Disc brake brake peristaltic flutter noise evaluation method
By using a unidirectional accelerometer and a machine learning model to evaluate brake creep and chatter noise in disc brakes, the problem of high cost and inconsistency in manual evaluation is solved, and automated, fast and accurate noise evaluation is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TONGJI UNIV
- Filing Date
- 2023-02-20
- Publication Date
- 2026-05-15
AI Technical Summary
In the existing technology, the assessment of disc brake creep and chatter noise relies on subjective human judgment, which leads to high cost, low efficiency and inconsistent results.
The raw acceleration signal at the brake caliper is obtained by a unidirectional accelerometer and evaluated by a machine learning classification model. Preprocessing and standardization are performed using eight physical features and three temporal features to construct a one-dimensional vector feature, and noise evaluation is performed using a support vector machine.
It enables automated, rapid, and accurate assessment of brake creep and flutter noise, reducing labor and time costs and improving the scientific rigor and consistency of the assessment.
Smart Images
Figure CN116166986B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of brake noise identification, and in particular to a method for assessing brake creep and chatter noise in disc brakes. Background Technology
[0002] The creeping noise generated during low-speed braking of automobiles is a significant contributor to urban traffic noise pollution, negatively impacting both passenger comfort and the surrounding environment. Testing brake creeping noise is a crucial part of automotive testing and a necessary step in friction pad selection. However, current brake creeping noise testing requires professional evaluators in the driver's seat to listen to the noise and score it subjectively to assess vehicle or friction pad performance. This method is time-consuming, labor-intensive, costly, and inefficient, and the subjective nature of human judgment can lead to differing results from different evaluators.
[0003] Therefore, how to develop a method that uses intelligent technology to replace manual evaluation of disc brake creep and chatter noise has become a technical problem that needs to be solved. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for evaluating the vibration noise of disc brakes.
[0005] The objective of this invention can be achieved through the following technical solutions:
[0006] A method for evaluating the noise of disc brake creep and chatter, the method comprising the following steps:
[0007] 1) Obtain the original acceleration signal at the brake caliper using a unidirectional accelerometer;
[0008] 2) Extract the eight physical characteristics corresponding to the occurrence of braking creep flutter;
[0009] 3) Preprocess the original acceleration signal to extract the corresponding three time-domain features;
[0010] 4) Construct a one-dimensional vector feature that includes all physical and temporal features, and then standardize it;
[0011] 5) Use the trained machine learning classification model to analyze and evaluate the constructed features to obtain a score for braking creep flutter noise.
[0012] Furthermore, the original acceleration signal is the vibration acceleration signal at the brake caliper when the vehicle experiences braking creep and flutter.
[0013] Furthermore, the eight physical characteristics include road slope, lower limit of braking pressure, upper limit of braking pressure, gear, driving direction, ambient temperature, disc temperature and ambient humidity; the three time-domain characteristics include the relative standard deviation of average peak amplitude, flutter time ratio and time-domain skewness.
[0014] Furthermore, the preprocessing specifically involves: passing the original acceleration signal through a fourth-order Butterworth high-pass filter, squaring the output signal and multiplying it by an amplification factor, and finally passing it through a first-order Butterworth low-pass filter to obtain a preprocessed acceleration signal, and then calculating the relative standard deviation of the average peak amplitude using the preprocessed acceleration signal.
[0015] Furthermore, the bandwidth of the fourth-order Butterworth high-pass filter is 0–10 kHz; the bandwidth of the first-order Butterworth low-pass filter is 0–550 Hz; and the amplification factor is shown in the following formula:
[0016]
[0017] Among them, f high For 10kHz, f low It is 550Hz.
[0018] Furthermore, the original acceleration signal is segmented according to a period of 0.25s.
[0019] Furthermore, the trained machine learning classification model is specifically constructed by: constructing a sample set through one-dimensional vector features, standardizing the sample set as input, and using the braking data corresponding to each acceleration signal as a label by subjective scoring by professional evaluators. The trained machine learning classification model is obtained by training the sample set. The standardization of the sample set is to subtract the average of the sample feature values of the training set from the sample feature values and then divide by the standard deviation of the sample feature values of the training set.
[0020] Furthermore, the machine learning classification model is a support vector machine.
[0021] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.
[0022] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0023] Compared with the prior art, the present invention has the following beneficial effects:
[0024] I. This invention uses machine learning to score braking creep flutter noise, and uses physical features and acceleration signal time-domain features as inputs to the machine learning training model. It can effectively solve the problem of different duration dimensions of braking creep flutter noise at different durations while ensuring recognition accuracy and iteration speed.
[0025] Second, the method of the present invention has a fast recognition speed, does not require human intervention, saves manpower and time costs, and has higher recognition accuracy and scientificity.
[0026] Third, the present invention processes the original acceleration signal through corresponding preprocessing methods, making it easier to process and simplifying subsequent feature extraction. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0028] Figure 2 This is a schematic diagram of signal feature acquisition according to the present invention. Detailed Implementation
[0029] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.
[0030] Example
[0031] like Figure 1 and 2 As shown, a method for evaluating the vibration noise of a disc brake during braking creep is described, the method comprising the following steps:
[0032] 1) Obtain the original acceleration signal at the brake caliper using a unidirectional accelerometer.
[0033] The raw acceleration signal is the vibration acceleration signal at the brake caliper when the vehicle experiences braking creep and vibration. The unidirectional accelerometer collects data based on the trigger and termination of the brake pedal signal, and saves the signal in formats such as WAV and Excel, as long as the signal is not distorted.
[0034] 2) Extract the eight physical characteristics corresponding to the occurrence of braking creep flutter.
[0035] The eight physical characteristics include road slope, lower limit of braking pressure, upper limit of braking pressure, gear position, driving direction, ambient temperature, disc temperature, and ambient humidity. Road slope and ambient humidity are expressed as decimals, such as 0.12 for 12% slope and 0.2 for 20% humidity. In gear position, forward gear is represented by 1, neutral gear by 0, and reverse gear by -1. In driving direction, forward direction is represented by 1 and reverse direction by -1.
[0036] 3) Preprocess the original acceleration signal to extract the corresponding three time-domain features.
[0037] The three time-domain features include the relative standard deviation of the average peak amplitude, the dithering time ratio, and the time-domain skewness. The preprocessing specifically involves passing the original acceleration signal through a fourth-order Butterworth high-pass filter, squaring the output signal and multiplying it by an amplification factor, and then passing it through a first-order Butterworth low-pass filter to obtain the preprocessed acceleration signal. The signal is then divided into multiple segments with a period of 0.25 seconds, and the relative standard deviation of the peak amplitude for each segment is extracted. These segments are then summed and averaged over the number of segments to obtain the average relative standard deviation of the peak amplitude. The dithering time ratio and time-domain skewness do not require preprocessing of the acceleration signal and can be directly calculated and extracted. The bandwidth of the fourth-order Butterworth high-pass filter is 0–10 kHz; the bandwidth of the first-order Butterworth low-pass filter is 0–550 Hz; and the amplification factor is shown in the following formula:
[0038]
[0039] Among them, f high For 10kHz, f low It is 550Hz.
[0040] 4) Construct a one-dimensional vector feature that includes all physical and temporal features, and then standardize it.
[0041] During the brake creep and chatter test, while acquiring each acceleration signal, a professional evaluator sat in the driver's cab, listened to the brake creep and chatter noise transmitted through a microphone, and made a subjective assessment and score; the score served as the label for the acceleration signal. For the already labeled acceleration signals, a 1D vector feature of 11 elements was extracted using the aforementioned feature extraction method. The data volume for each segment should be no less than 1000 data points for subsequent machine learning model training.
[0042] 5) Use the trained machine learning classification model to analyze and evaluate the constructed features to obtain a score for braking creep flutter noise.
[0043] The trained machine learning classification model is specifically constructed by: constructing a sample set through one-dimensional vector features, standardizing the sample set as input, and using the braking data corresponding to each acceleration signal as a label by subjective scoring by professional evaluators. The trained machine learning classification model is obtained by training the sample set. The standardization of the sample set is to subtract the average of the sample feature values of the training set from the sample feature values and then divide by the standard deviation of the sample feature values of the training set.
[0044] The machine learning classification model described is a Support Vector Machine (SVM). The model is built using the software's built-in SVM function package. The pre-scored features are standardized and used as input to the SVM. A small validation set is reserved, ensuring it includes all segments. The SVM parameters are iteratively selected multiple times, and the accuracy on the validation set is observed. Training ends when the validation set accuracy reaches its maximum, and the trained machine learning classification model is saved.
[0045] The trained machine learning classification model is used to score and predict the features of the brake creep flutter signal under test. The scores of brake creep flutter noise and the probabilities of different scores can be obtained. The highest probability corresponds to the predicted score. A threshold can be set. If the highest probability is lower than the threshold, it should be manually reviewed, while if it is higher than the threshold, the score is considered accurate and does not require manual review.
[0046] This invention is based on relatively mature machine learning signal processing technology. It uses the features of acceleration signals as input to the machine learning training model, which can effectively solve the problem of different feature dimensions of braking creep flutter noise at different durations while ensuring recognition accuracy and iteration speed.
[0047] The present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.
[0048] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0049] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.
Claims
1. A method for evaluating the noise of disc brake creep and chatter, characterized in that, The method includes the following steps: 1) Obtain the original acceleration signal at the brake caliper using a unidirectional accelerometer; 2) Extract the eight physical characteristics corresponding to the occurrence of braking creep and chatter, including road slope, lower limit of braking pressure, upper limit of braking pressure, gear, driving direction, ambient temperature, disc temperature and ambient humidity; 3) Preprocess the original acceleration signal to extract three corresponding time-domain features, including the relative standard deviation of the average peak amplitude, the flutter time ratio, and the time-domain skewness. Specifically, the preprocessing involves passing the original acceleration signal through a fourth-order Butterworth high-pass filter, squaring the output signal and multiplying it by an amplification factor, and then passing it through a first-order Butterworth low-pass filter to obtain the preprocessed acceleration signal. The relative standard deviation of the average peak amplitude is then calculated from the preprocessed acceleration signal. The bandwidth of the fourth-order Butterworth high-pass filter is 0–10 kHz; the bandwidth of the first-order Butterworth low-pass filter is 0–550 Hz; and the amplification factor is shown in the following formula: in, 10 kHz 550Hz; 4) Construct a one-dimensional vector feature that includes all physical and temporal features, and then standardize it; 5) Use the trained machine learning classification model to analyze and evaluate the constructed features to obtain a score for braking creep flutter noise.
2. The method for evaluating the vibration noise of a disc brake according to claim 1, characterized in that, The original acceleration signal is the vibration acceleration signal at the brake caliper when the car experiences braking creep and flutter.
3. The method for evaluating the vibration noise of a disc brake according to claim 1, characterized in that, The original acceleration signal is segmented according to a period of 0.25s.
4. The method for evaluating the vibration noise of a disc brake according to claim 1, characterized in that, The trained machine learning classification model is specifically constructed by: constructing a sample set through one-dimensional vector features, standardizing the sample set as input, and using the braking data corresponding to each acceleration signal as a label by subjective scoring by professional evaluators. The trained machine learning classification model is obtained by training the sample set. The standardization of the sample set is to subtract the average of the sample feature values of the training set from the sample feature values and then divide by the standard deviation of the sample feature values of the training set.
5. The method for evaluating the vibration noise of a disc brake according to claim 1, characterized in that, The machine learning classification model described is a support vector machine.
6. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 5.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.