Control method of clutch, electronic device and vehicle

Abstract

The application relates to a clutch control method, electronic equipment and a vehicle, relates to the technical field of vehicles, and aims to solve the problem of high manufacturing cost of a driving system in the prior art. The method comprises the following steps: acquiring driving data in a vehicle running process and state information of a driving device of the vehicle; determining an estimated pressure of a clutch based on the driving data and the state information; and controlling a hydraulic system to operate based on the estimated pressure, so that the pressure of the clutch reaches a target pressure from the estimated pressure, the algorithm replaces the pressure sensor, and the manufacturing cost of the driving system is reduced.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and more specifically to a clutch control method, electronic equipment, and vehicle. Background Technology

[0002] In a drive system, the pressure sensor is a core component for precise control. It converts the pressure of the hydraulic system into an electrical signal and transmits it to the clutch control unit to control the clutch engagement and disengagement. Pressure sensor data, as crucial feedback information, directly affects driving performance indicators such as the smoothness of vehicle power output and shift response speed. If the pressure sensor malfunctions, the system will lose vital pressure monitoring data, leading to abnormal power transmission, shift jerking, and other problems, seriously impacting driving safety and the driving experience.

[0003] In related technologies, pressure sensors collect real-time pressure data from the hydraulic system to provide a stable pressure feedback signal to the control unit, ensuring the normal operation of vehicle power transmission and gear shifting control. However, in the pursuit of cost reduction and efficiency improvement, the drive system faces the challenge of excessively high manufacturing costs. Summary of the Invention

[0004] This invention provides a clutch control method, electronic device, and vehicle to solve the problem of high manufacturing cost of drive systems in the prior art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, a clutch control method is provided, comprising: acquiring driving data and state information of the vehicle's drive device during vehicle operation; determining a predicted clutch pressure based on the driving data and state information; and controlling the operation of a hydraulic system based on the predicted pressure to make the clutch pressure reach a target pressure from the predicted pressure.

[0007] Based on the aforementioned technical means, the estimated clutch pressure is determined by driving data and status information, and control is achieved based on the estimated pressure. This realizes the function of replacing the pressure sensor to detect the clutch pressure, which not only meets the pressure detection requirements, but also reduces the cost of hardware, effectively balancing the performance and cost of the drive system.

[0008] Furthermore, the driving data includes at least one of the following: engine speed, throttle signal, brake signal, and mileage; the status information includes at least one of the following: clutch request current, clutch PI control parameters, clutch clutch status, clutch target pressure, clutch corresponding output shaft speed, main pressure request pressure, and transmission oil temperature.

[0009] Based on the aforementioned technical methods, driving data can reflect the macroscopic behavior of the vehicle, while status information can reflect the microscopic control of the drive system. The combination of these two can completely reconstruct the entire process of "driver operation → power request → clutch execution → system feedback," accurately capturing the dynamic correlation from driving intention to clutch pressure prediction. Simultaneously, multi-source data provides rich input feature dimensions, helping to improve the accuracy of clutch pressure prediction.

[0010] Furthermore, based on driving data and status information, the estimated pressure of the clutch is determined, including: inputting driving data and status information into the pressure prediction model, and determining the estimated pressure of the clutch through the pressure prediction model.

[0011] Based on the aforementioned technical means, the pressure prediction model can learn the complex mapping relationship between driving data, status information, and clutch pressure. By fusing driving data and status information, it can determine the estimated clutch pressure, reduce the error in the estimated clutch pressure, and improve the accuracy of determining the estimated clutch pressure.

[0012] Furthermore, the pressure prediction model is trained through the following steps: obtaining a training dataset; the training dataset includes multiple training samples, each of which includes: sample driving data, sample state information, and actual clutch pressure; inputting the sample driving data and sample state information from the training samples into the initial pressure prediction model, and obtaining the estimated pressure corresponding to the training sample through the pressure prediction model; training the initial pressure prediction model based on the estimated pressure and actual pressure corresponding to multiple training samples until the iteration stopping condition is met, thus obtaining the pressure prediction model.

[0013] Based on the above technical means, the pressure prediction model is trained with sample driving data and state information as input and actual pressure as output, allowing the model to directly learn the "input-output" mapping relationship without the need for manual pre-setting of complex rules, thereby improving the accuracy of the pressure prediction model. In turn, when determining the estimated pressure of the clutch through the pressure prediction model, the accuracy of determining the estimated pressure of the clutch is improved.

[0014] Furthermore, the stress prediction model is either a random forest model or an extreme random tree model.

[0015] Based on the aforementioned technical methods, a complex nonlinear relationship exists between driving data, state information, and pressure. Random forest models or extreme random tree models excel at capturing these nonlinear relationships, automatically learning higher-order interactions between variables without prior assumptions about data distribution. Furthermore, the input data for pressure prediction models includes multi-source heterogeneous data; random forest models or extreme random tree models can directly process discrete features, avoiding information loss caused by improper feature preprocessing in traditional linear models (such as regression).

[0016] Furthermore, the above method also includes: analyzing driving data and status information based on the SHAP algorithm and / or feature importance analysis to obtain importance scores; filtering based on the importance scores to obtain filtered driving data and status information; and retraining the pressure prediction model based on the filtered driving data and status information.

[0017] Based on the above technical means, the contribution of each feature in the driving data and status information to the clutch predicted pressure can be quantified by calculating the importance score. The influence weight of different features (input data) on the pressure prediction result can be clarified, and the model algorithm can be improved in a targeted manner, key input data can be screened, and the clutch control logic can be optimized based on feature importance, thereby improving the pressure prediction accuracy and system control performance.

[0018] Furthermore, the above-mentioned acquisition of the training dataset includes: determining the interfering factors of each training sample; selecting target training samples from multiple training samples corresponding to each interfering factor based on the interfering factors and their weights; and obtaining the training dataset based on the target training samples.

[0019] Based on the above technical means, the influence of each interference factor on the pressure prediction model is quantified by weighting, and samples with strong interference and a lot of noise are filtered out, so that the training data is more focused on the core driving factors of clutch pressure, reducing the interference of irrelevant variables on the model, thereby improving the accuracy of the pressure prediction model.

[0020] Furthermore, interfering factors include: manufacturing parameters, fluid cleanliness, driving habits, environmental information, and clutch operating status.

[0021] Based on the aforementioned technical methods, the interference factors cover five dimensions: mechanical manufacturing (manufacturing parameters), hydraulic system (oil cleanliness), human operation (driving habits), external environment (environmental information), and equipment status (operating status). This comprehensively maps the entire chain of influencing factors of the clutch in practical applications, avoiding prediction bias caused by the omission of key variables in the pressure prediction model. Furthermore, by incorporating multiple interference factors, a realistic training dataset can be constructed, enabling the model to learn the pressure change patterns under different scenarios and improving its applicability in real-world complex working conditions.

[0022] Secondly, a clutch control device is provided, comprising: an acquisition unit for acquiring driving data and state information of the vehicle's drive unit during vehicle operation; a processing unit for determining a predicted clutch pressure based on the driving data and state information; and controlling the operation of a hydraulic system based on the predicted pressure to make the clutch pressure reach a target pressure from the predicted pressure.

[0023] Thirdly, an electronic device is provided, comprising: a processor and a memory; the memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the methods of the first aspect and any possible implementation thereof.

[0024] Fourthly, a vehicle is provided, comprising: the electronic equipment described in the third aspect above.

[0025] Fifthly, a computer-readable storage medium is provided, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the methods described in the first aspect and any possible implementation thereof.

[0026] In a sixth aspect, a computer program product is provided, comprising computer instructions that, when executed on an electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0027] The beneficial effects of this invention are:

[0028] (1) By determining the estimated pressure of the clutch based on driving data and status information, and then controlling the clutch based on the estimated pressure, the function of replacing the pressure sensor to detect the clutch pressure is realized. This not only meets the pressure detection requirements, but also reduces the cost of hardware, effectively balancing the performance and cost of the drive system.

[0029] (2) Driving data can reflect the macroscopic behavior of the vehicle, while status information can reflect the microscopic control of the drive system. The combination of the two can completely reconstruct the entire process of "driver operation → power request → clutch execution → system feedback", accurately capturing the dynamic correlation from driving intention to clutch pressure prediction. At the same time, multi-source data provides rich input feature dimensions, which helps to improve the accuracy of clutch pressure prediction.

[0030] (3) The pressure prediction model is trained by taking sample driving data and state information as input and actual pressure as output, so that the model can directly learn the mapping relationship between "input and output" without the need for manual pre-setting of complex rules, which improves the accuracy of the pressure prediction model. In turn, when determining the estimated pressure of the clutch through the pressure prediction model, the accuracy of determining the estimated pressure of the clutch is improved.

[0031] (4) The pressure prediction model can learn the complex mapping relationship between driving data, status information and clutch pressure quality inspection. By integrating driving data and status information, it can determine the estimated pressure of the clutch, reduce the error of the estimated pressure of the clutch, and improve the accuracy of determining the estimated pressure of the clutch.

[0032] (5) There are complex nonlinear relationships between driving data, state information, and pressure. Random forest models or extreme random tree models are good at capturing these nonlinear relationships and can automatically learn higher-order interactions between variables without prior assumptions about data distribution. In addition, the input data of the pressure prediction model contains multi-source heterogeneous data. Random forest models or extreme random tree models can directly process discrete features, avoiding information loss caused by improper feature preprocessing in traditional linear models (such as regression).

[0033] (6) By calculating the importance score, the contribution of each feature in the driving data and status information to the clutch predicted pressure can be quantified, the influence weight of different features (input data) on the pressure prediction result can be clarified, the model algorithm can be improved in a targeted manner, key input data can be screened, and the clutch control logic can be optimized based on the feature importance, thereby improving the pressure prediction accuracy and system control performance.

[0034] (7) By quantifying the impact of each interference factor on the pressure prediction model through weights, samples with strong interference and a lot of noise are filtered out, so that the training data can focus more on the core driving factors of clutch pressure, reduce the interference of irrelevant variables on the model, and thus improve the accuracy of the pressure prediction model.

[0035] (8) The interference factors cover five dimensions: mechanical manufacturing (manufacturing parameters), hydraulic system (oil cleanliness), human operation (driving habits), external environment (environmental information), and equipment status (operating status). This comprehensively maps the entire chain of influencing factors of the clutch in actual application, avoiding prediction bias caused by the omission of key variables in the pressure prediction model. Furthermore, by incorporating multiple interference factors, a training dataset that closely resembles reality can be constructed, enabling the model to learn the pressure change patterns under different scenarios and improving the model's applicability in real and complex working conditions. Attached Figure Description

[0036] Figure 1 A schematic flowchart illustrating a clutch control method provided in an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of the pressure control process of a dual-clutch transmission provided in an embodiment of the present invention;

[0038] Figure 3A schematic flowchart illustrating another clutch control method provided in an embodiment of the present invention;

[0039] Figure 4 A schematic diagram illustrating the training process of a pressure prediction model provided in an embodiment of the present invention;

[0040] Figure 5 A schematic diagram illustrating the training process of another pressure prediction model provided in an embodiment of the present invention;

[0041] Figure 6 This is a schematic diagram of the structure of a clutch control device provided in an embodiment of the present invention;

[0042] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0043] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0044] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0045] Definitions:

[0046] The Shapley Additive Explanations (SHAP) algorithm is based on the Shapley value in cooperative game theory. It explains the model's decision-making logic by quantifying the contribution of each feature to the model's prediction results, thus addressing the interpretability problem of machine learning models (especially tree models and ensemble models). The SHAP value represents "the average contribution of a feature to the prediction result across all possible combinations of features," intuitively reflecting the positive or negative influence of each feature on the result, as well as the interactions between features. In this application, the SHAP value represents the contribution of driving data and state information to clutch pressure prediction.

[0047] In a drive system, the pressure sensor is a core component for precise control. It converts the pressure of the hydraulic system into an electrical signal and transmits it to the clutch control unit to control the clutch engagement and disengagement. Pressure sensor data, as crucial feedback information, directly affects driving performance indicators such as the smoothness of vehicle power output and shift response speed. If the pressure sensor malfunctions, the system will lose vital pressure monitoring data, leading to abnormal power transmission, shift jerking, and other problems, seriously impacting driving safety and the driving experience.

[0048] In related technologies, pressure sensors collect real-time pressure data from the hydraulic system to provide a stable pressure feedback signal to the control unit, ensuring the normal operation of vehicle power transmission and gear shifting control. However, in the pursuit of cost reduction and efficiency improvement, the drive system faces the challenge of excessively high manufacturing costs.

[0049] Based on this, this application proposes a clutch control method, electronic device, and vehicle, including: acquiring driving data and state information of the vehicle's drive unit during operation; determining the estimated clutch pressure based on the driving data and state information; and controlling the operation of the hydraulic system based on the estimated pressure to bring the clutch pressure from the estimated pressure to the target pressure. By determining the estimated clutch pressure based on driving data and state information, and thus controlling the clutch based on the estimated pressure, the function of detecting clutch pressure by replacing pressure sensors is realized. This not only meets the pressure detection requirements but also reduces hardware costs, effectively balancing drive system performance and cost.

[0050] Understandably, by using an algorithm to determine the estimated clutch pressure based on driving data and status information, a pressure sensor can be replaced. This means that no pressure sensor needs to be installed in the clutch itself; the clutch pressure value is estimated in real time solely based on vehicle driving data and drive unit status information. This purely software-based solution completely eliminates the need for a hardware pressure sensor, achieving "zero pressure sensor" at the hardware level, thereby reducing the cost of the drive system.

[0051] In some embodiments, the executing entity of the clutch control method provided in this invention is a clutch control device, which may be an electronic device. Optionally, the clutch control device may also be a vehicle including the aforementioned electronic device; or, the clutch control device may also be a functional module in the aforementioned electronic device used to execute the clutch control method, etc. This invention does not impose any limitations on these aspects.

[0052] like Figure 1 As shown, a clutch control method is provided, including:

[0053] S101. Obtain driving data and status information of the vehicle's drive system during vehicle operation.

[0054] The driving data includes at least one of the following: engine speed, throttle signal, brake signal, and mileage.

[0055] The status information includes at least one of the following: clutch request current, clutch PI control parameters, clutch status, clutch target pressure, clutch corresponding output shaft speed, main pressure request pressure, and transmission oil temperature.

[0056] As one possible implementation, driving data and status information are acquired through sensors installed in the vehicle. Specifically, driving data is mainly collected by sensors such as speed sensors and accelerator pedal sensors and converted into electrical signals, which are then transmitted to the control unit via a bus. Status information is mainly collected by components such as the clutch control unit and is systematically aggregated to the control unit via the bus system.

[0057] S102. Based on driving data and status information, determine the estimated pressure of the clutch.

[0058] Understandably, driving data (such as engine speed and throttle signal) reflects the driver's intentions and vehicle operating conditions (such as starting, accelerating, and braking), while status information (such as clutch request current and output shaft speed) characterizes the current control strategy and the mechanical state of the drive unit. By integrating these data, a mapping relationship between "operating conditions - control requirements - pressure parameters" can be established, enabling dynamic prediction of pressure.

[0059] One possible approach is to determine the estimated clutch pressure by establishing a working condition pressure mapping table. Specifically, a "working condition-pressure" mapping table is pre-established based on the vehicle's historical working conditions. By analyzing driving data and status information in real time to match the corresponding working conditions, the estimated pressure value is directly determined by querying the "working condition-pressure" mapping table based on the working conditions.

[0060] As another possible implementation, the estimated clutch pressure can be determined using a deep learning network model. Specifically, a deep learning network model is constructed and trained based on historical driving data, historical state information, and historical clutch pressure. The driving data and state information are then directly input into the deep learning network model, which outputs the estimated pressure.

[0061] As another possible approach, the initial estimated pressure is first calculated using a physical model of the clutch, and then the deviation is corrected using a machine learning algorithm to obtain the estimated pressure of the clutch.

[0062] S103. Control the operation of the hydraulic system based on the estimated pressure so that the pressure of the clutch reaches the target pressure from the estimated pressure.

[0063] One possible approach is to compare the estimated pressure of the clutch with the target pressure and dynamically adjust the clutch parameters. These parameters include the solenoid valve current and voltage. If the estimated pressure is lower than the target pressure, the current is increased; if it is higher than the target pressure, the current is decreased.

[0064] like Figure 2 As shown, Figure 2 The pressure control process for a dual-clutch transmission includes: inputting the odd-number clutch requested torque, and through interpolation, converting the odd-number clutch requested torque demand into odd-number clutch requested pressure. The odd-number clutch requested pressure is the "initial power request" of the control logic.

[0065] The Transmission Control Unit (TCU) acts as the "control center," receiving the requested pressure from the odd-numbered clutches and converting it into the first pressure (Demand pressure) of the odd-numbered clutches. It then calculates the requested current of the odd-numbered clutch solenoid valves through interpolation, thus converting pressure into actuation current. Based on the requested current of the odd-numbered clutches, it drives the odd-numbered clutch solenoid valves to actuate, changing the pressure of the odd-numbered clutches and thereby actuating the clutch pistons.

[0066] Pressure closed-loop feedback: The "virtual pressure sensor odd number" collects data such as the requested pressure, temperature, and mileage of the odd number clutch in real time to determine the estimated pressure of the odd number clutch and sends it back to the TCU software to form closed-loop control. If the deviation between the estimated pressure of the odd number clutch and the first pressure of the odd number clutch exceeds the threshold, the TCU software will readjust the PID parameters and correct the current output to ensure accurate pressure control.

[0067] Current closed-loop control: The current sensor collects the actual current of the odd clutch solenoid valve in real time. By comparing the actual current of the odd clutch solenoid valve with the requested current of the odd clutch solenoid valve, if the deviation between the actual current and the requested current exceeds the threshold, the TCU software will readjust the PID parameters to correct the actual current of the solenoid valve.

[0068] It should be understood that the above example of an odd-numbered clutch illustrates the pressure control process of an odd-numbered clutch. The pressure control process of an even-numbered clutch can be referred to the example of an odd-numbered clutch, and will not be repeated here.

[0069] It should be noted that the requested pressure and the first pressure are essentially different stages of the same signal. The requested pressure is the initial pressure demand command, which can be understood as a preliminary indication of "how much pressure the system needs from the clutch to transmit torque." The first pressure is the product of the requested pressure after "interpolation conversion" and "PID adjustment," and it is the precise pressure command actually used to drive the solenoid valve and control the clutch piston. Therefore, the status information of the drive device can also include: the first pressure of the clutch.

[0070] Therefore, by determining the estimated clutch pressure based on driving data and status information, and then controlling the clutch based on the estimated pressure, the system can replace the pressure sensor in detecting clutch pressure. This not only meets the pressure detection requirements but also significantly reduces the cost of hardware redundancy, effectively balancing system performance and cost.

[0071] In some embodiments, driving data and status information are input into a pressure prediction model, which can capture the nonlinear mapping relationship between "data" and "pressure" to determine the estimated pressure of the clutch. Therefore, S102 above includes: inputting driving data and status information into a pressure prediction model, and determining the estimated pressure of the clutch through the pressure prediction model.

[0072] As one possible implementation, the pressure prediction model is trained through the following steps: obtaining a training dataset; inputting the sample driving data and sample state information from the training samples into the initial pressure prediction model, and obtaining the estimated pressure corresponding to the training samples through the pressure prediction model; training the initial pressure prediction model based on the estimated pressure and actual pressure corresponding to multiple training samples until the iteration stopping condition is met, thus obtaining the pressure prediction model.

[0073] The training dataset includes multiple training samples, each of which includes: sample driving data, sample state information, and actual clutch pressure.

[0074] It should be noted that the training method of the pressure prediction model and the control method of the clutch can be deployed on the same electronic device or on different electronic devices; no specific restrictions are imposed here.

[0075] In one possible implementation, obtaining the training dataset includes: determining the perturbation factors for each training sample; selecting target training samples from multiple training samples corresponding to each perturbation factor based on the perturbation factors and their weights; and obtaining the training dataset based on the target training samples.

[0076] Interference factors include: manufacturing parameters, fluid cleanliness, driving habits, environmental information, and clutch operating status.

[0077] In another possible implementation, a training sample can have multiple interfering factors, and the interfering factors for each training sample can be determined. Based on the interfering factors and their weights, the comprehensive interference index of each training sample is calculated, a screening threshold is set, and samples with comprehensive interference indices below the threshold are selected to obtain target training samples. Based on the target training samples, the training dataset is obtained.

[0078] Understandably, by quantifying the impact of various interfering factors (such as samples with large fluctuations in manufacturing parameters or severe oil contamination) on the pressure prediction model, samples with strong interference and a lot of noise are filtered out, making the training data more focused on the core driving factors of clutch pressure, reducing the interference of irrelevant variables on the model, and thus improving the accuracy of the pressure prediction model.

[0079] Meanwhile, the interference factors cover five dimensions: mechanical manufacturing (manufacturing parameters), hydraulic system (oil cleanliness), human operation (driving habits), external environment (environmental information), and equipment status (operating status). This comprehensively maps the entire chain of influencing factors of the clutch in practical applications, avoiding prediction bias caused by the omission of key variables in the pressure prediction model. Furthermore, by incorporating multiple interference factors, a realistic training dataset can be constructed, enabling the model to learn the pressure change patterns under different scenarios and improving its applicability in real-world complex working conditions.

[0080] In some embodiments, the stress prediction model is a random forest model or an extreme random tree model.

[0081] It is understandable that there are complex nonlinear relationships between driving data, state information, and stress. Random forest models or extreme random tree models excel at capturing these nonlinear relationships, automatically learning higher-order interactions between variables without prior assumptions about data distribution. Furthermore, the input data for stress prediction models includes multi-source heterogeneous data; random forest models or extreme random tree models can directly handle discrete features, avoiding information loss caused by improper feature preprocessing in traditional linear models (such as regression).

[0082] Meanwhile, compared to deep learning models (such as neural networks), random forests / extreme random trees have faster inference speeds, which can meet the real-time requirements of clutch control.

[0083] Therefore, the clutch pressure is nonlinearly affected by temperature and wear. By inputting driving data and status information into the pressure prediction model, the model can determine the estimated clutch pressure. By learning from multiple training samples, the model can automatically generate the relationship between pressure and driving data and status information, thereby improving the accuracy of pressure prediction.

[0084] In some embodiments, such as Figure 3 As shown, the above method also includes:

[0085] S301. Analyze driving data and status information based on the SHAP algorithm and / or feature importance analysis method to obtain importance scores.

[0086] One possible implementation involves analyzing driving data and state information using the SHAP algorithm to obtain an importance score. Based on this, training samples are input into the stress prediction model, and the mean of the absolute values ​​of the SHAP values ​​for each sample's driving data and state information is calculated. This mean is used as the importance score. Here, SHAP represents the marginal contribution of each feature to the prediction result; the larger the absolute value of the SHAP value, the higher the importance.

[0087] As another possible implementation, the importance of features can be directly assessed through the reduction in the Gini coefficient or the ranking of importance built into the stress prediction model, resulting in an importance score for each feature. For example, the contribution of a feature to stress prediction can be measured by calculating the reduction in the Gini coefficient when a feature splits. The greater the reduction in the Gini coefficient, the more important the feature.

[0088] As another possible implementation, the driving data and status information are analyzed based on the SHAP algorithm to obtain the first importance score, and the driving data and status information are analyzed based on the feature importance analysis method to obtain the second importance score. The first importance score and the second importance score are weighted and summed to obtain the total importance score.

[0089] S302. Based on the importance score, filter the driving data and status information to obtain the filtered driving data and status information, and retrain the pressure prediction model based on the filtered driving data and status information.

[0090] As one possible approach, an importance score threshold is determined based on historical importance scores. Driving data and status information with scores below the importance score threshold are then removed, generating filtered driving data and status information. The stress prediction model is then retrained using the new driving data and status information.

[0091] As another possible implementation, we can directly select the top K features by importance score. Based on this, we sort the features in descending order of importance score, select the top K features, and construct new driving data and status information.

[0092] As another possible approach, the stress prediction model is trained using all features (driving data and status information), and the importance score of each feature is calculated. The feature with the lowest score is removed, and the model is retrained with the remaining features until the number of remaining features reaches a preset value or the model performance no longer improves.

[0093] As another possible approach, redundant features are eliminated by combining importance scores with feature correlations. Based on this, a feature correlation matrix is ​​calculated to identify highly correlated feature pairs. Among these highly correlated feature pairs, features with higher importance scores are retained, while redundant features with lower scores are eliminated.

[0094] Therefore, by calculating importance scores, the contribution of each feature in driving data and status information to the clutch predicted pressure can be quantified, the influence weight of different features (input data) on the pressure prediction results can be clarified, the model algorithm can be improved in a targeted manner, key input data can be screened, and the clutch control logic can be optimized based on feature importance, thereby improving the pressure prediction accuracy and system control performance.

[0095] In some embodiments, the stress prediction model can be an extreme random tree model. Extreme random trees and random forests both belong to tree-based ensemble learning models, and both optimize prediction results by constructing multiple decision trees. However, extreme random trees further enhance randomness during classifier generation, randomly selecting features and thresholds when splitting nodes, rather than splitting based on a local optimum strategy like random forests. This characteristic makes extreme random trees more robust, effectively reducing model variance, and often achieving higher prediction accuracy when dealing with complex data and noise interference.

[0096] However, extreme randomness also brings limitations: due to the high randomness of feature selection and splitting logic, traditional feature importance assessment methods based on Gini coefficient or node splitting gain are difficult to accurately reflect the true contribution of features in extreme random trees, resulting in weakened model interpretability and making it difficult to analyze the impact of features on prediction results.

[0097] To balance model performance and interpretability, an extreme randomized tree combined with the SHAP algorithm can be used: such as Figure 4 As shown, a training set (training dataset) is first obtained, and then the entire dataset is determined. A random splitting algorithm and features are used for training to obtain a trained stress prediction model and stress prediction results. Next, the SHAP algorithm is used to quantify the contribution of each feature to the stress prediction results and identify key influencing factors. The training set is adjusted based on feature importance, and the model is retrained using the new training set. This retains the predictive advantages of extreme random trees while reducing model complexity and improving the credibility of feature interpretation, providing a clear basis for subsequent clutch control strategy optimization and system improvement.

[0098] The extreme random tree model includes multiple trees, specifically: decision tree 1 to decision tree T.

[0099] like Figure 5 As shown below, the training process of the pressure prediction model is illustrated using a dual-clutch transmission as an example:

[0100] (1) Data Acquisition: The dataset mainly consists of a large amount of vehicle calibration data and operating data, including: requested current of even clutches, requested current of odd clutches, proportional-integral control parameters of even clutches, proportional-integral control parameters of odd clutches, throttle signal, brake signal, mileage, clutch status of even clutches, clutch status of odd clutches, main pressure requested pressure, output shaft speed corresponding to even clutches, output shaft speed corresponding to odd clutches, engine speed, transmission oil temperature, first pressure of even clutches, first pressure of odd clutches, requested pressure of odd clutches, requested pressure of even clutches, etc.

[0101] (2) Data batch processing: data transformation, data classification and dataset construction.

[0102] Data transformation: Convert the collected data into a uniform format. Data classification: Classify and organize the transformed data according to different operating parameters. Dataset construction: Divide the collected data into training, validation, and test sets, taking into account interference factors, to ensure the quality and diversity of the dataset.

[0103] The interfering factors include: N1. Manufacturing consistency: Differences in component manufacturing and assembly may lead to inconsistencies in system characteristics. N2. Oil cleanliness and solenoid valve performance: Deterioration in oil cleanliness may cause solenoid valve sticking; wear on the valve body and valve core or decrease in bolt torque may lead to leakage. N3. Driving habits: Different driving habits (such as throttle application and braking habits) will affect system pressure and clutch status. N4. Temperature and altitude: Temperature changes and altitude will affect oil viscosity, system pressure, and solenoid valve performance. N5. Operating conditions and malfunctions: Clutch status, pressure change slope, oil filling process, solenoid valve repeatability, and special operating conditions (such as PI and PT self-learning) will all affect the system.

[0104] (3) Training and validation of stress prediction model: including positive sample data augmentation, machine learning model training, Bayesian optimization of hyperparameters and determination of optimal feature parameters.

[0105] Positive sample data augmentation: Augmenting data according to its characteristics and actual needs (such as data resampling, lightweight neural network algorithms (such as support vector machines)).

[0106] In one possible implementation, for scenarios with insufficient positive sample data, data augmentation techniques are used to provide the model with richer and more diverse training samples. Data augmentation methods (such as oversampling, SMOTE algorithm, etc.) are introduced to generate new positive sample data. This ensures a reasonable data distribution, resulting in an augmented training dataset containing more effective positive samples.

[0107] Machine learning model training: Use the constructed dataset to train the model and adjust the parameters to improve the model performance.

[0108] In one possible implementation, a machine learning model is trained using an augmented dataset, enabling it to map from input features to output pressure. The dataset is divided into training and validation sets. A random forest or extreme random tree model is used for training to fit the relationship between input features and output pressure. The model's performance is evaluated using cross-validation (e.g., 5-fold cross-validation), resulting in a well-trained machine learning model.

[0109] Bayesian hyperparameter optimization: Based on the validation results, optimize the model structure and hyperparameters to improve the model's generalization ability.

[0110] In one possible implementation, a Bayesian optimization algorithm is used to automatically search for the model's hyperparameters to find the optimal parameter combination and improve model performance. The range of hyperparameters to be optimized (such as the number of trees, maximum depth, minimum number of samples, etc.) is defined, and the Bayesian optimization method is used to efficiently search for the optimal parameter combination. The performance of the optimized model is then verified to ensure accuracy and stability.

[0111] Determine feature parameters: refine feature parameters and determine the final optimized machine learning model.

[0112] One possible implementation involves identifying and extracting the most critical feature parameters for the model's predictions, simplifying model complexity and improving interpretability. Using the SHAP algorithm or feature importance analysis, the contribution of each feature to the model output is evaluated, and redundant or unimportant features are removed to reduce model complexity.

[0113] (4) Hardware-in-the-loop test verification: build test platform, deploy model, verify feasibility and analyze results.

[0114] Test Platform Setup: A hardware-in-the-loop test platform is built to simulate the real controller operating environment. Model Deployment: The trained model algorithm is deployed into the TCU software. Feasibility Verification: The feasibility and stability of the model and method in the hardware environment are verified, covering normal driving conditions, self-learning and adaptive operation of the vehicle powertrain, and canister cleaning conditions. Results Analysis: Experimental data is collected, and the model performance and its impact on the control strategy are analyzed.

[0115] (5) Vehicle testing and verification: Model deployment and vehicle testing and verification.

[0116] Model Deployment: Deploy the trained model algorithm into the TCU software. Vehicle Testing and Validation: Verify the feasibility and stability of the model and method in a vehicle environment, including not only normal driving conditions, self-learning and self-adaptation of the vehicle powertrain, and charcoal canister cleaning, but also winter, summer, and high-altitude tests.

[0117] The foregoing mainly describes the solutions provided by the embodiments of the present invention from a methodological perspective. To achieve the above functions, the clutch control device or electronic device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0118] According to the above method, the control device or electronic device of the clutch can be exemplarily divided into functional modules. For example, the control device or electronic device of the clutch may include functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in the embodiments of the present invention is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0119] Reference Figure 6 The clutch control device 600 provided in this embodiment of the invention includes: an acquisition unit 601 and a processing unit 602.

[0120] The acquisition unit 601 is used to acquire driving data and the status information of the vehicle's drive device during vehicle operation.

[0121] The processing unit 602 is used to determine the estimated pressure of the clutch based on driving data and status information; and to control the operation of the hydraulic system based on the estimated pressure so that the clutch pressure reaches the target pressure from the estimated pressure.

[0122] In some embodiments, driving data includes at least one of the following: engine speed, throttle signal, brake signal, and mileage; status information includes at least one of the following: clutch request current, clutch PI control parameters, clutch clutch status, clutch target pressure, clutch corresponding output shaft speed, main pressure request pressure, and transmission oil temperature.

[0123] In some embodiments, the processing unit 602 is specifically used to input driving data and status information into the pressure prediction model, and to determine the estimated pressure of the clutch through the pressure prediction model.

[0124] In some embodiments, the processing unit 602 is specifically used to acquire a training dataset; the training dataset includes multiple training samples, each training sample including: sample driving data, sample state information, and actual clutch pressure; the sample driving data and sample state information in the training samples are input into an initial pressure prediction model, and the estimated pressure corresponding to the training sample is obtained through the pressure prediction model; the initial pressure prediction model is trained according to the estimated pressure and actual pressure corresponding to multiple training samples until the iteration stopping condition is met, and the pressure prediction model is obtained.

[0125] In some embodiments, the stress prediction model is a random forest model or an extreme random tree model.

[0126] In some embodiments, the processing unit 602 is further configured to analyze driving data and status information based on the SHAP algorithm and / or feature importance analysis to obtain an importance score; filter the driving data and status information based on the importance score to obtain filtered driving data and status information; and retrain the pressure prediction model based on the filtered driving data and status information.

[0127] In some embodiments, the processing unit 602 is further configured to determine the interfering factors of each training sample; select target training samples from multiple training samples corresponding to each interfering factor according to the interfering factors and their weights; and obtain a training dataset according to the target training samples.

[0128] In some embodiments, interfering factors include: manufacturing parameters, fluid cleanliness, driving habits, environmental information, and clutch operating status.

[0129] like Figure 7 As shown, the electronic device 700 provided in this embodiment of the invention includes, but is not limited to, a processor 701 and a memory 702.

[0130] The memory 702 described above is used to store the executable instructions of the processor 701. It is understood that the processor 701 is configured to execute instructions to implement the methods in the above embodiments.

[0131] It should be noted that those skilled in the art will understand that Figure 7 The electronic device structure shown does not constitute a limitation on the electronic device; the electronic device may include, but is not limited to, other electronic devices. Figure 7 This may indicate more or fewer components, or a combination of certain components, or a different arrangement of components.

[0132] Processor 701 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in memory 702, and by calling data stored in memory 702, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Processor 701 may include one or more processing units. Optionally, processor 701 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 701.

[0133] The memory 702 can be used to store software programs and various data. The memory 702 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required by at least one functional module (such as a determination unit, processing unit, etc.), etc. Furthermore, the memory 702 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0134] In an exemplary embodiment, a vehicle is also provided, including the electronic equipment described above.

[0135] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 702 including instructions, which can be executed by a processor 701 of an electronic device 700 to implement the methods in the above embodiments.

[0136] In actual implementation, Figure 6 The functions of the acquisition unit 601 and the processing unit 602 can both be provided by Figure 7 The processor 701 calls the computer program stored in the memory 702 to implement the process. The specific execution process can be found in the description of the method section in the previous embodiment, and will not be repeated here.

[0137] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0138] In an exemplary embodiment, the present invention also provides a computer program product including one or more instructions, which can be executed by a processor 701 of an electronic device to perform the methods described above.

[0139] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of an electronic device, they implement the various processes of the above method embodiments and achieve the same technical effect as the above method. To avoid repetition, they will not be described again here.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0141] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0142] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0143] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0144] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of 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, ROM, RAM, magnetic disks, or optical disks.

[0145] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention.

Claims

1. A method for controlling a clutch, characterized in that, The clutch does not contain a pressure sensor; including: Acquire driving data and status information of the vehicle's drive system during vehicle operation. The driving data and status information are input into the pressure prediction model, and the estimated pressure of the clutch is determined by the pressure prediction model; the estimated pressure is used to replace the pressure detected by the pressure sensor. The hydraulic system is controlled to operate based on the estimated pressure, so that the pressure of the clutch reaches the target pressure from the estimated pressure. The step of controlling the operation of the hydraulic system based on the estimated pressure to bring the pressure of the clutch from the estimated pressure to the target pressure includes: The estimated pressure is compared with the target pressure to obtain a pressure comparison result, and the requested current of the solenoid valve is determined based on the pressure comparison result. The actual current of the solenoid valve is collected by a current sensor, and the actual current output to the solenoid valve is adjusted according to the comparison result between the actual current and the requested current, so that the pressure of the clutch reaches the target pressure from the estimated pressure.

2. The clutch control method according to claim 1, characterized in that, The driving data includes at least one of the following: engine speed, throttle signal, brake signal, and mileage; The status information includes at least one of the following: clutch request current, clutch PI control parameters, clutch status, clutch target pressure, clutch corresponding output shaft speed, main pressure request pressure, and transmission oil temperature.

3. The clutch control method according to claim 1, characterized in that, The pressure prediction model is trained through the following steps: Obtain a training dataset; the training dataset includes multiple training samples, each training sample including: sample driving data, sample state information, and actual clutch pressure; The driving data and state information of the training samples are input into the initial pressure prediction model, and the estimated pressure corresponding to the training samples is obtained through the pressure prediction model. The initial pressure prediction model is trained based on the estimated pressure and actual pressure corresponding to multiple training samples until the iteration stopping condition is met, thus obtaining the pressure prediction model.

4. The clutch control method according to claim 2 or 3, characterized in that, The stress prediction model is either a random forest model or an extreme random tree model.

5. The clutch control method according to claim 3, characterized in that, The method further includes: The driving data and the state information are analyzed based on the SHAP algorithm and / or feature importance analysis method to obtain an importance score; Based on the importance score, the driving data and status information are filtered to obtain the filtered driving data and status information, and the pressure prediction model is retrained based on the filtered driving data and status information.

6. The clutch control method according to claim 3, characterized in that, The acquisition of the training dataset includes: Identify the confounding factors for each training sample in the initial training dataset; Based on the interference factors and their weights, target training samples are selected from multiple training samples corresponding to each interference factor. The training dataset is obtained based on the target training samples.

7. The clutch control method according to claim 6, characterized in that, The interfering factors include: manufacturing parameters, fluid cleanliness, driving habits, environmental information, and clutch operating status.

8. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the method as described in any one of claims 1 to 7.

9. A vehicle, characterized in that, include: The electronic device according to claim 8.

10. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the vehicle's processor, the vehicle is able to perform the method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that, When the computer program product is run in a vehicle, it causes the vehicle to perform the method as described in any one of claims 1 to 7.

Images