M, N and M are both positive integers, the multi-scale initial features include first-scale initial features and second-scale initial features, each target data segment is input into the multi-scale feature extraction model for mapping, and the multi-scale initial features corresponding to each target data segment are obtained, and each multi-scale initial feature is subjected to multi-scale weighted fusion to obtain the target operation features corresponding to each target data segment, including but not limited to the following steps:

[0053] Step S210: for each target data segment, input the target data segment into a first feature extraction network for mapping to obtain a first-scale initial feature corresponding to the target data segment, and input the target data segment into a second feature extraction network for mapping to obtain a second-scale initial feature corresponding to the target data segment.

[0054] Step S220 : For each target data segment, multi-scale weighted fusion is performed on the first-scale initial features and the second-scale initial features corresponding to the target data segment to obtain the target operation features corresponding to the target data segment.

[0055] Among them, reference Figure 8 , the first feature extraction network and the second feature extraction network work in parallel.

[0056] In a possible implementation, multi-scale weighted fusion is performed on each multi-scale sample feature through a weighted fusion model.

[0057] It should be noted that the weighting process of the initial features of the first scale is as follows: one-dimensional global average pooling is performed on the initial features of the first scale, which is equivalent to averaging all elements of the initial features of the first scale, compressing all elements into an average value, determining the first mean, and then inputting the first mean into two fully connected layers for linear transformation to determine the first weight vector, and finally multiplying the first weight vector by the initial features of the first scale to determine the first weighted feature vector.

[0058] In addition, the weighting process of the second-scale initial features is as follows: one-dimensional global average pooling is performed on the second-scale initial features, which is equivalent to averaging all elements of the second-scale initial features, compressing all elements into an average value, determining the second mean, and then inputting the second mean into two fully connected layers for linear transformation to determine the second weight vector, and finally multiplying the second weight vector by the second-scale initial features to determine the second weighted feature vector.

[0059] Among them, the formula corresponding to one-dimensional global average pooling is as follows:

[0060]

[0061] Among them, before obtaining the first scale initial feature or the first scale initial feature, it is also necessary to go through maximum pooling, m j is the pooling kernel of the maximum pooling layer, j is the size of the pooling kernel, and i refers to the feature extraction network of the i-th scale in the multi-scale feature extraction model. For example, (i) = 1 represents the first feature extraction network, and (i) = 2 represents the second feature extraction network. When (i) = 1, z (i) is the first mean, L is the vector length of the initial feature of the first scale, is the nth element in the initial feature of the first scale. When (i) = 2, z (i) is the second mean, L is the vector length of the initial feature of the second scale, is the nth element in the initial feature of the second scale.

[0062] The formula for determining each element in the first weight vector or the second weight vector is as follows:

[0063]

[0064] Where i refers to the feature extraction network of the i-th scale in the multi-scale feature extraction model. For example, (i) = 1 represents the first feature extraction network, (i) = 2 represents the second feature extraction network, and when (i) = 1, s (i) is the first weight vector. When (i) = 2, s (i) is the second weight vector; as well as are the weight matrices corresponding to the two fully connected layers, as well as They are the biases corresponding to the two fully connected layers, σ is the sigmoid activation function, and ReLU is the ReLU activation function.

[0065] The formula for determining each element in the first weighted feature vector or the second weighted feature vector is as follows:

[0066]

[0067] Among them, before obtaining the first scale initial feature or the first scale initial feature, it is also necessary to go through maximum pooling, m j is the pooling kernel of the maximum pooling layer, j is the size of the pooling kernel, and i refers to the feature extraction network of the i-th scale in the multi-scale feature extraction model. For example, (i) = 1 represents the first feature extraction network, (i) = 2 represents the second feature extraction network, and when (i) = 1, s (i) is the first weight vector, is the nth element of the initial feature of the first scale, is the nth element of the first weighted eigenvector. When (i) = 2, s (i) is the second weight vector, is the nth element of the initial feature of the second scale, is the nth element of the second weighted eigenvector.

[0068] It should be noted that the matrix size of the first-scale initial feature is consistent with that of the second-scale initial feature. The first-scale initial feature and the second-scale initial feature are fused. The specific steps are to perform matrix multiplication or element-by-element multiplication of the first weighted feature vector and the second weighted feature vector to determine the target operation feature corresponding to the target data segment.

[0069] The formula for determining the target operation characteristics corresponding to the target data segment is as follows:

[0070]

[0071] Among them, x(n) is the nth element of the target operation feature, is the nth element of the first weighted eigenvector, is the nth element of the second weighted eigenvector.

[0072] Based on this, for each target data segment, the first-scale initial features are determined by a first feature extraction network with fewer convolution layers, which can reduce the amount of calculation and network complexity, and the second-scale initial features are determined by a second feature extraction network with more convolution layers, so that the second-scale initial features can represent more complex feature combinations. Then the first feature extraction network and the second feature extraction network understand the spatial features of the target data segment at two scales respectively, and then the first-scale initial features and the second-scale initial features are weighted and fused to obtain the target operation features corresponding to the target data segment, so that the target operation features can more comprehensively describe the spatial features corresponding to the target data segment, thereby avoiding the limitations of a single feature extraction network in understanding spatial features.

[0073] In addition, refer to Figure 3In one embodiment, the first feature extraction network includes a first convolutional layer, a second convolutional layer, and a first pooling layer connected in sequence, the second feature extraction network includes a third convolutional layer, a fourth convolutional layer, a second pooling layer, a fifth convolutional layer, a sixth convolutional layer, and a third pooling layer connected in sequence, the convolution kernel size of the first convolutional layer is greater than or equal to the convolution kernel size of the second convolutional layer, the convolution kernel size of the first convolutional layer is greater than the convolution kernel size of the third convolutional layer, the convolution kernel size of the third convolutional layer is greater than or equal to the convolution kernel size of the fourth convolutional layer, the convolution kernel size of the third convolutional layer is greater than the convolution kernel size of the fifth convolutional layer, and the convolution kernel size of the fifth convolutional layer is greater than or equal to the convolution kernel size of the sixth convolutional layer, the target data segment is input into the first feature extraction network for mapping, and the first scale initial feature corresponding to the target data segment is obtained, and the target data segment is input into the second feature extraction network for mapping, and the second scale initial feature corresponding to the target data segment is obtained, including but not limited to the following steps:

[0074] Step S310: Call the first convolution layer and the second convolution layer to perform convolution processing on the target data segment to obtain a first intermediate feature, call the first pooling layer to perform a pooling operation on the first intermediate feature to obtain a first-scale initial feature corresponding to the target data segment.

[0075] Step S320, calling the third convolution layer and the fourth convolution layer to perform convolution processing on the target data segment to obtain a second intermediate feature, calling the second pooling layer to perform a pooling operation on the second intermediate feature to obtain a third intermediate feature.

[0076] Step S330, calling the fifth convolution layer and the sixth convolution layer to perform convolution processing on the third intermediate feature to obtain a fourth intermediate feature, calling the third pooling layer to perform a pooling operation on the fourth intermediate feature to obtain a second-scale initial feature corresponding to the target data segment.

[0077] Among them, reference Figure 8 , the convolution kernels in the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, the fifth convolution layer and the sixth convolution layer can all be one-dimensional convolution kernels. The convolution kernel in the first feature extraction network is wider, and the convolution kernel in the second feature extraction network is narrower. For example, the convolution kernel size of the first convolution layer can be 1×32, the convolution kernel size of the second convolution layer can be 1×16, the convolution kernel size of the third convolution layer can be 1×16, the convolution kernel size of the fourth convolution layer can be 1×16, the convolution kernel size of the fifth convolution layer can be 1×5, and the convolution kernel size of the sixth convolution layer can be 1×5.

[0078] Among them, for the first convolutional layer in the feature extraction network of each size, the corresponding formula for feature extraction is as follows:

[0079]

[0080] Wherein, i refers to the feature extraction network of the i-th scale in the multi-scale feature extraction model. For example, (i)=1 represents the first feature extraction network, and (i)=2 represents the second feature extraction network. n =X (i) (n), h n is the nth element of the target runtime data, equivalent to X (i) (n) is the nth target data segment of the target runtime data, k j is the convolution kernel corresponding to the convolution layer, j is the size of the convolution kernel, is the feature extraction result of the nth target data segment, f is the convolution operation, δ is the activation function, and τ is the time step.

[0081] It should be noted that the first pooling layer, the second pooling layer, and the third pooling layer are all maximum pooling layers. Therefore, the dimensions of the first-scale initial features and the first intermediate features decrease successively, and the amount of data decreases successively; the dimensionality reduction corresponding to the second-scale initial features, the third intermediate features, and the second intermediate features decreases successively, and the amount of data decreases successively.

[0082] Among them, the formula corresponding to the maximum pooling layer is as follows:

[0083]

[0084] Among them, m j is the pooling kernel of the maximum pooling layer, j is the size of the pooling kernel, and i refers to the feature extraction network of the i-th scale in the multi-scale feature extraction model. For example, (i) = 1 represents the first feature extraction network, and (i) = 2 represents the second feature extraction network. When (i) = 1, is the nth element of the initial feature of the first scale. When (i) = 2, is the nth element of the initial feature of the second scale; is the feature extraction result of the nt-th target data segment. The maximum pooling layer is also set with a pooling window. t is the index variable of the pooling window of the maximum pooling layer. t is used to traverse each position of the pooling window. p is the size of the pooling window. max means taking the maximum value.

[0085] Based on this, the first feature extraction network extracts features of the target data segment through two wider convolutional layers, which can extract relatively overall features of the target data segment to obtain the first intermediate features, and then pools the first intermediate features to reduce the dimension of the first intermediate features, reduce the amount of calculation of the first feature extraction network, and obtain the first scale initial features corresponding to the target data segment. At the same time, the second feature extraction network first extracts features of the target data segment through two narrower convolutional layers with pooling operations, which can preliminarily extract relatively detailed features of the target data segment to obtain the second intermediate features, and then extracts features of the second intermediate features through two narrower convolutional layers with maximum pooling, and accumulates and combines the previously more detailed features into complex and hierarchical features to obtain the second scale initial features. Therefore, the first feature extraction network and the second feature extraction network can increase the diversity of features, improve the overall generalization ability, and improve the feature fusion results.

[0086] In addition, refer to Figure 4 In one embodiment, each target operation feature is input into the time series model for mapping to obtain the target time domain feature corresponding to each target data segment, including but not limited to the following steps:

[0087] Step S410: input each target operation feature into a time series model, and determine the forward hidden layer state and the reverse hidden layer state corresponding to each target operation feature based on the time series model.

[0088] Step S420: For each target operation feature, a weighted sum is performed based on the forward hidden layer state and the reverse hidden layer state corresponding to the target operation feature to determine the target time domain feature corresponding to the target data segment.

[0089] It should be noted that the timing model includes a forward hidden layer and a backward hidden layer. The time step of the forward hidden layer increases gradually from front to back. For each time step of the forward hidden layer, a forward sub-network is provided. The forward sub-network includes a first update module, a first reset module, a first candidate module and a first state update module. Each target operation feature corresponds to the input of the forward hidden layer from the first time step to the last time step. For each time step, the forward hidden layer state of each time step of the forward hidden layer is determined based on the target operation feature of the current time step and the forward hidden layer state of the previous time step.

[0090] Specifically, the forward hidden state of the previous time step of the first time step is pre-initialized as a zero vector, and the corresponding forward hidden state is determined sequentially from the first time step to the last time step. For each time step, the target operation feature of the current time step and the forward hidden state of the previous time step are inputted into the first update module and the first reset module corresponding to the current time step, the first activation value corresponding to the current time step is determined by the first update module, the second activation value corresponding to the current time step is determined by the first reset module, the second activation value, the target operation feature of the current time step and the forward hidden state of the previous time step are inputted into the first candidate module to obtain the first candidate output, and finally the first activation value, the first candidate output and the forward hidden state of the previous time step are inputted into the first state update module to determine the forward hidden state of the current time step.

[0091] The first activation value is used to determine the extent to which the forward hidden state of the previous time step is transferred to the forward hidden state of the current time step, and the second activation value is used to determine how much information of the forward hidden state of the previous time step is used to calculate the first candidate output of the current time step.

[0092] Among them, the formula for determining the forward hidden layer state is as follows:

[0093]

[0094] Among them, t is the time step number increasing from the front to the back, is the forward hidden state at the tth time step, e t is the target running feature input at the t-th time step, is the forward hidden state of the t-1 time step.

[0095] In addition, the time step of the backward hidden layer increases gradually from back to front. For each time step of the backward hidden layer, a backward subnetwork is set, and the backward subnetwork includes a second update module, a second reset module, a second candidate module and a second state update module. Each target operation feature corresponds to the input of the backward hidden layer from the last time step to the first time step. For each time step, the backward hidden layer state of each time step of the backward hidden layer is determined based on the target operation feature of the current time step and the backward hidden layer state of the previous time step.

[0096] Specifically, the backward hidden layer state of the previous time step of the first time step is pre-initialized as a zero vector, and the corresponding backward hidden layer state is determined in sequence from the th time step to the last time step. For each time step, the target operation feature of the current time step and the backward hidden layer state of the previous time step are inputted into the second update module and the second reset module corresponding to the current time step, the second activation value corresponding to the current time step is determined by the second update module, the second activation value corresponding to the current time step is determined by the second reset module, the second activation value, the target operation feature of the current time step and the backward hidden layer state of the previous time step are inputted into the second candidate module to obtain the second candidate output, and finally the second activation value, the second candidate output and the backward hidden layer state of the previous time step are inputted into the second state update module to determine the backward hidden layer state of the current time step.

[0097] Among them, the second activation value is used to determine the extent to which the backward hidden state of the previous time step is transferred to the backward hidden state of the current time step, and the second activation value is used to determine how much information of the backward hidden state of the previous time step is used to calculate the second candidate output of the current time step.

[0098] Among them, the formula for determining the backward hidden layer state is as follows:

[0099]

[0100] Among them, t is the time step number increasing from the back to the front, is the backward hidden layer state at the tth time step, e t is the target running feature input at the t-th time step, is the forward hidden state of the t-1 time step.

[0101] It should be noted that for each time step, a first weight matrix corresponding to the forward hidden layer state, a second weight matrix corresponding to the backward hidden layer state, and a hidden layer bias corresponding to the current time step are pre-set. The multiplication result of the first weight matrix and the forward hidden layer state, the multiplication result of the second weight matrix and the backward hidden layer state, and the hidden layer bias are summed to determine the target time domain features.

[0102] Among them, the formula for determining the target time domain characteristics is as follows:

[0103]

[0104] Among them, t is the time step number increasing from the front to the back, h t is the target time domain feature at the tth time step, is the forward hidden state at the t-th time step, is the backward hidden state of the tth time step, w tis the first weight matrix corresponding to the tth time step, v t is the second weight matrix corresponding to the tth time step, b t is the hidden layer bias corresponding to the tth time step.

[0105] Among them, reference Fig. 9 , change e1 to e t , that is, the target operation characteristics corresponding to the first time step to the tth time step are input into the time series model for mapping to obtain h1 to h t , that is, the target time domain features corresponding to the first time step to the tth time step.

[0106] Based on this, by inputting each target operation feature into the time series model, determining the forward hidden layer state and the reverse hidden layer state corresponding to each target operation feature based on the time series model, and then performing weighted summation based on the forward hidden layer state and the reverse hidden layer state corresponding to the target operation feature, the target time domain feature corresponding to the target data segment is determined. This can capture the data dependencies from the past to the future and from the future to the past at the same time, making the analysis of the data more comprehensive. The introduction of weighted summation can automatically adjust the weights during the training process to enhance the end-to-end learning ability.

[0107] In addition, refer to Figure 5 In one embodiment, each target time domain feature is processed by attention and then input into a classification model for classification processing to obtain a target fault detection result of the gearbox, including but not limited to the following steps:

[0108] Step S510, performing attention processing on each target time domain feature to obtain an attention feature.

[0109] Step S520, obtaining the working type identification of the gearbox, splicing the working type identification and the attention feature and inputting them into the classification model for classification processing, and obtaining the target fault detection result of the gearbox.

[0110] Among them, reference Fig. 9 , each target time domain feature is spliced ​​into a target time domain feature matrix, and then input into the multi-head attention module. Each attention head of the multi-head attention module has its own corresponding learnable query parameter matrix W Q , key-value parameter matrix W K , the numerical parameter matrix W V , and then the query parameter matrix W Q Multiply it with the input matrix to get the query matrix Q, and convert the key-value parameter matrix W K Multiply it with the input matrix to get the key matrix K, and convert the value parameter matrix W V Multiply it with the input matrix to get the value matrix V, and finally convert the query matrix Q and the transposed matrix K of the key value matrix KT Multiply them together to get the intermediate reference matrix, reduce the dimension of the intermediate reference matrix and normalize it, then multiply it with the numerical matrix V to get the corresponding attention sub-features, concatenate multiple attention sub-features and perform linear transformation to get the attention feature.

[0111] Among them, the corresponding calculation formulas for the input matrix, query matrix, key value matrix and value matrix are as follows:

[0112] X = BiGRU(x T )

[0113] Q=XW Q

[0114] K=XW K

[0115] V=XW V

[0116] Among them, x T is the input of the time series model, X is the input matrix, Q is the query matrix, and W Q is the query parameter matrix, K is the key value matrix, W K is the key-value parameter matrix, V is the numerical matrix, W V is the numerical parameter matrix.

[0117] In addition, the formula for determining each attention sub-feature is as follows:

[0118]

[0119] Among them, Q is the query matrix, K is the key value matrix, K T is the transposed matrix K of the key-value matrix K T , d k is the dimension of the key-value matrix K, V is the numerical matrix, Attention(Q,K,V) is the attention sub-feature of each attention head,

[0120] as well as:

[0121] head i =Attention(QW i Q ,KW i K ,VW i V )

[0122] Among them, head i is the i-th attention head, i is the attention head number, QW i Q is the query matrix of the ith attention head, KW iK is the query matrix of the ith attention head, VW i V ) is the query matrix of the i-th attention head.

[0123] In addition, after concatenating multiple attention sub-features and performing linear transformation, the formula for obtaining the attention feature is as follows:

[0124] MHA(Q,K,V)=Concat(head1,…,head h )W o

[0125] Among them, MHA(Q,K,V) is the attention feature, head1,…,head h is the attention sub-feature corresponding to the first attention head to the last attention head. For example, head1 is the attention sub-feature corresponding to the first attention head, head h is the attention sub-feature corresponding to the last attention head, W o is the linear transformation matrix.

[0126] as well as:

[0127] Y=MHA(XW Q ,XW K ,XW V )

[0128] Among them, Y is the attention feature, X is the input matrix, Q is the query matrix, and W Q is the query parameter matrix, K is the key value matrix, W K is the key-value parameter matrix, V is the numerical matrix, W V is the value parameter matrix.

[0129] It should be noted that the working type identification of the gearbox can be a string of numbers used to indicate the type or working condition of the gearbox. For example, the working type identification of the fixed-axis gearbox can be [0, 1, 1], the working type identification of the planetary gearbox can be [1, 1, 0], and the working type identification of the differential gearbox can be [0, 1, 0]. The embodiments of the present application are not limited here.

[0130] It can be understood that, according to the working type identification, the difficulty of failure of gearboxes of different types or under different working conditions can be reflected, which plays an auxiliary judgment role in fault detection.

[0131] Among them, the work type identifier and the attention feature can be directly spliced ​​together, or the work type identifier can be expanded to the same size as the attention feature and then spliced ​​together.

[0132] Based on this, by performing attention processing on each target time domain feature to obtain the attention feature, the dependency between different parts of the data and the unique features corresponding to different parts can be captured, the feature extraction capability can be enhanced, and information loss caused by too long data can be prevented, thereby improving the accuracy of the fault detection results. Then, the working type identification of the gearbox is obtained, and the working type identification and the attention feature are spliced ​​and input into the classification model for classification processing to obtain the target fault detection result of the gearbox. The splicing of the working type identification and the attention feature can enhance the richness of the feature representation, which helps the classification model to capture more complex data patterns and relationships, avoid the classification model from making a relatively single judgment only through the attention feature, and improve the accuracy of the fault detection results.

[0133] In addition, refer to Figure 6 In one embodiment, the work type identification and the attention feature are spliced ​​and input into the classification model for classification processing to obtain the target fault detection result of the gearbox, including but not limited to the following steps:

[0134] Step S610, obtaining the historical failure probability of the gearbox.

[0135] Step S620, input the attention features and the historical failure probability into the probability prediction model for regression processing to obtain the current failure probability of the gearbox.

[0136] Step S630, the work type identification, current fault probability and attention feature are spliced ​​and input into the classification model for classification processing to obtain the target fault detection result of the gearbox.

[0137] Among them, the historical failure probability can be the average failure of the gearbox from the past to the present, or it can be the current failure probability predicted by the probability prediction model last time, or it can be the failure probability calculated by other statistical methods or calculation methods. The embodiments of the present application are not limited here.

[0138] It can be understood that both the historical failure probability and the current failure probability are used to indicate the difficulty of the gearbox failure, and have an auxiliary judgment effect on fault detection. For example, when the current failure probability of the gearbox is large, that is, the gearbox is not prone to failure at present, the judgment of the corresponding fault detection result will tend to be normal or other fault categories with a higher probability of occurrence.

[0139] Based on this, by obtaining the historical failure probability of the gearbox, it is equivalent to obtaining the variable factors that affect the current failure probability. The attention feature and the historical failure probability are input into the probability prediction model for regression processing to obtain the current failure probability of the gearbox. Since the historical failure probability represents the difficulty of the failure of the gearbox in the past time, it can describe the current failure probability of the gearbox from one dimension, and the attention feature represents the dependency between different parts of the target operating timing data of the gearbox, which can describe the current failure probability of the gearbox from another dimension. Therefore, the attention feature and the historical failure probability both have a guiding role in determining the current failure probability and can increase the reliability of the current failure probability. Then, the work type identification, the current failure probability and the attention feature are spliced ​​and input into the classification model for classification processing to obtain the target fault detection result of the gearbox. The work type identification represents the type or working condition of the gearbox, which can affect the fault detection result in one dimension, and the current fault probability and the attention feature can affect the fault detection result in the other two dimensions. Therefore, the work type identification, the current failure probability and the attention feature have a guiding role in determining the fault detection result, avoiding the classification model from making a relatively single judgment only through the attention feature, which can improve the accuracy of the fault detection result.

[0140] In addition, refer to Figure 7 In one embodiment, before performing multi-scale feature extraction on each target data segment, the gearbox fault detection method further includes but is not limited to the following steps:

[0141] Step S710, obtaining sample operation timing data of the gearbox and sample detection labels corresponding to the sample operation timing data, and dividing the sample operation timing data into a plurality of sample data segments.

[0142] Step S720, input each sample data segment into the multi-scale feature extraction model for mapping, obtain the multi-scale sample features corresponding to each sample data segment, perform multi-scale weighted fusion on each multi-scale sample feature, and obtain the sample operation features corresponding to each target data segment.

[0143] Step S730: Input each sample operation feature into the time series model for mapping to obtain the sample time domain feature corresponding to each sample data segment.

[0144] Step S740, after attention processing, each sample time domain feature is input into the classification model for classification processing to obtain a sample fault detection result of the gearbox.

[0145] Step S750, determining the model loss based on the difference between the sample fault detection result and the sample detection label, and jointly training the multi-scale feature extraction model, the time series model, and the classification model based on the model loss.

[0146] Among them, the sample detection label is used to indicate that the sample running timing data belongs to a fault category or a normal type. For example, the fault category may include gear tooth breakage, gear defect, tooth surface wear, tooth root crack, bearing ball failure, inner and outer ring combined failure, outer ring failure and inner ring failure, etc.

[0147] Among them, the model loss is back-propagated to the classification model, timing model and multi-scale feature extraction model in turn, the parameter gradients in each model are calculated, and then the parameters of each model are adjusted according to the parameter gradients to minimize the model loss.

[0148] Based on this, firstly, by obtaining the sample operation timing data of the gearbox and the sample detection label corresponding to the sample operation timing data, the sample operation timing data is divided into multiple sample data segments, and then each sample data segment is input into the multi-scale feature extraction model for mapping, and the multi-scale sample features corresponding to each sample data segment are obtained. Multi-scale weighted fusion is performed on each multi-scale sample feature to obtain the sample operation features corresponding to each target data segment, and then each sample operation feature is input into the timing model for mapping to obtain the sample time domain features corresponding to each sample data segment, and then each sample time domain feature is input into the classification model after attention processing. Classification processing is performed to obtain the sample fault detection result of the gearbox. Finally, the model loss is determined based on the difference between the sample fault detection result and the sample detection label. Based on the model loss, the multi-scale feature extraction model, the timing model and the classification model are jointly trained, so that the multi-scale feature extraction model can learn to understand the local spatial information of the target operation timing data, the timing model can learn to understand the global time domain information of the target operation timing data, and the classification model can learn to understand the inherent connection between the local spatial information and the global time domain information of the target operation timing data. Then, in the reasoning stage, the target operation timing data of the gearbox is analyzed to obtain accurate fault detection results.

[0149] Specifically, refer to Fig.10 , Fig.10 An optional flowchart of the training process provided in an embodiment of the present application.

[0150] First, the data collected by the planetary gearbox experimental platform is obtained, and then the data collected by the planetary gearbox experimental platform is preprocessed, and then the preprocessed collected data is divided into a training data set and a test data set, so that the training data set includes sample running time data and sample detection labels corresponding to the sample running time data. For example, the data volume ratio of the training data set and the test data set is 8:2;

[0151] Then, all models and networks are initialized with parameters, and then the sample data segments of the sample running sequence data are input into the multi-scale feature extraction model for mapping, and each multi-scale sample feature is subjected to multi-scale weighted fusion to obtain the sample running features corresponding to each target data segment;

[0152] Then, each sample running feature is input into the time series model for first mapping, and the mapping result is input into the time series model for second mapping to obtain the sample time domain feature corresponding to each sample data segment;

[0153] Then, the sample time domain features are regularized, and the regularized results are processed by the multi-head attention model and input into the fully connected layer for mapping and normalization to obtain the sample fault detection results of the gearbox. Here, the fully connected layer is equivalent to the classification model.

[0154] Then, the model loss is determined based on the difference between the sample fault detection results and the sample detection labels, and the multi-scale feature extraction model, timing model and classification model are jointly trained based on the model loss.

[0155] In a possible implementation, multi-scale weighted fusion is performed on each multi-scale sample feature through a weighted fusion model, and attention is processed on each sample time domain feature through a multi-head attention module.

[0156] Among them, the model loss can also be back-propagated to the multi-head attention processing model and the weighted fusion model to calculate the parameter gradients of each model, and then adjust the parameters of each model according to the parameter gradients to minimize the model loss.

[0157] In addition, refer to Fig.11 The present application also provides a gearbox fault detection device 1100, comprising:

[0158] An acquisition module 1110 is used to acquire the running timing data of the gearbox;

[0159] The multi-scale feature extraction module 1120 is used to divide the running time series data into multiple data segments, perform multi-scale feature extraction on each data segment, obtain multi-scale initial features corresponding to each data segment, perform multi-scale weighted fusion on each multi-scale initial feature, and obtain target running features corresponding to each data segment;

[0160] The time domain feature extraction module 1130 is used to input each target operation feature into the time series model for mapping to obtain the time domain feature corresponding to each data segment;

[0161] The classification module 1140 processes each time domain feature and then inputs it into the classification model for classification processing to obtain the gearbox fault detection result. It can be understood that the specific implementation of the gearbox fault detection device 1100 is basically the same as the specific implementation of the above-mentioned gearbox fault detection method, which will not be repeated here.

[0162] In addition, refer to Fig.12 , Fig.12 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0163] The processor 1201 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0164] The memory 1202 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1202 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1202, and the processor 1201 is called to execute the gearbox fault detection method of the embodiment of the present application.

[0165] Input / output interface 1203, used to implement information input and output;

[0166] The communication interface 1204 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0167] A bus 1205 that transmits information between various components of the device (e.g., the processor 1201, the memory 1202, the input / output interface 1203, and the communication interface 1204);

[0168] The processor 1201 , the memory 1202 , the input / output interface 1203 and the communication interface 1204 are connected to each other in communication within the device via the bus 1205 .

[0169] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned gearbox fault detection method.

[0170] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0171] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0172] It can be understood by those skilled in the art that Figures 1 to 12 The technical solutions shown in the figure do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figure, or a combination of certain steps, or different steps.

[0173] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0174] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0175] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0176] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0177] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0178] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0179] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0180] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0181] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A gearbox fault detection method, characterized in that: include: Acquire target operating time data of the gearbox, and divide the target operating time data into a plurality of target data segments; Inputting each of the target data segments into a multi-scale feature extraction model for mapping, respectively, to obtain a multi-scale initial feature corresponding to each of the target data segments, and performing multi-scale weighted fusion on each of the multi-scale initial features, respectively, to obtain a target operation feature corresponding to each of the target data segments; Inputting each of the target operation characteristics into a time series model for mapping, and obtaining a target time domain characteristic corresponding to each of the target data segments; After attention processing, each of the target time domain features is input into a classification model for classification processing to obtain a target fault detection result of the gearbox.

2. The gearbox fault detection method according to claim 1, characterized in that: The multi-scale feature extraction model includes a first feature extraction network and a second feature extraction network, the first feature extraction network includes N first convolutional layers connected in sequence, the second feature extraction network includes M second convolutional layers connected in sequence, N is less than M, N and M are both positive integers, and the multi-scale initial features include first-scale initial features and second-scale initial features; The step of inputting each of the target data segments into a multi-scale feature extraction model for mapping to obtain a multi-scale initial feature corresponding to each of the target data segments, and performing multi-scale weighted fusion on each of the multi-scale initial features to obtain a target operation feature corresponding to each of the target data segments includes: For each of the target data segments, input the target data segment into the first feature extraction network for mapping to obtain the first-scale initial features corresponding to the target data segment, and input the target data segment into the second feature extraction network for mapping to obtain the second-scale initial features corresponding to the target data segment; For each of the target data segments, multi-scale weighted fusion is performed on the first-scale initial features and the second-scale initial features corresponding to the target data segment to obtain the target operation features corresponding to the target data segment.

3. The gearbox fault detection method according to claim 2, characterized in that: The first feature extraction network includes a first convolutional layer, a second convolutional layer, and a first pooling layer connected in sequence, the second feature extraction network includes a third convolutional layer, a fourth convolutional layer, a second pooling layer, a fifth convolutional layer, a sixth convolutional layer, and a third pooling layer connected in sequence, the convolution kernel size of the first convolutional layer is greater than or equal to the convolution kernel size of the second convolutional layer, the convolution kernel size of the first convolutional layer is greater than the convolution kernel size of the third convolutional layer, the convolution kernel size of the third convolutional layer is greater than or equal to the convolution kernel size of the fourth convolutional layer, the convolution kernel size of the third convolutional layer is greater than the convolution kernel size of the fifth convolutional layer, and the convolution kernel size of the fifth convolutional layer is greater than or equal to the convolution kernel size of the sixth convolutional layer; Inputting the target data segment into the first feature extraction network for mapping to obtain the first-scale initial features corresponding to the target data segment, and inputting the target data segment into the second feature extraction network for mapping to obtain the second-scale initial features corresponding to the target data segment, include: Calling the first convolution layer and the second convolution layer to perform convolution processing on the target data segment to obtain a first intermediate feature, and calling the first pooling layer to perform a pooling operation on the first intermediate feature to obtain the first-scale initial feature corresponding to the target data segment; Calling the third convolution layer and the fourth convolution layer to perform convolution processing on the target data segment to obtain a second intermediate feature, and calling the second pooling layer to perform a pooling operation on the second intermediate feature to obtain a third intermediate feature; The fifth convolution layer and the sixth convolution layer are called to perform convolution processing on the third intermediate feature to obtain a fourth intermediate feature, and the third pooling layer is called to perform a pooling operation on the fourth intermediate feature to obtain the second-scale initial feature corresponding to the target data segment.

4. The gearbox fault detection method according to claim 1, characterized in that: Inputting each of the target operation features into the time series model for mapping to obtain the target time domain features corresponding to each of the target data segments includes: Inputting each of the target operation characteristics into a time series model, and determining a forward hidden layer state and a reverse hidden layer state corresponding to each of the target operation characteristics based on the time series model; For each of the target operation features, a weighted sum is performed based on the forward hidden layer state and the reverse hidden layer state corresponding to the target operation feature to determine the target time domain feature corresponding to the target data segment.

5. The gearbox fault detection method according to claim 1, characterized in that: The attention processing of each of the target time domain features is input into the classification model for classification processing to obtain the target fault detection result of the gearbox, including: Performing attention processing on each of the target time domain features to obtain an attention feature; The working type identification of the gearbox is obtained, and the working type identification and the attention feature are spliced ​​and input into the classification model for classification processing to obtain the target fault detection result of the gearbox.

6. The gearbox fault detection method according to claim 5, characterized in that: The step of splicing the work type identifier and the attention feature and inputting them into a classification model for classification processing to obtain a target fault detection result of the gearbox includes: Obtaining a historical failure probability of the gearbox; Inputting the attention feature and the historical failure probability into a probability prediction model for regression processing to obtain a current failure probability of the gearbox; The working type identification, the current fault probability and the attention feature are spliced ​​and input into a classification model for classification processing to obtain a target fault detection result of the gearbox.

7. The gearbox fault detection method according to claim 1, characterized in that: Before performing multi-scale feature extraction on each of the target data segments, the gearbox fault detection method further includes: Acquire sample operation timing data of the gearbox and sample detection labels corresponding to the sample operation timing data, and divide the sample operation timing data into a plurality of sample data segments, wherein the sample detection labels are used to indicate that the sample operation timing data belongs to a fault category or a normal type; Inputting each of the sample data segments into a multi-scale feature extraction model for mapping, respectively, to obtain multi-scale sample features corresponding to each of the sample data segments, and performing multi-scale weighted fusion on each of the multi-scale sample features, respectively, to obtain sample operation features corresponding to each of the target data segments; Inputting each of the sample operation features into a time series model for mapping to obtain a sample time domain feature corresponding to each of the sample data segments; After performing attention processing on each of the sample time domain features, the features are input into a classification model for classification processing to obtain a sample fault detection result of the gearbox; A model loss is determined based on a difference between the sample fault detection result and the sample detection label, and the multi-scale feature extraction model, the timing model, and the classification model are jointly trained based on the model loss.

8. A gearbox fault detection device, characterized in that: include: An acquisition module, used for acquiring the running timing data of the gearbox; A multi-scale feature extraction module is used to divide the running time series data into multiple data segments, perform multi-scale feature extraction on each of the data segments, obtain multi-scale initial features corresponding to each of the data segments, and perform multi-scale weighted fusion on each of the multi-scale initial features to obtain target running features corresponding to each of the data segments; A time domain feature extraction module, used for inputting each of the target operation features into a time series model for mapping, and obtaining the time domain features corresponding to each of the data segments; The classification module performs attention processing on each of the time domain features and then inputs the features into the classification model for classification processing to obtain the fault detection result of the gearbox.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the gearbox fault detection method according to any one of claims 1 to 7 when executing the computer program.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the gearbox fault detection method according to any one of claims 1 to 7 is implemented.

Images