Mechanical composite fault data enhancement method, system, medium and equipment

Abstract

The invention relates to the field of data processing, and discloses a mechanical composite fault data enhancement method and system, a medium and equipment, and the method comprises the steps: obtaining various single fault data, and setting a Dirichlet distribution parameter alpha to control the shape of Dirichlet distribution; under the control of the parameter alpha, mixing coefficient distribution of each single fault data is sampled through Dirichlet distribution, so that a mixing coefficient beta i of each single fault data is generated; and determining a mixing coefficient beta i meeting a set condition, and superposing and mixing various single fault data according to the determined mixing coefficient beta i to form simulated composite fault data. The problem that the composite fault data is difficult to obtain is solved, and meanwhile the cost of obtaining the composite fault data is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method, system, medium and equipment for enhancing mechanical composite fault data. Background Art

[0002] In the field of fault diagnosis technology, whether it is the traditional diagnosis theory and method based on mechanism research, the fault diagnosis method based on signal processing and feature extraction, the diagnosis method based on knowledge expert system, or the latest diagnosis method based on deep learning, most of the research focuses on the single fault field. At present, compound faults have received more and more attention, and many scholars have published a large number of papers from mechanism research to diagnosis research based on deep neural networks. However, the compound faults of machinery have greatly increased the difficulty of fault diagnosis. On the one hand, the mechanism of the occurrence and evolution of compound faults is difficult to reveal from the perspective of cause and effect, and it is difficult to summarize the general rules. On the other hand, the relationship between compound faults and their corresponding single faults is closely related. Due to the large number of mechanical parts and fault types, the theoretical analysis modeling is complex, and it is impossible to study the combination of all compound faults in large quantities.

[0003] Moreover, the existing single fault data is relatively easy to obtain, while the compound fault data is relatively difficult to obtain. Especially when there are a large number of combinations of multiple single faults, it is impossible to obtain the compound fault data of all combinations in reality. Therefore, how to effectively obtain compound fault data has become a technical problem that needs to be solved urgently in the current fault diagnosis field. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a mechanical composite fault data enhancement method, system, medium and device, which solves the problem of difficulty in obtaining composite fault data and reduces the cost of obtaining composite fault data.

[0005] To achieve the above-mentioned purpose, in a first aspect, the technical solution adopted by the present invention is: a method for enhancing mechanical composite fault data, comprising: obtaining various single fault data, and setting a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution; under the control of the parameter α, sampling the mixing coefficient distribution of each single fault data through the Dirichlet distribution to generate a mixing coefficient β of each single fault data i ; Where i = 1, 2, ..., k, k is the total number of single faults; Determine the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i The simulated composite fault data is formed by superimposing and mixing, wherein the simulated composite fault data is: x = β 1 x 1 +β 2x 2 +…+β k x k , where x i is the single fault data, and x is the synthetic composite fault data.

[0006] Furthermore, the parameter α of the Dirichlet distribution is an array of k elements, each element corresponding to a single fault type.

[0007] Further, setting a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution includes: setting the value of the parameter α of the Dirichlet distribution to be greater than or equal to 1.

[0008] Furthermore, the setting conditions are: mixing coefficient β i is greater than 0, and

[0009] Furthermore, various single fault data are mixed according to the determined mixing coefficient β i Superposition and mixing are performed to form simulated composite fault data, including: when the degree of all single faults is equal or similar, a symmetrical Dirichlet distribution is adopted, the values ​​of all parameters α are equal, and all single faults are superimposed and mixed on average fairly to form simulated composite fault data.

[0010] Furthermore, various single fault data are mixed according to the determined mixing coefficient β i Superposition and mixing are performed to form simulated composite fault data, including: when the degree of a single fault is in a non-equivalent state or the difference exceeds a set range, an asymmetric Dirichlet distribution is used to adjust the value of the parameter α corresponding to the fault information far from the monitoring sensor to generate a mixing coefficient with a larger mean, and then superposition and mixing are performed to form simulated composite fault data.

[0011] Furthermore, after the composite fault data is formed, a step of label mixed setting is also included, which is:

[0012] Setting a label corresponding to a single fault of each compound fault to form a label of the compound fault;

[0013] The mixed simulated compound fault data and its labels are divided into a training set, a validation set and a test set together with the single fault data for training the neural network model. The trained model is used for the diagnosis of compound faults.

[0014] In a second aspect, the technical solution adopted by the present invention is: a mechanical composite fault data enhancement system, which includes: a data acquisition module, which acquires various single fault data and sets a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution; a mixing coefficient acquisition module, which samples the mixing coefficient distribution of each single fault data through the Dirichlet distribution under the control of the parameter α to generate a mixing coefficient β of each single fault data i ; Where i = 1, 2, ..., k, k is the total number of single faults; Mixing module, determine the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i The simulated composite fault data is formed by superimposing and mixing, wherein the simulated composite fault data is: x = β 1 x 1 +β 2 x 2 +…+β k x k , where x i is the single fault data, and x is the synthetic composite fault data.

[0015] In a third aspect, the technical solution adopted by the present invention is: a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by a computing device, the computing device executes any one of the above methods.

[0016] In a fourth aspect, the technical solution adopted by the present invention is: a computing device, comprising: one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the above methods.

[0017] The present invention adopts the above technical solution, which has the following advantages:

[0018] The present invention simulates compound faults of mechanical equipment by mixing single fault data collected from mechanical equipment or experimental environment, thereby solving the problem of difficulty in obtaining compound fault data and greatly reducing the cost of compound fault data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of a method for enhancing mechanical composite fault data in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of a mixed fitting compound fault in an embodiment of the present invention;

[0021] Figure 3It is a schematic diagram of a typical deep neural network fault diagnosis model in the prior art;

[0022] Figure 4 These are the three output modes used in fault diagnosis in the prior art. DETAILED DESCRIPTION

[0023] During the normal operation of mechanical equipment, parts may be damaged due to various reasons such as design, processing, and wear, which may cause the entire equipment to fail, causing serious equipment damage and even casualties. Therefore, the health monitoring, fault diagnosis and prediction of mechanical equipment, that is, predictive maintenance, has received more and more attention. The failures of these equipment can be divided into single failures and compound failures. Compared with single failures, compound failures are more common in actual engineering applications, more dangerous, and more difficult to diagnose. Therefore, achieving accurate compound fault diagnosis of mechanical equipment is a challenging task.

[0024] Generally, there are three main manifestations of compound failures: 1) Multiple failures occur simultaneously in the same part, such as pitting failures of the inner raceway and outer raceway of a rolling bearing. 2) Multiple parts in a component are damaged successively or simultaneously, such as bearing failure and gear failure in a gearbox. 3) Multiple components of a device are damaged successively or simultaneously, such as engine failure and gearbox failure occurring simultaneously in a complex system. The occurrence of compound failures is highly uncertain and has low predictability, which is more likely to cause serious equipment damage and bring significant economic losses.

[0025] The application of deep neural networks in the field of equipment fault diagnosis is constantly expanding and deepening. A large number of research papers have been published, and some have been verified for specific equipment. However, these studies mainly focus on single fault scenarios, and only a relatively small number of studies focus on compound fault scenarios. In the current research, typical deep neural network fault diagnosis models are as follows: Figure 3 As shown. Usually, the neural network consists of three modules: data preprocessing, feature extraction and fault classification. These modules have been widely and fully studied. For example, the data preprocessing module can convert the vibration data collected from the equipment into a frequency spectrum or envelope spectrum in the frequency domain, or use wavelet transform or short-time Fourier transform to convert it into the time-frequency domain; the feature extraction module can use convolutional networks, capsule networks, recurrent neural networks, etc.

[0026] For the output of compound faults, three methods are usually used, such as Figure 4As shown. Using softmax as the output activation function requires treating compound faults as a new single fault, because softmax only outputs one label at a time. Using sigmoid and multi-output methods, multiple labels can be output at a time, that is, the decoupling of compound faults can be achieved and the corresponding single faults can be output. All of these research works rely on training data, that is, fault data. Usually, it is difficult to obtain relevant data at the equipment operation site, especially labeled fault data. In most cases, existing research works obtain relevant fault data from experimental environments. In experimental environments, single fault data is relatively easy to obtain, while compound fault data is relatively difficult to obtain. Especially when there are a large number of combinations of multiple single faults, it is impossible to obtain compound fault data of all combinations in reality.

[0027] For example, there are two ways to obtain fault data: one is to obtain it from the working site of the equipment, and the other is to obtain it from the simulation environment of the laboratory. Taking vibration data as an example, no matter which method is used to obtain data, it is necessary to install a vibration sensor on the equipment being tested to obtain relevant data, and save the data for use in the neural network training model. There are limitations to the above two methods of obtaining training data for fault data: First, it is very difficult to obtain fault data from the working site of the equipment. First, it is rare to collect fault data from equipment that works normally on site, because many equipment use regular maintenance to replace parts to avoid failures as much as possible and interrupt normal production. Another point is that it is very difficult to obtain labeled data on site. Artificial intelligence data labeling requires a lot of meticulous work, and it is difficult for manufacturing companies to complete the relevant labeling work. Secondly, it is relatively easy to obtain data from the simulation environment of the laboratory. Collect some faulty parts, or process some faulty parts for data collection. One problem with this method is that when there are multiple single faults, taking bearings as an example, they are usually caused by bearing inner ring faults, outer ring faults, rolling element faults, and cage faults. If you want to collect composite fault data, you need When there are many single fault types, this is a very large combination and it is difficult to implement in reality. Even if there are a small number of combinations, the acquisition of experimental data requires a high experimental cost and a long experimental time.

[0028] For another example, in the prior art, a method of generating composite fault data using a generative adversarial network is used. The generative adversarial network is trained using composite fault data collected on-site or simulation data collected in the laboratory. After the training is completed, the generative network can be used to generate composite fault data. However, since this method requires composite fault data to train the generative adversarial network, the shortcomings of the above two fault data acquisition methods still exist. In addition, the generative adversarial model also has the disadvantage of being difficult to train.

[0029] In order to solve the problem in the prior art that the composite fault data of mechanical equipment faces great difficulties whether it is obtained from the field or from the experimental environment, and the cost of data acquisition is also very high, the present invention provides a mechanical composite fault data enhancement method, system, medium and equipment, which adopts single fault data obtained in experiments or on-site, and reduces the difficulty and cost of obtaining composite fault data by combining simulated composite fault data.

[0030] The present invention does not require a large number of combined experiments on single faults to obtain actual composite fault data, and does not require the implementation of composite fault experiments to obtain actual composite fault data, thereby solving the problem that composite fault data is difficult to obtain and reducing the demand for composite fault experimental data.

[0031] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0032] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0033] In one embodiment of the present invention, a method for enhancing mechanical composite fault data is provided. In this embodiment, Figure 1 , Figure 2 As shown, the method comprises the following steps:

[0034] 1) Obtain various single fault data and set the parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution.

[0035] 2) Under the control of parameter α, the mixing coefficient distribution of each single fault data is sampled through Dirichlet distribution to generate the mixing coefficient β of each single fault data i ; Wherein, i=1, 2, ..., k, k is the total number of single faults.

[0036] 3) Determine the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i Superimpose and mix to form simulated composite fault data.

[0037] Among them, the simulated composite fault data is:

[0038] x=β 1 x 1 +β 2 x 2 +…+β k x k ,

[0039] In the formula, x i is the single fault data, and x is the synthetic composite fault data.

[0040] When the present invention is used, single fault data are added together. For example, the bearing inner ring fault data and the outer ring fault data are added together to simulate the composite fault data of the inner ring and the outer ring failing at the same time; the bearing outer ring fault data and the flywheel imbalance data are added together to simulate the composite fault data between parts; or the bearing inner ring fault, the gear tooth breakage fault and the flywheel imbalance fault are added together to simulate the composite fault data of the three parts failing at the same time.

[0041] In the above step 1), the parameter α of the Dirichlet distribution is an array consisting of k elements, which is used to control the shape of the Dirichlet distribution, and each element corresponds to a single fault type.

[0042] In this embodiment, the value of parameter α of the Dirichlet distribution is set to be greater than or equal to 1. When the value of parameter α is less than 1, the distribution tends to the edge of simplicity, resulting in some single fault mixing coefficients being very small, thus losing the meaning of mixing.

[0043] In the above step 2), under the control of parameter α, the mixing coefficient distribution of each single fault data is sampled by Dirichlet distribution to generate the mixing coefficient β of each single fault data. i , as follows:

[0044] In this embodiment, when a composite fault occurs in the equipment, the severity of the single faults that make up the composite fault will be different, resulting in different feature strengths in the fault data. For example, the feature of the pitting fault of the bearing outer ring is weaker than the fault feature of the gear tooth breakage. The interweaving of faults of different severity is a feature of the composite fault. In order to synthesize composite fault combinations of various severity, the Dirichlet distribution is used to sample the mixing coefficient distribution to generate the mixing coefficient β of each single fault in the composite fault. i .

[0045] In the above step 3), the setting conditions are: mixing coefficient β i is greater than 0, and Among them, β~Dirichlet(α).

[0046] In the above step 3), various single fault data are mixed according to the determined mixing coefficient β i Superimpose and mix to form simulated composite fault data (such as Figure 2 shown), including:

[0047] 3.1) When all single faults are of equal or similar severity, a symmetric Dirichlet distribution is used, and the values ​​of all parameters α are equal, that is, α 1 =α 2 =…=α k At this time, all single faults are averaged and fairly superimposed to form a variety of simulated composite fault data. Among them, the larger the α value, the closer it is to uniform distribution. The specific value can be determined according to the actual situation. The final simulated composite fault data is:

[0048] x=β 1 x 1 +β 2 x 2 +…+β k x k ,

[0049] In the formula, x i is the single fault data, and x is the synthetic composite fault data.

[0050] 3.2) When the single fault degree is not equal or the difference exceeds the set range, the asymmetric Dirichlet distribution is used to adjust the value of the parameter α corresponding to the fault information far from the monitoring sensor to generate a mixing coefficient with a larger mean, and then superimpose and mix to form simulated composite fault data.

[0051] Specifically, for the working status data collected by mechanical equipment, such as vibration data, the amplitude of the information spectrum of some parts may be small due to the structure of the mechanical equipment. For example, although the defects on the inner and outer raceways of rolling bearings behave similarly, for defects of a given size, the spectrum amplitude of the inner raceway defect is usually much smaller. The reason is that the defects on the inner raceway enter the load zone only once per rotation, and the signal must pass through more structural interfaces, namely rolling elements, oil films, outer rings, and bearing seats before reaching the sensor position. The transmission path of the inner raceway fault is more difficult. For this reason, when synthesizing the data of the simulated composite fault, an asymmetric Dirichlet distribution is used, that is, the Dirichlet distribution parameter α is adjusted, and the α corresponding to the information with small spectrum amplitude is set to a larger value to generate a mixing coefficient with a larger mean. At this time, the proportion of this information in the simulated composite fault data is generally enhanced, so that the deep neural network model can better learn the relevant fault characteristics.

[0052] In this embodiment, symmetric Dirichlet distribution is used to generate fusion coefficients to mix single fault data into composite fault data; asymmetric Dirichlet distribution is used to generate fusion coefficients to enhance the strength of single faults with smaller spectrum amplitudes in simulating composite faults.

[0053] In the above embodiments, after the composite fault data is formed, a step of mixed label setting is also included, specifically:

[0054] 4) For mixed simulated composite faults, corresponding training labels are generated, that is, labels corresponding to the single faults of each composite fault are set to form labels of the composite faults; in this embodiment, multi-hot labels are used.

[0055] 5) The mixed simulated composite fault data and its labels are divided into a training set, a validation set and a test set together with the single fault data for training the neural network model. The trained model is used for the diagnosis of composite faults.

[0056] In one embodiment of the present invention, a mechanical compound fault data enhancement system is provided, comprising:

[0057] A data acquisition module acquires various single fault data and sets a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution;

[0058] The mixing coefficient acquisition module samples the mixing coefficient distribution of each single fault data through Dirichlet distribution under the control of parameter α to generate the mixing coefficient β of each single fault data. i ; Where i = 1, 2, ..., k, k is the total number of single faults;

[0059] Mixing module, determines the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i Perform superposition and mixing to form simulated composite fault data;

[0060] Among them, the simulated composite fault data is:

[0061] x=β 1 x 1 +β 2 x 2 +…+β k x k ,

[0062] In the formula, x i is the single fault data, and x is the synthetic composite fault data.

[0063] In the above embodiment, the parameter α of the Dirichlet distribution is an array consisting of k elements, each element corresponding to a single fault type. The value of the parameter α of the Dirichlet distribution is set to be greater than or equal to 1.

[0064] In the above embodiment, the setting conditions are: the mixing coefficient β i is greater than 0, and

[0065] In the above embodiment, various single fault data are obtained according to the mixed coefficient β i Superimpose and mix to form simulated composite fault data, specifically:

[0066] When the degree of all single faults is equal or similar, the symmetric Dirichlet distribution is used, the values ​​of all parameters α are equal, and all single faults are superimposed and mixed on average to form simulated composite fault data.

[0067] In the above embodiment, various single fault data are obtained according to the mixed coefficient β i Superimpose and mix to form simulated composite fault data, including:

[0068] When the single fault degree is in a non-equivalent state or the difference exceeds the set range, the asymmetric Dirichlet distribution is used to adjust the value of the parameter α corresponding to the fault information far from the monitoring sensor to generate a mixing coefficient with a larger mean, and then superimpose and mix to form simulated composite fault data.

[0069] In the above embodiment, after the composite fault data is formed, a step of label mixed setting is also included, which is:

[0070] Setting a label corresponding to a single fault of each compound fault to form a label of the compound fault;

[0071] The mixed simulated compound fault data and its labels are divided into a training set, a validation set and a test set together with the single fault data for training the neural network model. The trained model is used for the diagnosis of compound faults.

[0072] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0073] In one embodiment of the present invention, a computing device is provided, which may be a terminal, and may include: a processor, a communication interface, a memory, a display screen, and an input device. Among them, the processor, the communication interface, and the memory communicate with each other through a communication bus. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and an internal memory, and the non-volatile storage medium stores an operating system and a computer program. When the computer program is executed by the processor, the method in each of the above embodiments is implemented; the internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a management network, NFC (near field communication) or other technologies. The display screen may be a liquid crystal display screen or an electronic ink display screen, and the input device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computing device, or an external keyboard, touchpad or mouse, etc. The processor may call the logic instructions in the memory.

[0074] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0075] In one embodiment of the present invention, a computer program product is provided, wherein the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments.

[0076] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores server instructions, and the computer instructions enable a computer to execute the methods provided in the above embodiments.

[0077] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0078] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0079] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0080] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for enhancing mechanical composite fault data, characterized in that: include: Obtain various single fault data and set the parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution; Under the control of parameter α, the mixing coefficient distribution of each single fault data is sampled through Dirichlet distribution to generate the mixing coefficient β of each single fault data i ; Where i = 1, 2, ..., k, k is the total number of single faults; Determine the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i Perform superposition and mixing to form simulated composite fault data; Among them, the simulated composite fault data is: x=β1x1+β2x2+…+β k x k , In the formula, x i is the single fault data, and x is the synthetic composite fault data.

2. The mechanical composite fault data enhancement method according to claim 1, characterized in that: The parameter α of the Dirichlet distribution is an array of k elements, each element corresponding to a single fault type.

3. The mechanical compound fault data enhancement method according to claim 2, characterized in that: Setting a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution includes: setting the value of the parameter α of the Dirichlet distribution to be greater than or equal to 1.

4. The mechanical compound fault data enhancement method according to claim 1, characterized in that: Setting conditions: Mixing coefficient β i is greater than 0, and 5. The mechanical compound fault data enhancement method according to claim 1, characterized in that: Various single fault data are mixed according to the determined mixing coefficient β i Superimpose and mix to form simulated composite fault data, including: When the degree of all single faults is equal or similar, the symmetric Dirichlet distribution is used, the values ​​of all parameters α are equal, and all single faults are superimposed and mixed on average to form simulated composite fault data.

6. The mechanical compound fault data enhancement method according to claim 1, characterized in that: Various single fault data are mixed according to the determined mixing coefficient β i Superimpose and mix to form simulated composite fault data, including: When the single fault degree is in a non-equivalent state or the difference exceeds the set range, the asymmetric Dirichlet distribution is used to adjust the value of the parameter α corresponding to the fault information far from the monitoring sensor to generate a mixing coefficient with a larger mean, and then superimpose and mix to form simulated composite fault data.

7. The mechanical compound fault data enhancement method according to claim 1, characterized in that: After forming the composite fault data, the steps of label mixed setting are also included, which are: Setting a label corresponding to a single fault of each compound fault to form a label of the compound fault; The mixed simulated compound fault data and its labels are divided into a training set, a validation set and a test set together with the single fault data for training the neural network model. The trained model is used for the diagnosis of compound faults.

8. A mechanical composite fault data enhancement system, characterized in that: include: A data acquisition module acquires various single fault data and sets a parameter α of the Dirichlet distribution to control the shape of the Dirichlet distribution; The mixing coefficient acquisition module samples the mixing coefficient distribution of each single fault data through Dirichlet distribution under the control of parameter α to generate the mixing coefficient β of each single fault data. i ; Where i = 1, 2, ..., k, k is the total number of single faults; Mixing module, determines the mixing coefficient β that meets the set conditions i , various single fault data are mixed according to the determined mixing coefficient β i Perform superposition and mixing to form simulated composite fault data; Among them, the simulated composite fault data is: x=β1x1+β2x2+…+β k x k , In the formula, x i is the single fault data, and x is the synthetic composite fault data.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform any one of the methods of claims 1 to 7.

10. A computing device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any one of the methods described in claims 1 to 7.

Images