Hydraulic shock absorber unsupervised fault detection method, system, device and storage medium

Abstract

The present invention discloses an unsupervised fault detection method, system, equipment and storage medium for hydraulic shock absorbers, which are one-to-one corresponding schemes. In the scheme: an unsupervised training detection model is performed using data in a healthy state of the shock absorber. Specifically, sample consistency is measured by comparing the health center and healthy samples, and multi-sensor signal correlation of the health center and multi-sensor signal correlation of the healthy samples are compared to measure multi-sensor pattern consistency. Then, the working state of the shock absorber is directly modeled through the two types of consistency, and better detection results can be achieved without the participation of fault samples in training.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection of hydraulic shock absorbers, and in particular to an unsupervised fault detection method, system, equipment and storage medium for hydraulic shock absorbers. Background Art

[0002] Shock absorbers are devices used to measure vibration amplitude. Their health status is closely related to the stability of the system. Therefore, it is necessary to perform shock absorber fault detection to avoid unexpected faults that cause system crashes. During the shock absorber fault detection process, if the healthy state (normal state) is identified as a fault state, the shock absorber will be replaced prematurely, resulting in waste; and if the fault state is identified as a healthy state, the entire system may crash, posing a safety hazard.

[0003] In the Chinese invention patent application "Shock absorber fault detection method, device, equipment and storage medium" with publication number CN115578597A, the vibration performance differences between different sensors are used for comparative learning to achieve shock absorber fault detection under small samples. In the Chinese invention patent application "Fault detection method, device, electronic equipment and computer-readable storage medium" with publication number CN116067635A, a large amount of shock absorber operation data is collected to train a convolutional neural network to achieve shock absorber fault detection. However, these methods do not take into account the differences in the working modes of the shock absorber under different health states, that is, the changes in sensor signals and the changes in multi-sensor modes. Therefore, the detection performance needs to be improved.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a method, system, device and storage medium for unsupervised fault detection of hydraulic shock absorbers, which utilizes the data of the shock absorber in a healthy state to unsupervise the training detection model, thereby accurately detecting whether the shock absorber has a fault.

[0006] The objective of the present invention is achieved through the following technical solutions:

[0007] A method for unsupervised fault detection of a hydraulic shock absorber, comprising:

[0008] Collect the three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it as training sample data;

[0009] Construct an unsupervised fault detection model for hydraulic shock absorbers, including: a health center iterative update module, a sample consistency check module, and a multi-sensor mode consistency check module;

[0010] The unsupervised fault detection model of the hydraulic shock absorber is trained by using the training sample data, including: the health center iterative update module updates the current global health center by using the training sample data; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain the sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain the multi-sensor mode consistency check result; the unsupervised fault detection model of the hydraulic shock absorber is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result; after the training is completed, the final global health center is obtained and the health center iterative update module is removed;

[0011] The three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be tested is input into the trained unsupervised fault detection model of the hydraulic shock absorber to obtain the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center, so as to judge the working status of the hydraulic shock absorber.

[0012] A hydraulic shock absorber unsupervised fault detection system, comprising:

[0013] A training sample construction unit is used to collect three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it into training sample data;

[0014] A model building unit, used to build an unsupervised fault detection model for hydraulic shock absorbers, including: a health center iterative update module, a sample consistency check module, and a multi-sensor mode consistency check module;

[0015] A model training unit, used for training the hydraulic shock absorber unsupervised fault detection model using the training sample data, comprising: the health center iterative update module using the training sample data to update the current global health center; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain a sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain a multi-sensor mode consistency check result; the hydraulic shock absorber unsupervised fault detection model is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result; after the training is completed, the final global health center is obtained and the health center iterative update module is removed;

[0016] The fault detection unit is used to input the three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be detected into the trained unsupervised fault detection model of the hydraulic shock absorber, and obtain the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center, so as to judge the working state of the hydraulic shock absorber.

[0017] A processing device, comprising: one or more processors; a memory for storing one or more programs;

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.

[0019] A readable storage medium stores a computer program, which implements the above method when the computer program is executed by a processor.

[0020] It can be seen from the technical solution provided by the present invention that the detection model is unsupervisedly trained using data in the healthy state of the shock absorber. Specifically, the sample consistency is measured by comparing the health center and the healthy samples, and the multi-sensor signal correlation of the health center and the multi-sensor signal correlation of the healthy samples are compared to measure the multi-sensor pattern consistency. Then, the working state of the shock absorber is directly modeled through the two types of consistency, which can achieve better detection results without the participation of fault samples in training. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 A flow chart of an unsupervised fault detection method for a hydraulic shock absorber provided by an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of an unsupervised fault detection model for a hydraulic shock absorber provided in an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of an unsupervised fault detection model for a hydraulic shock absorber provided in an embodiment of the present invention;

[0025] Figure 4 A schematic diagram of experimental results provided by an embodiment of the present invention;

[0026] Figure 5A schematic diagram of an unsupervised fault detection system for a hydraulic shock absorber provided by an embodiment of the present invention;

[0027] Figure 6 A schematic diagram of a processing device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of the present invention.

[0029] First, the terms that may be used in this article are explained as follows:

[0030] The terms "include", "comprises", "contains", "has" or other descriptions with similar semantics should be interpreted as non-exclusive inclusion. For example, including certain technical feature elements (such as raw materials, components, ingredients, carriers, dosage forms, materials, dimensions, parts, components, mechanisms, devices, steps, procedures, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products or products, etc.) should be interpreted as including not only certain technical feature elements explicitly listed, but also other technical feature elements known in the art that are not explicitly listed.

[0031] The term "consisting of..." means excluding any technical feature elements not explicitly listed. If this term is used in a claim, it will make the claim closed, so that it does not contain technical feature elements other than the technical feature elements explicitly listed, except for the conventional impurities related to them. If this term only appears in a clause of a claim, it only limits the elements explicitly listed in the clause, and the elements recorded in other clauses are not excluded from the overall claim.

[0032] The following is a detailed description of a hydraulic shock absorber unsupervised fault detection method, system, device and storage medium provided by the present invention. The contents not described in detail in the embodiments of the present invention belong to the prior art known to professional and technical personnel in the field. If no specific conditions are specified in the embodiments of the present invention, they shall be carried out according to the conventional conditions in the field or the conditions recommended by the manufacturer.

[0033] Embodiment 1

[0034] The embodiment of the present invention provides a method for detecting unsupervised faults of a hydraulic shock absorber. Figure 1 As shown, it mainly includes the following steps:

[0035] Step 1: Collect the three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it as training sample data.

[0036] The preferred implementation of this step is as follows:

[0037] (1) Collecting the three-dimensional acceleration data of the sensor when the hydraulic shock absorber is in a healthy state , where each Represents acceleration data in one direction, and x, y, and z represent three directions in a three-dimensional rectangular coordinate system.

[0038] (2) According to the set sampling width, the three-dimensional acceleration data of the sensor in the healthy state of the hydraulic shock absorber is divided into samples , where each s represents a sample, the subscript is the sample number, N is the number of samples, and each sample is a three-dimensional acceleration data.

[0039] (3) Sample Transformed into a time-frequency matrix , each TF is the time-frequency matrix obtained by transforming the corresponding sample s, and the subscripts of the time-frequency matrix TF correspond one-to-one to the subscripts of the sample s.

[0040] (4) All time-frequency matrices obtained through transformation constitute the training sample data.

[0041] Step 2: Build an unsupervised fault detection model for hydraulic shock absorbers.

[0042] In the embodiment of the present invention, the unsupervised fault detection model of the hydraulic shock absorber mainly includes: a health center iterative update module, a sample consistency check module and a multi-sensor mode consistency check module.

[0043] Step 3: Use the training sample data to train the hydraulic shock absorber unsupervised fault detection model.

[0044] The training process is as follows: the health center iterative update module uses the training sample data to update the current global health center; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain a sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain a multi-sensor mode consistency check result; the hydraulic shock absorber unsupervised fault detection model is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result.

[0045] After training, the final global health center is obtained and the health center iterative update module is removed.

[0046] Step 4: Check the working status of the hydraulic shock absorber.

[0047] The detection method is as follows: the three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be detected is input into the trained unsupervised fault detection model of the hydraulic shock absorber, and the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center are obtained to judge the working status of the hydraulic shock absorber.

[0048] In order to more clearly demonstrate the technical solution and technical effects provided by the present invention, the method provided by the embodiment of the present invention is described in detail with reference to specific embodiments below.

[0049] 1. Data collection and processing.

[0050] (1) Data collection.

[0051] In the embodiment of the present invention, the data of the three-dimensional acceleration (i.e., the acceleration in three directions) of the shock absorber is collected in advance. , i.e., the three directions of x, y, and z in the three-dimensional rectangular coordinate system, x and y are the horizontal axis directions, and z is the axial direction. It should be noted that the present invention mainly collects and considers the sensor data of the shock absorber in a healthy state, and all kinds of sample data involved in the subsequent training process are data under the healthy state of the shock absorber.

[0052] (2) Data processing.

[0053] First, the 3D acceleration data is cut into several samples according to the set sampling width (for example, set to 1024). , N is the number of samples, each s is a sample, the subscript is the sample number, and each sample is a three-dimensional acceleration data, for example, represents the Nth sample, ,in , , is from arrive Acceleration data in the x, y, and z directions at all times.

[0054] Then, the sample Transformed into a time-frequency matrix , each TF is the time-frequency matrix obtained by transforming the corresponding sample s, and the subscripts of the time-frequency matrix TF correspond one-to-one to the subscripts of the sample s; for example, represents the time-frequency matrix of the Nth sample, , , , is from arrive The time-frequency matrix after the acceleration data in the x, y, and z directions is transformed.

[0055] Exemplarily, short-time Fourier transform may be used to implement the transformation of samples in time series form into a time-frequency matrix.

[0056] 2. Build and train an unsupervised fault detection model for hydraulic shock absorbers.

[0057] In the embodiment of the present invention, the hydraulic shock absorber unsupervised fault detection model mainly includes: a health center iterative update module, a sample consistency check module and a multi-sensor mode consistency check module; Figure 2 As shown, the main principles of the model are demonstrated.

[0058] 1. Iterative update module of health center.

[0059] In the embodiment of the present invention, the health center is randomly initialized, the health center is continuously approached through iterations during the training phase, and the health center is fixed during the testing phase.

[0060] (1) Randomly initialize the health center . Represent the acceleration time-frequency random initialization matrices in the x, y, and z directions respectively.

[0061] (2) During the training phase, the healthy center is continuously approached through iterations.

[0062] The number of training sample data in a batch in the current training is recorded as B, and the local health center of the training sample data in the batch is calculated by the following formula , expressed as:

[0063] ;

[0064] ;

[0065] ;

[0066] in, are the three acceleration time-frequency matrices of the local health center (i.e., the local health center in three directions), and x, y, and z represent the three directions in the three-dimensional rectangular coordinate system; is the jth training sample data in the batch, which is the acceleration time-frequency matrix in the x, y, and z directions.

[0067] (3) Utilize local health centers Update the current global health center , expressed as:

[0068] ;

[0069] ;

[0070] ;

[0071] in, is the global health center before the update. If this is the first training, then A randomly initialized health center , if it is not the first training, then Global Health Center updated for the last training; are the three acceleration time-frequency matrices in the global health center before updating, are the three acceleration time-frequency matrices in the current global health center.

[0072] 2. Sample consistency check module.

[0073] In the embodiment of the present invention, the sample consistency check module uses KL divergence to measure the difference between the training sample data and the health center under the same direction of acceleration, and uses it as a loss function to minimize the difference between the current health sample and the health center.

[0074] In the embodiment of the present invention, based on the first group of encoders, The time-frequency characteristics of , and based on the first set of encoders The time-frequency characteristics of Specifically, for the three acceleration directions, a first set of encoders is set in the sample consistency check module. , based on the first set of encoders, the time-frequency features of the current global health center and the time-frequency features of the training sample data are obtained. Considering the simplicity of the network structure, the health center and training samples of the same sensor (i.e., the same direction) share one encoder. At the same time, in order to reduce the loss of time-frequency information and avoid gradient disappearance, a residual structure is used to further improve the time-frequency feature extraction effect.

[0075] Based on the first set of encoders The process of obtaining each time-frequency feature is expressed as:

[0076] ;

[0077] ;

[0078] ;

[0079] Among them, the first group of encoders contains three encoders, whose subscripts correspond to the three directions of x, y, and z. ReLu is a rectified linear unit, which is a nonlinear activation function.

[0080] In the embodiment of the present invention, the first group of encoders It can be built by Transformer (transformation neural network) or CNN (convolutional neural network).

[0081] In an embodiment of the present invention, the consistency of the time-frequency features of the current training sample and the time-frequency features of the health center is measured, and the KL divergence is used as a metric, that is, the KL divergence is used to measure the distance between the time-frequency features of the current global health center and the time-frequency features of the training sample data in the same direction and accumulate them to obtain the sample consistency check result. Specifically, softmax (normalized exponential function) is used to convert the time-frequency feature matrix into a feature probability distribution. Since the KL divergence is asymmetric, cross calculation is required. In order to prevent the model from collapsing during training, a Stop mechanism is introduced, that is, the gradient of one of the items is ignored when calculating the KL divergence. The formalization is as follows:

[0082] ;

[0083] in, is the sample consistency check result (KL divergence loss), is the index symbol of the direction; Stop is the stopping mechanism, that is, the gradient inside the bracket after Stop is not calculated, and KL divergence is the relative entropy.

[0084] Considering the length of the article, Figure 2 The first group of encoders arranged in the sample consistency check module is not shown in FIG. Figure 3 As shown, taking a single direction (x direction) as an example, the working process of the sample consistency check module is presented.

[0085] 3. Multi-sensor mode consistency check module.

[0086] In an embodiment of the present invention, in a multi-sensor mode consistency check module, the correlation between the training sample data and the acceleration in different directions within the health center is first calculated respectively, and then the KL divergence is used to measure the internal correlation difference between the current health sample and the health center, and the KL divergence is used as a loss function to minimize the internal correlation difference between the current health sample and the health center.

[0087] In this embodiment of the present invention, the Pearson correlation function is selected To calculate the internal correlation, we describe the multi-sensor pattern.

[0088] Before calculating the internal correlation, it is necessary to obtain the time-frequency features and Therefore, a second set of encoders is provided in the multi-sensor mode consistency check module. Based on the second set of encoders, the time-frequency features are extracted again. The corresponding time-frequency features are extracted and recorded as , and based on the second set of encoders from the time-frequency features The corresponding time-frequency features are extracted and recorded as In this process, it is first encoded by the second set of encoders and then residually connected with the second set of encoder inputs, and then processed by a nonlinear activation function. For details, please refer to the process shown in the above three sets of formulas. The only difference is that the input information is different, so it will not be repeated.

[0089] In the embodiment of the present invention, the second group of encoders may be implemented using the same structure as the first group of encoders.

[0090] Afterwards, the internal correlation between the time-frequency features of the current global health center is calculated , expressed as:

[0091] ;

[0092] in, The corresponding representation and The correlation between and The correlation between and The correlation between them.

[0093] Calculate the internal correlation between the time-frequency features of the jth training sample data , expressed as:

[0094] ;

[0095] in, The corresponding representation and The correlation between and The correlation between and The correlation between them.

[0096] Then, the KL divergence is used to measure the distance between the internal correlation between the time-frequency characteristics of the current global health center and the internal correlation between the time-frequency characteristics of the training sample data to obtain the multi-sensor mode consistency check result, which is expressed as:

[0097] ;

[0098] in, is the result of the multi-sensor mode consistency check (KL divergence loss), Stop is the stopping mechanism, that is, the gradient of one item is ignored when calculating the KL divergence, KL divergence is the relative entropy; softmax is the normalized exponential function.

[0099] Construct the total loss function based on the above two consistency check results , which is expressed as follows:

[0100] ;

[0101] in, , and is a hyperparameter, represents the regularization loss, It is the parameters of the model, including the parameters of all modules within the model.

[0102] Based on the above total loss function, training is performed until convergence to obtain the final global health center. There is no need to update the global health center during the test, and the health center iterative update module can be removed.

[0103] Considering that the training process involved in this part can be implemented by referring to conventional technology, it will not be described in detail.

[0104] 3. Detection of the working status of hydraulic shock absorbers.

[0105] The monitoring threshold is set, and the shock absorber is inferred to be in a fault state according to the judgment criteria. In this process, the same calculation method as that in the sample consistency check module and the multi-sensor mode consistency check module is used, where the training sample data is replaced with the three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be tested (here it also needs to be converted into a time-frequency matrix), and the global health center is replaced with the final global health center obtained by training.

[0106] The sample consistency check result calculated in combination with the final global health center is recorded as , the consistency check result of multi-sensor mode is recorded as .

[0107] Afterwards, the fault score is calculated , expressed as:

[0108] ;

[0109] in, and is a hyperparameter.

[0110] Finally, the working state of the hydraulic shock absorber is determined according to the following formula:

[0111] ;

[0112] in, is the monitoring threshold set. is the judgment result, that is: When the hydraulic shock absorber is in the healthy state Health, when When , the hydraulic shock absorber is in fault state Fault.

[0113] In the embodiment of the present invention, the value of the monitoring threshold can be set according to the actual situation, and the theoretical value range is In practical applications, a value can be selected from [0,1], for example, 0.5.

[0114] The above-mentioned scheme provided by the embodiment of the present invention mainly has the following advantages: the existing methods do not utilize the differences in the working modes of the shock absorber in the healthy state and the faulty state, so they still need faulty samples to train the algorithm to find the differences between the samples in the healthy state and the faulty state. The scheme provided by the present invention uses KL divergence to compare the feature consistency between the current global health center and the training sample data, and compares the multi-sensor signal correlation of the global health center and the multi-sensor signal correlation of the current training sample data to measure the multi-sensor mode consistency, and then directly models the working state of the shock absorber through consistency, so that better detection results can be achieved without the participation of faulty samples in training.

[0115] In order to illustrate the performance of the above scheme of the present invention, an application experiment was conducted on a real shock absorber data set. The shock absorber data was collected from the solder paste printer in the surface mount production line. The health status data and fault data were collected for about one day respectively, with a sampling frequency of about 50HZ. The final fault detection accuracy was 0.992, the F1 score was 0.994, the recall rate was 0.997, and the precision rate was 0.990. Figure 4 Shown are the experimental results, the lower left corner corresponds to the fault state, and the upper right corner corresponds to the healthy state.

[0116] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by means of software plus necessary general hardware platforms. Based on such understanding, the technical solutions of the above embodiments can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0117] Embodiment 2

[0118] The present invention also provides a hydraulic shock absorber unsupervised fault detection system, which is mainly used to implement the method provided in the above embodiment, such as Figure 5 As shown, the system mainly includes:

[0119] A training sample construction unit is used to collect three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it into training sample data;

[0120] A model building unit, used to build an unsupervised fault detection model for hydraulic shock absorbers, including: a health center iterative update module, a sample consistency check module, and a multi-sensor mode consistency check module;

[0121] A model training unit, used for training the hydraulic shock absorber unsupervised fault detection model using the training sample data, comprising: the health center iterative update module using the training sample data to update the current global health center; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain a sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain a multi-sensor mode consistency check result; the hydraulic shock absorber unsupervised fault detection model is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result; after the training is completed, the final global health center is obtained and the health center iterative update module is removed;

[0122] The fault detection unit is used to input the three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be detected into the trained unsupervised fault detection model of the hydraulic shock absorber, and obtain the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center, so as to judge the working state of the hydraulic shock absorber.

[0123] Considering that the main technical details involved in the system have been introduced in detail in the previous embodiments, they will not be repeated here.

[0124] Technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above.

[0125] Embodiment 3

[0126] The present invention also provides a processing device, such as Figure 6As shown, it mainly includes: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the methods provided in the aforementioned embodiments.

[0127] Furthermore, the processing device also includes at least one input device and at least one output device; in the processing device, the processor, memory, input device, and output device are connected via a bus.

[0128] In the embodiment of the present invention, the specific types of the memory, input device and output device are not limited; for example:

[0129] The input device may be a touch screen, an image acquisition device, a physical button or a mouse, etc.;

[0130] The output device may be a display terminal;

[0131] The memory may be a random access memory (RAM) or a non-volatile memory such as a disk storage.

[0132] Embodiment 4

[0133] The present invention also provides a readable storage medium storing a computer program, which implements the method provided in the above embodiment when the computer program is executed by a processor.

[0134] In the embodiment of the present invention, the readable storage medium as a computer-readable storage medium can be set in the aforementioned processing device, for example, as a memory in the processing device. In addition, the readable storage medium can also be a U disk, a mobile hard disk, a read-only memory (ROM), a disk or an optical disk, etc., which can store program codes.

[0135] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for unsupervised fault detection of a hydraulic shock absorber, characterized in that: include: Collect the three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it as training sample data; Construct an unsupervised fault detection model for hydraulic shock absorbers, including: a health center iterative update module, a sample consistency check module, and a multi-sensor mode consistency check module; The unsupervised fault detection model of the hydraulic shock absorber is trained by using the training sample data, including: the health center iterative update module updates the current global health center by using the training sample data; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain the sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain the multi-sensor mode consistency check result; the unsupervised fault detection model of the hydraulic shock absorber is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result; after the training is completed, the final global health center is obtained and the health center iterative update module is removed; The three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be tested is input into the trained unsupervised fault detection model of the hydraulic shock absorber to obtain the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center, so as to judge the working status of the hydraulic shock absorber.

2. The unsupervised fault detection method for a hydraulic shock absorber according to claim 1, characterized in that: The collecting of the three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and constructing it as training sample data includes: Collect 3D acceleration data from sensors in a healthy state of hydraulic shock absorbers , where each Represents acceleration data in one direction, x, y, z represent three directions in the three-dimensional rectangular coordinate system; According to the set sampling width, the three-dimensional acceleration data of the sensor in the healthy state of the hydraulic shock absorber is divided into samples , where each s represents a sample, the subscript is the sample number, N is the number of samples, and each sample is a three-dimensional acceleration data; The sample Transformed into a time-frequency matrix , each TF is the time-frequency matrix obtained by transforming the corresponding sample s, and the subscripts of the time-frequency matrix TF correspond one-to-one to the subscripts of the sample s; All time-frequency matrices obtained by transformation constitute the training sample data.

3. A method for unsupervised fault detection of a hydraulic shock absorber according to claim 1 or 2, characterized in that: The updating of the current global health center by using the training sample data comprises: The number of training sample data in a batch in the current training is recorded as B, and the local health center of the training sample data in the batch is calculated by the following formula , expressed as: ； ； ； in, are the three acceleration time-frequency matrices of the local health center, and x, y, and z represent the three directions in the three-dimensional rectangular coordinate system; is the jth training sample data in the batch, which is the acceleration time-frequency matrix in the x, y, and z directions; Take advantage of local health centers Update the current global health center , expressed as: ； ； ； in, is the global health center before the update. If this is the first training, then is a randomly initialized health center. If this is not the first training, then Global Health Center updated for the last training; are the three acceleration time-frequency matrices in the global health center before updating, are the three acceleration time-frequency matrices in the current global health center.

4. A method for unsupervised fault detection of a hydraulic shock absorber according to claim 1 or 2, characterized in that: The sample consistency check module is provided with a first set of encoders , based on the first group of encoders, the time-frequency characteristics of the current global health center and the time-frequency characteristics of the training sample data are obtained; Among them, the current global health center is recorded as , are the three acceleration time-frequency matrices in the current global health center, x, y, and z represent the three directions in the three-dimensional rectangular coordinate system; the jth training sample data is recorded as , which is the acceleration time-frequency matrix in the x, y, and z directions; Based on the first set of encoders get The time-frequency characteristics of , and based on the first set of encoders get The time-frequency characteristics of , expressed as: ； ； ； Among them, the first group of encoders includes three encoders, whose subscripts correspond to the three directions of x, y, and z, and ReLu is a rectified linear unit.

5. The unsupervised fault detection method for a hydraulic shock absorber according to claim 4, characterized in that: The sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data, and obtains the sample consistency check result including: The KL divergence is used to measure the distance between the time-frequency characteristics of the current global health center and the time-frequency characteristics of the training sample data in the same direction and accumulate them to obtain the sample consistency check result, which is expressed as: ； in, is the sample consistency check result, is the index symbol of the direction; Stop is the stopping mechanism, that is, the gradient inside the bracket after Stop is not calculated, KL divergence is the relative entropy; softmax is the normalized exponential function; B is the number of training sample data in a batch in the current training.

6. The unsupervised fault detection method for a hydraulic shock absorber according to claim 4, characterized in that: The multi-sensor mode consistency check module calculates the internal correlation between the features of the current global health center and the internal correlation between the features of the training sample data, and measures the distance between the internal correlation between the features of the current global health center and the internal correlation between the features of the training sample data, and obtains the multi-sensor mode consistency check result including: The multi-sensor mode consistency check module is provided with a second group of encoders based on the time-frequency features of the second group of encoders. The corresponding time-frequency features are extracted and recorded as , and based on the second set of encoders from the time-frequency features The corresponding time-frequency features are extracted and recorded as ; After that, choose the Pearson correlation function To calculate the internal correlation, where: Calculate the internal correlation between the time-frequency characteristics of the current global health center , expressed as: ； Calculate the internal correlation between the time-frequency features of the jth training sample data , expressed as: ； in, The corresponding representation and The correlation between and The correlation between and The correlation between The corresponding representation and The correlation between and The correlation between and The correlation between Then, the KL divergence is used to measure the distance between the internal correlation between the time-frequency characteristics of the current global health center and the internal correlation between the time-frequency characteristics of the training sample data to obtain the multi-sensor mode consistency check result, which is expressed as: ； in, is the result of the multi-sensor mode consistency check, Stop is the stop mechanism, that is, the gradient inside the bracket after Stop is not calculated, KL divergence is the relative entropy; softmax is the normalized exponential function; B is the number of training sample data in a batch in the current training.

7. The unsupervised fault detection method for a hydraulic shock absorber according to claim 4, characterized in that: The obtaining of the sample consistency check result and the multi-sensor mode consistency check result calculated by combining the final global health center to judge the working state of the hydraulic shock absorber includes: The sample consistency check result calculated in combination with the final global health center is recorded as , the consistency check result of multi-sensor mode is recorded as ; Calculating Fault Score , expressed as: ； in, and is a hyperparameter; Determine the working status of the hydraulic shock absorber according to the following formula: ； in, is the monitoring threshold set. is the judgment result, that is: When the hydraulic shock absorber is in the healthy state Health, when When , the hydraulic shock absorber is in fault state Fault.

8. A hydraulic shock absorber unsupervised fault detection system, characterized in that: include: A training sample construction unit is used to collect three-dimensional acceleration data of the sensor of the hydraulic shock absorber in a healthy state and construct it into training sample data; A model building unit, used to build an unsupervised fault detection model for hydraulic shock absorbers, including: a health center iterative update module, a sample consistency check module, and a multi-sensor mode consistency check module; A model training unit, used for training the hydraulic shock absorber unsupervised fault detection model using the training sample data, comprising: the health center iterative update module using the training sample data to update the current global health center; the sample consistency check module measures the distance between the characteristics of the current global health center and the characteristics of the training sample data to obtain a sample consistency check result; the multi-sensor mode consistency check module calculates the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data, and measures the distance between the internal correlation between the internal correlation between the characteristics of the current global health center and the internal correlation between the characteristics of the training sample data to obtain a multi-sensor mode consistency check result; the hydraulic shock absorber unsupervised fault detection model is trained in combination with the sample consistency check result and the multi-sensor mode consistency check result; after the training is completed, the final global health center is obtained and the health center iterative update module is removed; The fault detection unit is used to input the three-dimensional acceleration data of the sensor of the hydraulic shock absorber to be detected into the trained unsupervised fault detection model of the hydraulic shock absorber, and obtain the sample consistency check result and the multi-sensor mode consistency check result calculated in combination with the final global health center, so as to judge the working state of the hydraulic shock absorber.

9. A processing device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

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

Images