Vehicle sheet metal abnormal noise detection method and system

Abstract

The present invention relates to a vehicle sheet metal abnormal noise detection method and system, belonging to the field of vehicle detection technology. The method includes: obtaining an enhanced noise signal through a signal filtering model according to a vehicle noise signal, obtaining a smoothed noise signal through a Gaussian filter according to the enhanced noise signal, determining a signal parameter according to the smoothed noise signal, obtaining a parameter initial value by initializing the signal parameter, obtaining a vehicle abnormal noise signal through abnormal noise detection according to the smoothed noise signal combined with the parameter initial value, obtaining a vehicle abnormal noise signal through feature extraction of the vehicle abnormal noise signal, obtaining a vehicle abnormal noise signal feature through a feature selection model according to the vehicle abnormal noise signal feature, obtaining an abnormal noise signal feature value according to the abnormal noise signal feature value, obtaining a feature value set according to the abnormal noise signal feature value, obtaining an abnormal noise category according to the abnormal noise category through a normalized linear classification model, and uploading the abnormal noise category to a terminal, thereby realizing the category judgment of the vehicle sheet metal abnormal noise and improving the accuracy and efficiency of sheet metal abnormal noise detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle detection, and in particular relates to a method and system for detecting abnormal noise of vehicle sheet metal. Background Art

[0002] In recent years, with the continuous development of the domestic automobile industry, the product categories have been continuously expanded and matured, and consumers' requirements for automobile quality have gradually increased. Various quality issues of automobiles have received increasing attention. There are many reasons and sources for abnormal noises in automobiles. Abnormal noises are mostly transient noises, which refer to abnormal sounds generated by the relative movement of adjacent parts of the automobile during driving. Among them, abnormal noises in automobile sheet metal are one of the common problems in automobile manufacturing and use. However, the source of abnormal noises in automobile sheet metal is often obscured by interior and exterior trims, which makes it difficult to find and repair the source of abnormal noises. In addition, abnormal noises in sheet metal not only affect the driver's information and driving experience, but may also have potential impacts on the safety and reliability of the vehicle.

[0003] At present, the traditional method of detecting abnormal noise in sheet metal mainly relies on manual experience to make preliminary judgments through listening, watching, touching, etc., and combined with road tests and other methods for verification. This method has low detection efficiency and is easily affected by the subjective judgment of the inspectors, resulting in inaccurate judgment of the location and cause of the abnormal noise. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides a vehicle sheet metal abnormal noise detection method and system, which collects vehicle noise signals, processes the vehicle noise signals through a signal filtering model to obtain an enhanced noise signal, processes the enhanced noise signals through a Gaussian filter to obtain a smoothed noise signal, determines signal parameters according to the smoothed noise signals, initializes the signal parameters to obtain initial parameter values, obtains vehicle abnormal noise signals through abnormal noise detection according to the smoothed noise signals combined with the initial parameter values, extracts features of the vehicle abnormal noise signals to obtain vehicle abnormal noise signal features, processes the vehicle abnormal noise signal features through a feature selection model to obtain abnormal noise signal feature values, obtains feature value sets according to the abnormal noise signal feature values, processes the feature value sets through a normalized linear classification model to obtain abnormal noise categories, and uploads the abnormal noise categories to a terminal, thereby realizing the detection of vehicle sheet metal abnormal noises and the judgment of abnormal noise categories, avoiding the influence of subjective factors in manual detection, improving the accuracy and efficiency of sheet metal abnormal noise detection, and reducing human labor costs.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A vehicle sheet metal abnormal noise detection method comprises the following steps:

[0007] S1: Collect vehicle noise signals;

[0008] S2: According to the vehicle noise signal, an enhanced noise signal is obtained by processing the signal filtering model, according to the enhanced noise signal, a smoothed noise signal is obtained by processing the enhanced noise signal by a Gaussian filter, a signal parameter is determined according to the smoothed noise signal, an initial value of the parameter is obtained by initializing the signal parameter, and a vehicle abnormal noise signal is obtained by abnormal noise detection according to the smoothed noise signal and the initial value of the parameter, wherein the signal parameters include a noise signal quantization vector, an abnormal noise signal quantization vector, a first sliding window and an initialization threshold;

[0009] S3: extracting features of the vehicle abnormal sound signal to obtain vehicle abnormal sound signal features, processing the vehicle abnormal sound signal features through a feature selection model to obtain abnormal sound signal feature values, obtaining feature value sets based on the abnormal sound signal feature values, processing the feature value sets through a normalized linear classification model to obtain abnormal sound categories, and uploading the abnormal sound categories to the terminal, the abnormal sound categories include vibration abnormal sound and knocking abnormal sound.

[0010] Preferably, the step S2 specifically includes the following steps:

[0011] S201: Setting a second sliding window, and obtaining the enhanced noise signal by processing the sliding window in combination with the vehicle noise signal through the signal filtering model, wherein the signal filtering model is expressed as: Wherein, f(n) represents the vehicle noise signal, g represents the second sliding window, n represents the signal length, PO(f(n)) represents the positive impact enhancement signal, ⊙ represents the corrosion operator, ⊕ represents the expansion operator, NE(f(n)) represents the negative impact enhancement signal, and EN(f(n)) represents the enhanced noise signal;

[0012] S202: Acquire the sampling frequency of the vehicle noise signal, set the length of the first sliding window to 0.05 times the sampling frequency, and calculate the vehicle noise signal according to the formula Calculate and obtain a signal mean, wherein Thr(1) represents the signal mean, N represents the signal length, and f(i) represents the vehicle noise signal, and set the signal mean as the initialization threshold;

[0013] S203: A difference operator is calculated according to the noise signal quantization vector and the abnormal sound signal quantization vector by using the formula △T=||D1-D2||, where △T represents the difference operator, D1 represents the noise signal quantization vector, D2 represents the abnormal sound signal quantization vector, and ||·| represents a modulo operation; a noise vector is obtained by feature extraction according to the smooth noise signal; an initial threshold value of the abnormal sound signal is set according to the noise vector in combination with the difference operator; and abnormal sound detection is performed on the smooth noise signal according to the initial threshold value of the abnormal sound signal in combination with the first sliding window to obtain the vehicle abnormal sound signal.

[0014] Preferably, the step S203 specifically includes the following steps:

[0015] The initial threshold of the abnormal sound signal is expressed as: TR(i)=D(i)+β△T, wherein TR(i) represents the initial threshold of the abnormal sound signal, D(i) represents the noise vector, β represents the adjustment factor, and △T represents the difference operator;

[0016] The smooth noise signal is windowed according to the first sliding window to obtain a signal within the window. When the signal within the window is greater than the initial threshold of the abnormal sound signal, the signal within the window is an abnormal sound signal segment. The signal peak of the abnormal sound signal segment is determined. The noise signal quantization vector and the abnormal sound signal quantization vector are updated according to the signal peak. The calculation formula is expressed as follows: Wherein, i represents the current signal length, D2(i) represents the abnormal sound signal quantization vector, D1(i) represents the noise signal quantization vector, α represents the signal peak value, ω represents the length of the first sliding window, D(i) represents the noise vector, and f(k) represents the signal in the window;

[0017] The difference operator is updated according to the updated noise signal quantization vector and the abnormal sound signal quantization vector, and step S203 is repeated. When the current signal length is greater than the signal length, the vehicle abnormal sound signal is obtained by integration according to the abnormal sound signal segments.

[0018] Preferably, the step S3 specifically includes the following steps:

[0019] S301: calling a standard abnormal sound signal in a sheet metal abnormal sound database, constructing the feature selection model through machine learning according to the standard abnormal sound signal, constructing a vehicle abnormal sound signal feature matrix according to the vehicle abnormal sound signal feature, and obtaining a matrix eigenvalue and a feature contribution rate according to the vehicle abnormal sound signal feature matrix through the feature selection model;

[0020] S302: Arrange the feature contribution rates from large to small. When the feature contribution rate is greater than 80%, the matrix eigenvalues ​​corresponding to the feature contribution rates are the vehicle abnormal sound signal eigenvalues. The vehicle abnormal sound signal eigenvalues ​​are integrated to obtain the eigenvalue set.

[0021] S303: Obtain a normalized information distance by processing the feature value set through the normalized linear classification model, and obtain the abnormal sound category according to the normalized information distance.

[0022] Preferably, the step S301 specifically includes the following steps:

[0023] According to the vehicle abnormal sound signal characteristic matrix, the formula The matrix eigenvalues ​​are calculated, where G i represents the eigenvalue of the matrix i, H represents the characteristic matrix of the vehicle abnormal sound signal, K represents the number of the vehicle abnormal sound signal features, Y i represents the i-th vehicle abnormal sound signal feature, θ(·) represents the kernel function, * represents the convolution symbol, and the matrix eigenvalue is calculated by the formula Calculate the eigenvector, where i = 1, 2, 3..., represents the i-th eigenvector, δ represents the coefficient factor;

[0024] According to the matrix eigenvalue and the eigenvector by formula The characteristic contribution rate is calculated, wherein con represents the characteristic contribution rate, L represents the total number of eigenvalues ​​of the matrix, and * represents a multiplication operator.

[0025] Preferably, the step S303 specifically includes the following steps:

[0026] The normalized linear classification model is expressed as: Where, d represents the normalized information distance, r h. represents a constant related to the classification standard, h is a positive integer, q represents a weight vector corresponding to the eigenvalue set, J represents a preset parameter, and v represents the eigenvalue set;

[0027] By formula Calculate the adjacent normalized information distance to obtain the adjacent information distance, Wi represents the adjacent information distance, d i represents the normalized information distance of the ith k represents the kth normalized information distance, i and k represent two adjacent positive integers, and D(·) is a preset calculation function;

[0028] When 0 < Wi < 4, output the abnormal sound category as the vibration abnormal sound; when Wi ≥ 4, output the abnormal sound category as the knocking abnormal sound.

[0029] A vehicle sheet metal abnormal sound detection system, comprising:

[0030] A signal acquisition module, configured to acquire vehicle noise signals;

[0031] An abnormal sound detection module, configured to process the vehicle noise signal through a signal filtering model to obtain an enhanced noise signal, process the enhanced noise signal through a Gaussian filter to obtain a smoothed noise signal, determine signal parameters according to the smoothed noise signal, obtain an initial parameter value through initializing the setting of the signal parameters, and obtain a vehicle abnormal sound signal through abnormal sound detection according to the smoothed noise signal in combination with the initial parameter value. The signal parameters include a noise signal quantization vector, an abnormal sound signal quantization vector, a first sliding window, and an initialization threshold;

[0032] An abnormal sound detection module, configured to extract features from the vehicle abnormal sound signal to obtain vehicle abnormal sound signal features, process the vehicle abnormal sound signal features through a feature selection model to obtain abnormal sound signal feature values, obtain a feature value set according to the abnormal sound signal feature values, process the feature value set through a normalized linear classification model to obtain an abnormal sound category, and upload the abnormal sound category to a terminal. The abnormal sound categories include friction abnormal sound, vibration abnormal sound, and knocking abnormal sound.

[0033] The beneficial effects of the present invention are as follows:

[0034] 1. Process the vehicle noise signal through a signal filtering model to obtain an enhanced noise signal, process the enhanced noise signal through a Gaussian filter to obtain a smoothed noise signal, determine signal parameters according to the smoothed noise signal, obtain an initial parameter value through initializing the setting of the signal parameters, and obtain a vehicle abnormal sound signal through abnormal sound detection according to the smoothed noise signal in combination with the initial parameter value. Since the vehicle noise composition is complex, and the vehicle abnormal sound noise is transient noise, with strong randomness and poor repeatability, the noise of the whole vehicle is collected, and the vehicle noise signal is smoothed and filtered through the signal filtering model and the Gaussian filter, effectively reducing the measurement error caused by external noise or the acquisition system, improving the quality of the vehicle noise signal, and helping to improve the subsequent signal processing accuracy. Obtain a vehicle abnormal sound signal through abnormal sound detection according to the smoothed noise signal in combination with the initial parameter value, realizing the separation of the vehicle abnormal sound signal;

[0035] 2. The vehicle abnormal noise signal is characterized by extracting the features of the vehicle abnormal noise signal, and the abnormal noise signal characteristic value is obtained by processing the vehicle abnormal noise signal through a feature selection model, and a characteristic value set is obtained according to the abnormal noise signal characteristic value, and the abnormal noise category is obtained by processing the characteristic value set through a normalized linear classification model, and the abnormal noise category is uploaded to the terminal. The vehicle abnormal noise signal separated from the vehicle noise signal is characterized by extracting the features of the vehicle abnormal noise signal in order to extract significant features from the vehicle abnormal noise signal, reduce the complexity and computational complexity of the model, and facilitate the identification of the type of abnormal noise. The category judgment of the vehicle sheet metal abnormal noise is realized by processing the characteristic value set through a normalized linear classification model, thereby improving the detection accuracy and efficiency of the vehicle sheet metal abnormal noise and reducing the labor cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0037] Figure 1 It is a schematic flow chart of the vehicle sheet metal abnormal noise detection method of the present invention. DETAILED DESCRIPTION

[0038] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0039] See also Figure 1 A vehicle sheet metal noise detection method comprises the following steps:

[0040] S1: Collect vehicle noise signals;

[0041] S2: According to the vehicle noise signal, an enhanced noise signal is obtained by processing the signal filtering model, according to the enhanced noise signal, a smoothed noise signal is obtained by processing the enhanced noise signal by a Gaussian filter, a signal parameter is determined according to the smoothed noise signal, an initial value of the parameter is obtained by initializing the signal parameter, and a vehicle abnormal noise signal is obtained by abnormal noise detection according to the smoothed noise signal and the initial value of the parameter, wherein the signal parameters include a noise signal quantization vector, an abnormal noise signal quantization vector, a first sliding window and an initialization threshold;

[0042] S3: extracting features of the vehicle abnormal sound signal to obtain vehicle abnormal sound signal features, processing the vehicle abnormal sound signal features through a feature selection model to obtain abnormal sound signal feature values, obtaining feature value sets based on the abnormal sound signal feature values, processing the feature value sets through a normalized linear classification model to obtain abnormal sound categories, and uploading the abnormal sound categories to the terminal, the abnormal sound categories include vibration abnormal sound and knocking abnormal sound.

[0043] Specifically, the step S2 comprises the following steps:

[0044] S201: Setting a second sliding window, and obtaining the enhanced noise signal by processing the sliding window in combination with the vehicle noise signal through the signal filtering model, wherein the signal filtering model is expressed as: Wherein, f(n) represents the vehicle noise signal, g represents the second sliding window, n represents the signal length, PO(f(n)) represents the positive impact enhancement signal, ⊙ represents the corrosion operator, represents the expansion operator, NE(f(n)) represents the negative impact enhancement signal, and EN(f(n)) represents the enhanced noise signal;

[0045] S202: Acquire the sampling frequency of the vehicle noise signal, set the length of the first sliding window to 0.05 times the sampling frequency, and calculate the vehicle noise signal according to the formula Calculate and obtain a signal mean, wherein Thr(1) represents the signal mean, N represents the signal length, and f(i) represents the vehicle noise signal, and set the signal mean as the initialization threshold;

[0046] S203: A difference operator is calculated according to the noise signal quantization vector and the abnormal sound signal quantization vector by using the formula △T=||D1-D2||, where △T represents the difference operator, D1 represents the noise signal quantization vector, D2 represents the abnormal sound signal quantization vector, and ||·|| represents a modulo operation; a noise vector is obtained by feature extraction according to the smooth noise signal; an initial threshold value of the abnormal sound signal is set according to the noise vector in combination with the difference operator; and abnormal sound detection is performed on the smooth noise signal according to the initial threshold value of the abnormal sound signal in combination with the first sliding window to obtain the vehicle abnormal sound signal.

[0047] Specifically, the step S203 includes the following steps:

[0048] The initial threshold of the abnormal sound signal is expressed as: TR(i)=D(i)+β△T, wherein TR(i) represents the initial threshold of the abnormal sound signal, D(i) represents the noise vector, β represents the adjustment factor, and △T represents the difference operator;

[0049] The smooth noise signal is windowed according to the first sliding window to obtain a signal within the window. When the signal within the window is greater than the initial threshold of the abnormal sound signal, the signal within the window is an abnormal sound signal segment. The signal peak of the abnormal sound signal segment is determined. The noise signal quantization vector and the abnormal sound signal quantization vector are updated according to the signal peak. The calculation formula is expressed as follows: Wherein, i represents the current signal length, D2(i) represents the abnormal sound signal quantization vector, D1(i) represents the noise signal quantization vector, α represents the signal peak value, ω represents the length of the first sliding window, D(i) represents the noise vector, and f(k) represents the signal in the window;

[0050] The difference operator is updated according to the updated noise signal quantization vector and the abnormal sound signal quantization vector, and step S203 is repeated. When the current signal length is greater than the signal length, the vehicle abnormal sound signal is obtained by integration according to the abnormal sound signal segments.

[0051] It should be further explained that the purpose of filtering and smoothing the vehicle noise signal is to reduce the measurement error caused by external noise or the acquisition system, improve the quality of the vehicle noise signal, and help improve the accuracy of subsequent signal processing.

[0052] Specifically, step S3 includes the following steps:

[0053] S301: calling a standard abnormal sound signal in a sheet metal abnormal sound database, constructing the feature selection model through machine learning according to the standard abnormal sound signal, constructing a vehicle abnormal sound signal feature matrix according to the vehicle abnormal sound signal feature, and obtaining a matrix eigenvalue and a feature contribution rate according to the vehicle abnormal sound signal feature matrix through the feature selection model;

[0054] S302: Arrange the feature contribution rates from large to small. When the feature contribution rate is greater than 80%, the matrix eigenvalues ​​corresponding to the feature contribution rates are the vehicle abnormal sound signal eigenvalues. The vehicle abnormal sound signal eigenvalues ​​are integrated to obtain the eigenvalue set.

[0055] S303: Obtain a normalized information distance by processing the feature value set through the normalized linear classification model, and obtain the abnormal sound category according to the normalized information distance.

[0056] Specifically, the step S301 includes the following steps:

[0057] According to the vehicle abnormal sound signal characteristic matrix, the formula The matrix eigenvalues ​​are calculated, where G i represents the eigenvalue of the matrix i, H represents the characteristic matrix of the vehicle abnormal sound signal, K represents the number of the vehicle abnormal sound signal features, Y i represents the i-th vehicle abnormal sound signal feature, θ(·) represents the kernel function, * represents the convolution symbol, and the matrix eigenvalue is calculated by the formula Calculate the eigenvector, where i = 1, 2, 3..., denote the i-th of the feature vectors, and δ denote the coefficient factor;

[0058] According to the matrix eigenvalue and the feature vector, through the formula calculate to obtain the feature contribution rate, where con represents the feature contribution rate, L represents the total number of the matrix eigenvalues, and * represents the multiplication operator.

[0059] Specifically, the step S303 specifically includes the following steps:

[0060] The normalized linear classification model is expressed as: where d represents the normalized information distance, r h. represents a constant related to the classification criterion, h is a positive integer, q represents the weight vector corresponding to the eigenvalue set, J represents a preset parameter, and v represents the eigenvalue set;

[0061] Through the formula calculate the adjacent normalized information distances to obtain the adjacent information distances, Wi represents the adjacent information distances, d i represents the i-th of the normalized information distances, d k represents the k-th of the normalized information distances, i and k represent two adjacent positive integers, and D(·) is a pre-designed calculation function;

[0062] When 0 < Wi < 4, output the abnormal sound category as the vibration abnormal sound, and when Wi ≥ 4, output the abnormal sound category as the knocking abnormal sound.

[0063] It should be further noted that for the vehicle abnormal sound signal separated from the vehicle noise signal, by performing feature extraction on the vehicle abnormal sound signal, the purpose is to extract significant features from the vehicle abnormal sound signal, reduce the complexity and computational amount of the model, facilitate the identification of the type of abnormal sound, improve the detection accuracy and detection efficiency of the vehicle sheet metal abnormal sound, and reduce the labor cost.

[0064] Furthermore, the present application also provides a vehicle sheet metal abnormal sound detection system, including:

[0065] A signal acquisition module, configured to acquire vehicle noise signals;

[0066] an abnormal noise detection module, configured to obtain an enhanced noise signal by processing the vehicle noise signal through a signal filtering model, obtain a smoothed noise signal by processing the enhanced noise signal through a Gaussian filter, determine a signal parameter according to the smoothed noise signal, obtain an initial value of the parameter by initializing the signal parameter, obtain a vehicle abnormal noise signal by performing abnormal noise detection according to the smoothed noise signal combined with the initial value of the parameter, wherein the signal parameters include a noise signal quantization vector, an abnormal noise signal quantization vector, a first sliding window, and an initialization threshold;

[0067] The abnormal noise detection module is used to obtain vehicle abnormal noise signal characteristics by extracting characteristics of the vehicle abnormal noise signal, obtain abnormal noise signal characteristic values ​​through feature selection model processing according to the vehicle abnormal noise signal characteristics, obtain characteristic value sets according to the abnormal noise signal characteristic values, obtain abnormal noise categories through normalized linear classification model processing according to the characteristic value sets, and upload the abnormal noise categories to the terminal, wherein the abnormal noise categories include friction abnormal noise, vibration abnormal noise and knocking abnormal noise.

[0068] The working principle and use process of the present invention:

[0069] The vehicle noise signal is collected through the signal acquisition module, and the vehicle noise signal is processed through a signal filtering model to obtain an enhanced noise signal, and the enhanced noise signal is processed through a Gaussian filter to obtain a smoothed noise signal, and the signal parameters are determined according to the smoothed noise signal, and the initial values ​​of the parameters are obtained by initializing the signal parameters, and the vehicle abnormal noise signal is obtained by abnormal noise detection according to the smoothed noise signal and the initial values ​​of the parameters, and the vehicle abnormal noise signal is obtained by extracting the features of the vehicle abnormal noise signal, and the abnormal noise signal feature is obtained by processing the abnormal noise signal feature through a feature selection model according to the vehicle abnormal noise signal feature, and the abnormal noise signal feature value is obtained according to the abnormal noise signal feature value. The feature value set is obtained according to the abnormal noise signal feature value, and the abnormal noise category is obtained by processing the abnormal noise category through a normalized linear classification model according to the feature value set, and the abnormal noise category is uploaded to the terminal.

[0070] The program code included in the method in the embodiment of the present invention can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, etc., or any suitable combination mentioned above. The computer program code for performing the operation of the present invention can be written with one or more programming languages ​​or their combination, and the programming language includes object-oriented programming languages--such as Java, Smalltalk, C++, and also includes conventional procedural programming languages--such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on the remote computer, or executed completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0071] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A vehicle sheet metal noise detection method, characterized in that: The following steps are involved: S1: Collect vehicle noise signals; S2: According to the vehicle noise signal, an enhanced noise signal is obtained by processing the signal filtering model, according to the enhanced noise signal, a smoothed noise signal is obtained by processing the enhanced noise signal by a Gaussian filter, a signal parameter is determined according to the smoothed noise signal, an initial value of the parameter is obtained by initializing the signal parameter, and a vehicle abnormal noise signal is obtained by abnormal noise detection according to the smoothed noise signal and the initial value of the parameter, wherein the signal parameters include a noise signal quantization vector, an abnormal noise signal quantization vector, a first sliding window and an initialization threshold; S3: extracting features of the vehicle abnormal sound signal to obtain vehicle abnormal sound signal features, processing the vehicle abnormal sound signal features through a feature selection model to obtain abnormal sound signal feature values, obtaining feature value sets based on the abnormal sound signal feature values, processing the feature value sets through a normalized linear classification model to obtain abnormal sound categories, and uploading the abnormal sound categories to the terminal, the abnormal sound categories include vibration abnormal sound and knocking abnormal sound.

2. The vehicle sheet metal abnormal noise detection method according to claim 1, characterized in that: The step S2 specifically includes the following steps: S201: Setting a second sliding window, and obtaining the enhanced noise signal by processing the sliding window in combination with the vehicle noise signal through the signal filtering model, wherein the signal filtering model is expressed as: Wherein, f(n) represents the vehicle noise signal, g represents the second sliding window, n represents the signal length, PO(f(n)) represents the positive impact enhancement signal, ⊙ represents the corrosion operator, represents the expansion operator, NE(f(n)) represents the negative impact enhancement signal, and EN(f(n)) represents the enhanced noise signal; S202: Acquire the sampling frequency of the vehicle noise signal, set the length of the first sliding window to 0.05 times the sampling frequency, and calculate the vehicle noise signal according to the formula Calculate and obtain a signal mean, wherein Thr(1) represents the signal mean, N represents the signal length, and f(i) represents the vehicle noise signal, and set the signal mean as the initialization threshold; S203: A difference operator is calculated according to the noise signal quantization vector and the abnormal sound signal quantization vector by using the formula △T=||D1-D2||, where △T represents the difference operator, D1 represents the noise signal quantization vector, D2 represents the abnormal sound signal quantization vector, and ||·|| represents a modulo operation; a noise vector is obtained by feature extraction according to the smooth noise signal; an initial threshold value of the abnormal sound signal is set according to the noise vector in combination with the difference operator; and abnormal sound detection is performed on the smooth noise signal according to the initial threshold value of the abnormal sound signal in combination with the first sliding window to obtain the vehicle abnormal sound signal.

3. The vehicle sheet metal abnormal noise detection method according to claim 2, characterized in that: The step S203 specifically includes the following steps: The initial threshold of the abnormal sound signal is expressed as: TR(i)=D(i)+β△T, wherein TR(i) represents the initial threshold of the abnormal sound signal, D(i) represents the noise vector, β represents the adjustment factor, and △T represents the difference operator; The smooth noise signal is windowed according to the first sliding window to obtain a signal within the window. When the signal within the window is greater than the initial threshold of the abnormal sound signal, the signal within the window is an abnormal sound signal segment. The signal peak of the abnormal sound signal segment is determined. The noise signal quantization vector and the abnormal sound signal quantization vector are updated according to the signal peak. The calculation formula is expressed as follows: Wherein, i represents the current signal length, D2(i) represents the abnormal sound signal quantization vector, D1(i) represents the noise signal quantization vector, α represents the signal peak value, ω represents the length of the first sliding window, D(i) represents the noise vector, and f(k) represents the signal in the window; The difference operator is updated according to the updated noise signal quantization vector and the abnormal sound signal quantization vector, and step S203 is repeated. When the current signal length is greater than the signal length, the vehicle abnormal sound signal is obtained by integration according to the abnormal sound signal segments.

4. The vehicle sheet metal abnormal noise detection method according to claim 1, characterized in that: The step S3 specifically comprises the following steps: S301: calling a standard abnormal sound signal in a sheet metal abnormal sound database, constructing the feature selection model through machine learning according to the standard abnormal sound signal, constructing a vehicle abnormal sound signal feature matrix according to the vehicle abnormal sound signal feature, and obtaining a matrix eigenvalue and a feature contribution rate according to the vehicle abnormal sound signal feature matrix through the feature selection model; S302: Arrange the feature contribution rates from large to small. When the feature contribution rate is greater than 80%, the matrix eigenvalues ​​corresponding to the feature contribution rates are the vehicle abnormal sound signal eigenvalues. The vehicle abnormal sound signal eigenvalues ​​are integrated to obtain the eigenvalue set. S303: Process the normalized information distance based on the eigenvalue set through the normalized linear classification model, and obtain the abnormal sound category based on the normalized information distance.

5. The vehicle sheet metal abnormal noise detection method according to claim 4, characterized in that: The specific steps of step S301 include the following steps: According to the vehicle abnormal sound signal characteristic matrix, the formula The matrix eigenvalues ​​are calculated, where G i represents the eigenvalue of the matrix i, H represents the characteristic matrix of the vehicle abnormal sound signal, K represents the number of the vehicle abnormal sound signal features, Y i represents the i-th vehicle abnormal sound signal feature, θ(·) represents the kernel function, * represents the convolution symbol, and the matrix eigenvalue is calculated by the formula Calculate the eigenvector, where i = 1, 2, 3..., represents the i-th eigenvector, δ represents the coefficient factor; According to the matrix eigenvalue and the eigenvector by formula The characteristic contribution rate is calculated, wherein con represents the characteristic contribution rate, L represents the total number of eigenvalues ​​of the matrix, and * represents a multiplication operator.

6. The vehicle sheet metal abnormal noise detection method according to claim 4, characterized in that: The specific steps of step S303 include the following steps: The normalized linear classification model is expressed as: Where, d represents the normalized information distance, r h. represents a constant related to the classification standard, h is a positive integer, q represents a weight vector corresponding to the eigenvalue set, J represents a preset parameter, and v represents the eigenvalue set; By formula Calculate the adjacent normalized information distance to obtain the adjacent information distance, Wi represents the adjacent information distance, d i represents the normalized information distance of the ith k represents the kth normalized information distance, i and k represent two adjacent positive integers, and D(·) is a preset calculation function; When 0 < Wi < 4, output the abnormal sound category as the vibration abnormal sound; when Wi ≥ 4, output the abnormal sound category as the knocking abnormal sound.

7. A vehicle sheet metal noise detection system, the vehicle sheet metal noise detection system adopts the vehicle sheet metal noise detection method according to claim 1, characterized in that: Include: A signal acquisition module for acquiring vehicle noise signals; An abnormal sound detection module for processing the vehicle noise signal through a signal filtering model to obtain an enhanced noise signal, processing the enhanced noise signal through a Gaussian filter to obtain a smoothed noise signal, determining signal parameters based on the smoothed noise signal, obtaining an initial parameter value through an initial setting of the signal parameters, and obtaining a vehicle abnormal sound signal through abnormal sound detection based on the smoothed noise signal in combination with the initial parameter value. The signal parameters include a noise signal quantization vector, an abnormal sound signal quantization vector, a first sliding window, and an initial threshold; An abnormal sound detection module for extracting the characteristics of the vehicle abnormal sound signal to obtain the characteristics of the vehicle abnormal sound signal, processing the characteristics of the vehicle abnormal sound signal through a feature selection model to obtain abnormal sound signal eigenvalue, obtaining an eigenvalue set based on the abnormal sound signal eigenvalue, processing the eigenvalue set through a normalized linear classification model to obtain the abnormal sound category, and uploading the abnormal sound category to the terminal. The abnormal sound category includes friction abnormal sound, vibration abnormal sound, and knocking abnormal sound.

Images