No-load wet clutch driving torque prediction method, system, equipment and medium

Abstract

The invention discloses a no-load wet clutch driving torque prediction method, system and device and a medium, and relates to the technical field of power consumption prediction. The method comprises the following steps: acquiring to-be-tested data; the to-be-measured data comprises the number of friction pairs, the rotating speed of a friction plate, oil temperature, oil supply flow, a friction pair gap, the number of grooves, the width of the grooves and the depth of the grooves; constructing a belt row torque hybrid prediction model; the belt row torque hybrid prediction model comprises a low-speed mechanism model and a high-speed data driving prediction model; the high-speed data driving prediction model adopts a GA-RBF neural network model; the low-speed mechanism model divides the belt row torque of the friction pair clearance rotating flow field with the radial grooves into two parts, namely a groove area and a non-groove area; and performing normalization processing on the to-be-measured data, inputting the normalized data into the belt row torque hybrid prediction model for processing, and outputting a belt row torque prediction value. According to the method, the prediction precision and the calculation efficiency of the belt discharge torque in the wide speed range of the wet clutch can be improved.

Description

Technical Field

[0001] This invention relates to the field of power consumption prediction technology, and in particular to a method, system, device and medium for predicting the torque of a wet clutch under no-load conditions. Background Technology

[0002] Wet clutches offer advantages such as smooth shifting, stable performance, strong heat dissipation, and high torque transmission, leading to their widespread use in automotive transmission systems. Under no-load conditions, a certain gap is maintained between the friction plates and steel plates of a wet clutch, allowing them to rotate independently. When a speed difference exists between the friction plates and steel plates, the lubricating oil is subjected to viscous shear, generating pull-out torque. This pull-out torque not only consumes engine power but also causes a sharp rise in the temperature of the friction pair and lubricating oil, negatively impacting the reliability and fuel economy of the transmission system. Therefore, in-depth research into the pull-out torque (power loss) characteristics of wet clutches is of significant engineering importance for improving the performance of automatic transmissions in China and enhancing vehicle power transmission efficiency. Currently, there is a lack of technical solutions for quickly and accurately predicting the pull-out torque of wet clutches across a wide speed range. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, device and medium for predicting the discharge torque of an unloaded wet clutch, aiming to solve or improve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A method for predicting the torque of a wet clutch under no-load conditions includes:

[0006] Acquire the data to be tested; the data to be tested includes the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width and groove depth;

[0007] A hybrid prediction model for radial groove torque is constructed; the hybrid prediction model for radial groove torque includes a low-speed mechanism model and a high-speed data-driven prediction model; the high-speed data-driven prediction model adopts a GA-RBF neural network model; the low-speed mechanism model divides the radial groove torque of the rotating flow field of the friction pair gap into two parts: the groove region and the non-groove region.

[0008] The measured data is normalized, and the normalized data is input into the belt-driven torque hybrid prediction model for processing, and the belt-driven torque prediction value is output.

[0009] Optionally, the expression for the low-speed mechanism model is:

[0010]

[0011] Where Z represents the number of friction pairs, Ng Indicates the number of slots, θ l The angle r represents the area corresponding to the slot. o The outer radius of the friction pair is represented by r. i μ represents the inner radius of the friction pair. l Let θ represent the viscosity of the flow field outside the groove region, w represent the angular velocity of the friction plate, r represent the radius at any point in the rotating flow field with the gap, h0 represent the gap between the friction pairs, and θ represent the velocity of the friction plate. g μ represents the angle corresponding to the groove region. g h represents the viscosity of the flow field in the tank region. g Indicates the depth of the radial groove.

[0012] Optionally, the expression for the high-speed GA-RBF neural network model is:

[0013]

[0014] Among them, h j Let l be the radial basis functions, and w be the number of radial basis functions. j Let X = [x1, x2, ..., xj] be the weights of the j-th basis function. d ] T Let d be the input vector of the neural network, σ be the number of neurons in the input layer, and σ be the input vector of the neural network. j C is the center width of the Gausky function of the j-th neuron in the hidden layer. j =[c j1 c j2 c jn ] T Let ||XC| be the center vector of the j-th neuron in the hidden layer. j || is the vector XC j The Euclidean norm of X represents the relationship between X and C. j The distance between them, T is the torque value of the drive system during the high-speed stage.

[0015] Optionally, the training process of the high-speed data-driven prediction model is as follows:

[0016] The main architecture is based on an RBF neural network, and the center vector C of the hidden layer nodes of the main architecture is obtained by using the GA algorithm. j σ, the center width of the Gaussian function j and the connection weights w between the hidden layer and the output layer j Optimize the training to determine the best parameters for the network.

[0017] Optionally, the training process of the GA algorithm specifically includes:

[0018] Using real number encoding, the center vector C j and the center width σ of the Gaussian function jPerform cross-arrangement and assign weights w j Arrange them last to obtain chromosome codes;

[0019] Based on the chromosome encoding and according to the set value range of each parameter variable, a certain number of chromosome strings are randomly generated. Each chromosome string includes different parameter combinations to form an initial population.

[0020] Construct a fitness function and a genetic operator; the fitness function is the reciprocal of the root mean square error between the actual value and the predicted value; the genetic operator is determined by roulette wheel selection, three-point arithmetic crossover, and mutation operation;

[0021] The process iterates based on the initial population and the genetic operator, and calculates the fitness value according to the fitness function until the maximum number of iterations is reached or the set convergence condition of optimal fitness value is met. Then the iteration ends, and the individual with the optimal fitness value in the population is selected as the optimal solution.

[0022] The present invention also provides a no-load wet clutch belt torque prediction system, comprising:

[0023] The data acquisition unit is used to acquire the data to be measured, including the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth.

[0024] The model building unit is used to construct a hybrid prediction model for radial groove torque. The hybrid prediction model for radial groove torque includes a low-speed mechanism model and a high-speed data-driven prediction model. The high-speed data-driven prediction model adopts a GA-RBF neural network model. The low-speed mechanism model divides the radial groove torque of the rotating flow field of the friction pair gap into two parts: the groove region and the non-groove region.

[0025] The model prediction unit is used to normalize the data to be measured, and input the normalized data into the hybrid prediction model of the belt-driven torque for processing, and output the predicted value of the belt-driven torque.

[0026] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to cause the electronic device to perform the above-described method for predicting the belt torque of an unloaded wet clutch.

[0027] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the no-load wet clutch belt torque prediction method as described above.

[0028] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0029] This invention discloses a method, system, device, and medium for predicting the torque of a wet clutch under no-load conditions. The method includes acquiring test data, which includes the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth. A hybrid prediction model for the torque of the wet clutch is constructed, comprising a low-speed mechanism model and a high-speed data-driven prediction model. The high-speed data-driven prediction model employs a GA-RBF neural network model. The low-speed mechanism model divides the torque of the wet clutch in the rotating flow field of the friction pair clearance with radial grooves into grooved and non-grooved regions. The test data is normalized, and the normalized data is input into the hybrid prediction model for processing, outputting the predicted torque value. This invention improves the prediction accuracy and computational efficiency of the torque of the wet clutch over a wide speed range. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of the RBF neural mesh structure in this embodiment;

[0032] Figure 2 This is a schematic diagram of the chromosome structure of the genetic algorithm in this embodiment;

[0033] Figure 3 This is a flowchart of the GA-RBF neural network algorithm in this embodiment;

[0034] Figure 4 This is a graph showing the torque prediction curve of the hybrid model in this embodiment;

[0035] Figure 5 This is a flowchart illustrating the no-load wet clutch torque prediction method in this embodiment.

[0036] Figure 6 This is a schematic diagram illustrating the computational efficiency of the model in this embodiment. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] The purpose of this invention is to provide a method, system, device and medium for predicting the discharge torque of an unloaded wet clutch, aiming to solve or improve at least one of the above-mentioned technical problems.

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] like Figures 1-6 As shown, the present invention provides a method for predicting the torque of a wet clutch under no-load conditions, comprising:

[0041] Step 100: Obtain the data to be tested; the data to be tested includes the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width and groove depth.

[0042] Step 200: Construct a hybrid prediction model for radial groove torque; the hybrid prediction model for radial groove torque includes a low-speed mechanism model and a high-speed data-driven prediction model; the high-speed data-driven prediction model adopts a GA-RBF neural network model; the low-speed mechanism model divides the radial groove torque of the rotating flow field of the friction pair gap into two parts: the groove region and the non-groove region.

[0043] Step 300: Normalize the data to be measured, and input the normalized data into the belt-driven torque hybrid prediction model for processing, and output the belt-driven torque prediction value.

[0044] Based on the above technical solution, the following embodiments are provided.

[0045] Firstly, the prediction of viscous shear torque in the low-speed stage is addressed by using the finite volume method to solve the unified control equation of the rotating flow field in the wet clutch friction pair clearance. Based on the principle of mass flow conservation and the equations of pressure and viscous pressure, the density and viscosity of the rotating flow field in the clearance are obtained. Then, according to Newton's law of internal friction, the expression for calculating the wet clutch belt torque is obtained by integration. Secondly, the prediction of the recovery torque in the high-speed stage is addressed by obtaining the wet clutch belt torque under different operating conditions and design parameters through experiments. The obtained belt torque data is used as a training set, and appropriate model parameters such as the number of neurons, maximum number of iterations, error threshold, crossover probability, and mutation probability are selected to construct a GA-RBF neural network model. The model is continuously iterated and optimized to achieve accurate prediction of the belt torque in the high-speed stage.

[0046] For low-speed mechanism models:

[0047] The unified governing equations of the rotating flow field in the friction pair clearance of a wet clutch are solved using the finite volume method. The density and viscosity of the clearance flow field are obtained by utilizing mass flow conservation and the pressure and viscous pressure equations. Then, by integration based on Newton's law of internal friction, the calculation expression for the belt torque of the wet clutch is obtained. This invention divides the belt torque of the rotating flow field in the clearance of a friction pair with radial grooves into two parts: the grooved region and the non-grooved region. Therefore, the expression for the belt torque of the friction pair is:

[0048]

[0049] For high-speed GA-RBF neural network models:

[0050] 1) RBF neural network model:

[0051] The main factors affecting belt torque during the operation of a wet clutch are the number of friction pairs, friction plate speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth. Therefore, this paper uses the number of friction pairs, friction plate speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth as the input signals of the RBF neural network, and the belt torque as the output result of the RBF neural network. A schematic diagram of the RBF neural network structure is shown below. Figure 1 As shown, where X = [x1, x2, ..., x...] d ] T Here, d represents the input vector of the neural network, h represents the number of neurons in the input layer, and d represents the number of neurons in the input layer. j Here, l represents the radial basis functions, which are the number of radial basis functions, i.e., the number of neurons in the hidden layer. This number is determined through adjustment during training. j Let be the weight of the j-th basis function. A Gaussian function is used as the radial basis function, and its calculation formula is shown in equation (2). For the linear combination output value T of the RBF neural network... d The calculation is as shown in equation (3).

[0052]

[0053] In equations (2) and (3): C j =[c j1 c j2 c jn ] T σ is the center vector of the j-th neuron in the hidden layer; j Let XC be the width parameter of the j-th neuron in the hidden layer; j || is the vector XC j The Euclidean norm of X represents the relationship between X and C. j The distance between them.

[0054] Because the torque of a wet clutch is affected by multiple parameters, and these parameters have different units and orders of magnitude, directly inputting these parameters will lead to problems such as reduced accuracy and non-convergence in the prediction model. Therefore, the data needs to be normalized to facilitate the training of the RBF neural network. The normalization formula is:

[0055]

[0056] In the formula, y represents the normalized data; x represents the experimental data; x max and x min These represent the maximum and minimum values ​​for a given parameter.

[0057] 2) GA algorithm:

[0058] Genetic algorithms are optimization algorithms that simulate biological evolution mechanisms. They encode potential solutions to a problem as individuals, thereby constructing an initial population. During the algorithm's iteration process, the quality of individuals is evaluated based on a fitness function. Individuals with higher fitness values ​​are selected and subjected to crossover and mutation operations to generate a new generation of individuals. After multiple generations of genetic evolution, the individual with the best fitness is obtained as the optimal solution for the target value.

[0059] Analysis of the RBF neural network reveals that the most crucial step in network design is determining the center value C of the hidden layer nodes. j σ, the center width of the Gaussian function j and the connection weights w between the hidden layer and the output layer j Choosing the right parameter values ​​is crucial to achieving optimal prediction accuracy for the RBF neural network. Since genetic algorithms possess excellent global optimization capabilities, robustness, and parallelizability, they are introduced to calculate the optimal parameters for the RBF neural network. The specific optimization steps are as follows (e.g., ...). Figure 3 As shown):

[0060] (1) Chromosome coding:

[0061] The center value C of the hidden layer node j and basis function width σ j The changes in the two parameters are closely related; therefore, the genetic algorithm will affect C. j σ j The two parameters are cross-arranged, and the weights w are... j Arranged last, this increases the probability that the center value and basis function width of the hidden layer nodes change simultaneously. Using real-number encoding effectively simplifies the encoding and decoding process, improving algorithm speed. The chromosome structure of the genetic algorithm is as follows: Figure 2 As shown.

[0062] (2) Initialize the population:

[0063] Based on the value range of each variable, a certain number of chromosome strings are randomly generated. Each chromosome string includes different parameter combinations and is used as part of the initial population to explore a wide search space.

[0064] (3) Construct the fitness function:

[0065] The purpose of optimizing the RBF neural network is to improve the prediction accuracy of the network. Therefore, the reciprocal of the root mean square error between the actual value and the predicted value is selected as the fitness function, and the calculation formula is shown in Equation (5):

[0066]

[0067] In the formula, Y i and Y i ' represents the true output value and the predicted output value obtained from the input data of the i-th sample, respectively. This fitness function can effectively distinguish the performance of different individuals.

[0068] (4) Genetic operators:

[0069] ① Selection operation: A roulette wheel selection method is used. The fitness value of each individual is proportionally mapped onto a roulette wheel, and individuals are randomly selected by spinning the roulette wheel. The probability of individual i being selected is P. i The specific expression is:

[0070]

[0071] In the formula: F i Let F be the fitness value of the i-th parent individual. i The sum of fitness values ​​of all parent individuals is given, where m is the population size. Individuals with higher fitness values ​​have a greater chance of being replicated in the next generation, thus passing on their advantageous traits.

[0072] ② Crossover operation: The method used for crossover operation is the three-point arithmetic crossover method, and the crossover probability is P. c In each round of the cycle, two parent chromosomes are randomly selected and arithmetic crossover is performed to generate two offspring chromosomes:

[0073]

[0074] In the formula, p i p j For the selection of two different paternal chromosomes; o i o j α represents the two offspring chromosomes generated; α is a random number between [0,1] used to control the crossover ratio of genes.

[0075] ③ Mutation operation: Gene mutation can effectively expand the search range of the algorithm and enhance population diversity. The mutation probability is P. m In each round of the cycle, a certain number of individuals are first selected as the targets of the mutation operation. Then, specific gene loci on the chromosome are randomly selected, and the genes at those loci are mutated, i.e., from C... j σ j With w j A new random value is generated within the allowed range of values ​​to replace the original gene value.

[0076] The initial population size for the genetic algorithm is set to 50, the maximum number of generations is 200, and the crossover probability P in the genetic operators is set to... c The value is 0.65, and the mutation probability P m Set the fitness value to 0.1. Repeat the above steps until the maximum number of iterations is reached or the optimal convergence condition of the set fitness value is met. When the algorithm ends, select the individual with the best fitness value in the population as the optimal solution, and output the optimal parameters to construct the GA-RBF neural network model.

[0077] 3) GA-RBF neural network model:

[0078] The prediction results of the hybrid model are compared with those of the traditional mechanistic model and experimental results, such as... Figure 4 As shown, it is evident that the corrected prediction of belt torque is in better agreement with the experimental results, with the average relative error decreasing from 36.74% to 15.23%. This indicates that improving and correcting the model by combining intelligent optimization algorithms with RBF neural networks can effectively enhance the prediction accuracy of wet clutch belt torque.

[0079] As a second aspect, the present invention also provides a no-load wet clutch belt torque prediction system, comprising:

[0080] The data acquisition unit is used to acquire the data to be measured, which includes the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth.

[0081] The model building unit is used to construct a hybrid prediction model for belt-driven torque; the hybrid prediction model for belt-driven torque includes a low-speed mechanism model and a wet clutch belt-driven torque prediction model; the wet clutch belt-driven torque prediction model adopts a GA-RBF neural network model; the low-speed mechanism model divides the belt-driven torque of the rotating flow field of the friction pair clearance with radial grooves into two parts: groove region and non-groove region;

[0082] The model prediction unit is used to normalize the data to be measured, and input the normalized data into the hybrid prediction model of belt and exhaust torque for processing, and output the predicted value of belt and exhaust torque.

[0083] As a third aspect, this embodiment also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to cause the electronic device to perform the above-described no-load wet clutch belt torque prediction method.

[0084] As a fourth aspect, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the no-load wet clutch belt torque prediction method as described above.

[0085] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0086] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for predicting the torque of a wet clutch under no-load conditions, characterized in that, include: Acquire the data to be tested; the data to be tested includes the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width and groove depth; A hybrid prediction model for radial groove torque is constructed; the hybrid prediction model for radial groove torque includes a low-speed mechanism model and a high-speed data-driven prediction model; the high-speed data-driven prediction model adopts a GA-RBF neural network model; the low-speed mechanism model divides the radial groove torque of the rotating flow field of the friction pair gap into two parts: the groove region and the non-groove region. The measured data is normalized, and the normalized data is input into the belt-driven torque hybrid prediction model for processing, and the belt-driven torque prediction value is output.

2. The method for predicting the torque of a wet clutch under no-load conditions according to claim 1, characterized in that, The expression for the low-speed mechanism model is: Where Z represents the number of friction pairs, N g Indicates the number of slots, θ l The angle r represents the area corresponding to the slot. o The outer radius of the friction pair is represented by r. i μ represents the inner radius of the friction pair. l Let θ represent the viscosity of the flow field outside the groove region, w represent the angular velocity of the friction plate, r represent the radius at any point in the rotating flow field with the gap, h0 represent the gap between the friction pairs, and θ represent the velocity of the friction plate. g μ represents the angle corresponding to the groove region. g h represents the viscosity of the flow field in the tank region. g Indicates the depth of the radial groove.

3. The method for predicting the torque of a wet clutch under no-load conditions according to claim 1, characterized in that, The expression for the high-speed GA-RBF neural network model is: Among them, h j Let l be the radial basis functions, and w be the number of radial basis functions. j Let X = [x1, x2, ..., xj] be the weights of the j-th basis function. d ] T Let d be the input vector of the neural network, σ be the number of neurons in the input layer, and σ be the input vector of the neural network. j C is the center width of the Gausky function of the j-th neuron in the hidden layer. j =[c j1 c j2 c jn ] T Let ||XC| be the center vector of the j-th neuron in the hidden layer. j || is the vector XC j The Euclidean norm of X represents the relationship between X and C. j The distance between them, T is the torque value of the drive system during the high-speed stage.

4. The method for predicting the torque of a wet clutch under no-load conditions according to claim 1, characterized in that, The training process of the high-speed data-driven prediction model is as follows: The main architecture is based on an RBF neural network, and the center vector C of the hidden layer nodes of the main architecture is obtained by using the GA algorithm. j σ, the center width of the Gaussian function j and the connection weights w between the hidden layer and the output layer j Optimize the training to determine the best parameters for the network.

5. The method for predicting the torque of a wet clutch under no-load conditions according to claim 4, characterized in that, The training process of the GA algorithm specifically includes: Using real number encoding, the center vector C j and the center width σ of the Gaussian function j Perform cross-arrangement and assign weights w j Arrange them last to obtain chromosome codes; Based on the chromosome encoding and according to the set value range of each parameter variable, a certain number of chromosome strings are randomly generated. Each chromosome string includes different parameter combinations to form an initial population. Construct a fitness function and a genetic operator; the fitness function is the reciprocal of the root mean square error between the actual value and the predicted value; the genetic operator is determined by roulette wheel selection, three-point arithmetic crossover, and mutation operation; The process iterates based on the initial population and the genetic operator, and calculates the fitness value according to the fitness function until the maximum number of iterations is reached or the set convergence condition of optimal fitness value is met. Then the iteration ends, and the individual with the optimal fitness value in the population is selected as the optimal solution.

6. A no-load wet clutch torque prediction system, characterized in that, include: The data acquisition unit is used to acquire the data to be measured, including the number of friction pairs, friction plate rotation speed, oil temperature, oil supply flow rate, friction pair clearance, number of grooves, groove width, and groove depth. The model building unit is used to construct a hybrid prediction model for radial groove torque. The hybrid prediction model for radial groove torque includes a low-speed mechanism model and a high-speed data-driven prediction model. The high-speed data-driven prediction model adopts a GA-RBF neural network model. The low-speed mechanism model divides the radial groove torque of the rotating flow field of the friction pair gap into two parts: the groove region and the non-groove region. The model prediction unit is used to normalize the data to be measured, and input the normalized data into the hybrid prediction model of the belt-driven torque for processing, and output the predicted value of the belt-driven torque.

7. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the no-load wet clutch belt torque prediction method according to any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the no-load wet clutch belt torque prediction method as described in any one of claims 1-6.

Images