A method for detecting wear of a brake pad of a vehicle

Abstract

The application provides a kind of automobile brake pad wear detection method, it is related to brake pad detection technical field.Brake pad is carried out brake test, collects the surface temperature distribution graph after brake pad braking and temperature-time sequence of cooling stage;According to the temperature-time sequence of cooling stage, exponential decay module is fitted to obtain thermal inertia attenuation coefficient;Surface temperature distribution graph is carried out feature extraction, and feature data is obtained;Characteristic data and thermal inertia attenuation coefficient are carried out weighting processing, and input data is obtained;Random forest network model is constructed, random network model is optimized using bitter fish optimization algorithm, and input experimental data is trained, and optimal network model is obtained;Input data is input into optimal network model, and detection result is obtained.The application is optimized and trained by constructing random forest network model to network model, obtains optimal network model, and input data is input therein, improves the accuracy of brake pad detection, and reduces the security risk.

Description

Technical Field

[0001] The present invention relates to the technical field of brake pad detection, and in particular to a method for detecting wear of automobile brake pads. Background Art

[0002] With the development of the automotive industry, the braking system, a critical component of vehicle safety, has become increasingly important, and its performance and reliability directly impact driving safety. Brake pads, a crucial component of the braking system, play a crucial role in the vehicle's braking performance, driving stability, and safety. Therefore, timely and accurate detection of brake pad wear has become a crucial research topic for ensuring vehicle safety. The primary function of brake pads is to generate braking force through friction with the brake disc, slowing or stopping the vehicle. Their primary function is to generate friction with the brake disc during braking, converting kinetic energy into heat, thereby slowing or stopping the vehicle. Brake pads gradually wear over time, which not only affects braking performance but can also pose a serious safety threat. Different driving habits significantly affect the degree of brake pad wear. This wear process not only impairs braking performance but can also cause braking noise and vibration, and even damage other components of the braking system. Promptly replacing severely worn brake pads ensures the proper functioning of the braking system, thereby improving driving safety, preventing wear-related degradation of braking performance, and reducing traffic accidents caused by brake failure.

[0003] Brake pad inspection usually relies on manual inspection, which is prone to misjudgment and leads to failure to replace the brake pads in time, affecting braking performance. Summary of the Invention

[0004] The present invention provides a method for detecting wear of automobile brake pads, which is used to solve the problem that manual inspection is required during the existing technology for detecting brake pads, which is prone to misjudgment and affects the braking performance.

[0005] In one aspect, the present invention provides a method for detecting wear of an automobile brake pad, comprising the following steps: performing a braking test on the brake pad, collecting a surface temperature distribution diagram of the brake pad after braking and a temperature-time series during a cooling stage;

[0006] The thermal inertia attenuation coefficient is obtained by fitting an exponential decay model according to the temperature-time series of the cooling stage;

[0007] Performing feature extraction on the surface temperature distribution map to obtain feature data; performing weighted processing on the feature data and the thermal inertia attenuation coefficient to obtain input data;

[0008] Constructing a random forest network model, optimizing the random forest network model using a bitterfish optimization algorithm, and inputting experimental data for training to obtain an optimal network model;

[0009] Input the input data into the optimal network model to obtain the detection results.

[0010] According to a method for detecting wear of a vehicle brake pad provided by the present invention, the specific steps of performing a braking test on the brake pad include:

[0011] Clean the brake pad to remove surface grease;

[0012] Set the target speed, drive the test bench to the target speed, apply braking pressure to the brake pad, and keep the deceleration of the test bench constant until the speed drops to zero.

[0013] According to a method for detecting wear of an automobile brake pad provided by the present invention, the specific steps of obtaining the thermal inertia attenuation coefficient include:

[0014] Smoothing the temperature-time series data in the cooling stage using a sliding window method to obtain smoothed data;

[0015] The thermal inertia attenuation coefficient was calculated by fitting the smoothed data using nonlinear least squares.

[0016] According to a method for detecting wear of automobile brake pads provided by the present invention, the specific steps of extracting features from the surface temperature distribution map include:

[0017] Extracting the temperature data along the width of the brake pad from the temperature distribution graph, finding the maximum and minimum temperature values, and calculating the maximum temperature difference;

[0018] extracting temperature data along the friction contact direction of the brake pad from the temperature distribution diagram, drawing a line along the friction contact direction of the brake pad, and recording temperature values ​​at different positions of the line to obtain a temperature sequence, and then fitting the temperature sequence using a regression method to obtain a temperature decay rate along the friction contact direction of the brake pad;

[0019] The average temperature of the temperature distribution graph is calculated, a temperature threshold is set, and then temperature points with temperatures above the average temperature are filtered out based on the threshold. The areas of the temperature points are measured and summarized to obtain the total area of ​​the temperature points, and the proportion of areas with temperatures above the average temperature is calculated.

[0020] According to a method for detecting wear of automobile brake pads provided by the present invention, the weighted processing specifically comprises the following steps:

[0021] Normalizing the thermal inertia attenuation coefficient and the characteristic data to obtain normalized data;

[0022] The weight distribution is set according to actual experience, and the normalized data is weighted to obtain input data.

[0023] According to a method for detecting automobile brake pad wear provided by the present invention, the specific steps of constructing a random forest network model include:

[0024] Construct multiple decision trees, and set the maximum depth of the multiple decision trees and the minimum sample input required for the classification point to obtain an initial network model;

[0025] The initial network is trained using the experimental data to obtain a random forest network model.

[0026] According to a method for detecting wear of automobile brake pads provided by the present invention, the specific steps of acquiring experimental data include:

[0027] A braking experiment was conducted using different brake pads, and the thermal inertia attenuation coefficients of the different brake pads, the maximum temperature difference in the width direction of the brake pad, the temperature attenuation rate in the friction contact direction of the brake pad, the proportion of areas with temperatures higher than the average temperature, and the corresponding wear levels and abnormality types were recorded to obtain data 1; the data 1 was preprocessed to obtain experimental data.

[0028] According to a method for detecting wear of automobile brake pads provided by the present invention, the specific steps of preprocessing the data 1 include:

[0029] Cleaning the data one, removing outliers and duplicate values ​​in the data one, and replacing missing values ​​therein with the mean to obtain cleaned data;

[0030] The cleaning data is normalized to obtain experimental data.

[0031] According to a method for detecting wear of automobile brake pads provided by the present invention, the specific steps of optimizing the random forest network using the bitter fish optimization algorithm include:

[0032] Use mean square error as the fitness function of the random forest network model;

[0033] Initialize the population and randomly generate the positions of individual bitter fish. Each individual bitter fish represents a hyperparameter group of the random forest network model.

[0034] Set up an iterative mechanism. Each time a new population is formed, the perception range and hunger step threshold of the individual bitter fish are set according to the environmental adaptation mechanism.

[0035] Calculate the fitness function of the individual bitter fish in the initialized population, and update the position of each individual bitter fish according to the behavioral strategy to obtain population one;

[0036] The fitness value of each individual in the population is calculated. If the output fitness value is higher than the preset fitness threshold, the optimal hyperparameter is output. Otherwise, the system continues to iterate until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached to obtain the optimal hyperparameter.

[0037] According to a method for detecting wear of automobile brake pads provided by the present invention, the behavior strategy includes:

[0038] The individual bitter fish compares its own fitness with the average fitness value of the individual bitter fish within the sensing range. If the fitness value is better than the average fitness value of the individual bitter fish within the sensing range, the position of the individual bitter fish is updated according to local foraging.

[0039] If the number of hungry steps of the bitter fish individual reaches the hunger step threshold, the bitter fish individual will update its position according to random migration;

[0040] If the individual bitter fish does not meet the conditions of local foraging and random migration, the individual bitter fish will update their individual position through group collaboration. When all the individual bitter fish have updated their positions, we get population one.

[0041] The present invention provides a method for detecting automobile brake pad wear. This method constructs a random forest network model, optimizes it using the bitterfish optimization algorithm, and trains it using experimental data to obtain an optimal network model. The experimentally obtained thermal inertia attenuation coefficient and characteristic data are then weighted to obtain input data, which is then fed into the optimal network model to obtain detection results. This method improves brake pad detection accuracy and reduces safety hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Fig. 1 The present invention provides a flow chart of a method for detecting wear of automobile brake pads.

[0044] Fig. 2 This is a diagram showing the specific steps for extracting features from a surface temperature distribution diagram of a method for detecting wear of an automobile brake pad provided by an embodiment of the present invention;

[0045] Fig. 3 These are specific steps for optimizing a random forest network using a bitterfish optimization algorithm in a method for detecting automobile brake pad wear provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0047] See also Figs. 1-3 , an embodiment of the present invention provides a method for detecting wear of a vehicle brake pad, the method comprising:

[0048] A braking experiment was performed on the brake pad, and a surface temperature distribution diagram of the brake pad after braking and a temperature-time series in the cooling stage were collected.

[0049] In this embodiment, after the brake pad has completed braking, a refined temperature monitoring process can be further implemented. Specifically, a high-resolution infrared thermal imager can be used to perform non-contact temperature field scanning of the brake pad surface, generating a two-dimensional temperature distribution map. A data logger can be used to record the temperature changes of the brake pad during the cooling phase. Timed sampling can be set to record temperature values ​​at regular intervals to obtain detailed sequential data on temperature changes over time.

[0050] The specific steps of performing a brake test on the brake pad include:

[0051] Clean the brake pad to remove surface oil and dirt.

[0052] In this embodiment, to clean the brake pads, you can use a clean, lint-free, soft cloth or brush to gently wipe the surface of the brake pad. This step primarily removes dust and loose dirt. Be careful when wiping, and avoid applying excessive force to avoid damaging the brake pad surface. This helps maintain the brake pad's performance and extend its lifespan. Alternatively, you can use an ultrasonic cleaner. Place the brake pads in the cleaner and add an appropriate amount of water and detergent. Turn on the ultrasonic cleaning machine for several minutes, allowing the ultrasonic vibrations to act on the brake pads, removing grease and dirt. This method is particularly suitable for cleaning hard-to-reach areas. After cleaning, remove the brake pads and rinse them with clean water to ensure all detergent and residue are removed. Finally, thoroughly dry the brake pads with a dry cloth to prevent any residual moisture. Alternatively, use a pneumatic spray gun or air pump to gently blow the brake pad surface to remove dust and particulate matter. Alternatively, spray brake pad cleaner on the brake pad surface, apply evenly, and allow it to soak for several minutes to fully break down grease and dirt. When using a cleaner, avoid oil-based detergents to avoid affecting braking performance.

[0053] A target speed is set, the test bench is driven to the target speed, a brake pressure is applied to the brake pad, and the test bench deceleration is kept constant until the speed is reduced to zero.

[0054] In this embodiment, the target speed can be set to 60 km / h, and the brake pressure is applied by the brake system. According to the design of the brake system, a predetermined brake pressure value can be set. Ensure that the applied brake pressure is appropriate to effectively reduce the test bench speed under the condition of keeping constant deceleration. The value of this brake pressure should be set in advance according to the design specification of the brake system and the experimental requirements, and its appropriateness should be ensured to avoid the undesirable brake effect caused by too high or too low. When the test bench reaches the set target speed, the predetermined brake pressure is immediately applied. The brake pressure can be selected according to the experimental design requirements, or it can be adjusted by the brake control system.

[0055] The thermal inertia decay coefficient is calculated by fitting an exponential decay model to the temperature-time sequence of the cooling phase.

[0056] The specific steps for obtaining the thermal inertia decay coefficient include:

[0057] The data in the temperature-time sequence of the cooling phase is smoothed using the sliding window method to obtain smoothed data.

[0058] In this embodiment, the window size can be selected as 5. The selection of window size will affect the smoothing effect, a larger window will result in smoother results, but some details may be lost. Starting from the first data point of the temperature-time sequence, set the position of the current window. Calculate the average value of the temperature values in the current window. For points at the front end of the window, the actual available data points can be used for average calculation. Add the calculated average value to the new list. Move the window one data point and repeat the average calculation process until the entire temperature data sequence is traversed. At the beginning and end of the sequence, due to insufficient data points to fill the entire window, the following two strategies can be used: for points at the front end of the window, all available data points can be used for averaging; at the back end of the window, use the actual available data points. At the front and back ends of the window, fill with fixed values to ensure that the length of the output list is consistent with the original data.

[0059] The thermal inertia decay coefficient is calculated by fitting the smoothed data using nonlinear least squares.

[0060] In this embodiment, the nonlinear least squares method is used for fitting. In the fitting process, the optimization goal is to minimize the error between the fitting model and the actual observed temperature. Through iterative calculation, the model parameters are adjusted until the best fitting effect is found. The thermal inertia decay coefficient is obtained in the model parameters.

[0061] The surface temperature distribution map is subjected to feature extraction to obtain feature data, and the feature data and the thermal inertia attenuation coefficient are subjected to weighted processing to obtain input data.

[0062] The specific steps of extracting features from the surface temperature distribution map include:

[0063] Extract the temperature data along the width of the brake pad from the temperature distribution graph, find the maximum and minimum temperature values, and calculate the maximum temperature difference.

[0064] In this example, the desired width direction is determined, typically the centerline of the brake pad or a specific width position. All temperature data along this direction is then extracted. Within the extracted temperature data array, the maximum and minimum temperature values ​​are found using a simple traversal or built-in functions.

[0065] Temperature data along the friction contact direction of the brake pad is extracted from the temperature distribution diagram, a line is drawn along the friction contact direction of the brake pad, and the temperature values ​​at different positions of the line are recorded to obtain a temperature sequence. The temperature sequence is then fitted using a regression method to obtain the temperature attenuation rate along the friction contact direction of the brake pad.

[0066] In this embodiment, the temperature decay rate is expressed as follows:

[0067]

[0068] in, is the temperature decay rate, is the temperature change value, is the distance change value.

[0069] The average temperature of the temperature distribution graph is calculated, a temperature threshold is set, and then temperature points with temperatures above the average temperature are filtered out based on the threshold. The areas of the temperature points are measured and summarized to obtain the total area of ​​the temperature points, and the proportion of areas with temperatures above the average temperature is calculated.

[0070] In this implementation, the expression formula for the average temperature is:

[0071]

[0072] in, is the average temperature, is the total number of temperature points, For the The temperature value of the point.

[0073] The specific steps of weighted processing include:

[0074] The thermal inertia attenuation coefficient and the characteristic data are normalized to obtain normalized data.

[0075] In this embodiment, a minimum-maximum normalization method may be used, and the expression formula is:

[0076]

[0077] in, is the normalized data, is the original data point, is the minimum value in the original data set, is the maximum value in the original data set.

[0078] The weight distribution is set according to actual experience, and the normalized data is weighted to obtain input data.

[0079] A random forest network model was constructed, and the bitter fish optimization algorithm was used to optimize the random forest network model. Experimental data was input for training to obtain the optimal network model.

[0080] The specific steps to build a random forest network model include:

[0081] Construct multiple decision trees, set the maximum depth of the decision trees and the minimum sample input required for the classification point, and obtain the initial network model.

[0082] In this embodiment, to prevent overfitting, the maximum depth of each decision tree can be set to 10. This means that when constructing a decision tree, further node splitting will cease when the tree depth reaches 10. The depth of a decision tree determines the number of data feature combinations it can fit. When the depth is too large, the tree may learn noise or accidental patterns in the training data, leading to overfitting. By limiting the maximum depth, the model is forced to ignore high-order feature interactions and instead focus on more generalizable low-order features. A shallower tree structure means fewer nodes and splitting conditions, significantly reducing the computational complexity of the training and prediction phases. Each node must contain at least five samples when splitting. This mandates that splitting operations must be based on sufficient data support to avoid statistical instability caused by too few samples. By requiring a minimum number of samples, the model is forced to make splitting decisions based on a wider range of samples, thereby learning more representative feature patterns. The minimum sample limit and the maximum depth jointly constrain the growth of the tree. Even if the maximum depth allows the tree to continue growing, if the number of samples at a node is less than 5, the split will terminate prematurely. This dual restriction ensures that the tree structure is neither too deep nor blindly split when the sample size is insufficient. By limiting the maximum depth and minimum number of samples, the model avoids overfitting the noise in the training data and instead focuses on the essential patterns in the data.

[0083] The initial network is trained using the experimental data to obtain a random forest network model.

[0084] The specific steps of obtaining the experimental data include:

[0085] Braking tests were conducted using different brake pads, and the thermal inertia attenuation coefficients, maximum temperature differences in the width direction of the brake pads, temperature attenuation rates in the friction contact direction of the brake pads, percentages of areas with temperatures above the average temperature, and corresponding wear levels and abnormality types of the different brake pads were recorded to obtain Data 1. Data 1 was preprocessed to obtain experimental data.

[0086] In this example, braking tests were conducted on different brake pads one by one. After each test, the performance of each brake pad was recorded, along with temperature changes at specific points in time. Multiple temperature sensors were used to measure the temperature of the brake pad at different locations. These sensors covered the width of the brake pad and the friction contact direction, allowing for comprehensive temperature distribution data.

[0087] The specific steps of preprocessing the data include:

[0088] The data one is cleaned to remove outliers and duplicate values ​​in the data one, and the mean is used to replace the missing values ​​therein to obtain cleaned data.

[0089] In this embodiment, by drawing a box plot, outliers in the data set that are outside the upper and lower quartiles can be intuitively identified. These points are usually considered extreme values. The Z-score of each data point can also be calculated. Usually, a threshold is set. If the Z-score of a data point exceeds the set range, it can be considered an outlier. Use the unique identifier of each data in the data set to determine which data are duplicates. Based on these identifiers, duplicate entries are removed to ensure that each record is unique. During the cleaning process, the mean is used to replace missing data. The mean is obtained by calculating the average of the valid values ​​in the cleaned data set. Replacing missing values ​​with the mean can effectively reduce the deviation caused by missing values. When integrating and cleaning data, ensure the consistency of the format of all data and unify the temperature values ​​in degrees Celsius to facilitate subsequent analysis. At the same time, check the correctness of the numerical type and ensure that all numbers are stored as numerical types, not text or other types. Generate a new cleaned data set that does not contain outliers and duplicates, and the missing values ​​are filled with the mean, and all data formats are consistent.

[0090] The cleaning data is normalized to obtain experimental data.

[0091] In this embodiment, the cleaned data can be normalized using min-max normalization. The normalized data is converted to the same scale, eliminating the impact of dimensional inconsistency between different features. In deep learning models, normalization can make the loss function smoother, thereby accelerating the learning process. Especially when processing extreme value data, the original features can cause numerical stability problems. By scaling the data to between 0 and 1, the stability of numerical operations can be improved. This helps to avoid overfitting of the model. By limiting the training data to the same range, the generalization ability of the model on unknown data can be improved.

[0092] The specific steps of the Bitter Fish Optimization Algorithm for optimizing the random forest network include:

[0093] The mean square error is used as the fitness function of the random forest network model.

[0094] In this embodiment, the mean square error is expressed as:

[0095]

[0096] in, is the mean square error, is the actual value, i.e. The true output in samples, is the predicted value, i.e., the model The predicted output on samples.

[0097] Initialize the population and randomly generate individual positions of bitter fish. Each bitter fish individual represents a hyperparameter group of the random forest network model.

[0098] An iterative mechanism is set up. Each time a new population is formed, the perception range and hunger step threshold of the individual bitter fish are set according to the environmental adaptation mechanism.

[0099] In this embodiment, the environmental adaptation mechanism is to gradually narrow the perception range of individual bitter fish as the number of iterations increases. This improves the selectivity of individual bitter fish in resource-scarce situations and improves their efficiency in surviving in the environment. The perception range can be set to decrease by a certain percentage each iteration, so that individuals can still effectively capture resources when adapting to a smaller resource area. In each iteration, the hunger step threshold is gradually increased. This means that when searching for food, individual bitter fish will increase the number of steps they move to ensure longer-term activity in a resource-scarce environment. An increasing ratio can be set to extend the time individuals spend swimming in the environment.

[0100] Calculate the fitness function of the individual bitter fish in the initialized population, and update the position of each individual bitter fish according to the behavioral strategy to obtain population one.

[0101] The fitness value of each individual in the population is calculated, and if the output fitness value is higher than the preset fitness threshold, the optimal hyperparameter is output, otherwise, the iteration continues until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached, and the optimal hyperparameter is obtained.

[0102] The behavior strategy includes:

[0103] The bitterfish individual compares its fitness with the average fitness value of the bitterfish individuals within its perception range, and if it is better than the average fitness value of the bitterfish individuals within its perception range, the bitterfish individual updates its position according to local foraging.

[0104] In this embodiment, the local foraging expression formula is:

[0105]

[0106] wherein, is the updated position of the bitterfish individual, is the current position of the bitterfish individual, is a step factor, is a random number between 0 and 1, is the position of the individual with the optimal fitness within the perception range of the bitterfish individual.

[0107] If the hunger step number of the bitterfish individual reaches the hunger step threshold, the bitterfish individual updates its position according to random migration.

[0108] In this embodiment, the random migration expression formula is:

[0109]

[0110] wherein, is the updated position of the bitterfish individual, and are the upper and lower bounds of the search space, is a random number between 0 and 1.

[0111] If the bitterfish individual does not meet the conditions of local foraging and random migration, the bitterfish individual updates its position through group cooperation. After the positions of all bitterfish individuals are updated, the population one is obtained.

[0112] In this embodiment, the group cooperation expression formula is:

[0113]

[0114] wherein, is the updated position of the bitterfish individual, is a cooperation factor, is a random number between 0 and 1, is the historical global optimal position, For the bitter fish in the individual's current location.

[0115] Input the input data into the optimal network model to obtain the detection results.

[0116] Example 1: Use a cleaner and a brush to remove oil, dirt, and dust from the brake pad surface. Set the test bench to 60 km / h. After reaching the target speed, apply brake pressure and maintain a deceleration of -5 m / s² until the speed drops to zero. Record the surface temperature distribution of the brake pad after braking and the temperature-time series during the cooling phase. Smooth the collected temperature-time series using a sliding window method with a window size of 5 seconds. Apply a nonlinear least squares fit to the smoothed data to obtain a thermal inertia attenuation coefficient of 0.85. Extract temperature data along the width of the temperature distribution, finding the maximum temperature of 350°C and the minimum temperature of 200°C. Calculate the maximum temperature difference to be 150°C. Extract temperature data along the friction contact direction, plot the temperature series, and apply regression to obtain a temperature attenuation rate of 0.5°C / m. Calculate the average value of the temperature distribution (275°C). Set a temperature threshold of 300°C to exclude points with temperatures above this threshold. Measure the total area of ​​the temperature points and calculate the area percentage as 20%. The thermal inertia attenuation coefficient was 0.85 and the feature data were normalized. After normalization, the thermal inertia attenuation coefficient was 0.5 and the feature data was 0.6. The weights were set according to experience, with the thermal inertia attenuation coefficient weight being 0.4 and the feature data weight being 0.6. After weighted processing, the input data was 0.54. Ten decision trees were constructed, with a maximum depth of 10 and a minimum number of sample splits of 5. The initial network was trained using the collected experimental data to obtain a random forest network model. The experimental data was cleaned to remove outliers, records with temperature values ​​exceeding 500°C, and duplicate values. The missing value 0.7 was replaced by the mean value, which was 0.75 of the feature. The cleaned data was normalized to obtain the final experimental data. The mean square error was used as the fitness function, and the population was initialized to generate 10 bitter fish individuals, each representing a set of hyperparameters. An iterative mechanism is set up to calculate the fitness of each individual and update the position until a fitness value is found that is higher than the preset threshold (0.95) or the maximum number of iterations (100) is reached to obtain the optimal hyperparameters.

[0117] The present invention constructs a random forest network model, optimizes it using the bitterfish optimization algorithm, and inputs experimental data for training to obtain an optimal network model. The thermal inertia attenuation coefficient and characteristic data obtained from the brake pad experiment are weighted to obtain input data, which is then fed into the optimal network model to obtain a test result. This improves the accuracy of brake pad testing and reduces safety hazards.

[0118] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0119] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for detecting wear of automobile brake pads, characterized in that: include: Perform a braking test on the brake pad, and collect a surface temperature distribution diagram of the brake pad after braking and a temperature-time series during the cooling stage; The thermal inertia attenuation coefficient is obtained by fitting an exponential decay model according to the temperature-time series of the cooling stage; Extracting features from the surface temperature distribution map to obtain feature data, wherein the feature data includes a maximum temperature difference, a temperature attenuation rate, and a region ratio; and performing weighted processing on the feature data and a thermal inertia attenuation coefficient to obtain input data. Constructing a random forest network model, optimizing the random forest network model using a bitterfish optimization algorithm, and inputting experimental data for training to obtain an optimal network model; The specific steps of the Bitter Fish Optimization Algorithm for optimizing the random forest network include: Use mean square error as the fitness function of the random forest network model; Initialize the population and randomly generate the positions of individual bitter fish. Each individual bitter fish represents a hyperparameter group of the random forest network model. Set up an iterative mechanism. Each time a new population is formed, the perception range and hunger step threshold of the individual bitter fish are set according to the environmental adaptation mechanism. Calculate the fitness function of the individual bitter fish in the initialized population, and update the position of each individual bitter fish according to the behavioral strategy to obtain population one; The fitness value of each individual in the population is calculated. If the output fitness value is higher than a preset fitness threshold, the optimal hyperparameter is output. Otherwise, the system continues to iterate until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached, and the optimal hyperparameter is obtained. The behavioral strategies include: The individual bitter fish compares its own fitness with the average fitness value of the individual bitter fish within the sensing range. If the fitness value is better than the average fitness value of the individual bitter fish within the sensing range, the position of the individual bitter fish is updated according to local foraging. The local foraging expression formula is: in, The updated position of the bitter fish individual, For the bitter fish in the individual's current position, is the step size factor, is a random number between 0 and 1, The position of the individual with the best fitness within the perception range of the bitter fish; If the number of hungry steps of the bitter fish individual reaches the hunger step threshold, the bitter fish individual will update its position according to random migration; The random migration expression formula is: in, The updated position of the bitter fish individual, and are the upper and lower bounds of the search space, A random number between 0 and 1; If the bitter fish individual does not meet the conditions of local foraging and random migration, the bitter fish individual will update the individual position through group collaboration. When the positions of all bitter fish individuals are updated, the population is obtained. The group collaboration expression formula is: in, The updated position of the bitter fish individual, is the collaboration factor, is a random number between 0 and 1, is the historical global optimal position, For the bitter fish in the individual's current location; Input the input data into the optimal network model to obtain the detection results.

2. The method for detecting wear of automobile brake pads according to claim 1, characterized in that: The specific steps of performing a brake test on the brake pad include: Clean the brake pad to remove surface grease; Set the target speed, drive the test bench to the target speed, apply braking pressure to the brake pad, and keep the deceleration of the test bench constant until the speed drops to zero.

3. The method for detecting wear of automobile brake pads according to claim 1, characterized in that: The specific steps of obtaining the thermal inertia attenuation coefficient include: Smoothing the temperature-time series data in the cooling stage using a sliding window method to obtain smoothed data; The thermal inertia attenuation coefficient was calculated by fitting the smoothed data using nonlinear least squares.

4. The method for detecting wear of automobile brake pads according to claim 1, characterized in that: The specific steps of extracting features from the surface temperature distribution map include: Extracting the temperature data in the width direction of the brake pad from the temperature distribution diagram to find the maximum and minimum temperature values, and calculating the maximum temperature difference; extracting temperature data along the friction contact direction of the brake pad from the temperature distribution diagram, drawing a line along the friction contact direction of the brake pad, and recording temperature values ​​at different positions of the line to obtain a temperature sequence, and then fitting the temperature sequence using a regression method to obtain a temperature decay rate along the friction contact direction of the brake pad; The average temperature of the temperature distribution map is calculated, a temperature threshold is set, and then the temperature points with higher than the average temperature are screened out according to the threshold, and the areas of the temperature points are measured and summarized to obtain the total area of ​​the temperature points, and the proportion of the area with temperature higher than the average temperature is calculated.

5. The method for detecting wear of automobile brake pads according to claim 1, characterized in that: The specific steps of weighted processing include: Normalizing the thermal inertia attenuation coefficient and the characteristic data to obtain normalized data; The weight distribution is set according to actual experience, and the normalized data is weighted to obtain input data.

6. The method for detecting wear of automobile brake pads according to claim 1, characterized in that: The specific steps to build a random forest network model include: Construct multiple decision trees, and set the maximum depth of the multiple decision trees and the minimum number of samples required for classification points to obtain an initial network model; The initial network is trained using the experimental data to obtain a random forest network model.

7. The method for detecting wear of automobile brake pads according to claim 4, characterized in that: The specific steps of obtaining the experimental data include: A braking experiment was conducted using different brake pads, and the thermal inertia attenuation coefficients of the different brake pads, the maximum temperature difference in the width direction of the brake pad, the temperature attenuation rate in the friction contact direction of the brake pad, the proportion of areas with temperatures higher than the average temperature, and the corresponding wear levels and abnormality types were recorded to obtain data 1; the data 1 was preprocessed to obtain experimental data.

8. The method for detecting wear of automobile brake pads according to claim 5, characterized in that: The specific steps of preprocessing the data include: Cleaning the data one, removing outliers and duplicate values ​​in the data one, and replacing missing values ​​therein with the mean to obtain cleaned data; The cleaning data is normalized to obtain experimental data.

Images