An automatic testing method and system for air suspension

Abstract

The application belongs to the technical field of testing, and discloses an automatic testing method and system for an air suspension, the method comprising: obtaining a vibration test signal of the air suspension; performing multi-modal decomposition on the vibration test signal based on an empirical mode decomposition algorithm to obtain a plurality of intrinsic modal components; performing screening processing on the plurality of intrinsic modal components to obtain a plurality of processed intrinsic modal components; constructing key features based on the plurality of processed intrinsic modal components; inputting the key features into a suspension fault test model for fault testing to obtain a test result, the test result comprising a probability of a fault occurring in the air suspension and a corresponding fault type; and wherein the suspension fault test model is constructed based on transfer learning. The application has the advantages of strong real-time performance, the ability to locate early-stage hidden faults, a model with strong generalization ability, high testing efficiency, and high accuracy of fault testing.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of testing, and particularly relates to an automatic testing method and system for an air suspension. BACKGROUND

[0002] With the rapid development of the automobile industry towards intelligence, electrification and high-end, the air suspension system has become one of the core configurations of high-end passenger cars and commercial vehicles due to its excellent dynamic adjustment capability (body height, stiffness self-adaptive adjustment). The system replaces the traditional steel spring with an air spring, and combines an electronic control unit (ECU) and a continuously adjustable damper (CDC), which significantly improves the ride comfort, handling stability and passability of the vehicle. However, the complex mechatronic structure (including air path, valve body, rubber air bag, sensor, etc.) also introduces higher fault risks, such as air spring leakage / breakage leading to loss of control of the body height, damper damping failure causing abnormal body shaking, guide mechanism bushing wear producing abnormal noise and positioning misalignment, and air path blockage or electromagnetic valve failure causing loss of adjustment function.

[0003] Such faults not only directly affect the safety and comfort of driving and riding, but also may cause a chain damage of the chassis system, greatly increasing the maintenance cost. The traditional fault diagnosis method mainly relies on manual experience troubleshooting, static air pressure detection or special bench disassembly testing, which has problems such as low efficiency, poor real-time performance and difficulty in positioning early hidden faults. Especially for slowly developing chronic faults (such as small air leakage and bushing gradual aging), the traditional method often cannot identify the faults until they are significantly deteriorated. SUMMARY

[0004] The purpose of the present application is to provide an automatic testing method and system for an air suspension, which solves the problems of low efficiency, poor real-time performance and difficulty in positioning early hidden faults in the prior art.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] In a first aspect, the present application provides an automatic testing method for an air suspension, which comprises:

[0007] obtaining a vibration test signal of the air suspension;

[0008] performing multi-modal decomposition on the vibration test signal based on an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components;

[0009] performing screening processing on the plurality of intrinsic mode components to obtain a plurality of processed intrinsic mode components;

[0010] constructing a key feature based on the plurality of processed intrinsic mode components;

[0011] inputting the key features into a suspension fault test model for fault test to obtain a test result, the test result including a probability of air suspension failure and a corresponding fault type; wherein the suspension fault test model is constructed based on transfer learning.

[0012] Preferably, the method further comprises: constructing a suspension fault test model, comprising:

[0013] obtaining a first sample data set and a second sample data set, the first sample data set containing a plurality of known-labeled suspension vibration signals, and the second sample data set being a suspension vibration signal to be tested, the suspension vibration signal to be tested including a normal vibration signal or a vibration signal of an unknown fault type;

[0014] training the deep learning model using the first sample data set to obtain a trained deep learning model;

[0015] based on a transfer learning algorithm, transferring model parameters of the trained deep learning model to a learning model to be trained to obtain a transferred model;

[0016] updating the model parameters of the transferred model based on the second sample data set to obtain a suspension fault test model.

[0017] Preferably, the deep learning model is a convolutional neural network or a recurrent neural network.

[0018] Preferably, before the vibration test signal is decomposed into multiple modes based on the empirical mode decomposition algorithm, the method further comprises: dividing the vibration test signal into a plurality of segment signals; after the vibration test signal is decomposed into multiple modes based on the empirical mode decomposition algorithm, each segment signal corresponds to at least one intrinsic mode component.

[0019] Preferably, the vibration test signal is decomposed into multiple modes based on the empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components, comprising:

[0020] For any one segment signal, the segment signal is expanded into two groups to obtain a first segment test signal and a second segment test signal;

[0021] adding positive random white noise to the first segment test signal to obtain a first signal;

[0022] adding negative random white noise to the second segment test signal to obtain a second signal;

[0023] decomposing the first signal and the second signal based on the empirical mode decomposition algorithm to obtain a plurality of first components and a plurality of second components;

[0024] averaging the kth first component and the kth second component to obtain a kth intrinsic modal component of the segment signal, wherein k=1, 2, …, K, and K is a total order of the empirical mode decomposition algorithm.

[0025] Preferably, the plurality of intrinsic modal components are screened to obtain a plurality of processed intrinsic modal components, including:

[0026] determining a screening coefficient between each segment signal and each intrinsic modal component corresponding to the segment signal;

[0027] screening each intrinsic modal component corresponding to the segment signal based on the screening coefficient to filter out unnecessary intrinsic modal components, to obtain necessary intrinsic modal components corresponding to each segment signal, and taking the necessary intrinsic modal components as the processed intrinsic modal components.

[0028] Preferably, the determination of the screening coefficient between each segment signal and each intrinsic modal component corresponding to the segment signal includes:

[0029] extracting a first feature corresponding to each segment signal at each sampling time in a sampling period and a second feature of each intrinsic modal component corresponding to each segment signal;

[0030] based on the first feature value and the second feature value, using a similarity algorithm to calculate a similarity distance between the first feature and the second feature at each sampling time;

[0031] based on the number of sampling times in the sampling period and the similarity distance between the first feature and the second feature at each sampling time, determining a distance sum value corresponding to each intrinsic modal component in the sampling period;

[0032] based on the distance sum value, determining the screening coefficient.

[0033] Preferably, the similarity algorithm is a cosine similarity algorithm.

[0034] In a second aspect, the present application provides an automatic test system of an air suspension, which is used to implement the automatic test method of the air suspension, and the system includes:

[0035] a signal acquisition module, configured to acquire a vibration test signal of the air suspension;

[0036] a signal decomposition module, configured to perform multi-modal decomposition on the vibration test signal based on an empirical mode decomposition algorithm to obtain a plurality of intrinsic modal components;

[0037] a component screening module, configured to screen the plurality of intrinsic modal components to obtain a plurality of processed intrinsic modal components;

[0038] a feature construction module, configured to construct key features based on a plurality of processed intrinsic modal components;

[0039] a fault test module, configured to input the key features into a suspension fault test model for fault test, to obtain a test result, the test result including a probability of failure of the air suspension and a corresponding fault type; wherein the suspension fault test model is constructed based on transfer learning.

[0040] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the automatic test method of the air suspension when executing the computer program.

[0041] In a fourth aspect, the present application provides a computer readable storage medium, having a computer program stored thereon, wherein the program is executed by a processor to implement the automatic test method of the air suspension.

[0042] Advantages:

[0043] 1. The present application can quickly determine whether the air suspension will fail and the fault type of the failure by collecting the vibration test signal of the air suspension, constructing key features based on the vibration test signal, and inputting the key features into the suspension fault test model, and has the advantages of strong real-time performance and early positioning of hidden faults.

[0044] 2. The suspension fault test model of the present application is constructed based on transfer learning, and has high generalization ability and test efficiency.

[0045] 3. The vibration test signal of the present application is decomposed by the empirical mode decomposition algorithm, and the multiple intrinsic modal components are screened and processed to extract the intrinsic modal components useful for fault test, reduce the interference of noise, and improve the accuracy of the suspension fault test model for fault test. BRIEF DESCRIPTION OF DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the embodiments of the present application together with the following specific embodiments, but do not constitute a limitation on the embodiments of the present application. In the drawings:

[0047] Figure 1 is a flowchart of the automatic test method of the air suspension provided by an embodiment of the present application;

[0048] Figure 2 is a block diagram of the automatic test system of the air suspension provided by an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be briefly introduced below in connection with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the drawings is only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor. It should be noted that the description of these embodiments is used to help understand the present application, but does not constitute a limitation on the present application.

[0050] Embodiment one

[0051] Figure 1 A flowchart of an automatic test method of an air suspension provided by an embodiment of the present application, which can be executed by a computer device with certain computing resources, such as a cloud server, a personal computer (PC, referring to a multi-purpose computer suitable for personal use in size, price and performance; desktop computers, notebook computers to small notebook computers and tablet computers, and ultrabooks, etc.), a smart phone, a personal digital assistant (PDA), or a wearable device, etc. As shown in Figure 1 The embodiment provides an automatic test method of an air suspension, which comprises the following steps:

[0052] Step S10: obtaining a vibration test signal of the air suspension.

[0053] In the embodiment, any component (air spring, shock absorber, guide mechanism bushing, connecting piece, air path, compressor, etc.) of the air suspension has a fault (such as air leakage, damping failure, wear, looseness, blockage, rupture), which will change the mass, stiffness, damping characteristics of the entire suspension system. The change of these characteristics will directly affect the vibration transmission characteristics and natural frequency of the system. The vibration response (amplitude, frequency component, energy distribution, time domain characteristics) under the fault state will inevitably have detectable differences compared with the healthy state.

[0054] The embodiment adopts an accelerometer to collect the vibration test signal of the air suspension, and the installation position of the accelerometer is as follows: 1, non-sprung mass (close to the wheel center / control arm), 2, sprung mass (close to the suspension mounting point on the vehicle body side), 3, air spring body / shock absorber cylinder, and 4, guide mechanism (such as control arm).

[0055] The embodiment can be tested by road test, bench test, and pulse excitation / force hammer knocking, etc. The road test is to drive on different road profiles (such as rough road, Belgian road, and deceleration zone) in a real vehicle or test field to collect vibration test signals, which is closest to the actual working condition. The bench test is to simulate various road surface spectrum inputs on a four-column vibration table or a multi-axis excitation table to accurately control the excitation conditions and has good repeatability. The pulse excitation / force hammer knocking is to obtain the frequency response function (FRF) of the system and identify the changes of the natural frequency and damping ratio.

[0056] Step S20: performing multi-modal decomposition on the vibration test signal based on the empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components.

[0057] As a further optimization of the embodiment, before performing multi-modal decomposition on the vibration test signal based on the empirical mode decomposition algorithm, the method further includes: dividing the vibration test signal into a plurality of segment signals.

[0058] For step S20, the specific steps of performing multi-modal decomposition on the vibration test signal based on the empirical mode decomposition algorithm are as follows:

[0059] Step S201: for any one segment signal, the segment signal is expanded into two groups to obtain a first segment test signal and a second segment test signal; wherein, the segment signal can be copied to obtain two groups of identical segment signals as the first segment test signal and the second segment test signal, or another segment signal most similar to the segment signal can be found from all segment signals to be respectively as the first segment test signal and the second segment test signal.

[0060] Step S202: adding positive random white noise to the first segment test signal to obtain a first signal; wherein, the function expression of the first signal is:

[0061] ;

[0062] In the formula, is the first segment test signal, t is the sampling time, is the first signal, is the i-th added positive random white noise, is the noise amplitude.

[0063] Step S203: adding negative random white noise to the second segment test signal to obtain a second signal; wherein, the function expression of the second signal is:

[0064] ;

[0065] In the formula, is the second signal, Test the signal for the second segment, for the i-th added negative random white noise, for the noise amplitude.

[0066] Step S204: based on the empirical mode decomposition algorithm, the first signal and the second signal are decomposed respectively to obtain a plurality of first components and a plurality of second components.

[0067] The function relationship of the first component is:

[0068] ;

[0069] In the formula, for the i-th k-th first component, for the first residual wave after the i-th empirical mode decomposition algorithm decomposition;

[0070] The function relationship of the second component is:

[0071] ;

[0072] In the formula, for the i-th k-th second component, for the second residual wave after the i-th empirical mode decomposition algorithm decomposition.

[0073] Step S205: the k-th first component and the k-th second component are averaged to obtain the k-th intrinsic mode component of the segment signal; wherein k=1, 2, …, K, K is the total order of the empirical mode decomposition algorithm.

[0074] The function expression of the k-th intrinsic mode component is:

[0075] ;

[0076] In the formula, for the k-th intrinsic mode component, I is the total number of noise addition.

[0077] In this embodiment, by adding negative random white noise and positive random white noise to the two groups of segment signals, after the average processing of step S205, the noise in the original signal can be eliminated, and the noise interference of the later fault test can be reduced.

[0078] Step S30: a plurality of intrinsic mode components are screened to obtain a plurality of processed intrinsic mode components.

[0079] In the embodiment, after the vibration test signal is decomposed into multiple modes based on the empirical mode decomposition algorithm, each segment signal corresponds to at least one intrinsic mode component; since there are components helpful for fault test and components useless for fault test in the at least one intrinsic mode component of each segment signal, the multiple intrinsic mode components need to be screened.

[0080] Specifically, the step of screening the multiple intrinsic mode components is as follows.

[0081] Step S301: determining a screening coefficient between each segment signal and each intrinsic mode component corresponding to the segment signal.

[0082] In the embodiment, the screening coefficient between each segment signal and each intrinsic mode component corresponding to the segment signal is determined, including the following steps.

[0083] Step a1: extracting a first feature corresponding to each segment signal at each sampling time in a sampling period and a second feature of each intrinsic mode component corresponding to each segment signal; wherein the first feature and the second feature can be amplitude, frequency component, energy distribution, time domain feature, etc.

[0084] Step a2: based on the first feature value and the second feature value, using a similarity algorithm, calculating the similarity distance between the first feature and the second feature at each sampling time, the similarity algorithm in the embodiment can be a cosine similarity algorithm, and the function expression of the similarity distance between the first feature and the second feature at each sampling time is as follows:

[0085] ;

[0086] In the formula, is the similarity distance between the first feature and the second feature at the jth sampling time, is a cosine similarity function, is the nth feature value of the first feature corresponding to the jth sampling time, and N is the total number of feature values of the first feature, is the mth feature value of the second feature corresponding to the jth sampling time, and M is the total number of feature values of the second feature.

[0087] Step a3: based on the number of sampling times in the sampling period and the similarity distance between the first feature and the second feature at each sampling time, determining a distance sum value corresponding to each intrinsic mode component in the sampling period.

[0088] The function expression of the distance sum value in the embodiment is as follows:

[0089] ;

[0090] In the formula, D is a distance sum value, and J is a number of sampling time points.

[0091] Step a4: determining a screening coefficient based on the distance sum value.

[0092] The function expression of the screening coefficient of the embodiment is:

[0093] ;

[0094] In the formula, D is a distance sum value, and J is a number of sampling time points. is the screening coefficient.

[0095] Step S302: screening each eigenmodal component corresponding to the segment signal based on the screening coefficient, to screen out unnecessary eigenmodal components, to obtain necessary eigenmodal components corresponding to each segment signal, and taking the necessary eigenmodal components as processed eigenmodal components.

[0096] In the embodiment, when the screening coefficient of a certain eigenmodal component in the segment signal is lower than a preset value, the eigenmodal component is an unnecessary eigenmodal component and needs to be screened out; when the screening coefficient of a certain eigenmodal component is greater than or equal to the preset value, the eigenmodal component is a necessary eigenmodal component and needs to be retained; the processing mode can reduce the influence of noise and also reduce the complexity of vibration test signal processing, and multiple eigenmodal components have good stationarity and linear characteristics.

[0097] Step S40: constructing a key feature based on the processed eigenmodal components, and the key feature of the embodiment can be one or more of amplitude, frequency component, energy distribution, time domain feature, etc.

[0098] Step S50: inputting the key feature into a suspension fault test model for fault test to obtain a test result, and the test result includes a probability of air suspension failure and a corresponding fault type; wherein the suspension fault test model is constructed based on transfer learning.

[0099] As a further optimization of the embodiment, the method further includes: constructing a suspension fault test model, including:

[0100] Step B10: acquiring a first sample data set and a second sample data set, the first sample data set containing a plurality of known label suspension vibration signals, and the second sample data set being a suspension vibration signal to be tested, the suspension vibration signal to be tested containing a normal vibration signal or a vibration signal of an unknown fault type; therefore, taking the first sample data set as a source domain data set of transfer learning and taking the second sample data set as a target domain data set of transfer learning.

[0101] Step B20: training the deep learning model using the first sample dataset to obtain a trained deep learning model; in this embodiment, the deep learning model is a convolutional neural network or a recurrent neural network; the trained deep learning model is a source domain model;

[0102] Step B30: based on a transfer learning algorithm, migrating the model parameters of the trained deep learning model to a learning model to be trained to obtain a migrated model; the model parameters in this embodiment can be the weights and biases of the convolutional layers.

[0103] Step B40: updating the model parameters of the migrated model based on the second sample dataset to obtain a suspension fault test model; after migrating the model parameters from the source domain to the target domain, fine-tuning is performed using the target domain dataset to obtain the final suspension fault test model.

[0104] In this embodiment, the steps of training the deep learning model using the first sample dataset are as follows:

[0105] The first sample dataset is used as the input parameter of the deep learning model, forward propagation is performed, and the output of the model is calculated;

[0106] The loss function can be cross-entropy loss or mean square error loss, and the loss function is used to measure the error between the model prediction result and the actual result;

[0107] Back propagation is performed to calculate the gradient of the loss function on the model parameters, and an optimization algorithm (such as the gradient descent algorithm or the Adam algorithm) is used to update the model parameters according to the gradient;

[0108] The performance of the model is evaluated using the validation dataset, such as calculating the accuracy, recall, F1 value and other indicators, and the model structure and parameters are adjusted according to the evaluation results, such as increasing or decreasing the convolutional layers, adjusting the learning rate, etc.; to obtain a trained deep learning model.

[0109] The present application can quickly determine whether the air suspension will fail and the failure type of the failure by collecting the vibration test signal of the air suspension, constructing key features based on the vibration test signal, and inputting the key features into the suspension fault test model, which has the advantages of strong real-time and early positioning of hidden faults. And, the suspension fault test model is constructed using transfer learning, which has high generalization ability and test efficiency; secondly, after the vibration test signal is decomposed by the empirical mode decomposition algorithm, the multiple intrinsic mode components are screened and processed to extract the intrinsic mode components useful for fault test, reduce the interference of noise, and improve the accuracy of the suspension fault test model for fault test.

[0110] Embodiment two

[0111] Figure 2 is a block diagram of an automatic test system of an air suspension provided by an embodiment of the present application. As shown in Figure 2 the embodiment, the automatic test system of the air suspension is provided for implementing the automatic test method of the air suspension in Embodiment One, and the system comprises:

[0112] a signal acquisition module configured to acquire a vibration test signal of the air suspension;

[0113] a signal decomposition module configured to perform multi-modal decomposition on the vibration test signal based on an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components;

[0114] a component screening module configured to perform screening processing on the plurality of intrinsic mode components to obtain a plurality of processed intrinsic mode components;

[0115] a feature construction module configured to construct key features based on the plurality of processed intrinsic mode components;

[0116] a fault test module configured to input the key features into a suspension fault test model for fault testing to obtain a test result, the test result comprising a probability of failure of the air suspension and a corresponding fault type; wherein the suspension fault test model is constructed based on transfer learning.

[0117] The embodiment also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the automatic test method of the air suspension in Embodiment One when executing the computer program.

[0118] The embodiment also provides a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the automatic test method of the air suspension in Embodiment One.

[0119] The present application can quickly determine whether the air suspension will fail and the fault type of the failure by acquiring the vibration test signal of the air suspension, constructing key features based on the vibration test signal, and inputting the key features into a suspension fault test model, and has the advantages of strong real-time performance and the ability to locate early implicit faults. In addition, the suspension fault test model is constructed using transfer learning, which has high generalization ability and test efficiency. Furthermore, the vibration test signal is decomposed by the empirical mode decomposition algorithm, and the plurality of intrinsic mode components are screened to extract the intrinsic mode components useful for fault testing, thereby reducing the interference of noise and improving the accuracy of the suspension fault test model for fault testing.

[0120] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system, or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, and the like) embodying computer readable program code.

[0121] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions of the flow diagrams and / or block diagrams as Figure 1 one or more functions Figure 1 one or more functions

[0122] The above merely provides an example of the present application and is not intended to limit the present application. The present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the scope of claims of the present application.

Claims

1. An automatic testing method for air suspension, characterized in that, The method includes: Obtain vibration test signals for the air suspension; The vibration test signal is decomposed into multiple modes based on the empirical mode decomposition algorithm to obtain multiple intrinsic mode components; Multiple intrinsic mode components are filtered to obtain several processed intrinsic mode components; Key features are constructed based on several processed intrinsic mode components; Key features are input into the suspension failure test model for failure testing to obtain test results, which include the probability of air suspension failure and the corresponding failure type; wherein, the suspension failure test model is constructed based on transfer learning; Before performing multimodal decomposition on the vibration test signal based on the empirical mode decomposition algorithm, the method further includes: dividing the vibration test signal into multiple segment signals; after performing multimodal decomposition on the vibration test signal based on the empirical mode decomposition algorithm, each segment signal corresponds to at least one intrinsic mode component; Multimodal decomposition of vibration test signals is performed based on empirical mode decomposition algorithm to obtain multiple intrinsic mode components, including: For any segment of signal, the segment of signal is expanded into two groups to obtain the first segment test signal and the second segment test signal; Positive random white noise is added to the first segment of the test signal to obtain the first signal; Negative random white noise is added to the second segment of the test signal to obtain the second signal; The first signal and the second signal are decomposed based on the empirical mode decomposition algorithm to obtain several first components and several second components. The k-th first component and the k-th second component are averaged to obtain the k-th intrinsic mode component of the signal segment; where k = 1, 2, ..., K, and K is the total order of the empirical mode decomposition algorithm; Multiple intrinsic mode components are filtered to obtain several processed intrinsic mode components, including: Determine the screening coefficients between each signal segment and its corresponding intrinsic mode components; Based on the screening coefficient, the intrinsic mode components corresponding to each segment of the signal are screened to filter out unnecessary intrinsic mode components, and the necessary intrinsic mode components corresponding to each segment of the signal are obtained. The necessary intrinsic mode components are used as the processed intrinsic mode components. Determine the screening coefficients between each signal segment and its corresponding intrinsic mode components, including: Extract the first feature corresponding to each sampling time of each segment signal and the second feature of each intrinsic mode component corresponding to each segment signal; Based on the first and second feature values, a similarity algorithm is used to calculate the similarity distance between the first feature and the second feature at each sampling time. Based on the number of sampling moments within the sampling period and the similarity distance between the first and second features at each sampling moment, the sum of distances corresponding to each intrinsic mode component within the sampling period is determined. The selection coefficient is determined based on the sum of the distances. The functional expression for the similarity distance between the first feature and the second feature at each sampling time is: In the formula, Dis j Let V be the similarity distance between the first and second features at sampling time j, and cos be the cosine similarity function. jn V represents the nth feature value of the first feature at the j-th sampling time, where N is the total number of feature values ​​of the first feature. jm Let M be the m-th feature value of the second feature corresponding to the j-th sampling time, and M be the total number of feature values ​​of the second feature; Accordingly, the functional expression for the screening coefficient is: In the formula, δ is the screening coefficient, and D is the sum of distances.

2. The automatic testing method for air suspension according to claim 1, characterized in that, The method further includes: constructing a suspension failure test model, including: Obtain a first sample dataset and a second sample dataset. The first sample dataset contains several suspension vibration signals with known labels, and the second sample dataset contains suspension vibration signals to be tested. The suspension vibration signals to be tested include normal vibration signals or vibration signals with unknown fault types. The deep learning model is trained using the first sample dataset to obtain a trained deep learning model. Based on the transfer learning algorithm, the model parameters of the trained deep learning model are transferred to the learning model to be trained, and the transferred model is obtained. The model parameters of the migrated model are updated based on the second sample dataset to obtain the suspension failure test model.

3. The automatic testing method for air suspension according to claim 2, characterized in that, The deep learning model is a convolutional neural network or a recurrent neural network.

4. The automatic testing method for air suspension according to claim 1, characterized in that, The similarity algorithm is the cosine similarity algorithm.

5. An automatic testing system for air suspension, used to implement the automatic testing method for air suspension according to any one of claims 1-4, characterized in that, The system includes: The signal acquisition module is used to acquire vibration test signals of the air suspension. The signal decomposition module is used to perform multimodal decomposition on the vibration test signal based on the empirical mode decomposition algorithm to obtain multiple intrinsic mode components; The component filtering module is used to filter multiple intrinsic mode components to obtain several processed intrinsic mode components. The feature construction module is used to construct key features based on several processed intrinsic mode components; The fault testing module is used to input key features into the suspension fault testing model to conduct fault testing and obtain test results. The test results include the probability of air suspension failure and the corresponding fault type. The suspension fault testing model is constructed based on transfer learning.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the automatic testing method for the air suspension as described in any one of claims 1-4.

Images