Transmission fault diagnosis method of electronic control brake system based on multi-domain feature fusion of vibration signals

Abstract

The present invention discloses a method for diagnosing transmission faults of an electric control brake system by fusing multi-domain features of vibration signals. The present invention simulates the braking process of an actual vehicle, collects the vibration signal and the output displacement signal of an automobile electric control brake booster motor when it is working, separates the fault feature data, and fuses three methods, namely, time domain eigenvalue analysis, time-frequency domain short-time Fourier analysis, and order analysis based on displacement resampling. The method diagnoses the transmission faults of automobile electric control brake booster motors based on pattern recognition of support vector machines, judges the eligibility of automobile electric control brake booster motors, and prevents automobile electric control brake booster system components with potential safety hazards from entering the market.

Description

Technical Field

[0001] The invention relates to the field of detection, and in particular to a transmission fault diagnosis method for an electric control brake system by fusing multi-domain features of vibration signals. Background Art

[0002] Transmission fault diagnosis is widely used in the fields of main reducer, gearbox, motor, engine, etc. For example, Zhao Zhijun of Hefei University of Technology studied the quality of automobile main reducer based on order analysis and vibration velocity method; Wang Yuancheng of Tongji University proposed the application of order analysis in gear transmission fault diagnosis; Zhang Xiang of Wuhan University of Science and Technology analyzed the vibration signal and fault diagnosis of fan gearbox based on order analysis; Zheng Di of Northeastern University used the vibration and noise analysis technology of automobile power transmission system based on multi-domain hybrid modeling; Liu Yingjie of Jilin University studied the matching and control of NVH performance of passenger car powertrain, etc. The main reducer, gearbox, motor and engine are all only rotating motion, and vibration data can be collected continuously, and the speed signal can be directly collected for order analysis, but the motor of the automobile electric control brake booster system will convert the rotational motion of the motor into linear motion through the ball screw. Because the screw limits the range of motion, only limited data can be collected, and the speed data of the motor cannot be collected, which is not applicable to the traditional order analysis method. The electric booster includes two forms of motion, rotational motion and linear motion, and there is currently no corresponding method for transmission fault diagnosis.

[0003] The motor of the electric control brake assist system of the automobile will generate vibrations of various frequencies during operation. Different structural units will have different vibration components, and the vibration signals will be different in different working states. When there is a fault in the mechanical structure, the motor of the electric control brake assist system of the automobile will generate new vibration components or the amplitude of some vibration components will increase.

[0004] There are also specialized vibration and noise analysis software in the industry. For example, the LMS Test.Lab software of the Belgian LMS company integrates functions such as data acquisition, digital signal processing, structural testing, rotating machinery analysis, acoustic and environmental testing, and can directly analyze and obtain vibration signal order data. However, the software needs to continuously collect vibration data. The motor of the automotive electronic control brake assist system is based on the demand for rapid braking. In emergency braking, the working time from the start of the brake pedal to the full travel of the brake pedal to achieve braking is less than 1 second. The large gear of the brake transmission mechanism rotates only a few times, and the effective data collected is very limited. The collected data cannot meet the test requirements of the LMS Test.Lab software in terms of quantity. In addition, LMS Test.Lab is relatively expensive and cannot achieve online real-time diagnosis at industrial sites. It is suitable for laboratory analysis and processing. The DEWESOFT X3 software of the Austrian DEWESOFT company integrates functions such as signal acquisition, FFT transform analysis, STFT transform analysis, and order tracking, but order tracking requires the simultaneous acquisition of speed and vibration signals, and FFT transform is only applicable to analysis under the condition of uniform rotation and known rotation period. The movement of the motor in the automotive electronic brake assist system is non-uniform and non-steady, and the speed signal cannot be collected.

[0005] In summary, conventional rotating machinery such as gearboxes, motors, and engines can collect data continuously, and a large amount of vibration data can be collected for analysis. The braking process of the motor in the automotive electric control brake assist system is fast and the travel is limited. The collected vibration data is small, and it is not easy to diagnose transmission faults. Although the vibration and noise analysis software in the industry integrates many functions such as collection and analysis, it cannot provide simulated loads, and the test conditions are harsh, requiring synchronous collection of speed signals. The motor of the automotive electric control brake assist system converts the motor's rotational motion into linear motion through a screw rod. The motor and transmission mechanism are enclosed in the brake component housing, and the output is performed by a linear motion push rod. The speed data cannot be collected, so the conventional order analysis method cannot be used.

[0006] The abnormal vibration of automobile parts is mainly caused by damage to the internal mechanical device or abnormal working state. Common fault parts are gears, bearings, shafts, box fasteners, etc. The most important fault part of automobile parts is the gear. For electric boosters, the main fault part is also the gear. When the electric booster is running, due to manufacturing errors, local damage, poor assembly and other factors, the gears of the electric booster interact with external interference forces to generate complex vibration signals. Common forms of gear damage include: tooth surface bonding, tooth surface wear, gear tooth fracture, tooth surface contact fatigue, etc. Therefore, there is a need for a method that can detect whether the transmission gear is normal when the motor is rotating. Summary of the invention

[0007] The purpose of the present invention is to solve the above-mentioned technical problems and provide a method for diagnosing transmission faults of an electric control brake system by fusing multi-domain features of vibration signals. The present invention simulates the braking process of a real vehicle, collects the vibration signal and output displacement signal of the automobile electric control brake booster system when it is working, separates the fault feature data, and integrates three methods: time domain eigenvalue analysis, time-frequency domain short-time Fourier analysis, and order analysis based on displacement resampling. The pattern recognition based on support vector machine is used to diagnose the motor transmission fault of the automobile electric control brake booster system, judge the qualification of the automobile electric control brake booster system, and prevent boosters with safety hazards from entering the market.

[0008] The technical solution adopted by the present invention to solve the above technical problems is: a transmission fault diagnosis method of an electronically controlled braking system by fusion of multi-domain features of vibration signals, comprising the following steps:

[0009] Step 1: According to the transmission fault information of the electric control brake assist system of the automobile, the transmission fault is divided into N categories according to the mechanical structure defects, which are recorded as {x1, x2, ..., x N Then, a total of m' faulty parts and m' qualified parts, totaling 2m' product samples, including N types of faults, were used for experimental testing. The servo loading system drove the electric control brake assist system, and the vibration sensor signal arranged on the shell of the electric control brake assist system and the displacement sensor signal arranged at the output end of the electric control brake assist system were synchronously collected to realize the acquisition of time signal. Vibration signal Output displacement signal Synchronous acquisition of:

[0010]

[0011] Where n is the number of samples;

[0012] Step 2: Calculate the root mean square value tv1, peak factor tv2 and kurtosis value tv3 of the vibration signal, and construct the time domain statistical feature vector of the vibration signal

[0013]

[0014] Step 3: Construct the energy ridge feature vector of the vibration signal in the time-frequency domain: set the transformation window scale and window overlap width, and perform short-time Fourier transform on the vibration signal; based on the energy ridge of the gear meshing frequency of the brake system transmission mechanism, extract the local maximum value data of the short-time Fourier spectrum value of the electric control brake assist system at the start and end stages to construct the energy ridge feature vector in the time-frequency domain Where tf1 and tf2 are the local maximum values ​​of the short-time Fourier spectrum at the start and end stages of the electronically controlled brake assist system respectively;

[0015] Step 4: Based on the time domain brake system output displacement signal and vibration signal, combined with the linear interpolation method, reconstruct the equiangular vibration signal; construct the vibration signal angular frequency order spectrum feature vector based on the Fourier spectrum value of the equiangular vibration signal

[0016] Step 5: Transform the time domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Based on the principal component analysis method, the data redundancy and data conflict problems between multi-domain features are eliminated to achieve the fusion of multi-domain feature vectors;

[0017] Step 6: Construct a support vector machine classification decision model and transform the multi-domain principal component feature vectors of 2m' samples and product qualified / faulty type labels, train the decision model of the support vector machine, find the optimal decision function for normal parts and faulty parts, and complete the training of the support vector machine classification decision model;

[0018] Step 7: Take any electronically controlled brake assist system product and repeat steps 1 to 5 to construct the multi-domain principal component feature vector of the product. The feature vector Input the support vector machine classification decision model to complete the fault diagnosis of the electronically controlled brake assist system product.

[0019] Preferably, step 4 comprises the following steps:

[0020] Step 1: Determine the equal angle value Δθ of the resampled vibration signal, and determine the output displacement value Δs of the electric control brake assist system corresponding to the equal angle value Δθ based on the transmission ratio L of the ball screw of the electric control brake assist system:

[0021]

[0022] Step 2: Determine the output displacement points of the equal-angle vibration signal resampling points

[0023] in,

[0024] In the formula, i is an integer, and s′ m is not greater than s n The maximum displacement value of

[0025] Step 3: Based on the synchronously acquired time signal And output displacement signal At the displacement point s iIn the neighborhood of ′, the piecewise second-order polynomial fitting method is used to fit the relationship between the output displacement and time of the electronically controlled braking assist system:

[0026] d i (t) = a i +b i t+c i t 2 , where i = 1, ..., m;

[0027] Step 4: Let s i ′=d i (t), find the time point value of the equal angle resampling point i=1,...,m;

[0028] Step 5: Based on the synchronously acquired time signal And output displacement signal Use linear interpolation method to get the time point value The corresponding equal-angle resampled vibration signal value The specific method is as follows:

[0029]

[0030] where t j ,t j+1 For time signal Middle and t i 'The two closest time points;

[0031] Step 6: Right The signal is transformed by Fourier to obtain the vibration signal order spectrum data, and the local maximum values ​​of the three order segments of the vibration signal order spectrum are extracted in a segmented manner to construct the angular frequency order spectrum feature vector

[0032]

[0033] Among them, af1, af2, and af3 are the local maxima of the low, medium, and high order segments of the resampled vibration signal order spectrum, respectively.

[0034] Preferably, step five comprises the following steps:

[0035] Step A: Based on time-domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Constructing multi-domain feature vectors The details are as follows:

[0036]

[0037] Step B: Construct a sample set from the multi-domain feature vectors of the vibration signal of 2m' samples Based on the Gaussian kernel function, the sample set M is transformed into a new feature space to obtain the matrix K:

[0038]

[0039] Step C: Set the principal component contribution rate threshold, and determine the principal component eigenvectors of all sample multi-domain eigenvectors by principal component analysis based on the sample feature space matrix K Let the multi-domain feature vector of the i-th sample be The principal component eigenvectors of Where l is the number of principal components, i = 1,...,2m'.

[0040] The present invention has the following beneficial effects:

[0041] 1. The present invention simulates the braking process of a real vehicle, collects the vibration signal and output displacement signal of the motor of the automobile electric control brake assist system when it is working, separates the fault feature data, integrates the three methods of time domain eigenvalue analysis, time-frequency domain short-time Fourier analysis, and order analysis based on displacement resampling, and diagnoses the transmission fault of the motor of the automobile electric control brake assist system based on pattern recognition of support vector machine, judges the qualification of the motor transmission mechanism of the automobile electric control brake assist system, and prevents the electric control brake assist system components with safety hazards from entering the market.

[0042] 2. The present invention proposes a separation method based on extreme value interval positioning-difference feature point extraction to separate the data in the braking process, calculate the characteristic value in the time domain, and make a preliminary judgment; in the time-frequency domain, based on short-time Fourier analysis, the characteristic energy ridge is extracted to judge the qualification of the product; in the angular frequency domain, an order analysis algorithm based on displacement resampling is proposed to extract the characteristic order to determine whether the product has a fault; and the Gaussian kernel space transformation method is used to fuse the characteristic values ​​of the time domain, time-frequency domain, and angular frequency order domain, and the pattern recognition algorithm of the support vector machine is used to fuse the results of the three methods for training, thereby improving the accuracy of transmission fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0044] The present invention is further described below in conjunction with the accompanying drawings and implementation modes.

[0045] like Figure 1 As shown, the electric control brake system transmission fault diagnosis method of the present invention by fusing multi-domain features of vibration signals comprises the following steps:

[0046] Step 1 S1: According to the transmission fault information of the electric control brake assist system of the automobile, the transmission fault is divided into N categories according to the mechanical structure defects, which are recorded as {x1, x2, ..., x N Then, a total of m' faulty parts and m' qualified parts, totaling 2m' product samples, including N types of faults, were used for experimental testing. The servo loading system drove the electric control brake assist system, and the vibration sensor signal arranged on the shell of the electric control brake assist system and the displacement sensor signal arranged at the output end of the electric control brake assist system were synchronously collected to realize the acquisition of time signal. Vibration signal Output displacement signal Synchronous acquisition of:

[0047]

[0048] Where n is the number of samples;

[0049] Transmission failures are divided into tooth profile error, tooth surface wear, tooth breakage, and tooth surface bonding according to mechanical structure defects.

[0050] Step 2 S2: Calculate the root mean square value tv1, peak factor tv2 and kurtosis value tv3 of the vibration signal, and construct the time domain statistical feature vector of the vibration signal

[0051]

[0052] Step 3 S3: Construct the energy ridge feature vector of the vibration signal in the time-frequency domain: set the transformation window scale and window overlap width, and perform short-time Fourier transform on the vibration signal; based on the energy ridge of the gear meshing frequency of the transmission mechanism of the brake system, extract the local maximum value data of the short-time Fourier spectrum value of the electric control brake assist system at the start and end stages to construct the energy ridge feature vector in the time-frequency domain Among them, tf1 and tf2 are the local maximum values ​​of the short-time Fourier spectrum value at the start and end stages of the electric control brake assist system respectively;

[0053] For the motor of the electric control brake assist system of the automobile, the energy output by the motor is transmitted to the box. The periodic vibration caused by the gear defects will cause the output energy to increase. The root mean square value is usually used to calculate the overall energy loss. The root mean square value represents the average value of the signal and is very representative in describing the overall strength of power and energy.

[0054] The root mean square value is the average of time, and can make appropriate evaluations of abnormalities with irregular vibration waveforms such as surface cracks. However, it is not applicable to faults with instantaneous impact vibrations such as missing teeth. For such faults, the peak factor can be used to make statistical judgments on the defect details. It characterizes the obviousness of the signal peak, and a higher peak factor means a more obvious peak.

[0055] In gear fault diagnosis, it is necessary to know the numerical statistics of the vibration distribution characteristics. Kurtosis represents the flatness of the waveform and is used to describe the distribution of variables. It is particularly sensitive to impact signals and is particularly suitable for the diagnosis of surface damage faults of motor gears in automotive electronic control brake assist systems.

[0056] In summary, the low-frequency signals generated by gear faults such as wear and cracks can be characterized by the root mean square value; the high-frequency signals generated by faults such as missing teeth can be characterized by the peak factor; the kurtosis can reflect the fault forms such as wear, cracks, and missing teeth. The test system first extracts the vibration signal of the motor housing of the automotive electronic control brake assist system. The automotive electronic control brake assist system motor can only reflect the fault when it is in motion, so the outbound and return data during the movement process are extracted from the original data, and then the root mean square value, peak factor and kurtosis are calculated respectively. The product process department statistically analyzes the thresholds of these time domain characteristic values ​​through a large amount of experimental data. The test value is considered unqualified if it exceeds the upper or lower limit.

[0057] Step 4 S4: Based on the time domain brake system output displacement signal and vibration signal, combined with the linear interpolation method, reconstruct the equiangular vibration signal; construct the vibration signal angular frequency order spectrum feature vector based on the Fourier spectrum value of the equiangular vibration signal Order analysis based on displacement resampling does not require speed data. Instead, the vibration data is resampled at equal intervals based on the displacement data to transform the non-stationary signal into a stationary signal. Then, Fourier transform is performed to obtain the order spectrum, and the vibration components of each order can be clearly observed. The motor of the automotive electronically controlled brake assist system is an integrated electronic servo assist system that converts rotational motion into linear motion. Because the interior is sealed, the motion state of the rotating body cannot be seen, and the speed signal cannot be measured. Therefore, angle-based hardware resampling and computational resampling are not applicable. In response to this, an order analysis algorithm based on displacement resampling is proposed.

[0058] The motor of the electric control brake assist system of the automobile converts the rotational motion into linear motion through the ball screw, and the rotation angle is proportional to the output displacement. The order analysis method based on displacement resampling simultaneously collects displacement data and vibration signals, and then resamples the data sampled at equal time intervals with equal displacement, and finally performs Fourier transform on the resampled vibration data to obtain the order spectrum. The first step is to collect the output displacement data of the motor of the electric control brake assist system of the automobile and fit the displacement curve equation; the second step is to calculate the cumulative rotation angle based on the output displacement data, take a point at every fixed angle, calculate the corresponding time, and determine the phase detection time scale; the third step is to extract the vibration data corresponding to the phase detection time scale; the fourth step is to interpolate the missing data at some moments; the fifth step is to perform Fourier transform on the interpolated vibration data to obtain the characteristic order spectrum of vibration.

[0059] The specific steps include:

[0060] Step 1: Determine the equal angle value Δθ of the resampled vibration signal, and determine the output displacement value Δs of the electric control brake assist system corresponding to the equal angle value Δθ based on the transmission ratio L of the ball screw of the electric control brake assist system:

[0061]

[0062] Step 2: Determine the output displacement points of the equal-angle vibration signal resampling points

[0063] in,

[0064] In the formula, i is an integer, and s′ m is not greater than s n The maximum displacement value of

[0065] Step 3: Based on the synchronously acquired time signal And output displacement signal At the displacement point s i In the neighborhood of ′, the piecewise second-order polynomial fitting method is used to fit the relationship between the output displacement and time of the electronically controlled braking assist system:

[0066] d i (t) = a i +b i t+c i t 2 , where i = 1, ..., m;

[0067] Step 4: Let s i ′=d i (t), find the time point value of the equal angle resampling point i=1,...,m;

[0068] Step 5: Based on the synchronously acquired time signal And output displacement signal Use linear interpolation method to get the time point value The corresponding equal-angle resampled vibration signal value The specific method is as follows:

[0069]

[0070] where t j ,t j+1 For time signal Middle and t i 'The two closest time points;

[0071] Step 6: Right The signal is Fourier transformed to obtain the vibration signal order spectrum data. The local maximum values ​​of the three order segments of the vibration signal order spectrum are extracted in a segmented manner to construct the angular frequency order spectrum feature vector

[0072]

[0073] Where af1, af2, and af3 are the local maxima of the low, medium, and high order segments of the resampled vibration signal order spectrum, respectively.

[0074] Step 5 S5: Transform the time domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Based on the principal component analysis method, the data redundancy and data conflict problems between multi-domain features are eliminated to achieve the fusion of multi-domain feature vectors;

[0075] The specific steps include:

[0076] Step A: Based on time-domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Constructing multi-domain feature vectors The details are as follows:

[0077]

[0078] Step B: Construct a sample set from the multi-domain feature vectors of the vibration signal of 2m' samples ( is the multi-domain feature vector of the i-th sample), and based on the Gaussian kernel function, the sample set M is transformed into a new feature space to obtain the matrix K:

[0079]

[0080] Step C: Set the principal component contribution rate threshold, and determine the principal component eigenvectors of all sample multi-domain eigenvectors by principal component analysis based on the sample feature space matrix K Let the multi-domain feature vector of the i-th sample be The principal component eigenvectors of Where l is the number of principal components, i = 1,...,2m'.

[0081] Step 6 S6: Construct a support vector machine classification decision model and transform the multi-domain principal component feature vectors of 2m' samples and product qualified / faulty type labels, train the decision model of the support vector machine, find the optimal decision function for normal parts and faulty parts, and complete the training of the support vector machine classification decision model;

[0082] The quality inspection of the motor transmission fault of the electric control brake booster system of the automobile only needs to distinguish between normal parts and faulty parts, without pointing out the cause of the faulty parts, so the pattern recognition method of the support vector machine can be used. The qualified products of the motor of the electric control brake booster system of the automobile are regarded as one category, and the faulty parts are regarded as another category. All samples are divided into two parts, one as the training set and the other as the test set. The time domain eigenvalue, the time-frequency domain STFT characteristic frequency and the angular frequency domain characteristic order are used as the data input of each sample, and the product qualification given by the manufacturer is used as the type mark. The qualified is recorded as +1 and the unqualified is recorded as -1. The training is based on the algorithm of the support vector machine to find the optimal hyperplane and decision function of the normal parts and the faulty parts. Finally, the data of the test set samples are imported into the decision function to judge the qualification of the product. The actual qualification is compared with the predicted qualification to obtain the accuracy of the diagnosis.

[0083] Step 7 S7: Take any electronically controlled brake assist system product and repeat steps 1 to 5 to construct a multi-domain principal component feature vector of the product The feature vector Input the support vector machine classification decision model to complete the fault diagnosis of the electronically controlled brake assist system product.

[0084] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0085] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for diagnosing transmission faults of an electronically controlled braking system based on multi-domain feature fusion of vibration signals, characterized in that: The following steps are involved: Step 1: According to the transmission fault information of the electric control brake assist system of the automobile, the transmission fault is divided into N categories according to the mechanical structure defects, which are recorded as {x1, x2, ..., x N Then, a total of m' faulty parts and m' qualified parts, totaling 2m' product samples, including N types of faults, were used for experimental testing. The servo loading system drove the electric control brake assist system, and the vibration sensor signal arranged on the shell of the electric control brake assist system and the displacement sensor signal arranged at the output end of the electric control brake assist system were synchronously collected to realize the acquisition of time signal. Vibration signal Output displacement signal Synchronous acquisition of: Where n is the number of samples; Step 2: Calculate the root mean square value tv1, peak factor tv2 and kurtosis value tv3 of the vibration signal, and construct the time domain statistical feature vector of the vibration signal Step 3: Construct the energy ridge feature vector of the vibration signal in the time-frequency domain: set the transformation window scale and window overlap width, and perform short-time Fourier transform on the vibration signal; based on the energy ridge of the gear meshing frequency of the brake system transmission mechanism, extract the local maximum value data of the short-time Fourier spectrum value of the electric control brake assist system at the start and end stages to construct the energy ridge feature vector in the time-frequency domain Among them, tf1 and tf2 are the local maximum values ​​of the short-time Fourier spectrum value at the start and end stages of the electric control brake assist system respectively; Step 4: Based on the time domain brake system output displacement signal and vibration signal, combined with the linear interpolation method, reconstruct the equiangular vibration signal; construct the vibration signal angular frequency order spectrum feature vector based on the Fourier spectrum value of the equiangular vibration signal Step 5: Transform the time domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Based on the principal component analysis method, the data redundancy and data conflict problems between multi-domain features are eliminated to achieve the fusion of multi-domain feature vectors; Step 6: Construct a support vector machine classification decision model and transform the multi-domain principal component feature vectors of 2m' samples Where i = 1, ..., 2m', and the type label of product qualification / failure, the decision model of the support vector machine is trained to find the optimal decision function of normal parts and faulty parts, and the training of the support vector machine classification decision model is completed; Step 7: Take any electronically controlled brake assist system product and repeat steps 1 to 5 to construct the multi-domain principal component feature vector of the product. The feature vector Input the support vector machine classification decision model to complete the fault diagnosis of the electronically controlled brake assist system product.

2. The electric control brake system transmission fault diagnosis method based on multi-domain feature fusion of vibration signals according to claim 1 is characterized in that: Step 4 includes the following steps: Step 1: Determine the equal angle value Δθ of the resampled vibration signal, and determine the output displacement value Δs of the electric control brake assist system corresponding to the equal angle value Δθ based on the transmission ratio L of the ball screw of the electric control brake assist system: Step 2: Determine the output displacement points of the equal-angle vibration signal resampling points in, In the formula, i is an integer, and s′ m is not greater than s n The maximum displacement value of Step 3: Based on the synchronously acquired time signal And output displacement signal At the displacement point s i In the neighborhood of ′, the piecewise second-order polynomial fitting method is used to fit the relationship between the output displacement and time of the electronically controlled braking assist system: d i f(t) = a i + b i t + c i t 2 , where i = 1,..., m; Step 4: Let s i ′=d i (t), find the time point value of the equal angle resampling point i=1,...,m; Step 5: Based on the synchronously acquired time signal And output displacement signal Use linear interpolation method to get the time point value The corresponding equal-angle resampled vibration signal value The specific method is as follows: where t j ,t j+1 For time signal Middle and t i 'The two closest time points; Step 6: Right The signal is transformed by Fourier to obtain the vibration signal order spectrum data, and the local maximum values ​​of the three order segments of the vibration signal order spectrum are extracted in a segmented manner to construct the angular frequency order spectrum feature vector Among them, af1, af2, and af3 are the local maxima of the low, medium, and high order segments of the resampled vibration signal order spectrum, respectively.

3. The method for diagnosing transmission faults of an electronically controlled brake system by fusion of multi-domain features of vibration signals according to claim 1 or 2, characterized in that: Step five includes the following steps: Step A: Based on time-domain statistical feature vector Energy ridge eigenvector in time-frequency domain Angular frequency order spectrum eigenvector Constructing multi-domain feature vectors The details are as follows: Step B: Construct a sample set from the multi-domain feature vectors of the vibration signal of 2m' samples Based on the Gaussian kernel function, the sample set M is transformed into a new feature space to obtain the matrix K: Step C: Set the principal component contribution rate threshold, and determine the principal component eigenvectors of all sample multi-domain eigenvectors by principal component analysis based on the sample feature space matrix K Let the multi-domain feature vector of the i-th sample be The principal component eigenvectors of Where l is the number of principal components, i = 1,...,2m'.

Images