A method for acquiring information from wet clutches based on fuzzy neural networks

Abstract

This invention relates to the technical field of clutch control design, specifically to a method for acquiring information about a wet clutch based on a fuzzy neural network. The method includes: analyzing a preset set of partial engagement point state parameters to generate an optimal feature subset one; generating a set of slip friction state parameters based on preset slip friction characteristic parameters, and analyzing and selecting the optimal feature subset two; generating the clutch state based on the optimal feature subset one, optimal feature subset two, and a preset fuzzy neural network model; acquiring operating condition information; determining whether the learning conditions are met based on the operating condition information; if the learning conditions are met, acquiring the clutch state based on the fuzzy neural network model; and generating control information based on the clutch state. Using this method, the clutch state of a wet clutch can be analyzed based on the fuzzy neural network model, thereby executing different control operations and improving the control accuracy of the wet clutch.

Description

[0001] This application is a divisional application of Chinese Patent Application 202111272009.4, "A Wet Clutch State Recognition Method Based on Fuzzy Neural Network", filed on October 29, 2021. Technical Field

[0002] This invention relates to the technical field of clutch control design, specifically to a method for acquiring information about a wet clutch based on a fuzzy neural network. Background Technology

[0003] Research on wet clutches requires the establishment of an accurate torque model. The engagement process of a wet clutch is a complex dynamic process, with the transmitted torque determined by both hydraulic fluid characteristics and friction pair characteristics. This torque is related to numerous influencing factors such as hydraulic oil viscosity, oil temperature, slip between the driving and driven plates, friction coefficient, and control pressure. In existing technologies, to improve the accuracy of wet clutch torque models, more influencing factors are typically introduced. By increasing the dimensionality of these influencing factors, the accuracy of the torque model is enhanced.

[0004] However, in practical applications, it has been found that some influencing factors are difficult to obtain, such as the oil film thickness between the driving and driven plates. The difficulty in obtaining these influencing factors makes it impossible to introduce them, thereby failing to improve the accuracy of the torque model and consequently failing to accurately control the wet clutch. Summary of the Invention

[0005] The present invention aims to provide a wet clutch information acquisition method based on fuzzy neural network to solve the technical problem of low accuracy of wet clutch torque model in the prior art.

[0006] This invention provides the following basic solution: The wet clutch information acquisition method based on fuzzy neural network includes the following: Analyze the entire set of pre-set semi-binding point state parameters to generate the optimal feature subset one; Generate a complete set of friction state parameters based on preset friction feature parameters, and analyze the complete set of friction state parameters to select and generate the optimal feature subset two; The clutch state is generated based on the first optimal feature subset, the second optimal feature subset, and the preset fuzzy neural network model.

[0007] Beneficial effects of the basic scheme: The engagement process of a wet clutch includes three states: complete disengagement, slippage, and complete engagement. The torque transmitted in each state differs, and correspondingly, the control parameters executed differ in each state. The slippage state is the transitional state between complete disengagement and complete engagement. The transition between these states is achieved by gradually increasing or decreasing the frictional torque between the friction pairs. The starting and ending points of the slippage state are defined as the half-engagement point state and the micro-slippage state, respectively.

[0008] Many factors influence the bonding process, but not every factor has a significant impact. Therefore, a subset is generated from the entire set of parameters based on requirements. The complete set of semi-bonding point state parameters represents the set of factors affecting the semi-bonding point state, while the first optimal feature subset is the set of factors selected from the complete set of semi-bonding point state parameters. Similarly, the complete set of slip friction state parameters represents the set of factors affecting the micro-slip friction state, while the second optimal feature subset is the set of factors selected from the complete set of slip friction state parameters.

[0009] By analyzing the clutch state of the wet clutch using a fuzzy neural network model, different control operations can be executed. By identifying different clutch states, precise control of the clutch can be achieved, thereby improving the control accuracy of the wet clutch.

[0010] Furthermore, the optimal feature subset is generated by analyzing the entire set of pre-defined semi-binding point state parameters, specifically including the following: Obtain the screening conditions for the wet clutch in the half-engagement state, analyze the complete set of half-engagement state parameters based on the screening conditions, and select the optimal feature subset one, which includes engine torque.

[0011] Beneficial effect: Compared with the complete set of state parameters at the semi-junction point, the parameters of the optimal feature subset one can better reflect the characteristics related to the physical essence of the signal. Selection is based on screening criteria. For example, the complete set of state parameters at the semi-junction point includes engine speed. However, in the idling state, engine speed changes due to factors such as temperature and pressure; in the non-idling state, engine speed changes due to factors such as throttle opening and load. It is difficult to form a certain pattern in the semi-junction point state and before and after it. Therefore, engine speed cannot be used as a state parameter for the semi-junction point; that is, engine speed is not included in the optimal feature subset one.

[0012] Furthermore, a complete set of friction state parameters is generated based on preset friction characteristic parameters, specifically including the following: The preset friction feature parameters are called to search and optimize the boundaries between the friction feature parameters to generate a complete set of friction state parameters.

[0013] Beneficial effects: After a wet clutch enters the slipping state, as the clutch pressure increases, the transmitted torque increases, and the speed difference of the driven plates gradually decreases. When the speed difference is sufficiently small, the wet clutch can transmit almost all the torque from the engine, which is the micro-slipping state. The two states before and after micro-slipping are defined as the excessive slipping state and the insufficient slipping state, respectively, thus constituting the three slipping states of a wet clutch.

[0014] By searching and optimizing the boundaries between the friction feature parameters, the friction feature parameters are optimized and their dimensionality reduced. The best friction feature parameters are selected to generate a complete set of friction state parameters. The optimal feature subset is then selected from this set for state recognition. This avoids excessive index dimensionality and reduces the calculation of redundant information, thereby improving the reaction speed of state recognition.

[0015] Furthermore, the boundaries between the slip friction feature parameters are searched and optimized to generate a complete set of slip friction state parameters, which specifically includes the following: The intelligent search algorithm searches and optimizes the boundaries between the slip characteristic parameters to generate a complete set of slip state parameters.

[0016] Beneficial effects: During the search and optimization process, a global search operation is mainly performed to achieve multi-parameter and multi-objective optimization. This solution uses an intelligent search algorithm to achieve fast search.

[0017] Furthermore, the optimal feature subset two is generated by analyzing the entire set of slip friction state parameters, specifically including the following: The entire set of slip friction state parameters is searched using the floating search algorithm to generate the second optimal feature subset.

[0018] Beneficial effects: The second optimal feature subset is generated based on the complete set of slip state parameters. At this time, the local search operation is mainly performed. This scheme adopts the floating search algorithm, which has strong search capabilities for small and medium-sized datasets, shortens the search time, and avoids the situation where the calculation time is too long and local optima are easily generated.

[0019] Furthermore, the clutch state is generated based on the first optimal feature subset, the second optimal feature subset, and the preset fuzzy neural network model, specifically including the following: Normalize the first and second optimal feature subsets to generate normalization results, and obtain the clutch state output by the fuzzy neural network model based on the normalization results.

[0020] Beneficial effects: This scheme normalizes the first and second optimal feature subsets. Normalization eliminates the calculation bias caused by the different dynamic ranges of the parameters. At the same time, without changing the original distribution characteristics of the parameters, it makes each parameter have the same status in the identification calculation.

[0021] Furthermore, the fuzzy neural network model includes a fuzzy layer, which performs fuzzy processing on the input of the fuzzy neural network model.

[0022] Beneficial effects: In this scheme, the model input is fuzzed through a fuzzy layer. Compared with intermediate fuzziness or output fuzziness, the structure is simpler, the reasoning is more convenient, and it is easier to carry out secondary development.

[0023] Furthermore, it also includes the following: The system acquires operating condition information and determines whether the learning conditions are met. If the learning conditions are met, the system acquires the clutch state based on the fuzzy neural network model and generates control information based on the clutch state.

[0024] Beneficial effects: In wet clutch testing, changes in vehicle status can easily lead to changes in test results, such as clutch wear. Therefore, in this solution, self-learning is used to improve control accuracy when learning conditions are met.

[0025] Furthermore, the clutch status includes under-clutch and over-clutch, and the control information includes decreasing pressure and increasing pressure; When the clutch is in an under-clutch state, the control information generated based on the clutch state is to reduce the pressure; when the clutch is in an over-clutch state, the control information generated based on the clutch state is to increase the pressure.

[0026] Beneficial effects: When the clutch is in an insufficient state, it means that the desired state has not been achieved, such as an insufficient engagement state; the control information is the adjustment information for controlling the corresponding equipment, such as increasing the voltage corresponding to the clutch when the clutch is in an insufficient state.

[0027] Furthermore, the learning conditions include half-junction point learning conditions and torque learning conditions. When the learning condition is the half-engagement point learning condition, the control information generated based on the clutch state is to increase or decrease the clutch half-engagement point pressure. When the learning condition is torque learning condition, the control information generated based on the clutch state is to increase or decrease the control pressure of the clutch.

[0028] Beneficial effects: The learning conditions are used to determine whether to enter the self-learning mode. Self-learning is performed for the semi-engagement point state and the micro-slipping state, thereby continuously correcting parameters and improving the accuracy of clutch control. Attached Figure Description

[0029] Figure 1 This is a flowchart of an embodiment of the wet clutch information acquisition method based on fuzzy neural network of the present invention; Figure 2This is a flowchart of Embodiment 2 of the wet clutch information acquisition method based on fuzzy neural network of the present invention; Figure 3 The results are from the torque map test of the hardware-in-the-loop simulation test of the wet dual-clutch assembly of this invention. Detailed Implementation

[0030] The following detailed description illustrates the specific implementation method: Example 1 A method for acquiring information from wet clutches based on fuzzy neural networks, as shown in the appendix. Figure 1 As shown, it includes the following: Step 1: Analyze the preset set of semi-bonding point state parameters to generate the optimal feature subset one; generate the set of sliding friction state parameters based on the preset sliding friction feature parameters, and analyze the set of sliding friction state parameters to generate the optimal feature subset two.

[0031] The optimal feature subset is generated by analyzing the entire set of pre-defined semi-binding point state parameters, specifically including the following: A complete set of semi-engagement point state parameters is preset. In this embodiment, the parameters in the complete set of semi-engagement point state parameters include engine speed, clutch driven part speed, clutch driving and driven part speed difference, engine torque, and engine torque change rate.

[0032] Obtain the screening conditions for the wet clutch in the half-engagement state, analyze the complete set of half-engagement state parameters based on the screening conditions, and select the optimal feature subset one.

[0033] The selection criteria are the influencing factors of each parameter in the complete set of semi-engagement point state parameters. Parameters in the complete set of semi-engagement point state parameters that do not exhibit a consistent pattern at or before the semi-engagement point are excluded. For example, the complete set of semi-engagement point state parameters includes engine speed. However, in idling mode, engine speed is affected by factors such as temperature and pressure; in non-idling mode, engine speed is affected by throttle opening and load. It is difficult to establish a consistent pattern at or before the semi-engagement point, therefore engine speed cannot be used as a semi-engagement point state parameter. Conversely, parameters from the first optimal feature subset are selected. In this embodiment, the first optimal feature subset includes engine torque.

[0034] In this scheme, parameters are selected based on the application scenarios, working principles, and semi-engagement point states of wet clutches to generate the optimal feature subset one.

[0035] A complete set of friction state parameters is generated based on preset friction feature parameters, and the optimal feature subset is generated by analyzing and filtering the complete set of friction state parameters; specifically, it includes the following: The preset slip friction characteristic parameters are invoked. In this embodiment, the slip friction characteristic parameters are shown in Table 1 below.

[0036] Table 1. Friction Characteristic Parameters

[0037] The boundaries between the friction feature parameters are searched and optimized to generate a complete set of friction state parameters. Specifically, an intelligent search algorithm is used to search and optimize the boundaries between the friction feature parameters to generate a complete set of friction state parameters. In this embodiment, an adaptive genetic algorithm is selected as the intelligent search algorithm.

[0038] The optimal feature subset two is generated by analyzing the complete set of friction state parameters. Specifically, the optimal feature subset two is generated by searching the complete set of friction state parameters using a floating search algorithm.

[0039] An adaptive genetic algorithm is used as the outer loop search algorithm, and a floating search algorithm is used as the inner loop search algorithm. Specifically, the boundary between the slippage feature parameters is searched. An initial population is randomly generated, and the slippage state parameters of each individual in the population are calculated to generate a complete set. A candidate feature subset one is generated according to the floating search algorithm. It is then determined whether the floating search termination condition is met. If not, a second candidate feature subset is generated according to the floating search algorithm. The two candidate feature subsets are compared, and the optimal candidate feature subset is selected and inherited to the next search. Otherwise, if the condition is met, the genetic algorithm termination condition is checked. If the genetic algorithm termination condition is not met, the adaptive genetic algorithm is used for calculation. The objective function of the genetic algorithm is as follows: 1=W A R acc +W F N SW (1)

[0040] In equation (1), 1 represents the objective function value of the genetic algorithm, W A In this embodiment, the weight for accuracy is set to 0.99, R. acc To ensure accuracy, N SW W represents the number of parameters in the optimal feature subset two. F The weight of the number of parameters in the optimal feature subset two is 0.01 in this embodiment.

[0041] According to Equation (1), the fitness of each individual in the initial population can be calculated. Crossover and mutation operations are performed based on the fitness. Individuals with fitness higher than the upper limit of fitness are crossover and mutation operations according to the first probability. Individuals with fitness lower than the lower limit of fitness are crossover and mutation operations according to the second probability. The first probability is less than the second probability.

[0042] The crossover probability formula is as follows: (2)

[0043] The mutation probability formula is as follows: (3)

[0044] In equations (2) and (3), The fitness of the individual with higher fitness among the two individuals to be crossed. The average fitness of all individuals in the current population. The fitness of the individual with the highest fitness in the current population.

[0045] Crossover and mutation operations are performed according to the crossover probability formula and mutation probability formula to obtain a new population. The sliding friction state parameters of each individual in the population are then recalculated until the genetic algorithm terminates. The current candidate feature subset is then considered the second optimal feature subset. In this embodiment, the second optimal feature subset includes n... d LN IS η n1-n2 and η r3 .

[0046] Step Two: Generate the clutch state based on the first and second optimal feature subsets and the preset fuzzy neural network model. This specifically includes the following: normalizing the first and second optimal feature subsets to generate a normalized result, and obtaining the clutch state output by the fuzzy neural network model based on the normalized result.

[0047] In this embodiment, a linear method is used for normalization, and the processing formula is as follows: (k=1,2,...m; n=1,2,3,4) (4)

[0048] In equation (4), This is the result of normalizing the nth parameter in the kth sample. For the nth parameter in the kth sample, , These are the maximum and minimum values ​​of the nth parameter in the sample, respectively, and m is the number of samples.

[0049] The fuzzy neural network model includes an input layer, a fuzzy layer, a hidden layer, and an output layer. The input layer is used to pass the normalized optimal feature subset one and optimal feature subset two of the input to the next layer.

[0050] The fuzzy layer is used to calculate the membership function values ​​of each input component belonging to each fuzzy set, that is, to perform fuzzy processing on the input of the fuzzy neural network model. It calculates the fuzzy values ​​of each input component. The membership function is as follows:

[0051] (5)

[0052] In equation (5), and (i=1,2,3;n=1,2,3,4) represent the mean and variance of the nth feature in the i-th class of states, respectively.

[0053] The hidden layer is used to obtain the output of the fuzzy layer, calculate the fitness of the fuzzy rules according to the fuzzy calculation formula, and output the calculation result to the output layer. The fuzzy calculation formula is as follows: (6)

[0054] In equation (6), For weights, This is a fuzzy calculation function for the weights.

[0055] The output layer is used to calculate the fuzzy function result based on the calculation formula of the output value. The calculation formula of the output value is as follows: (7)

[0056] The output layer outputs the calculation results of the fuzzy function.

[0057] Acquire typical data for various clutch operating conditions, such as experimental data corresponding to the partial engagement point state. Input the experimental data into a fuzzy neural network model and obtain the output of the fuzzy neural network model based on the experimental data for typical operating conditions. Define the output as a state threshold and record the state threshold corresponding to each typical operating condition. When recording, the state threshold values ​​are arranged and stored in order.

[0058] During state identification, the clutch state is generated by comparing the fuzzy function calculation result with a state threshold. Specifically, in this embodiment, the state threshold with the largest value is placed first, and the state threshold with the smallest value is placed last. The fuzzy function calculation result and the state threshold are compared sequentially. When the fuzzy function calculation result is greater than the state threshold, it meets the typical working condition corresponding to the state threshold, and the state of the typical working condition corresponding to the state threshold is selected as the clutch state. Otherwise, the fuzzy function calculation result is compared with the next state threshold.

[0059] Example 2 The difference between this embodiment and Embodiment 1 is that: A method for acquiring information from wet clutches based on fuzzy neural networks, as shown in the appendix. Figure 2 As shown, it also includes the following: The system acquires operating condition information and determines whether the learning conditions are met. If the conditions are met, it obtains the clutch state based on the fuzzy neural network model and generates control information accordingly. Specifically, this includes the following: Acquire operating condition information, including usage information, braking information, action information, and gear information. Taking driving a vehicle as an example, usage information is used to characterize the vehicle's current state, such as starting, driving, or stopping; braking information is used to characterize whether the vehicle is currently braking; action information is used to characterize the driver's current action, such as shifting gears; and gear information is used to characterize the vehicle's gear changes, such as shifting the lever into D gear.

[0060] In this embodiment, the clutch state includes under-engagement and over-engagement, and according to the clutch engagement process, it is further divided into under-engagement state, under-slipping state, over-engagement state, and over-slipping state. Control information includes decreasing pressure and increasing pressure, and is further divided into adjusting the pressure at the half-engagement point and adjusting the control pressure at the torque point. In this embodiment, for ease of distinction, the control information controlling the pressure at the half-engagement point is defined as control information one, and the control information controlling the pressure corresponding to the torque point is defined as control information two.

[0061] The system determines whether the learning conditions are met based on the operating condition information. These conditions include the semi-engagement learning condition and the torque learning condition. The semi-engagement learning condition occurs when the engine is idling and unloaded, but the clutch driven part is loaded, transitioning from a fully disengaged state to a semi-engagement state. The torque learning condition occurs when the vehicle maintains a certain throttle opening while driving on a flat road, without any gear shifting. For example, when starting, pressing the brake and shifting the lever into Drive (D), the semi-engagement learning condition is met. Similarly, the torque learning condition is met when the vehicle maintains a certain throttle opening while driving on a flat road, without any gear shifting. Only when the operating condition information meets all the learning conditions can the required state recognition be achieved.

[0062] The system determines whether the operating condition information meets the half-engagement point learning conditions. If it does, it reads the position information of the half-engagement point and generates control information one based on the position information. The wet clutch is then controlled to operate at the half-engagement point pressure based on control information one. The half-engagement point state of the wet clutch is identified using a fuzzy neural network model. In this embodiment, the half-engagement point state includes under-engagement, over-engagement, and half-engagement point states.

[0063] When the learning condition is the half-engagement point learning condition, the control information generated based on the clutch state is to increase or decrease the clutch half-engagement point pressure. Specifically, when the clutch state is under-engaged, control update information 1 that increases the half-engagement point pressure is generated; when the clutch state is over-engaged, control update information 1 that decreases the half-engagement point pressure is generated. The clutch half-engagement point pressure is then updated based on this control update information 1.

[0064] When the operating condition information does not meet the half-engagement point learning condition, or when the clutch is in a half-engagement point state, it is determined whether the operating condition information meets the torque learning condition. If it does, the current torque is obtained by searching the preset torque map, the clutch pressure required to transmit the current torque is calculated, control information two is generated based on the clutch pressure, and the control pressure corresponding to the torque point of the wet clutch is controlled according to control information two. The slippage state of the wet clutch is identified based on the fuzzy neural network model. In this embodiment, the slippage state includes insufficient slippage state, excessive slippage state, and micro-slippage state.

[0065] When the learning condition is torque learning condition, the control information generated based on the clutch state is to increase or decrease the clutch control pressure. Specifically, when the clutch state is under-slip, control update information two is generated to increase the control pressure corresponding to that torque point; when the clutch state is over-slip, control update information two is generated to decrease the control pressure corresponding to that torque point. The control pressure corresponding to that torque point of the clutch is updated according to control update information two.

[0066] When the operating condition information does not meet the half-engagement point learning conditions, the self-learning of the half-engagement point state ends. When the operating condition information does not meet the torque learning conditions, the self-learning of the slipping friction state ends. When the clutch state is a micro-slipping friction state, it indicates that no correction is needed.

[0067] In other embodiments, before obtaining the clutch state based on the fuzzy neural network model, the following is also included: Clutch information is collected based on the first and second optimal feature subsets. The clutch information is then used to determine whether the preset collection conditions are met. If they are met, the clutch state is obtained based on the fuzzy neural network model. Otherwise, the clutch information is collected again.

[0068] The process of determining whether preset acquisition conditions are met based on clutch information includes the following: acquiring continuous clutch information, which includes multiple parameter data; determining whether the continuous parameter data meets a preset parameter change threshold; if it does, the clutch information meets the acquisition conditions; otherwise, the clutch information does not meet the acquisition conditions. The parameter change threshold is the maximum change between continuous data during normal clutch use. The change in continuous parameter data is calculated, and it is determined whether the change exceeds the parameter change threshold; if not, the parameter change threshold is met; otherwise, it is not met.

[0069] If the clutch information does not meet the acquisition conditions, it indicates that the acquired clutch information has changed abnormally. At this time, the clutch may have a major problem, or the clutch information acquisition equipment may be damaged. Abnormal clutch information should not be learned to avoid reducing the clutch control accuracy and causing improper clutch control.

[0070] In this scheme, a wet dual-clutch based on dsPACE was tested. Hardware-in-the-loop simulation tests were conducted on three sets of wet dual-clutch assemblies. The test results are shown in Table 2.

[0071] Table 2 Comparison of Self-Learning Results at Semi-Joining Points

[0072] According to the test results, the maximum error between the learned value and the actual value of the half-engagement point pressure of the wet clutch does not exceed 2.7%.

[0073] Torque Map test results are as follows Figure 3 As shown, the learned torque map is close to the actual torque map, with a maximum error of no more than 4.47%. The experimental results show that the model can achieve precise control of the clutch.

[0074] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for acquiring information from a wet clutch based on a fuzzy neural network, characterized in that, Includes the following: Analyze the entire set of pre-set semi-binding point state parameters to generate the optimal feature subset one; Generate a complete set of friction state parameters based on preset friction feature parameters, and analyze the complete set of friction state parameters to select and generate the optimal feature subset two; The clutch state is generated based on the first optimal feature subset, the second optimal feature subset, and the preset fuzzy neural network model. The fuzzy neural network model includes an input layer, a fuzzy layer, a hidden layer, and an output layer. The input layer is used to pass the normalized optimal feature subset one and optimal feature subset two of the input to the next layer. The fuzzy layer is used to calculate the membership function of each input component belonging to each fuzzy set, and to calculate the fuzzy value of each input component. The membership function is as follows: In the formula, and (i=1,2,3;n=1,2,3,4) represent the mean and variance of the nth feature in the i-th class of states, respectively; The hidden layer is used to obtain the output of the fuzzy layer, calculate the fitness of the fuzzy rules according to the fuzzy calculation formula, and output the calculation result to the output layer. The fuzzy calculation formula is as follows: In the formula, For weights, The fuzzy calculation function for the weights; The output layer is used to calculate the fuzzy function result based on the calculation formula of the output value. The calculation formula of the output value is as follows: The output layer outputs the fuzzy function calculation results; Before obtaining the clutch state based on the fuzzy neural network, the process includes: collecting clutch information based on the first and second optimal feature subsets; determining whether the clutch information meets the preset collection conditions; if it does, obtaining the clutch state based on the fuzzy neural network model; otherwise, collecting the clutch information again. The system determines whether the clutch information meets the preset acquisition conditions, including: acquiring continuous clutch information, which includes multiple parameter data, and determining whether the continuous parameter data meets the preset parameter change threshold. If it does, the clutch information meets the acquisition conditions; otherwise, the clutch information does not meet the acquisition conditions. The parameter change threshold is the maximum change between continuous data during normal clutch use. The change in continuous parameter data is calculated, and it is determined whether the change exceeds the parameter change threshold. If not, the parameter change threshold is met; otherwise, it is not met.

2. The wet clutch information acquisition method based on fuzzy neural network according to claim 1, characterized in that: The optimal feature subset is generated by analyzing the entire set of pre-defined semi-binding point state parameters, specifically including the following: Obtain the screening conditions for the wet clutch in the half-engagement state, analyze the complete set of half-engagement state parameters based on the screening conditions, and select the optimal feature subset one, which includes engine torque.

3. The wet clutch information acquisition method based on fuzzy neural network according to claim 1, characterized in that: Generate a complete set of friction state parameters based on preset friction characteristic parameters, specifically including the following: The preset friction feature parameters are called to search and optimize the boundaries between the friction feature parameters to generate a complete set of friction state parameters.

4. The wet clutch information acquisition method based on fuzzy neural network according to claim 3, characterized in that: The boundaries between the slip friction feature parameters are searched and optimized to generate a complete set of slip friction state parameters, which includes the following: The intelligent search algorithm searches and optimizes the boundaries between the slip characteristic parameters to generate a complete set of slip state parameters.

5. The wet clutch information acquisition method based on fuzzy neural network according to claim 3, characterized in that: The optimal feature subset two is generated by analyzing and filtering the complete set of slip friction state parameters, specifically including the following: The entire set of slip friction state parameters is searched using the floating search algorithm to generate the second optimal feature subset.

6. The wet clutch information acquisition method based on fuzzy neural network according to claim 1, characterized in that: The clutch state is generated based on the first optimal feature subset, the second optimal feature subset, and a pre-defined fuzzy neural network model, specifically including the following: Normalize the first and second optimal feature subsets to generate normalization results, and obtain the clutch state output by the fuzzy neural network model based on the normalization results.

7. The wet clutch information acquisition method based on fuzzy neural network according to claim 1 or 6, characterized in that: A fuzzy neural network model includes a fuzzy layer, which performs fuzzy processing on the input of the fuzzy neural network model.

8. The wet clutch information acquisition method based on fuzzy neural network according to claim 7, characterized in that, Also includes the following: The system acquires operating condition information and determines whether the learning conditions are met. If the learning conditions are met, the system acquires the clutch state based on the fuzzy neural network model and generates control information based on the clutch state.

9. The wet clutch information acquisition method based on fuzzy neural network according to claim 8, characterized in that: Clutch status includes under-clutch and over-clutch, and control information includes decreasing pressure and increasing pressure. When the clutch is in an under-clutch state, the control information generated based on the clutch state is to reduce the pressure; when the clutch is in an over-clutch state, the control information generated based on the clutch state is to increase the pressure.

10. The wet clutch information acquisition method based on fuzzy neural network according to claim 8, characterized in that: The learning conditions include the half-junction point learning condition and the torque learning condition. When the learning condition is the half-engagement point learning condition, the control information generated based on the clutch state is to increase or decrease the clutch half-engagement point pressure. When the learning condition is torque learning condition, the control information generated based on the clutch state is to increase or decrease the control pressure of the clutch.

Images