A method, apparatus, equipment and medium for determining PID control parameters of a clutch
By employing machine learning methods, combining Pearson correlation coefficient and K-means clustering algorithm with SVM model, the inaccuracy problem in determining PID control parameters of AMT clutches in commercial vehicles was solved, achieving efficient and reliable parameter prediction and improving the overall comfort and power performance of AMT systems in commercial vehicles.
Patent Information
- Application Number
- CN202311012548.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-11
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-08-11
AI Technical Summary
In the existing technology, the methods for determining the PID control parameters of AMT clutches in commercial vehicles mainly rely on trial and error and empirical methods, which are time-consuming, labor-intensive, and have unsatisfactory results, lacking reliability and consistency.
Machine learning methods are employed, using Pearson correlation coefficient to analyze data correlation, K-means clustering algorithm to determine cluster centers, and SVM model to train the system, thereby accurately predicting clutch PID control parameters.
It enables accurate prediction of clutch PID control parameters for new AMT models, saving manpower and time costs, improving AMT R&D efficiency, and optimizing the operating status of existing models.
Smart Images

Figure CN116894311B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of AMT control technology, specifically to a method, device, equipment, and medium for determining clutch PID control parameters. Background Technology
[0002] AMT (Automated Manual Transmission) models in commercial vehicles are still in the development stage in China. For AMT control systems, clutch PID control is an extremely important component. Determining the clutch PID control parameters is a crucial prerequisite for applying control logic to different vehicle models. Appropriate clutch PID control parameters not only improve overall vehicle comfort and performance but also affect the lifespan of clutch-related mechanical components. To determine suitable clutch PID control parameters, a reliable calculation method is needed to replace the traditional subjective setting method.
[0003] Currently, the determination of PID control parameters in AMT control systems for commercial vehicles primarily employs trial-and-error and empirical methods. The trial-and-error method involves manually setting and continuously optimizing the clutch PID control parameters during test drives, ultimately determining a set based on subjective experience and data feedback. The disadvantages of this method are its time-consuming and labor-intensive nature, and the final results are not always ideal, exhibiting significant uncertainty. The empirical control method uses the clutch PID control parameters of similar vehicle models as a basic reference, setting the parameters for new models based on personal experience. The disadvantages of this method are its strong reliance on experience, making it less suitable for industry newcomers, and the results are also subject to bias due to subjective feelings.
[0004] To overcome the shortcomings of the two methods mentioned above and the unsatisfactory results they lead to, it is essential to design a cluster-centric method for determining the PID control parameters of the AMT clutch in commercial vehicles. Summary of the Invention
[0005] This invention provides a method, device, equipment, and medium for determining PID control parameters of a clutch. It uses machine learning to predict and set parameters, thereby replacing subjective human setting and avoiding undesirable results caused by human setting.
[0006] In a first aspect, the present invention provides a method for determining PID control parameters for a clutch, comprising the following steps: Select clutch PID control parameter data and related data of different types of commercial vehicles, and perform data preprocessing on the selected data; In the preprocessed data, data related to PID control parameters are selected, and correlation analysis is performed on each of the selected data and the PID control parameter data to determine the data related to PID control parameters whose correlation is greater than the set value. The number of clusters is determined by the silhouette coefficient, and the data related to the PID control parameters with a correlation greater than the set value are clustered. The data related to the PID control parameters with a known correlation greater than the set value are divided into a training dataset and a prediction dataset. The training dataset is then trained using an SVM model based on the clustering results to obtain the training classification results. The prediction dataset is classified, and PID control parameters are selected based on the classification results.
[0007] As a further limitation of the technical solution of the present invention, the steps of selecting clutch PID control parameter data and related data of different types of commercial vehicles and performing data preprocessing on the selected data include: Collect and summarize vehicle configuration and related parameter data for different AMT models; Based on the AMT clutch PID control logic, various data are categorized and filtered to select relevant data that affect the PID control parameters: Image data is converted into digital data by selecting plotted points.
[0008] As a further limitation of the technical solution of the present invention, the step of selecting data related to PID control parameters from the preprocessed data, performing data correlation analysis on each pair of the selected data and the PID control parameter data, and determining the data related to PID control parameters with a correlation greater than the set value includes: In the preprocessed data, PID control parameter data and data related to PID control parameters are selected; and the two parameter data are combined to form an input matrix. Calculate the Pearson correlation coefficient between the two selected parameters; After permuting and combining the various parameters, the Pearson correlation coefficients between the parameters are obtained, and the correlations are visualized.
[0009] As a further limitation of the technical solution of the present invention, the step of calculating the Pearson correlation coefficient between the two selected parameters includes: The Pearson correlation coefficient between the two selected parameters is defined as the quotient of the covariance and standard deviation of the two parameters: By estimating the sample covariance and standard deviation, the Pearson correlation coefficient can be obtained. : By The Pearson correlation coefficient is obtained by estimating the mean of the standard scores of the sample points. : in , and They are respectively for The standard score, sample mean, and sample standard deviation of the sample.
[0010] As a further limitation of the technical solution of the present invention, the step of clustering the data related to the PID control parameters with a correlation greater than a set value by using the silhouette coefficient to determine the number of clusters includes: Calculate the silhouette coefficient for each vector in the cluster; where, the first... The contour coefficients of each object are: In the formula express The distance from a vector to all other points in the cluster it belongs to. express The average distance from a vector to all points within a cluster that does not contain it; By assigning values to the definition of the number of clusters, the clusters are selected as defined. The contour coefficient has an optimal solution as the final number of clusters; Clustering of AMT clutch control data based on the number of clusters; in, In the formula, For the first There are 3 cluster sets, with cluster centers being 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... ; For the first Clusters The number of samples contained in it; the minimum value can be obtained by selecting the criterion function: Right now The cluster center of a class is chosen as the mean of the samples in that class.
[0011] As a further limitation of the technical solution of this invention, the steps of dividing the data related to the PID control parameters with a correlation greater than the set value into a training dataset and a prediction dataset, and training the training dataset using an SVM model based on the clustering results to obtain the training classification results include: After shuffling the data related to the PID control parameters whose correlation is greater than the set value, the datasets are divided into training datasets and prediction datasets according to a set ratio. For each data point in the training dataset, we search for a classification decision boundary, where the optimal decision boundary, or classification margin, can be expressed as: in, For kernel function, For the target category, For non-negative Lagrange multipliers, For offset terms; kernel function In the formula The center of the kernel function, For vectors and Euclidean distance; The nonlinear SVM model used in conjunction with the clustering results is selected to obtain the classification margin with the largest margin, thus obtaining the training classification result.
[0012] As a further limitation of the technical solution of the present invention, the step of classifying the prediction dataset and selecting PID control parameters based on the classification results includes: The distribution of landing points is determined by predicting the dataset, and the distribution accuracy is calculated based on the training classification results. The classification accuracy is calculated by dividing the number of correctly predicted results by the total number of results. The number of correctly identified positive samples, the number of falsely reported negative samples, the number of correctly identified negative samples, and the number of missed positive samples are selected. The horizontal axis is selected as the predicted category and the vertical axis as the actual category to construct a confusion matrix, which is then visualized using a heatmap.
[0013] Secondly, the present invention provides a clutch PID control parameter determination device, including a data collection and preprocessing module, a data correlation analysis module, a data clustering module, a training processing module, and a prediction processing module; The data collection and preprocessing module is used to select clutch PID control parameter data and related data of different types of commercial vehicles and to preprocess the selected data. The data correlation analysis module is used to select data related to PID control parameters from the preprocessed data, perform data correlation analysis on each pair of the selected data and the PID control parameter data, and determine the data related to PID control parameters whose correlation is greater than the set value. The data clustering module is used to determine the number of clusters by using the silhouette coefficient, and to cluster the data related to the PID control parameters whose correlation is greater than the set value. The training processing module is used to divide the data related to the PID control parameters with a certain correlation greater than the set value into training datasets and prediction datasets. The training dataset is then trained using an SVM model based on the clustering results to obtain the training classification results. The prediction processing module is used to classify the prediction dataset and select PID control parameters based on the classification results.
[0014] As a further limitation of the technical solution of the present invention, the data collection and preprocessing module includes a collection and summarization unit, a classification and filtering unit, and a data preprocessing unit; The data collection and aggregation unit is used to collect and aggregate vehicle configuration and related parameter data for different AMT models; The classification and filtering unit is used to classify and filter various data based on the AMT clutch PID control logic, selecting relevant data that affects the PID control parameters: The data preprocessing unit is used to select plotted points in image data and convert them into digital data.
[0015] As a further limitation of the technical solution of the present invention, the data correlation analysis module includes a selection processing unit, a coefficient calculation unit, and a correlation visualization unit; A processing unit is selected to select PID control parameter data and data related to PID control parameters from the preprocessed data; and to form an input matrix from the two parameter data. The coefficient calculation unit is used to calculate the Pearson correlation coefficient between two selected parameters. The correlation visualization unit is used to obtain the Pearson correlation coefficient between various parameters after permuting and combining them, and then visualize the correlation.
[0016] As a further limitation of the technical solution of the present invention, the coefficient calculation unit is used to define the Pearson correlation coefficient between two selected parameters as the quotient of the covariance and standard deviation between the two parameters: By estimating the sample covariance and standard deviation, the Pearson correlation coefficient can be obtained. : By The Pearson correlation coefficient is obtained by estimating the mean of the standard scores of the sample points. : in , and They are respectively for The standard score, sample mean, and sample standard deviation of the sample.
[0017] As a further limitation of the technical solution of the present invention, the data clustering module includes a silhouette coefficient calculation unit, a cluster number determination unit, and a clustering unit; The silhouette coefficient calculation unit is used to calculate the silhouette coefficient of each vector in the cluster; where, the first... The contour coefficients of each object are: In the formula express The distance from a vector to all other points in the cluster it belongs to. express The average distance from a vector to all points within a cluster that does not contain it; The cluster number determination unit is used to select clusters defined by assigning values to the cluster number definition. The contour coefficient has an optimal solution as the final number of clusters; Clustering unit, used to cluster AMT clutch control data according to the number of clusters; in, In the formula, For the first There are 3 cluster sets, with cluster centers being 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... ; For the first Clusters The number of samples contained in it; the minimum value can be obtained by selecting the criterion function: Right now The cluster center of a class is chosen as the mean of the samples in that class.
[0018] As a further limitation of the technical solution of the present invention, the training processing module includes a dataset partitioning unit, a classification interval determination unit, and a training processing output unit. The dataset partitioning unit is used to shuffle the data related to the PID control parameters with a certain correlation greater than the set value and then divide the dataset into a training dataset and a prediction dataset according to a set ratio. The classification margin determination unit is used to find the classification decision surface for the data points in the training dataset. The optimal decision surface, i.e., the classification margin, can be expressed as: in, For kernel function, For the target category, For non-negative Lagrange multipliers, For offset terms; kernel function In the formula The center of the kernel function, For vectors and Euclidean distance; The training output unit is used to select the classification margin with the largest margin from the nonlinear SVM model to obtain the training classification result.
[0019] As a further limitation of the technical solution of the present invention, the prediction processing module includes a prediction calculation unit and a prediction result processing unit; The prediction calculation unit is used to distribute the landing points through the prediction dataset and calculate the distribution accuracy based on the training classification results. The classification accuracy is calculated by dividing the number of correctly predicted results by the total number of results. The prediction result processing unit is used to select the number of correctly identified positive samples, the number of falsely reported negative samples, the number of correctly identified negative samples, and the number of missed positive samples. It selects the predicted category as the horizontal axis and the actual category as the vertical axis to construct a confusion matrix and visualizes it using a heatmap.
[0020] Thirdly, the present invention also provides an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory storing computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the clutch PID control parameter determination method as described in the first aspect.
[0021] Fourthly, the present invention also provides a non-transitory computer-readable storage medium that stores computer instructions that cause the computer to execute the clutch PID control parameter determination method as described in the first aspect.
[0022] As can be seen from the above technical solutions, the present invention has the following advantages: it can accurately predict the clutch PID control parameters required for newly configured AMT models, and judge the feasibility of the clutch PID control parameters through calculation, saving both manpower and time costs, which is of great significance for the forward development of AMT; and the method can also be applied to existing AMT models on the market to conduct feasibility analysis of clutch PID control parameters, further optimizing the current batch of models' operating status.
[0023] Furthermore, the design principle of this invention is reliable, the structure is simple, and it has a very wide range of application prospects.
[0024] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.
[0027] Figure 2 The figure shows the contour coefficients for an example.
[0028] Figure 3 This is a visualization of the clustering results under certain dimensions.
[0029] Figure 4 A visualization of the confusion matrix.
[0030] Figure 5 This is a schematic block diagram of an apparatus according to an embodiment of the present invention. Detailed Implementation
[0031] This invention discloses a method for determining PID control parameters of AMT clutches in commercial vehicles based on cluster centers. This method is applicable to the electro-pneumatic clutch control mode of Sinotruk AMT control systems currently on the market. It also solves the shortcomings of traditional PID control algorithms, which mainly rely on trial-and-error or empirical data methods for parameter determination, resulting in inaccurate parameter determination, strong subjectivity, and high time costs. This method first uses Pearson coefficients to verify data correlation, then performs clustering using the K-means clustering algorithm to determine the PID parameters at the cluster centers, then divides the dataset for SVM (Support Vector Machine) model training, and finally predicts the target PID parameters. Furthermore, this method uses real vehicle-related data for training, making the final parameters more suitable for practical applications. Finally, simulated data under the same conditions verifies the effectiveness and advanced nature of the method, making it very suitable for PID control under the current Sinotruk AMT control system. To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention. Figure 1This is a schematic flowchart illustrating the clutch PID control parameter determination method provided in this embodiment of the invention. The embodiments of this invention define the following parameters: clutch characteristic curve, clutch travel (clutch_dt), clutch disengagement force (clutch_df), engine reference torque (engine_rt), engine P2 point (engine_p2), average gearbox speed ratio range (trans_pi), and clutch booster cylinder type (clutch_bct). Simultaneously, the clutch PID control parameters are defined as follows: Pos_max, Pos_min, Kpmax_op, Kpmin-op, Kpmax_cls, Kpmin-cls, Ki, and Kd. The method includes the following steps: Step 1: Select clutch PID control parameter data and related data for different types of commercial vehicles, and preprocess the selected data. Step 2: In the preprocessed data, select the data related to the PID control parameters, and perform data correlation analysis on each pair of the selected data and the PID control parameter data to determine the data related to the PID control parameters whose correlation is greater than the set value. Step 3: Use the silhouette coefficient to determine the number of clusters, and cluster the data related to the PID control parameters whose correlation is greater than the set value; Step 4: Divide the data related to the PID control parameters with a correlation greater than the set value into a training dataset and a prediction dataset. Combine the clustering results and train the training dataset using an SVM model to obtain the training classification results. Step 5: Classify the prediction dataset and select PID control parameters based on the classification results.
[0032] Pearson coefficients are used to verify data correlation. Next, K-means clustering is employed to determine the PID parameters at the cluster centers. The dataset is then divided for SVM (Support Vector Machine) model training, ultimately predicting the target PID parameters. Furthermore, this method uses real-vehicle data for training, making the final parameters more suitable for practical applications. Finally, simulated data under identical conditions validates the effectiveness and advanced nature of the proposed method, making it highly suitable for PID control in the current Sinotruk AMT control system.
[0033] In some embodiments, the step of selecting PID control data for AMT clutches of multiple types of commercial vehicles and preprocessing the selected data includes: Step 11: Collect and summarize vehicle configuration and related parameter data for different AMT models; collect and summarize vehicle configuration and related parameter data for the three different AMT models currently used in the market; Step 12: Based on the AMT clutch PID control logic, classify and filter the data, and select the relevant data that affect the PID control parameters: clutch characteristic curve, engine reference torque engine_rt, engine P2 point engine_p2, transmission speed ratio range average value trans_pi, and clutch booster cylinder type clutch_bct as input; Step 13: Select plotting points in the image data and convert it into digital data. Perform data preprocessing on the clutch characteristic curve, using Origin to digitize the curve, transforming the image into formatted data consisting of clutch travel (clutch_dt) and clutch disengagement force (clutch_df).
[0034] We collect and summarize vehicle configuration and related parameter data for different vehicle models; based on the PID control logic of commercial vehicle clutches, we classify and filter the data, select relevant parameters that affect PID control as inputs, and in this case, we select factors that have a significant impact on PID control after parameter changes; we preprocess some image parameters and digitize the curves to convert the images into formatted data.
[0035] In some embodiments, the step of selecting data related to PID control parameters from the preprocessed data, performing pairwise correlation analysis between the selected data and the PID control parameter data, and determining the data related to PID control parameters with a correlation greater than a set value includes: Step 21: Select PID control parameter data and data related to PID control parameters from the preprocessed data; and form an input matrix from the two parameter data. Step 22: Calculate the Pearson correlation coefficient between the two selected parameters; Step 23: After permuting and combining the various parameters, obtain the Pearson correlation coefficients between the parameters and visualize the correlation.
[0036] Specifically, the steps for calculating the Pearson correlation coefficient between the two selected parameters include: The Pearson correlation coefficient between the two selected parameters is defined as the quotient of the covariance and standard deviation of the two parameters: By estimating the sample covariance and standard deviation, the Pearson correlation coefficient can be obtained. : By The Pearson correlation coefficient is obtained by estimating the mean of the standard scores of the sample points. : in , and They are respectively for The standard scores, sample mean, and sample standard deviation of the sample are used to determine the strength of the correlation between the parameters. Here, as the linear relationship between the two parameters strengthens, the Pearson correlation coefficient tends to be 1 or -1. After permuting and combining the parameters, the Pearson correlation coefficients between the parameters are obtained, and the correlation is visualized.
[0037] In some embodiments, the step of clustering data related to PID control parameters with a correlation greater than a set value by using a silhouette coefficient to determine the number of clusters includes: Step 31: Calculate the silhouette coefficient for each vector in the cluster; where, the first... The contour coefficients of each object are: In the formula express The distance from a vector to all other points in the cluster it belongs to. express The average distance from a vector to all points within a cluster that does not contain it; this allows for the calculation of the silhouette coefficient of each vector within the cluster; by assigning values to the defined number of clusters, a value is selected that is defined as... The contour coefficient has an optimal solution as the final number of clusters (in the example). ,Right now Figure 3 (Profile coefficient diagram of the embodiment shown).
[0038] Step 32: By assigning values to the defined number of clusters, select the clusters defined as follows: The silhouette coefficient has an optimal solution as the final number of clusters; the silhouette coefficient is introduced to judge the quality of clustering and determine the number of clusters. ; Step 33: Cluster the AMT clutch control data based on the number of clusters; Based on the above determination There are several pattern classes. In the formula, For the first There are 3 cluster sets, with cluster centers being 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... ; For the first Clusters The number of samples contained in it; the minimum value can be obtained by selecting the criterion function: Right now The cluster center of a class is chosen as the mean of the samples in that class.
[0039] A silhouette coefficient is introduced to judge the clustering effect of clusters, based on the basic parameter data related to PID control parameters. Perform K-means clustering, where For the first There are 3 cluster sets, with cluster centers being 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... The clutch PID control parameters for the cluster center are determined through a heuristic iterative process. The PID parameters corresponding to the cluster center are used as the standard for selecting PID parameters for this category.
[0040] In some embodiments, the steps of dividing the data related to PID control parameters with a known correlation greater than a set value into a training dataset and a prediction dataset, and then using an SVM model to train the training dataset based on the clustering results to obtain the training classification results include: Step 41: After shuffling the data related to the PID control parameters with a correlation greater than the set value, divide the dataset into a training dataset and a prediction dataset according to a set ratio; in this embodiment, 70% is divided into a training dataset and 30% into a prediction dataset. Step 42: For the data points in the training dataset, find the classification decision boundary, where the optimal decision boundary, i.e., the classification margin, can be expressed as: in, For kernel function, For the target category, For non-negative Lagrange multipliers, For offset terms; kernel function In the formula The center of the kernel function, For vectors and Euclidean distance; Step 43: The nonlinear SVM model is used to select the classification margin with the largest margin to obtain the training classification result.
[0041] The nonlinear SVM model used in this embodiment maps from a low-dimensional space to a high-dimensional space, based on a Gaussian kernel function. To select the classification interval with the largest margin, where The center of the kernel function, For vectors and The Euclidean distance between two vectors, the Gaussian kernel function, decreases monotonically as the distance between them increases. Control the range of the Gaussian kernel function; this will give us the support vectors we need, which is the training classification result.
[0042] The dataset is divided into training and prediction datasets according to a certain ratio, with the training dataset usually being larger than the prediction dataset. For the data points in the training dataset, a classification decision boundary is sought. For each class that can be divided, there is an optimal decision boundary, i.e., the classification margin. The classification margin with the largest margin, i.e., the support vector, is found through the constrained Lagrange multiplier method. Thus, the training classification result is obtained.
[0043] In some embodiments, the step of classifying the prediction dataset and selecting PID control parameters based on the classification results includes: Step 51: Distribute the landing points using the prediction dataset and calculate the distribution accuracy based on the training classification results. The classification accuracy is calculated as the number of correctly predicted results / the total number of results. Step 52: Select the number of correctly identified positive samples, the number of falsely reported negative samples, the number of correctly identified negative samples, and the number of missed positive samples. Select the horizontal axis as the predicted category and the vertical axis as the actual category to construct a confusion matrix and visualize it using a heatmap.
[0044] The distribution of landing points is predicted using the dataset, and the distribution accuracy is calculated based on the training classification results. The feasibility of this classification is determined based on the distribution accuracy. A confusion matrix is constructed to visualize the evaluation metrics of the classification algorithm results. Figure 5 The confusion matrix constructed for this embodiment is used to visualize the classification results.
[0045] This invention provides a clutch PID control parameter determination device, including a data collection and preprocessing module, a data correlation analysis module, a data clustering module, a training processing module, and a prediction processing module; The data collection and preprocessing module is used to select clutch PID control parameter data and related data of different types of commercial vehicles and to preprocess the selected data. The data correlation analysis module is used to select data related to PID control parameters from the preprocessed data, perform data correlation analysis on each pair of the selected data and the PID control parameter data, and determine the data related to PID control parameters whose correlation is greater than the set value. The data clustering module is used to determine the number of clusters by using the silhouette coefficient, and to cluster the data related to the PID control parameters whose correlation is greater than the set value. The training processing module is used to divide the data related to the PID control parameters with a certain correlation greater than the set value into training datasets and prediction datasets. The training dataset is then trained using an SVM model based on the clustering results to obtain the training classification results. The prediction processing module is used to classify the prediction dataset and select PID control parameters based on the classification results.
[0046] In some embodiments, the data collection and preprocessing module includes a collection and summarization unit, a classification and filtering unit, and a data preprocessing unit; The data collection and aggregation unit is used to collect and aggregate vehicle configuration and related parameter data for different AMT models; The classification and filtering unit is used to classify and filter various data based on the AMT clutch PID control logic, and select relevant data that affect the PID control parameters. The data preprocessing unit is used to select plotted points in image data and convert them into digital data.
[0047] In some embodiments, the data correlation analysis module includes a selection processing unit, a coefficient calculation unit, and a correlation visualization unit; A processing unit is selected to select two parameter data from the parameters related to the clutch PID control parameters to form an input matrix; The coefficient calculation unit is used to calculate the Pearson correlation coefficient between two selected parameters. The correlation visualization unit is used to obtain the Pearson correlation coefficient between various parameters after permuting and combining them, and then visualize the correlation.
[0048] In some embodiments, the coefficient calculation unit is used to define the Pearson correlation coefficient between two selected parameters as the quotient of the covariance and standard deviation between the two parameters: By estimating the sample covariance and standard deviation, the Pearson correlation coefficient can be obtained. : By The Pearson correlation coefficient is obtained by estimating the mean of the standard scores of the sample points. : in , and They are respectively for The standard score, sample mean, and sample standard deviation of the sample.
[0049] In some embodiments, the data clustering module includes a silhouette coefficient calculation unit, a cluster number determination unit, and a clustering unit; The silhouette coefficient calculation unit is used to calculate the silhouette coefficient of each vector in the cluster; where, the first... The contour coefficients of each object are: In the formula express The distance from a vector to all other points in the cluster it belongs to. express The average distance from a vector to all points within a cluster that does not contain it; The cluster number determination unit is used to select clusters defined by assigning values to the cluster number definition. The contour coefficient has an optimal solution as the final number of clusters; Clustering unit, used to cluster AMT clutch control data according to the number of clusters; in, In the formula, For the first There are cluster sets, with cluster centers being . ; For the first Clusters The number of samples contained in it; the minimum value can be obtained by selecting the criterion function: Right now The cluster center of a class is chosen as the mean of the samples in that class.
[0050] In some embodiments, the training processing module includes a dataset partitioning unit, a classification margin determination unit, and a training processing output unit; Data set partitioning unit is used to shuffle the data and divide it into training dataset and prediction dataset according to a set ratio; The classification margin determination unit is used to find the classification decision surface for the data points in the training dataset. The optimal decision surface, i.e., the classification margin, can be expressed as: in, For kernel function, For the target category, For non-negative Lagrange multipliers, For offset terms; kernel function In the formula The center of the kernel function, For vectors and Euclidean distance; The training output unit is used to select the classification margin with the largest margin from the nonlinear SVM model to obtain the training classification result.
[0051] In some embodiments, the prediction processing module includes a prediction calculation unit and a prediction result processing unit; The prediction calculation unit is used to distribute the landing points through the prediction dataset and calculate the distribution accuracy based on the training classification results. The classification accuracy is calculated by dividing the number of correctly predicted results by the total number of results. The prediction result processing unit is used to select the number of correctly identified positive samples, the number of falsely reported negative samples, the number of correctly identified negative samples, and the number of missed positive samples. It selects the predicted category as the horizontal axis and the actual category as the vertical axis to construct a confusion matrix and visualizes it using a heatmap.
[0052] This invention also provides an electronic device, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other via the communication bus. The communication bus can be used for information transmission between the electronic device and sensors. The processor can call logical instructions in the memory to execute the following method: Step 1: Select clutch PID control parameter data and related data of different types of commercial vehicles, and preprocess the selected data; Step 2: Select data related to PID control parameters from the preprocessed data, perform data correlation analysis on the selected data and PID control parameter data pairwise, and determine the data related to PID control parameters with a correlation greater than a set value; Step 3: Use silhouette coefficient to determine the number of clusters, and cluster the data related to PID control parameters with a correlation greater than a set value; Step 4: Divide the data related to PID control parameters with a correlation greater than a set value into a training dataset and a prediction dataset, and train the training dataset using an SVM model based on the clustering results to obtain the training classification result; Step 5: Perform classification operation on the prediction dataset, and select PID control parameters based on the classification result.
[0053] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0054] This invention provides a non-transitory computer-readable storage medium storing computer instructions that cause a computer to execute the method provided in the above-described method embodiments. For example, the method includes: Step 1: Selecting clutch PID control parameter data and related data for different types of commercial vehicles, and preprocessing the selected data; Step 2: Selecting data related to PID control parameters from the preprocessed data, performing pairwise correlation analysis between the selected data and the PID control parameter data, and determining the data related to PID control parameters with a correlation greater than a set value; Step 3: Determining the number of clusters using silhouette coefficients, and clustering the determined data related to PID control parameters with a correlation greater than a set value; Step 4: Dividing the determined data related to PID control parameters with a correlation greater than a set value into a training dataset and a prediction dataset, and training the training dataset using an SVM model based on the clustering results to obtain training classification results; Step 5: Performing a classification operation on the prediction dataset, and selecting PID control parameters based on the classification results.
[0055] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should also be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.
Claims
1. A method for determining PID control parameters for a clutch, characterized in that, Includes the following steps: Select clutch PID control parameter data and related data of different types of commercial vehicles, and perform data preprocessing on the selected data; In the preprocessed data, data related to PID control parameters are selected, and Pearson correlation analysis is performed on each of the selected data and the PID control parameter data to determine the data related to PID control parameters whose correlation is greater than the set value. The number of clusters is determined by the silhouette coefficient. The data related to the PID control parameters with a correlation greater than the set value are subjected to K-means clustering to obtain the PID control parameters corresponding to each cluster and its cluster center. The data related to the PID control parameters with a known correlation greater than the set value are divided into a training dataset and a prediction dataset. The clustering results are used as category labels. The training dataset is then trained using an SVM model based on the clustering results to obtain the training classification results. The predicted dataset is classified, and the PID control parameters corresponding to the cluster centers of the categories are selected based on the classification results.
2. The method for determining clutch PID control parameters according to claim 1, characterized in that, The steps for selecting clutch PID control parameter data and related data for different types of commercial vehicles and performing data preprocessing on the selected data include: Collect and summarize vehicle configuration and related parameter data for different AMT models; Based on the AMT clutch PID control logic, various data are categorized and filtered to select relevant data that affect the PID control parameters: Image data is converted into digital data by selecting plotted points.
3. The method for determining clutch PID control parameters according to claim 2, characterized in that, The steps involved in selecting data related to PID control parameters from the preprocessed data, performing Pearson correlation analysis on each pair of selected data and PID control parameter data, and determining the data related to PID control parameters with a correlation greater than the set value include: In the preprocessed data, PID control parameter data and data related to PID control parameters are selected; and the two parameter data are combined to form an input matrix. Calculate the Pearson correlation coefficient between the two selected parameters; After permuting and combining the various parameters, the Pearson correlation coefficients between the parameters are obtained, and the correlations are visualized.
4. The method for determining clutch PID control parameters according to claim 3, characterized in that, The steps for calculating the Pearson correlation coefficient between the two selected parameters include: The Pearson correlation coefficient between the two selected parameters is defined as the quotient of the covariance and standard deviation of the two parameters: By estimating the sample covariance and standard deviation, the Pearson correlation coefficient can be obtained. : By The Pearson correlation coefficient is obtained by estimating the mean of the standard scores of the sample points. : in , and They are respectively for The standard score, sample mean, and sample standard deviation of the sample.
5. The method for determining clutch PID control parameters according to claim 4, characterized in that, The steps for determining the number of clusters using silhouette coefficients, performing K-means clustering on the data related to PID control parameters with correlations greater than a set value, and obtaining the PID control parameters corresponding to each cluster and its cluster center include: Calculate the silhouette coefficient for each vector in the cluster; where, the first... The contour coefficients of each object are: In the formula express The distance from a vector to all other points in the cluster it belongs to. express The average distance from a vector to all points within a cluster that does not contain it; By assigning values to the definition of the number of clusters, the clusters are selected as defined. The contour coefficient has an optimal solution as the final number of clusters; Clustering of PID control parameter data with correlation greater than a set value is performed based on the number of clusters. in, There are several pattern classes. In the formula, For the first There are 3 cluster sets, with cluster centers being 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... ; For the first Clusters The number of samples contained in the sample; obtained by selecting the minimum value through the criterion function: Right now The cluster center of a class is chosen as the mean of the samples in that class.
6. The method for determining clutch PID control parameters according to claim 5, characterized in that, The steps for dividing the data related to PID control parameters with a known correlation greater than the set value into training and prediction datasets, using the clustering results as category labels, and training the training dataset using an SVM model based on the clustering results to obtain the training classification results include: After shuffling the data related to the PID control parameters whose correlation is greater than the set value, the datasets are divided into training datasets and prediction datasets according to a set ratio. For each data point in the training dataset, we search for a classification decision boundary, where the optimal decision boundary, or classification margin, can be expressed as: in, For kernel function, For the target category, For non-negative Lagrange multipliers, Offset term; kernel function In the formula The center of the kernel function, For vectors and European distance, For bandwidth parameters; The nonlinear SVM model used in conjunction with the clustering results is selected to obtain the classification interval with the largest margin, thus obtaining the training classification result.
7. The method for determining clutch PID control parameters according to claim 6, characterized in that, The steps for classifying the prediction dataset and selecting the PID control parameters for the corresponding cluster centers based on the classification results include: The distribution of landing points is determined by predicting the dataset, and the distribution accuracy is calculated based on the training classification results. The classification accuracy is calculated by dividing the number of correctly predicted results by the total number of results. The number of correctly identified positive samples, the number of falsely reported negative samples, the number of correctly identified negative samples, and the number of missed positive samples are selected. The horizontal axis is selected as the predicted category and the vertical axis as the actual category to construct a confusion matrix, which is then visualized using a heatmap.
8. A clutch PID control parameter determination device, characterized in that, It includes a data collection and preprocessing module, a data correlation analysis module, a data clustering module, a training processing module, and a prediction processing module; The data collection and preprocessing module is used to select clutch PID control parameter data and related data of different types of commercial vehicles and to preprocess the selected data. The data correlation analysis module is used to select data related to PID control parameters from the preprocessed data, perform Pearson correlation analysis on each pair of selected data and PID control parameter data, and determine the data related to PID control parameters whose correlation is greater than the set value. The data clustering module is used to determine the number of clusters using the silhouette coefficient, and to perform K-means clustering on the data related to the PID control parameters whose correlation is greater than the set value, so as to obtain the PID control parameters corresponding to each cluster and its cluster center. The training processing module is used to divide the data related to the PID control parameters with a certain correlation greater than the set value into training datasets and prediction datasets. The clustering results are used as category labels. The training dataset is trained through an SVM model in combination with the clustering results to obtain the training classification results. The prediction processing module is used to classify the prediction dataset and select the PID control parameters for the corresponding cluster centers based on the classification results.
9. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the clutch PID control parameter determination method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the clutch PID control parameter determination method as described in any one of claims 1 to 7.
Citation Information
Patent Citations
Metalearning-based vehicle OD point clustering method, system and device and storage medium
CN112215261A
Operation scheduling method, system and device of power system and medium
CN114219216A
