Brake pad running state real-time monitoring and early warning method based on data analysis

Abstract

The present application relates to the technical field of data processing, and especially relates to a brake pad operation state real-time monitoring and early warning method based on data analysis, which comprises the following steps: obtaining a brake pad braking vibration signal and segmenting to obtain multiple signal sequences; calculating an instability index of each signal sequence; determining a decomposition layer number of wavelet transformation and a position of a target base function in a preset base function library based on the instability index, and selecting the target base function from the preset base function library according to the position; performing wavelet transformation on each signal sequence based on the decomposition layer number and the target base function to obtain a wavelet transformation result of each signal sequence; and calculating an abnormal value based on the wavelet transformation result of each signal sequence, and responding to an operation state abnormality when the abnormal value is greater than a set threshold value. The present application determines a decomposition layer number and a target base function suitable for different signal sequences based on an instability index, realizes adaptive analysis of the signal sequences, and improves monitoring accuracy.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for real-time monitoring and early warning of brake pad operating status based on data analysis. Background Technology

[0002] As a core safety component of the automotive braking system, brake pads' friction performance, wear resistance, and resistance to heat fade directly affect vehicle driving safety. During brake pad research and development testing, or in actual road operation, the braking interface is under intense frictional coupling. The contact state between the brake disc and brake pads is extremely complex and susceptible to the effects of temperature, pressure, and wear debris. Abnormal wear of the brake pads or abrupt changes in the properties of the friction layer material can lead to brake shudder, brake squeal, or even brake failure.

[0003] In existing technologies, braking process signals are typically acquired by installing a single vibration or temperature sensor on the brake test bench or caliper, and then using wavelet transform algorithms to perform time-frequency analysis on the signals to extract features characterizing the braking state. However, existing wavelet transform application methods have inherent limitations: their analysis results are highly dependent on the pre-setting of two key parameters: the basis function and the number of decomposition levels. In practice, these parameters are often statically configured once by technicians based on their personal experience.

[0004] This static configuration is ill-suited to handle braking processes with highly nonlinear and non-stationary characteristics. For example, when brake pads transition from the initial break-in phase to the thermal decay stage, or when surface material detachment causes a sudden change in the coefficient of friction, the impact characteristics and complexity of the resulting signals change drastically. In such cases, the previously fixed analysis parameters can no longer optimally match and capture the new signal features, leading to distortion in feature extraction. Furthermore, for weak abnormal vibrations induced by minor porosity or localized hard spots, the resulting transient impact characteristics are often obscured by complex conventional vibration signals. Without adaptive feature extraction methods, it is difficult to separate effective fault information from the coupled vibration waveforms, easily resulting in missed or false alarms, affecting the sensitivity and accuracy of the monitoring system, and failing to meet the higher monitoring requirements of high-safety braking systems. Summary of the Invention

[0005] To address the technical problems of low sensitivity and susceptibility to false alarms caused by the inability of statically configured wavelet transform parameters to adapt to non-stationary braking processes, and the insufficient ability of a single monitoring signal to detect early, weak faults, this invention provides a real-time monitoring and early warning method for brake pad operating status based on data analysis. This method includes the following steps:

[0006] Vibration signals from brake pad braking during the monitoring period are acquired and segmented into multiple signal sequences. An instability index is calculated for each signal sequence, which is positively correlated with both the kurtosis and information entropy of the corresponding signal sequence. The instability index is used to determine the number of decomposition levels for wavelet transform of each signal sequence, which is positively correlated with the instability index. Based on the instability index, the position of the target basis function in a preset basis function library is determined, and a target basis function is selected from the preset basis function library according to its position. The basis functions in the preset basis function library are arranged in a preset order according to their waveform complexity. Wavelet transform is performed on each signal sequence using the decomposition level and the target basis function to obtain the wavelet transform results for each signal sequence. Based on the wavelet transform results, the temporal characteristics of the state of each signal sequence are determined. Anomalies are determined based on the changes in these temporal characteristics between adjacent signal sequences. When an anomaly exceeds a set threshold, the response is an abnormal operating state.

[0007] This invention assesses signal non-stationarity in real time by constructing an instability index related to the kurtosis and information entropy of a signal sequence. Based on this instability index, it adaptively adjusts the decomposition level and basis functions of the wavelet transform, ensuring that signals under different states can be analyzed adaptively. Stable signals are adapted to fewer decomposition levels and smoother basis functions to improve efficiency, while abnormal signals are adapted to more decomposition levels and complex basis functions to capture high-frequency features. This analysis method enables matching the instantaneous characteristics of the signal and accurately capturing early and weak fault information from complex background noise, improving the sensitivity, accuracy, and intelligence of monitoring, and solving the problems of missed and false alarms caused by fixed parameters in existing technologies.

[0008] Preferably, the instability index Satisfying the relation:

[0009] ;

[0010] in, It is the first The kurtosis of a signal sequence; It is the first Information entropy of a signal sequence; It is the maximum kurtosis of all signal sequences in the vibration signals of the historical period; It is the maximum information entropy of all signal sequences in the vibration signals of a historical period.

[0011] This invention integrates two indicators: kurtosis and information entropy. Kurtosis characterizes the impulse characteristics of a signal, while information entropy characterizes the complexity of a signal. By integrating the two, the brake pad state can be comprehensively reflected, avoiding the one-sidedness of evaluation by a single indicator. Ultimately, it provides a reliable basis for the adaptive selection of the number of wavelet transform decomposition layers and the target basis function.

[0012] Preferably, the The method for obtaining the vibration signal of the brake pad during a historical normal braking process includes: acquiring the vibration signal of the brake pad during a historical braking process and dividing it into several historical signal sequences; calculating the kurtosis of each historical signal sequence to obtain a historical kurtosis set, and denoting the maximum value in the historical kurtosis set as... The The method for obtaining the information entropy includes: calculating the information entropy of each historical signal sequence to obtain a set of historical information entropy, and taking the maximum value in the set of historical information entropy as the information entropy value. .

[0013] Preferably, the number of decomposition layers Satisfying the relation:

[0014] ;

[0015] in, It is the first The instability index of a signal sequence; It is the minimum instability index of all signal sequences in the vibration signals of a historical period; It is the maximum instability index of all signal sequences in the vibration signals of a historical period; It is the preset minimum number of decomposition levels; This is the preset maximum number of decomposition levels; It is a rounding function.

[0016] This invention maps the number of decomposition layers within a reasonable range through normalization and rounding functions. This linear mapping ensures that the larger the instability index, the more decomposition layers there are, so that the ability to capture high-frequency components is enhanced synchronously with the degree of signal anomaly. It can improve analysis efficiency with fewer decomposition layers when the signal is stable, and accurately extract high-frequency features with more decomposition layers when the signal is abnormal, thus realizing the on-demand configuration of the number of decomposition layers.

[0017] Preferably, the position of the target basis function in the preset basis function library Satisfying the relation:

[0018] ;

[0019] in, It is the first The instability index of a signal sequence; , These are the minimum and maximum instability indices of all signal sequences in the vibration signals during historical periods, respectively. It is the total number of base functions in the preset base function library; It is the round-up symbol. It is a minimum value function.

[0020] This invention first normalizes the instability index and maps it to the index range of a preset basis function library. Then, it ensures that the index is an integer by rounding up and uses a minimum value constraint to prevent the index from exceeding the library range, thus guaranteeing stable and effective basis function selection. This invention combines the feature of the preset basis function library being arranged in descending order of information entropy, so that signal sequences with larger instability indices are more likely to select target basis functions with larger indices and more complex waveforms, adapting to the abrupt changes in abnormal signals; when the signal is stable, it selects basis functions with smooth waveforms, allowing signals in different states to match the optimal basis function, improving the targeting and accuracy of feature extraction.

[0021] Preferably, the step of determining the temporal characteristics of each signal sequence state based on the wavelet transform result, and determining outliers based on the changes of the temporal characteristics between adjacent signal sequences, includes: obtaining the energy values ​​of all components in the wavelet transform result of each signal sequence to form an energy sequence; recording the signal sequence at the analysis time as the current signal sequence; calculating the dynamic time warping distance between the current signal sequence and the energy sequence of the previous signal sequence in the period to be monitored; and recording the dynamic time warping distance as the outlier in the period to be monitored.

[0022] This invention employs a dynamic time warping algorithm to calculate the distance between adjacent periodic energy sequences. The dynamic time warping algorithm can assess the similarity between adjacent periodic signal sequences. Even subtle frequency distribution changes in abnormal signals can be sensitively identified by increasing the DTW distance, avoiding missed detections of anomalies due to minor signal shifts.

[0023] Preferably, the waveform of each basis function in the preset basis function library is obtained, and an analog signal sequence consistent with the waveform is generated; the information entropy of each analog signal sequence is calculated, and all basis functions in the preset basis function library are arranged in descending order of information entropy according to the information entropy.

[0024] Preferably, the step of responding to an abnormal operating state when the abnormal value is greater than a set threshold includes: outputting a warning signal when the abnormal values ​​during the monitoring period of a preset duration are all greater than the set threshold.

[0025] Preferably, the step of segmenting the signal into multiple signal sequences includes: performing non-overlapping segmentation of the vibration signal during the period to be monitored with a preset period length.

[0026] Preferably, the preset basis function library includes Coiflet wavelet basis functions, Dobessie wavelet basis functions, Haar wavelet basis functions, and Symlet wavelet basis functions.

[0027] The beneficial effects of this invention are as follows: This invention constructs an instability index combining kurtosis and information entropy to evaluate the non-stationarity of vibration signals in real time. Not only does it determine the number of decomposition layers applicable to different signal sequences based on this instability index, but it also selects target basis functions from a pre-sorted basis function library based on information entropy. This invention achieves adaptive analysis of signal sequences, improving the sensitivity and accuracy of monitoring early, subtle faults in brake pads, and solving the problem of insufficient monitoring accuracy caused by fixed parameters in existing technologies. This invention transforms the wavelet transform results into energy sequences for each sequence, and then uses a dynamic time warping algorithm to calculate the distance between adjacent periodic energy sequences. This enables the evaluation of the similarity between adjacent periodic signal sequences, avoiding abnormal omissions caused by small signal shifts. Attached Figure Description

[0028] Figure 1 A flowchart of a data analysis-based real-time monitoring and early warning method for brake pad operating status provided in an embodiment of the present invention;

[0029] Figure 2 A visualization curve of the calculation results of the signal sequence anomaly value within the monitoring period provided in the embodiments of the present invention. Detailed Implementation

[0030] This invention provides a method for real-time monitoring and early warning of brake pad operating status based on data analysis, such as... Figure 1 As shown, the method includes steps S100-S500:

[0031] Step S100: Obtain the vibration signal of the brake pad braking during the period to be monitored, and divide it into multiple signal sequences.

[0032] It should be noted that during braking tests or actual road operation, the state of the friction coupling system formed by the brake pads and the brake disc, such as thermal fading of the friction layer, surface glazing, or localized hard spot detachment, is dynamically changing. These changes are directly reflected in the vibration signals generated during braking. In order to capture these dynamic changes in real time and accurately, the continuously acquired vibration signals need to be segmented and processed to form a series of time-discrete signal sequences, laying the data foundation for subsequent adaptive analysis for different friction states.

[0033] Specifically, a triaxial accelerometer is placed on the brake test bench or at the position of the vehicle brake caliper. The triaxial accelerometer is used to collect vibration signals during the braking friction process, forming a continuous vibration signal time series.

[0034] In addition, for ease of processing, the vibration signal of the period to be monitored is segmented into non-overlapping segments with a preset period length. Non-overlapping segmentation can avoid repeated statistics of friction features in the same time period, reduce computational redundancy, and at the same time ensure the independence of each signal sequence.

[0035] The preset cycle length needs to be compatible with the brake disc's rotational speed and sampling frequency. As a preferred implementation, the cycle length can be set to 0.1 seconds. This duration covers the complete friction characteristics of the brake disc rotating several revolutions while ensuring real-time monitoring resolution. If the total duration of the monitoring period is not an integer multiple of the segmented cycle length, the last segment of the signal sequence needs to be padded with zeros or truncated to ensure that all signal sequences have the same length. In this way, the continuous vibration signal of the monitoring period is processed into multiple signal sequences of consistent length and independent time.

[0036] Thus, multiple signal sequences divided into periods were obtained within the monitoring period.

[0037] Step S200: Calculate the instability index of each signal sequence.

[0038] It should be noted that when brake pads are under ideal friction conditions, the resulting vibration signal typically exhibits a stable, broadband random signal. However, when abnormalities occur on the brake pad surface, such as cracks, chipping, hard spots, or thermal degradation, the friction interface generates severe transient impacts or complex self-excited vibrations, leading to increased non-Gaussianity and complexity of the signal waveform. Kurtosis effectively measures the strength of the impact component in the signal waveform, while information entropy characterizes the complexity or uncertainty of the signal. Therefore, combining kurtosis and information entropy to construct an instability index can comprehensively assess the stability of the braking friction state reflected by each signal sequence.

[0039] Specifically, in order to calculate the instability index, a benchmark needs to be established first. This benchmark is obtained by acquiring the vibration signal of the brake pad during its historical normal operation under the same brake pad model and braking conditions as the current monitoring. Then, the vibration signal is divided into several historical signal sequences according to the same period length as in step S100.

[0040] Then, the kurtosis and information entropy of each historical signal sequence are calculated. To establish a reference upper limit for normalization, the maximum kurtosis among all signal sequences in the historical data is recorded as the historical maximum kurtosis. Simultaneously, the maximum value of the information entropy among all signal sequences in the historical data is recorded as the historical maximum information entropy. The calculation of kurtosis and information entropy of sequences are existing techniques and will not be elaborated upon here.

[0041] After obtaining historical benchmarks, the instability index for each segment of the monitoring period can be calculated. The logic behind this is to extract the statistical characteristics of the vibration signal sequence within the current period and select the feature values ​​that best reflect the degree of anomaly for comparison with historical peak levels. Therefore, the instability index satisfies the following relationship:

[0042] ;

[0043] in, It is the first The instability index of a signal sequence, It is the first The kurtosis of a signal sequence; It is the first Information entropy of a signal sequence; It is the maximum kurtosis of all signal sequences in the vibration signals of the historical period, and its value is greater than 0; It is the maximum information entropy of all signal sequences in the vibration signals of the historical period, and its value is greater than 0.

[0044] In this relation, Part will be the first The kurtosis of a signal sequence is normalized relative to the historical maximum kurtosis, reflecting the relative intensity of the signal's impulse. The larger this value, the more pronounced the oscillations of the signal sequence within the current period. The part will be the first Normalizing the information entropy of a signal sequence reflects the relative complexity of the signal. A larger value indicates a more complex vibration pattern within the current period; a smaller value indicates a simpler vibration pattern. This relationship, by multiplying these two normalization indices, yields a comprehensive instability index. This index reflects the stability of the braking process; a larger value indicates a more unstable braking process.

[0045] Thus, an instability index that can characterize the steady state of each signal sequence has been obtained.

[0046] Step S300: Determine the number of decomposition layers for wavelet transform of each signal sequence using the instability index.

[0047] It should be noted that wavelet transform, as a classic time-frequency analysis tool, can decompose time-domain signals into components of different frequency scales. It not only preserves the time information of the signal, but also accurately divides the frequency range. It can simultaneously take into account the time-frequency localization characteristics and effectively capture the frequency components in the signal that change with time.

[0048] The number of decomposition levels in wavelet transform is the core parameter that determines the frequency resolution of signal analysis: the more decomposition levels, the finer the frequency division, and the stronger the ability to capture high-frequency detail components; the fewer decomposition levels, the coarser the frequency division, and the more suitable it is for focusing on low-frequency fundamental features.

[0049] Specifically, for the steady vibration signal generated when the brake pads are in a stable braking state, its energy is mainly concentrated in the low-frequency part reflecting changes in braking pressure. In this case, a smaller number of decomposition levels is sufficient to extract effective features and improve computational efficiency. However, when the braking system experiences abnormalities such as brake pad thermal fade, surface glazing, or physical damage, the friction interface will generate nonlinear abnormal vibrations. The vibration signal will contain rich high-frequency noise components or modulation sidebands. This type of feature information is often hidden in high-frequency details and requires deeper decomposition for accurate capture. Therefore, the larger the signal instability index, i.e., the more unstable the signal, the more wavelet transform decomposition levels are required, and the two are positively correlated. This invention adaptively adjusts the number of wavelet transform decomposition levels according to the signal sequence instability index, enabling adaptive analysis of signals under different braking states, ultimately improving the effectiveness and specificity of abnormal feature extraction during brake pad operation.

[0050] First, a minimum and maximum decomposition level need to be preset. For example, based on experience, the minimum decomposition level can be set to 3 and the maximum decomposition level to 6. Then, the instability index of all signal sequences in the historical running data needs to be obtained, and their minimum and maximum values ​​need to be determined as the benchmark for the range of variation of the instability index.

[0051] Based on the above logic, the first Number of wavelet transform decomposition levels for a signal sequence Satisfying the relation:

[0052] ;

[0053] in, It is the first The instability index of a signal sequence; It is the minimum instability index of all signal sequences in the vibration signals of a historical period; It is the maximum instability index of all signal sequences in the vibration signals of a historical period; It is the preset minimum number of decomposition levels; This is the preset maximum number of decomposition levels; It is a rounding function.

[0054] In this relation, the second part For the first The instability index of each signal sequence was min-max normalized, and its value was mapped to... Interval. An adjustable range for the number of decomposition levels is set. In this relationship, the normalized instability index is multiplied by this range and then added to the minimum number of decomposition levels to achieve the desired result. and A linear mapping between them. When The larger it is, the more you get The larger the value, the smaller the value, and vice versa, thus achieving adaptive determination of the number of decomposition layers.

[0055] Thus, the number of wavelet transform decomposition levels matching the instability of each signal sequence has been determined.

[0056] Step S400: Determine the position of the target basis function in the preset basis function library based on the instability index, and select the target basis function from the preset basis function library according to the position.

[0057] It should be noted that wavelet transform is based on wavelet basis functions. Different wavelet basis functions have different responses to signal features of different shapes. For example, smooth basis functions are more suitable for processing stationary signal sequences, while basis functions with sharp shapes are more suitable for processing signal sequences with abrupt changes. Therefore, this invention selects the most suitable wavelet basis function based on the instability index of each signal sequence.

[0058] Specifically, the first step is to construct a pre-defined basis function library. This library contains various commonly used wavelet basis functions, such as Coiflet wavelet, Dobessie wavelet, Haar wavelet, and Symlet wavelet basis functions. Implementers can add basis functions to the library as needed. To achieve adaptive selection, the basis functions in the library need to be arranged in a pre-defined order according to their waveform complexity. As a preferred implementation, firstly, the waveform of each basis function in the pre-defined basis function library is obtained, and an analog signal sequence completely identical to that waveform is generated. Next, the information entropy of each analog signal sequence is calculated. Then, based on the calculated information entropy, all basis functions in the pre-defined basis function library are arranged in descending order of information entropy.

[0059] After the preset basis function library is constructed and sorted, the process of selecting the target basis function corresponding to the signal sequence is also based on the instability index. Similar to the logic of determining the number of decomposition layers, the instability index is mapped to a specific location in the preset basis function library.

[0060] Based on the above logic, the first The position of the target basis function corresponding to each signal sequence in the preset basis function library Satisfying the relation:

[0061] ;

[0062] in, It is the first The instability index of a signal sequence; , These are the minimum and maximum instability indices of all signal sequences in the vibration signals during historical periods, respectively. It is the total number of base functions in the preset base function library; It is the round-up symbol. It is a minimum value function.

[0063] In this relation, the maximum-minimum normalization is used to... Mapped to The interval, multiplied by the total number of basis functions in the preset basis function library. This maps the relative magnitude of the instability index to a position in a predefined basis function library. A smaller instability index corresponds to a smaller position, allowing for the selection of a basis function with a smooth waveform to suit a stable braking signal; conversely, a larger instability index corresponds to a larger position, allowing for the selection of a basis function with a more complex waveform suitable for detecting transient shocks to suit abnormal braking signals.

[0064] At this point, a target basis function has been selected from the preset basis function library for each signal sequence.

[0065] Step S500: Perform wavelet transform on each signal sequence using the decomposition level and target basis function to obtain the wavelet transform results of each signal sequence; determine the temporal characteristics of the state of each signal sequence based on the wavelet transform results; determine outliers based on the changes of the temporal characteristics between adjacent signal sequences; when the outlier is greater than a set threshold, the response is an abnormal operating state.

[0066] It should be noted that the signal sequences generated by brake pads are different under different braking conditions. In order to fully extract the features of the signal sequences under different friction states for monitoring the operating status of brake pads, this invention performs wavelet transform on each signal sequence by determining the target basis function and the number of decomposition layers.

[0067] Specifically, wavelet transform is used to process each signal sequence, employing the target basis function and decomposition level obtained in the steps described above. After performing the wavelet transform, the components of the signal sequence at different frequency scales are obtained, and these components constitute the wavelet transform result of the signal sequence. Wavelet transform is an existing technology; this invention only optimizes the logic for obtaining the decomposition level and target basis function, without any other modifications, which will not be elaborated upon here.

[0068] It should be noted that, as mentioned earlier, the vibration signal of the monitored period is segmented without overlap using a preset period length, resulting in multiple signal sequences. During stable braking, the signal characteristics within adjacent time periods typically exhibit strong continuity and similarity. However, when the brake pad's operating state becomes abnormal, this continuity is disrupted, leading to differences in the signal characteristics of adjacent periods. Therefore, this invention constructs an outlier index based on the differences between the wavelet transform results of signal sequences from adjacent periods, for real-time monitoring of the brake pad's health status.

[0069] Specifically, for the wavelet transform result of each signal sequence, firstly, the energy values ​​of all its components are obtained, and these energy values ​​are arranged in order from low frequency to high frequency to form an energy sequence. The acquisition of component energy values ​​is existing technology and will not be elaborated upon here. Then, based on the energy sequence at each time point, the outlier value of the brake pad operating state is obtained. This outlier value satisfies the following relationship:

[0070] ;

[0071] in, It is the [number]th period to be monitored Abnormal values ​​of brake pad operating status in a signal sequence; , It is the [number]th period to be monitored The, the The energy sequence corresponding to each signal sequence; It is the Dynamic Time Warped Distance Function. The DTW algorithm uses dynamic programming to find the optimal matching path between two time series and calculates the cumulative distance on the path to evaluate their similarity. It is an existing technology and will not be elaborated on here.

[0072] In this relationship, when the braking friction state is normal and stable, the energy distribution of adjacent period signal sequences should be relatively consistent, thus the DTW distance between them is small; when the braking friction state undergoes abnormal changes, such as brake pad glazing leading to a change in vibration mode, the energy distribution of adjacent period signal sequences should be relatively consistent, thus the DTW distance between them is small. The, the The difference in energy sequences corresponding to each signal sequence will increase, and the DTW between them will also increase. Therefore, the larger the outlier, the more likely the brake pad's operating state is to deviate from normal.

[0073] It should be noted that after obtaining an outlier, the system also needs to perform an anomaly response. This invention compares the calculated outlier with a preset threshold. If the current outlier is greater than this threshold, the system initially determines that the current brake pad's operating state is abnormal. The threshold can be determined through statistical analysis of historical data, for example, set as the 95th percentile of the historical outlier distribution. Implementers can also set it according to their needs. Figure 2 The figure shows a visualization of the calculation results of the signal sequence outliers during the monitoring period. In the figure, the horizontal axis is the period length in seconds, representing the period division of the signal sequence; the vertical axis is the outlier calculated from the signal sequence within the corresponding period. The solid line is the distribution trend of the outlier in the signal sequence within different periods, the dashed line is the preset threshold, and the dots are the outliers where the outlier is greater than the threshold, indicating that there is an abnormality in the brake pad operation at the corresponding time.

[0074] As a preferred implementation, to improve monitoring reliability and avoid false alarms caused by single-point noise, a continuous judgment logic can be further set. For example, when the abnormal values ​​during the monitoring period of a preset duration are all greater than the preset threshold, the system confirms that the brake pads are in a continuous abnormal state and outputs a control command to stop the braking inertia test bench, thereby avoiding damage to the test equipment or preserving fault samples for analysis. The preset duration can be set to 10 seconds, which means that if the abnormal value calculated for each signal sequence is greater than the preset threshold within 10 seconds, the current operating state of the brake pads is considered abnormal.

[0075] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring and early warning of brake pad operating status based on data analysis, characterized in that, include: The vibration signal of the brake pad braking during the period to be monitored is acquired and segmented to obtain multiple signal sequences; Calculate the instability index of each signal sequence The instability index is positively correlated with both the kurtosis and information entropy of the corresponding signal sequence; The instability index is used to determine the number of wavelet transform decomposition levels for each signal sequence. , ; It is the minimum instability index of all signal sequences in the vibration signals of a historical period; It is the maximum instability index of all signal sequences in the vibration signals of a historical period; It is the preset minimum number of decomposition levels; This is the preset maximum number of decomposition levels; It is a rounding function, and the number of decomposition levels is positively correlated with the instability index; ;in, It is the first The kurtosis of a signal sequence; It is the first Information entropy of a signal sequence; It is the maximum kurtosis of all signal sequences in the vibration signals of the historical period; It is the maximum information entropy of all signal sequences in the vibration signals of a historical period; The position of the target basis function in the preset basis function library is determined based on the instability index. ; It is the total number of base functions in the preset base function library; It is the round-up symbol. It is a minimum value function, and the target basis function is selected from a preset basis function library according to its position; the basis functions in the preset basis function library are arranged in a preset order according to their waveform complexity; Wavelet transform is performed on each signal sequence using the decomposition level and objective basis function to obtain the wavelet transform results for each signal sequence. The temporal characteristics of the state of each signal sequence are determined based on the wavelet transform results, and outliers are determined based on the changes in the temporal characteristics between adjacent signal sequences. This includes: obtaining the energy values ​​of all components in the wavelet transform results of each signal sequence to form an energy sequence; recording the signal sequence at the analysis time as the current signal sequence; calculating the dynamic time warping distance between the current signal sequence and the energy sequence of the previous signal sequence in the monitoring period; recording the dynamic time warping distance as the outlier in the monitoring period; and responding to an abnormal operating state when the outlier exceeds a set threshold.

2. The method for real-time monitoring and early warning of brake pad operating status based on data analysis according to claim 1, characterized in that, The Methods for obtaining [the information] include: Vibration signals of brake pads during normal braking processes were acquired and divided into several historical signal sequences. Calculate the kurtosis of each historical signal sequence to obtain a historical kurtosis set, and denote the maximum value in the historical kurtosis set as . ; The The method for obtaining the information entropy includes: calculating the information entropy of each historical signal sequence to obtain a set of historical information entropy, and taking the maximum value in the set of historical information entropy as the information entropy value. .

3. The method for real-time monitoring and early warning of brake pad operating status based on data analysis according to claim 1, characterized in that, The basis functions in the preset basis function library are arranged in a preset order according to their waveform complexity, including: Obtain the waveform of each basis function in the preset basis function library, and generate an analog signal sequence consistent with the waveform; Calculate the information entropy of each of the analog signal sequences, and arrange all basis functions in the preset basis function library in descending order of information entropy based on the information entropy.

4. The method for real-time monitoring and early warning of brake pad operating status based on data analysis according to claim 1, characterized in that, The response when the abnormal value exceeds a set threshold is an abnormal operating state, including: When all abnormal values ​​during the monitoring period of the preset duration exceed the set threshold, an early warning signal is output.

5. The method for real-time monitoring and early warning of brake pad operating status based on data analysis according to claim 1, characterized in that, The process of segmenting the signal into multiple signal sequences includes: The vibration signal of the period to be monitored is segmented without overlap using a preset period length.

6. The method for real-time monitoring and early warning of brake pad operating status based on data analysis according to claim 1, characterized in that, The preset basis function library includes Coiflet wavelet basis functions, Dobessie wavelet basis functions, Haar wavelet basis functions, and Symlet wavelet basis functions.

Images