A vehicle brake pad wear degree prediction method and device based on working condition recognition

Abstract

This invention discloses a method and device for predicting vehicle brake pad wear based on operating condition identification, belonging to the field of brake pad wear prediction. It addresses the technical problems of poor accuracy and robustness in brake pad friction prediction in existing vehicle braking systems, hindering efficient and economical wear prediction and optimization of braking system performance. The method includes: performing clustering selection based on difference maximization on the vehicle speed sequence to obtain typical operating conditions; performing wavelet transform processing on the original vehicle speed sequences under different operating conditions to obtain reconstructed and denoised vehicle speed signals; performing matrix transformation on multiple sub-sequences in the typical operating condition vehicle speed signal related to Gram angle fields to obtain a two-dimensional image; extracting and calculating the spatial and temporal features in the two-dimensional image using a hybrid neural network model to obtain the predicted wear value of the brake pads in the vehicle braking system at the current moment; and adjusting the braking gain of the vehicle braking system to obtain the final braking torque.

Description

Technical Field

[0001] This application relates to the field of brake pad wear prediction, and in particular to a method and device for predicting vehicle brake pad wear based on operating condition identification. Background Technology

[0002] Brake pads in a vehicle's braking system are critical components related to driving safety. Their wear level determines braking performance and lifespan, and therefore needs to be inspected during vehicle maintenance. Assessing brake pad wear during vehicle use remains a challenge.

[0003] In recent years, researchers have begun exploring ways to indirectly infer brake pad wear based on existing vehicle operating data (such as vehicle speed, acceleration, braking pressure, and ambient temperature), reducing the need for direct inspection of the brake pads themselves. Current technologies often require sensors to predict brake pad wear, which increases system cost and complexity. Furthermore, brake pad performance degradation affects braking system performance, posing a safety hazard to drivers. Additionally, current brake pad prediction methods struggle to improve the accuracy and robustness of wear predictions, making it difficult to utilize prediction results to compensate for and optimize braking system performance. Summary of the Invention

[0004] This application provides a method and device for predicting the wear degree of vehicle brake pads based on operating condition identification, which is used to solve the following technical problem: In the existing vehicle braking system, the friction prediction of brake pads is difficult to predict and evaluate during the use stage, making it difficult to improve the accuracy and robustness of brake pad wear degree prediction, and is not conducive to completing wear prediction and optimizing braking system performance efficiently and economically.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] On one hand, this application provides a method for predicting the wear level of vehicle brake pads based on driving condition identification, including: performing clustering selection under difference maximization on the vehicle speed sequence according to different types of driving scenarios to obtain the original vehicle speed sequence; performing wavelet transform processing on the original vehicle speed sequence under different driving conditions to obtain the reconstructed and denoised vehicle speed signal; performing matrix transformation on multiple sub-sequences in the vehicle speed signal related to Gram angle fields to obtain a two-dimensional image; using a pre-trained CNN+LSTM hybrid neural network model to output and calculate the feature vector sequence in the two-dimensional image to obtain the predicted value of brake pad wear in the vehicle braking system at the current moment; performing deviation scaling processing on the predicted wear value related to load correction weights to determine the current cumulative wear; and adjusting the braking gain of the vehicle braking system according to the current cumulative wear to obtain the final braking torque to complete the braking control of the vehicle.

[0007] This application utilizes existing vehicle speed data for wear prediction, avoiding additional hardware installation and reducing system cost and complexity. The DBSCAN density clustering algorithm is employed to classify the generated operating condition data, selecting the most differentiated typical operating conditions to ensure the diversity and representativeness of the dataset and improve the model's generalization ability. A hybrid neural network model is formed by combining a convolutional neural network (CNN) to extract spatial features from GASF images and a long short-term memory network (LSTM) to model time dependencies, significantly improving the accuracy and robustness of wear prediction. Through a data feedback mechanism, the system can continuously collect data from actual use for model retraining and optimization, ensuring the long-term effectiveness and adaptability of the prediction system. Simultaneously, through adaptive gain compensation, even in the presence of abnormal wear, the system can maintain the vehicle's deceleration response to the maximum extent, avoiding a sharp decline in braking performance due to wear.

[0008] In one feasible implementation, based on different types of driving scenarios, a clustering selection method maximizing the difference of vehicle speed sequences is performed to obtain the original vehicle speed sequences. Specifically, this includes: discretizing the vehicle speed data according to different types of driving scenarios to obtain multiple speed intervals; constructing a speed transition probability matrix based on actual road test data; and performing state transition processing from the current moment to the next moment on the multiple speed intervals using the speed transition probability matrix to obtain a series of vehicle speed sequences; wherein, the series of vehicle speed sequences represents the speed changes of the vehicle under different operating conditions; and extracting the series of vehicle speeds. The speed features in the sequence are analyzed, and a speed feature vector is constructed based on the speed features. The speed features include at least: average vehicle speed, vehicle speed variance, and acceleration frequency. The Euclidean distance between the speed feature vectors in different types of driving scenarios is calculated, and all the series of vehicle speed sequences are clustered according to the DBSCAN parameters defined by the Euclidean distance to obtain vehicle speed clustering results. The typical operating condition type with the strongest cluster center representativeness and the greatest inter-cluster difference is selected, and the original vehicle speed sequence corresponding to the typical operating condition type is determined. The typical operating condition type covers different types of driving scenarios.

[0009] In one feasible implementation, wavelet transform processing is performed on the original vehicle speed sequence under different operating conditions to obtain a reconstructed and denoised vehicle speed signal. Specifically, this includes: performing a soft-threshold transform on the original vehicle speed sequence using a preset Daubechies wavelet to obtain soft-threshold detail coefficients; and according to... The vehicle speed signal is obtained. Among them, c j (t) represents the approximation coefficients of the wavelet transform; d′ j (t) represents the soft threshold detail coefficient; j represents the number of decomposition layers; J represents the total number of decomposition layers.

[0010] In one feasible implementation, a matrix transformation related to the Gram angle field is performed on multiple subsequences in the vehicle speed signal to obtain a two-dimensional image. Specifically, this includes: segmenting the vehicle speed sequence in the vehicle speed signal into segments of fixed length to obtain the multiple subsequences; wherein each subsequence index contains a segmented vehicle speed value; according to... Obtain each vehicle speed value V′ k,i Angle φ mapped to polar coordinates k,i ; where V′ min With V′ max Subsequence S k Minimum and maximum values; V′ k,i The vehicle speed value in each segment; according to G k,ij=cos(φ k,i +φ k,j ), i,j∈{1,2,…,N}, to obtain the row and column element relationship G of the GASF matrix. k,ij Where i is a row element of the GASF matrix, and j is a column element of the GASF matrix; G k,ij It also represents the relationship between the vehicle speed sequence at different time steps; through the relationship between the row and column elements, the GASF matrix corresponding to the vehicle speed value is mapped to a grayscale image to obtain a two-dimensional image based on the vehicle speed signal.

[0011] In one feasible implementation, before calculating the output of the feature vector sequence in the two-dimensional image using a pre-trained CNN+LSTM hybrid neural network model to obtain the predicted wear value of the brake pads in the vehicle braking system at the current moment, the method further includes: constructing a comprehensive wear evaluation index D based on D = w1·ΔD + w2·Δμ + w3·ΔT; where ΔD is the comprehensive thickness change of the brake pads; Δμ is the mechanical performance change of the brake pads at different wear stages; w1, w2, and w3 are all weighting coefficients; and ΔT is the thermal decay performance change due to friction.

[0012] In one feasible implementation, a pre-trained CNN+LSTM hybrid neural network model is used to calculate the output of the feature vector sequence in the two-dimensional image to obtain the predicted wear value of the brake pads in the vehicle braking system at the current moment. Specifically, this includes: extracting spatial features from the two-dimensional image using a convolutional neural network to obtain a feature vector sequence; inputting and combining a comprehensive wear evaluation index into a long short-term memory network; and performing output prediction processing on the feature vector sequence based on time dependencies using the long short-term memory network. The hybrid neural network model includes the convolutional neural network and the long short-term memory network. The predicted wear amount of the brake pads is obtained. Among them, W fc b represents the weights of the fully connected layers in the Long Short-Term Memory network. fc For bias terms; h K This represents the final hidden state of the Long Short-Term Memory network; according to

[0013] The minimum loss function MSE related to brake pad wear is obtained; where D i This represents the actual wear and tear. To predict wear, n is the sample size; according to The determination coefficient R of the brake pad wear amount is obtained. 2 ;in, The average value of the actual wear is given. Based on the minimized loss function and the determination coefficient, the optimizer in the hybrid neural network model is updated with parameters, and based on the optimized long short-term memory network, the predicted wear value of the brake pads in the vehicle braking system at the current moment is output.

[0014] In one feasible implementation, the predicted wear value is subjected to deviation scaling processing related to load correction weights to determine the current cumulative wear amount, specifically including: based on... The fusion result of historical load and current load is obtained. Where λ is the smoothing coefficient; M(t) is the load capacity; This represents the average load after smoothing over the previous time interval; according to The load correction weight w is obtained. L (t); where M full It is the full load mass; γ is the sensitivity coefficient; R is a real number; according to The current cumulative wear amount D is obtained. current ;in, D is the predicted wear amount at the current moment; previous This represents the increase in wear and tear at a historical moment.

[0015] In one feasible implementation, the braking gain of the vehicle braking system is adjusted based on the current accumulated wear to obtain the final braking torque. Specifically, this includes: if the current accumulated wear is greater than or equal to a safety threshold, then an adaptive compensation coefficient for the braking torque is constructed based on the ratio between the current accumulated wear and the safety threshold; according to T... raw (t)=G0·u(t), to obtain the original braking torque T raw (t); where u(t) represents the driver's pedal displacement input to the brake pedal at time t; G0 is the standard braking gain of the vehicle under normal brake pad conditions; according to β(t)·T raw (t), to obtain the adaptive amplification torque Where β(t) is the adaptive compensation coefficient; according to The final braking torque T is obtained cmd (t); where T max Based on the maximum permissible braking torque under the comprehensive hardware factors of the vehicle, the braking force compensation control of the vehicle braking system is performed on the brakes using the final braking torque to complete the braking control of the vehicle.

[0016] In one feasible implementation, if the current cumulative wear is greater than or equal to a safety threshold, the generated brake pad maintenance information is displayed visually on the vehicle screen as a reminder; and the vehicle's safe driving level is generated based on the numerical range of the adaptive compensation coefficient.

[0017] On the other hand, embodiments of this application also provide a vehicle brake pad wear prediction device based on operating condition identification, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to execute a vehicle brake pad wear prediction method based on operating condition identification as described in any of the above embodiments.

[0018] This application provides a method, device, and medium for predicting the wear degree of vehicle brake pads based on operating condition identification. Compared with the prior art, the embodiments of this application have the following beneficial technical effects:

[0019] 1. No need to install sensors on the brakes: By using the vehicle's existing speed data for wear prediction, additional hardware installation is avoided, reducing system cost and complexity.

[0020] 2. DBSCAN-based differential working condition screening: The generated working condition data is classified using the density clustering algorithm DBSCAN, and the most differential typical working conditions are screened out to ensure the diversity and representativeness of the dataset and improve the generalization ability of the model.

[0021] 3. The time series is converted into two-dimensional images using the GASF method, which fully leverages the advantages of CNN in image feature extraction.

[0022] 4. Combination of CNN and LSTM deep learning models: Combining convolutional neural networks (CNN) to extract spatial features of GASF images and long short-term memory networks (LSTM) to model time dependencies, a hybrid neural network model is formed, which significantly improves the accuracy and robustness of wear prediction.

[0023] 5. The system can predict wear in real time during actual vehicle operation and automatically trigger maintenance suggestions based on the prediction results, thereby realizing intelligent braking system health management and braking performance optimization.

[0024] 6. Continuous learning and model optimization mechanism: Through the data feedback mechanism, the system can continuously collect data from actual use, retrain and optimize the model, and ensure the long-term effectiveness and adaptability of the prediction system.

[0025] 7. Based on predicted data, adaptive gain compensation can maintain the vehicle's deceleration response to the maximum extent even in the presence of abnormal wear, thus avoiding a sharp decline in braking performance due to wear. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0027] Figure 1 A flowchart of a method for predicting the wear degree of vehicle brake pads based on working condition identification is provided in an embodiment of this application;

[0028] Figure 2 A schematic diagram of a working speed curve provided in an embodiment of this application;

[0029] Figure 3 Schematic diagrams illustrating the wear prediction error effect under various typical working conditions provided in the embodiments of this application;

[0030] Figure 4 This is a schematic diagram of a vehicle brake pad wear prediction device based on operating condition identification, provided in an embodiment of this application. Detailed Implementation

[0031] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0032] This application provides a method for predicting the wear level of vehicle brake pads based on operating condition identification, such as... Figure 1 As shown, a method for predicting the wear level of vehicle brake pads based on operating condition identification specifically includes steps S101-S106:

[0033] S101. Based on different types of driving scenarios, perform clustering selection on the vehicle speed sequence to maximize the difference, and obtain the original vehicle speed sequence.

[0034] Specifically, the vehicle speed data needs to be discretized according to different driving scenarios to obtain multiple speed ranges. Then, a speed transition probability matrix is ​​constructed based on actual road test data.

[0035] Furthermore, using the speed transition probability matrix, state transition processing is performed on multiple speed intervals from the current time to the next time interval to obtain a series of vehicle speed sequences. These vehicle speed sequences represent the speed changes of the vehicle under different operating conditions.

[0036] In one embodiment, the vehicle speed V is first discretized into S speed intervals {V1, V2, ..., V...} S The driving conditions are divided into sub-segments. Based on actual road test data, a speed transition probability matrix P is constructed, where each element represents the probability of transitioning from state i to state j in the next moment. This satisfies the following condition: Then, using the aforementioned transition probability matrix, a series of vehicle speed sequences are generated. This allows for the simulation of vehicle speed changes under different operating conditions.

[0037] Furthermore, speed features are extracted from the series of vehicle speed sequences, and a speed feature vector is constructed based on these features. The speed features include at least: average vehicle speed, vehicle speed variance, and acceleration frequency.

[0038] Furthermore, it is necessary to calculate the Euclidean distance of the speed feature vectors in different types of driving scenarios, and to perform clustering processing on all series of vehicle speed sequences according to the DBSCAN parameters defined by the Euclidean distance to obtain the vehicle speed clustering results.

[0039] Furthermore, the typical operating condition type with the strongest representativeness of the cluster center point and the greatest inter-cluster differences was selected, and the original vehicle speed sequence corresponding to the typical operating condition type was determined. Among them, the typical operating condition type covers driving scenarios under different types.

[0040] In one embodiment, features, including but not limited to average vehicle speed, are extracted from the generated series of vehicle speed sequences. Vehicle speed variance acceleration frequency f A etc., constitute the feature vector Then, the distance metric between samples is calculated, typically using Euclidean distance d(x). i ,x j ): Where, x i and x jLet K be the feature vectors of the i-th and j-th samples, respectively, and K be the feature dimension. Next, DBSCAN parameters are set: the radius parameter ∈ is set to define the maximum distance between sample neighborhoods. The minimum number of samples, MinPts, is set to define the minimum number of samples required to form a cluster. For each sample point, all its neighbors within its ∈ neighborhood are found. If the number of points in the neighborhood is greater than or equal to MinPts, then the point is marked as a core point, and cluster formation begins. The core point and all points in its neighborhood are grouped into the same cluster. This process is repeated until all points are classified or marked as noise.

[0041] In one embodiment, several driving condition categories with the strongest cluster center representativeness and the greatest inter-cluster differences are selected from the DBSCAN clustering results (vehicle speed clustering results) to ensure that the selected driving conditions cover different driving scenarios (such as high-speed stable driving conditions, frequent start-stop driving conditions, etc.). The original vehicle speed sequence corresponding to each typical driving condition is recorded. Used for subsequent wear testing and model training.

[0042] S102. Perform wavelet transform processing on the original vehicle speed sequences under different operating conditions to obtain the reconstructed and denoised vehicle speed signals.

[0043] Specifically, the original vehicle speed sequence is first processed by a soft threshold transformation under the relevant detail coefficients using a preset Daubechies wavelet to obtain the soft threshold detail coefficients.

[0044] Furthermore, according to Get vehicle speed signal Among them, c j (t) represents the approximation coefficients of the wavelet transform. d′ j (t) represents the soft-threshold detail coefficient; j represents the number of decomposition layers; J represents the total number of decomposition layers.

[0045] In one embodiment, multiple braking tests can be conducted in a laboratory or closed test site according to a sampled sequence of typical operating conditions, and the vehicle speed V during each braking process can be recorded. t Braking force F b Data such as thermal degradation performance change T are collected. Then, the cumulative wear amount D corresponding to each operating condition sequence is used as a label to establish a correspondence between operating condition data and wear amount. Next, to remove high-frequency noise from the vehicle speed data, wavelet transform is used for multi-scale decomposition and thresholding. Let the original vehicle speed time series be... The Daubechies wavelet is chosen for wavelet decomposition, and approximation coefficients {c} are obtained based on existing classical algorithms. j (t)} and detail coefficients {d j (t)}, where j represents the decomposition level. For the detail coefficients d j (t) Thresholding is performed using a soft thresholding method to determine the soft threshold detail coefficient d′.j (t):

[0046] Where λ is the threshold and sgn(□) is the sign function. Finally, the reconstructed and denoised vehicle speed signal is calculated.

[0047] S103. Perform matrix transformation on multiple subsequences in the vehicle speed signal related to the Gramian Angular Summation Field (GASF) to obtain a two-dimensional image.

[0048] Specifically, Figure 2 A schematic diagram of a working speed curve is provided for an embodiment of this application, such as... Figure 2 As shown, the vehicle speed signal sequence also needs to be segmented into segments of fixed length to obtain multiple subsequences. Each subsequence index contains a segmented vehicle speed value.

[0049] In one embodiment, the noise-reduced vehicle speed sequence Divide into segments of a fixed length N to obtain multiple...

[0050] Subsequence S k =[V′ k,1 ,V′ k,2 ,…,V′ k,N ], where k is the subsequence index, and

[0051] Furthermore, according to Obtain each vehicle speed value V′ k,i Angle φ mapped to polar coordinates k,i Among them, V′ min With V′ max Subsequence S k Minimum and maximum values; V′ k,i The vehicle speed value for each segment.

[0052] Furthermore, according to G k,ij =cos(φ k,i +φ k,j ), i,j∈{1,2,…,N}, to obtain the row and column element relationship G of the GASF matrix. k,ij Where i is a row element of the GASF matrix, and j is a column element of the GASF matrix; G k,ij It also indicates the relationship between the vehicle speed sequence at different time steps.

[0053] Furthermore, by using the row and column element relationships, the GASF matrix corresponding to the vehicle speed value is mapped to a grayscale image to obtain a two-dimensional image based on the vehicle speed signal.

[0054] In one embodiment, the GASF matrix G k Treating it as a two-dimensional grayscale image, to adapt to the input requirements of CNNs, matrix G is... k Map to a grayscale image, i.e., a two-dimensional image, while setting the pixel value range to [0,1].

[0055] S104. Using the pre-trained CNN+LSTM hybrid neural network model, the feature vector sequence in the two-dimensional image is output and calculated to obtain the predicted value of the wear of the brake pads in the vehicle braking system at the current moment.

[0056] Specifically, it is necessary to first construct the comprehensive wear evaluation index D based on D = w1·ΔD + w2·Δμ + w3·ΔT. Here, ΔD represents the comprehensive thickness change of the brake pad; Δμ represents the change in mechanical properties of the brake pad at different wear stages; w1, w2, and w3 are all weighting coefficients; and ΔT represents the change in thermal degradation performance due to friction.

[0057] In one embodiment, to achieve a quantitative assessment of brake pad wear, it is necessary to comprehensively consider changes in brake pad thickness and mechanical properties, as well as changes in thermal degradation performance caused by friction, to construct a quantitative wear evaluation index. Thickness change measurement: The thickness D of the brake pad is measured periodically using precision measuring equipment, and the thickness value D at different time points is recorded. t Mechanical performance measurement: At different wear stages, measure the mechanical performance parameters of the brake pads, such as the coefficient of friction μ, and record the change Δμ.

[0058] Furthermore, spatial features in the two-dimensional image are extracted and processed using a convolutional neural network to obtain a feature vector sequence. The comprehensive wear evaluation index is then input and combined into a long short-term memory network. The long short-term memory network then performs output prediction processing based on time dependencies on the feature vector sequence. The hybrid neural network model includes both a convolutional neural network and a long short-term memory network.

[0059] In one embodiment, the Convolutional Neural Network (CNN) is responsible for extracting spatial features from the GASF image. The architecture is as follows: Input layer: Accepts image data of size N×N×C, where C is the number of channels (e.g., 1 or 3). Convolutional layer: Applies multiple convolutional kernels K. l Perform convolution operations to extract feature map F l : Where σ is the activation function (e.g., ReLU), M is the number of input channels, and b l This is the bias term. Pooling layer: Max pooling or average pooling is used to reduce the feature map size and retain the main features. Activation function: ReLU activation function is used: ReLU(x) = max(0,x). Flattening layer: The multidimensional feature map is flattened into a one-dimensional vector v.k .

[0060] In one embodiment, the Long Short-Term Memory (LSTM) network is responsible for modeling the temporal dependencies of the feature vector sequence, with the following architecture:

[0061] ① Input layer: Accepts the feature vector sequence {v1,v2,…,v} extracted by CNN K}, where K is the number of time steps.

[0062] ②LSTM unit: Captures long-term dependencies through a memory gate mechanism. Its core calculation formula is as follows:

[0063] f t =σ(W f ·[h t-1 ,v t ]+b f );

[0064] i t =σ(W i ·[h t-1 ,v t ]+b i );

[0065]

[0066] o t =σ(W o ·[h t-1 ,v t ]+b o );

[0067] h t =o t ⊙tanh(C t );

[0068] Where: f t Forgotten Gate; i t Input gate; Candidate memory unit; C t : Current memory unit; o t Output gate; h t : Current hidden state; σ: Sigmoid activation function; ⊙: Element-wise product;

[0069] W f W i W C W o : Weight matrices of each gate in the LSTM; b f ,b i ,b C ,b o : Bias vectors of each gate in the LSTM; vt : The input feature vector at the t-th time step.

[0070] ③ Fully connected layer: The final hidden state h of the LSTM K Mapped to the output layer, the predicted value of brake pad wear is obtained.

[0071] Furthermore, according to Obtain the predicted wear amount of brake pads. Among them, W fc b represents the weights of the fully connected layers in a Long Short-Term Memory (LSTM) network. fc This is a bias term. K This represents the final hidden state of the Long Short-Term Memory (LSTM) network.

[0072] Furthermore, the mean squared error (MSE) is used as the loss function: according to The minimum loss function MSE related to brake pad wear is obtained. Where, D i This represents the actual wear and tear. To predict wear, n is the number of samples.

[0073] Furthermore, the Adam optimizer is used for parameter updates, with the optimization objective being to minimize the loss function. GASF images from the training set are input into a CNN to extract spatial features, and then the feature vector sequence is input into an LSTM for temporal modeling, ultimately outputting the wear prediction value. Model parameters are optimized through multiple iterations (epochs), during which model performance is monitored on the validation set to prevent overfitting. When the validation set performance is optimal, the current model parameters are saved as the optimal model. The model's predictive performance is evaluated on the test set using mean squared error (MSE) and coefficient of determination (R²). 2 Evaluation based on indicators such as ) The coefficient of determination R for brake pad wear was obtained. 2 .in, This represents the average value of the actual wear.

[0074] Furthermore, based on minimizing the loss function and the coefficient of determination, the optimizer in the hybrid neural network model is updated with parameters, and based on the optimized long short-term memory network, the predicted value of the wear of the brake pads in the vehicle braking system at the current moment is output.

[0075] As a possible implementation method, the generated GASF image G t Input the pre-trained CNN model and extract the spatial feature vector v t Given a sequence of K consecutive feature vectors {v} t-K+1 ,v t-K+2 ,…,v tInput an LSTM model and output the predicted brake pad wear at the current time step.

[0076] S105. Perform deviation scaling on the wear prediction value with relevant load correction weights to determine the current cumulative wear amount.

[0077] Specifically, M(t) is first smoothed or weighted and fused to obtain... according to The fusion result of historical load and current load is obtained. Where λ is the smoothing coefficient, which can be selected according to the need for sensitivity to real-time changes; M(t) is the load capacity; This represents the average load after smoothing over the previous time interval.

[0078] Furthermore, since the pre-trained model is based on full-load conditions, the wear prediction will be appropriately reduced when the actual vehicle load is lower than full load; if the actual vehicle load exceeds full load (overload), the wear will increase accordingly. A correction function with a sensitivity coefficient γ is proposed: Obtain the load correction weight w L (t). Where M full γ is the full load mass; γ is the sensitivity coefficient; R is a real number. Using a linear form, γ can be conveniently used to control the adjustment of wear resistance when the load changes, without overshooting. γ is calibrated based on actual experimental data, bench tests, or historical operational data.

[0079] Furthermore, since the CNN+LSTM hybrid neural network model was pre-trained using a fully loaded vehicle, wear is relatively reduced when the load is low. Therefore, the model needs to have some adjustability in its sensitivity to load. In the t-th time interval, the original predicted wear value is denoted as... Historical cumulative wear and tear is D previous The load correction can then be incorporated into the following formula:

[0080] Get the current cumulative wear amount D current .in, D is the predicted wear amount at the current moment; previous This represents the increase in wear and tear at a historical moment.

[0081] S106. Based on the current accumulated wear, adjust the braking gain of the vehicle braking system to obtain the final braking torque, thereby completing the braking control of the vehicle.

[0082] Specifically, if the current cumulative wear is greater than or equal to the safety threshold, an adaptive compensation coefficient for the braking torque is constructed based on the ratio between the current cumulative wear and the safety threshold.

[0083] Furthermore, it is also necessary to consider T raw (t)=G0·u(t), to obtain the original braking torque T raw (t). Where u(t) represents the driver's pedal displacement input to the brake pedal at time t. G0 is the standard braking gain of the vehicle under normal brake pad conditions.

[0084] Furthermore, it is also necessary to... Obtain adaptive amplification torque Where β(t) is the adaptive compensation coefficient.

[0085] Furthermore, utilizing The final braking torque T is obtained cmd (t). Where T max This refers to the maximum permissible braking torque based on the vehicle's overall hardware factors.

[0086] Furthermore, by combining the final braking torque, the braking force compensation control of the vehicle braking system is performed on the brakes to complete the braking control of the vehicle.

[0087] In one embodiment, to maintain the original braking effect within a certain range, when D is detected... current ≥D threshold At that time, the constructed adaptive compensation coefficient β(t) is used to amplify and correct the braking gain. The specific definition is as follows: When wear and tear intensifies, D current Significantly higher than D threshold When β(t) exceeds 1, the brake pedal input is appropriately amplified. If the vehicle is in an extremely low friction coefficient or severely worn condition, an upper limit β(t) is set to prevent excessive amplification leading to loss of control. max ,β nax After design or experimental research, calibration is performed to ensure that the gain amplification is within a safe and controllable range.

[0088] In one embodiment, let u(t) represent the driver's pedal displacement input to the brake pedal at time t, and G0 be the nominal braking gain of the vehicle under normal brake pad conditions. The original braking command (braking torque) can typically be written as: T raw (t) = G0·u(t). To compensate for insufficient braking force caused by brake wear, when β(t) exceeds 1, the original torque is adaptively amplified: Through the aforementioned amplification, the vehicle can maintain braking performance close to the original design even during moderate or light wear stages. However, when wear is severe, simply increasing the braking force may lead to brake overload or tire lock-up. Based on the compensated torque, a maximum permissible braking torque T is set. max Amplitude limiting is applied to obtain the final braking torque T.cmd (t): Among them, T max It can be set according to comprehensive factors such as vehicle structure, load, and thermal degradation.

[0089] As a feasible implementation method, if the current cumulative wear is greater than or equal to a safety threshold, the generated brake pad maintenance information will be displayed visually on the in-vehicle screen as a reminder. Based on the numerical range of the adaptive compensation coefficient, a vehicle safety driving level will be generated.

[0090] In one embodiment, Figure 3 This application provides schematic diagrams illustrating the wear prediction error effects under various typical operating conditions in its embodiments. Figure 3 As shown in the example experiment: the experimental data is divided into training set, validation set, and test set. A portion of the data (60%) is used as the training set; GASF images are extracted and input into a CNN. The feature vector sequence output by the CNN is then input into an LSTM for time-dependent training. Another portion of the data (20%) is used as the validation set for model hyperparameter adjustment and early stop strategy monitoring to prevent overfitting. The remaining data (approximately 20%) is reserved as the test set to evaluate the model's final predictive performance. The experiment is conducted on a chassis dynamometer based on actual vehicle operating conditions (i.e., collecting brake wear data while testing vehicle fuel economy). DBSCAN clusters the actual operating condition segments into 20 typical operating conditions. Each typical operating condition in the test set contains 5 to 20 test data points, including the operating condition data and its corresponding actual brake wear. The method of this patent is applied to the test set composed of real data for testing, and the results are as follows... Figure 3 As shown in the figure, 20 bins are grouped, each corresponding to the prediction error distribution under a typical operating condition. The upper and lower edges of the bins represent the upper and lower quartiles of the data for that group, respectively, and the horizontal line inside the bin represents the median. The vertical dashed line shows the maximum and minimum observed values ​​within the normal range. The results show that the prediction error for vehicle brake wear under various operating conditions is within ±30%, and most are within ±20%. The prediction effect is best for typical operating conditions 1, 2, 6, and 8, with errors within ±15%. The experiment demonstrates that the prediction results of the method of this invention have good consistency with the actual results, and the method has high accuracy in predicting high-wear brakes. This method can achieve the goal of predicting vehicle brake pad wear based on operating conditions.

[0091] In addition, this application also provides a vehicle brake pad wear prediction device based on operating condition identification, such as... Figure 4 As shown, the vehicle brake pad wear prediction device 400 based on operating condition identification specifically includes:

[0092] At least one processor 401. And a memory 402 communicatively connected to the at least one processor 401. The memory 402 stores instructions executable by the at least one processor 401, enabling the at least one processor 401 to execute:

[0093] Based on different types of driving scenarios, the vehicle speed sequences are clustered to maximize differences, thus obtaining the original vehicle speed sequences.

[0094] Wavelet transform processing is performed on the original vehicle speed sequences under different operating conditions to obtain the reconstructed and denoised vehicle speed signals.

[0095] A two-dimensional image is obtained by performing matrix transformations on multiple subsequences of the vehicle speed signal related to the Gram angle field.

[0096] The pre-trained CNN+LSTM hybrid neural network model is used to calculate the output of the feature vector sequence in the two-dimensional image to obtain the predicted value of the wear of the brake pads in the vehicle braking system at the current moment.

[0097] The wear prediction value is adjusted by scaling down the deviation of the load correction weight to determine the current cumulative wear.

[0098] Based on the current accumulated wear, the braking gain of the vehicle's braking system is adjusted to obtain the final braking torque, thereby completing the braking control of the vehicle.

[0099] The embodiments of this application also include:

[0100] 1. DBSCAN is introduced to cluster a large number of vehicle speed sequences, which can automatically identify and select "typical operating conditions" with significant differences between clusters. Compared with traditional screening methods based on manual methods or a small number of cluster centers, DBSCAN does not require pre-specifying the number of clusters and can better cover diverse driving scenarios (such as high-speed stability, frequent start-stop, etc.).

[0101] 2. It not only measures the thickness change of brake pads, but also takes into account the impact of the wear process on braking performance and thermal fade performance; by using weighting coefficients, it obtains a more quantitative index that reflects the true degree of wear, which is more accurate and comprehensive.

[0102] 3. Sample each type of "typical working condition" in the laboratory or closed test site and conduct multiple braking experiments to collect information on vehicle speed and wear degree at different time periods. Match the final cumulative wear amount with the working condition sequence to construct a dataset with the advantages of high quality and supervision.

[0103] 4. Visual representation using Gram angles and field methods can simultaneously preserve the global correlations of time series on a single two-dimensional graph, making it possible for deep convolutional neural networks to extract spatial structural features. Mapping one-dimensional data into two-dimensional space can capture more intra-temporal correlation patterns.

[0104] 5. Since Gram angle and field image matrix capture the interrelationship of the working condition time series at different time steps, this time series dependency is re-extracted at the two-dimensional image level. The latent feature patterns can be effectively extracted through convolutional neural networks.

[0105] 6. After extracting features from each segment of the Gram angle and field image using a convolutional neural network, a set of two-dimensional feature vectors is obtained. These vectors are then fed into an LSTM network in chronological order. The hybrid model can better balance the dependence on local spatial structure and time steps, and has higher prediction accuracy for phenomena such as vehicle brake wear that accumulate over time.

[0106] 7. The amount of brake pad wear is quantified into an input parameter that can be controlled in real time. The braking torque is corrected in real time according to the degree of wear, thereby compensating for the impact of wear on the vehicle's braking performance, effectively delaying or mitigating the driving safety risks caused by severe brake pad wear, and also triggering maintenance and replacement prompts.

[0107] 8. By utilizing existing vehicle speed data for wear prediction, additional hardware modifications are avoided, reducing system cost and complexity.

[0108] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0109] The foregoing has described specific embodiments of this application. In some cases, the actions or steps described in the specification may be performed in a different order than those shown in the embodiments and the desired results may still be achieved. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] The above description is merely an embodiment of this application and is not intended to limit this application. For those skilled in the art, various modifications and variations can be made to the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of this application should be included within the specification of this application.

[0111] It should be noted that this application was supported by the Shandong Natural Science Foundation (ZR2023QE208), the China Postdoctoral Science Foundation (2024M751577), and the Shandong Provincial Key Research and Development Program (2022CXGC020302) to conduct research on the dynamic performance of vehicles during braking.

Claims

1. A method for predicting the wear degree of vehicle brake pads based on operating condition identification, characterized in that, The method includes: Based on different types of driving scenarios, the vehicle speed sequences are clustered to maximize differences, thus obtaining the original vehicle speed sequences. Wavelet transform processing is performed on the original vehicle speed sequences under different operating conditions to obtain the reconstructed and denoised vehicle speed signals. Multiple subsequences in the vehicle speed signal are subjected to matrix transformations related to the Gram angle field to obtain a two-dimensional image; The pre-trained CNN+LSTM hybrid neural network model is used to calculate the output of the feature vector sequence in the two-dimensional image to obtain the predicted value of the wear of the brake pads in the vehicle braking system at the current moment. The predicted wear value is subjected to deviation scaling processing related to load correction weights to determine the current cumulative wear amount, specifically including: according to The combined results of historical load and current load are obtained. ;in, For smoothing coefficients; Load capacity; This represents the average load after smoothing over the previous time interval. according to The load correction weight is obtained. ;in, This is the full load mass; This is the sensitivity coefficient; It is a real number; according to The current cumulative wear amount is obtained. ;in, This is the predicted wear amount at the current moment; This represents the cumulative historical wear and tear. Based on the current accumulated wear, the braking gain of the vehicle braking system is adjusted to obtain the final braking torque, thereby completing the braking control of the vehicle.

2. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, Based on different types of driving scenarios, the vehicle speed sequences are clustered to maximize differences, resulting in the original vehicle speed sequences, specifically including: Based on different types of driving scenarios, the vehicle speed data is discretized to obtain multiple speed ranges; and a speed transition probability matrix is ​​constructed based on actual road test data. Using the speed transition probability matrix, state transition processing is performed on the multiple speed intervals from the current time to the next time to obtain a series of vehicle speed sequences; wherein, the series of vehicle speed sequences represent the speed changes of the vehicle under different operating conditions; Extract speed features from the series of vehicle speed sequences, and construct a speed feature vector based on the speed features; wherein the speed features include at least: average vehicle speed, vehicle speed variance, and acceleration frequency; Calculate the Euclidean distance between the speed feature vectors in different types of driving scenarios, and perform clustering processing on all the series of vehicle speed sequences according to the DBSCAN parameters defined by the Euclidean distance to obtain the vehicle speed clustering results; The typical operating condition type with the strongest representativeness of the cluster center point and the greatest difference between clusters is selected, and the original vehicle speed sequence corresponding to the typical operating condition type is determined; wherein, the typical operating condition type covers driving scenarios under different types.

3. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, Wavelet transform processing is performed on the original vehicle speed sequences under different operating conditions to obtain reconstructed and denoised vehicle speed signals, specifically including: The original vehicle speed sequence is subjected to soft threshold transformation processing with relevant detail coefficients using a preset Daubechies wavelet to obtain soft threshold detail coefficients. according to The vehicle speed signal is obtained. ;in, These are the approximation coefficients of the wavelet transform; The soft threshold detail coefficient; j Indicates the number of decomposition levels; This indicates the total number of decomposition levels.

4. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, The vehicle speed signal is subjected to matrix transformations related to the Gram angle field to obtain a two-dimensional image, specifically including: The vehicle speed sequence in the vehicle speed signal is segmented for a fixed length to obtain the plurality of subsequences; wherein, each subsequence index contains a segmented vehicle speed value; according to Get each vehicle speed value Angle mapped to polar coordinates ;in, and Subsequences The minimum and maximum values; The vehicle speed value in each segment; according to The row and column element relationships of the GASF matrix are obtained. ;in, i For the row elements of the GASF matrix, j These are the column elements of the GASF matrix; It also indicates the relationship between the vehicle speed sequence at different time steps; By using the row and column element relationships, the GASF matrix corresponding to the vehicle speed value is mapped to a grayscale image to obtain a two-dimensional image based on the vehicle speed signal.

5. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, Before calculating the predicted wear value of the brake pads in the vehicle braking system at the current moment by outputting the feature vector sequence in the two-dimensional image through a pre-trained CNN+LSTM hybrid neural network model, the method further includes: according to A comprehensive wear evaluation index was constructed. ;in, This refers to the overall thickness variation of the brake pads; The mechanical properties of the brake pads change at different wear stages; , , All are weighting coefficients; This refers to the change in thermal degradation performance caused by friction.

6. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, The pre-trained CNN+LSTM hybrid neural network model is used to calculate the output of the feature vector sequence in the two-dimensional image to obtain the predicted wear value of the brake pads in the vehicle braking system at the current moment, specifically including: The spatial features in the two-dimensional image are extracted and processed by a convolutional neural network to obtain a feature vector sequence; The comprehensive wear evaluation index is input and combined into the long short-term memory network; The long short-term memory network is used to perform output prediction processing on the feature vector sequence under time dependence; wherein, the hybrid neural network model includes: the convolutional neural network and the long short-term memory network; according to The predicted wear amount of the brake pads is obtained. ;in, These are the weights of the fully connected layers in the Long Short-Term Memory network; For bias terms; This represents the final hidden state of the Long Short-Term Memory network; according to The loss function that minimizes brake pad wear is obtained. ;in, This represents the actual wear and tear. To predict wear, The number of samples; according to The determination coefficient of the wear amount of the brake pads is obtained. ;in, This represents the average value of the actual wear amount; Based on the minimized loss function and the determination coefficient, the optimizer in the hybrid neural network model is updated with parameters, and based on the optimized long short-term memory network, the predicted wear value of the brake pads in the vehicle braking system at the current moment is output.

7. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, Based on the current accumulated wear, the braking gain of the vehicle braking system is adjusted to obtain the final braking torque, specifically including: If the current cumulative wear is greater than or equal to the safety threshold, then an adaptive compensation coefficient for the braking torque is constructed based on the ratio between the current cumulative wear and the safety threshold. according to The original braking torque is obtained. ;in, Indicates the driver at the time t Input the pedal displacement of the brake pedal; This is the standard braking gain of the vehicle when the brake pads are in normal operating condition. according to To obtain the adaptive amplification torque ;in, The adaptive compensation coefficient is... according to The final braking torque is obtained. ;in, The maximum permissible braking torque is determined based on the vehicle's overall hardware factors. The final braking torque is used to perform braking force compensation control on the vehicle braking system to complete the braking control of the vehicle.

8. The method for predicting vehicle brake pad wear based on operating condition identification according to claim 1, characterized in that, If the current cumulative wear is greater than or equal to the safety threshold, the generated brake pad maintenance information will be displayed on the vehicle screen as a visual reminder; and the vehicle's safe driving level will be generated based on the numerical range of the adaptive compensation coefficient.

9. A vehicle brake pad wear prediction device based on operating condition identification, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform a vehicle brake pad wear prediction method based on operating condition identification according to any one of claims 1-8.

Images