A method and computer device for quickly predicting the temperature of friction plates of wet dual clutches

Abstract

The present invention proposes a method and computer device for quickly predicting the temperature of a wet dual clutch friction plate, which solves the problem of clutch failure caused by excessive friction plate temperature during the operation of the wet dual clutch. The method uses a wet dual clutch heat transfer simulation model to obtain training samples, establish a clutch temperature prediction model, and quickly predict the temperature of the wet dual clutch friction plate according to the temperature of the wet dual clutch friction plate, dual steel plate, and clutch surface cooling oil under the clutch operating conditions and operating environment. The method can be used to monitor the clutch temperature on a real vehicle, and return the temperature result to the clutch cooling system controller, thereby controlling the cooling system according to the predicted value of the friction plate temperature, so that the wet dual clutch always operates within a normal temperature range.

Description

Technical Field

[0001] The invention relates to a wet dual clutch of an automobile, and in particular to a prediction technology for a friction plate, a dual steel plate, and the surface oil temperature of the dual steel plate of the wet dual clutch. Background Art

[0002] Due to the advantages of good acceleration continuity, low fuel consumption, smooth shifting, and high transmission efficiency, DCT automatic transmission is widely used in passenger cars. As the core component of DCT automatic transmission, the relative sliding between the dual steel plate and the friction plate during the engagement process of the wet dual clutch will generate a large amount of friction heat. Part of the generated heat is taken away by the cooling oil, and part enters the clutch body, causing the temperature of each clutch component to rise, among which the temperature increase of the friction plate is the most obvious. If the heat dissipation capacity of the clutch cooling system is insufficient and the heat cannot be taken away in time, the temperature of the friction plate will rise, causing the friction plate to burn and other faults, resulting in clutch failure.

[0003] Due to the characteristics of the clutch structure, it is difficult to arrange temperature sensors on the friction plate and the dual steel plate during the operation of the car, so the temperature of the wet dual clutch friction plate cannot be monitored by sensors. At present, the temperature monitoring of the wet dual clutch is mainly to return the temperature of the oil thrown out of the clutch oil outlet through the clutch oil outlet oil temperature sensor, and indirectly reflect the temperature of the friction plate and the dual steel plate inside the clutch by the temperature of the oil thrown out of the oil outlet, or to calculate the friction plate temperature by constructing a functional relationship between the friction plate and the clutch outlet cooling oil temperature.

[0004] Patent document 201911226122.1 discloses a control method and device based on the surface temperature of a wet dual clutch. The temperature of the clutch friction plate and the dual steel plate is indirectly calculated using the clutch outlet cooling oil temperature. An outlet cooling oil temperature sensor needs to be installed, which is costly and causes cumulative errors due to multiple conversions in the middle. Patent document 201410217118.X discloses a method for measuring the temperature of a DCT wet clutch based on neural network prediction. A clutch temperature prediction model is established based on a neural network. The main problem to be solved is that the oil temperature at the wet clutch outlet collected by the oil temperature sensor in the prior art has a delay, and the temperature of the clutch is predicted by the oil outlet temperature. However, a large number of experiments are required to obtain oil outlet oil temperature samples during the model training process, which is costly and time-consuming. Summary of the invention

[0005] The technical problem to be solved by the present invention is to solve the problem of clutch failure caused by excessive friction plate temperature during the operation of a wet dual clutch. A fast prediction method for the friction plate temperature of a wet dual clutch is proposed. The training samples are obtained by using a wet dual clutch heat transfer simulation model, and a clutch temperature prediction model is established. The temperature of the wet dual clutch friction plate, the dual steel plate, and the clutch surface cooling oil under the clutch operating conditions and operating environment can be used to quickly predict the temperature of the wet dual clutch friction plate. This method can be used to monitor the clutch temperature on a real vehicle and return the temperature result to the clutch cooling system controller, thereby controlling the cooling system according to the predicted value of the friction plate temperature, so that the wet dual clutch always operates within a normal temperature range.

[0006] In order to solve the technical problems in the present invention, the technical solutions adopted are as follows:

[0007] In one aspect, the present invention provides a method for quickly predicting the temperature of a friction plate of a wet dual clutch, comprising the following steps:

[0008] Step A, online acquisition of the outer clutch oil pressure, slip difference, inner clutch oil pressure, slip difference, clutch input speed, inlet cooling oil flow and inlet cooling oil temperature of the wet dual clutch, and a transmission gear position signal connected to the clutch.

[0009] Step B, calculate the inner and outer clutch transmission torques according to the three-dimensional difference table of the inner and outer clutch transmission torques, as well as the inner and outer clutch oil pressures and slip differences, and calculate the inner and outer clutch slip work based on the transmission torque calculation results and the slip difference.

[0010] Step C, normalize the calculated external clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and input them into the temperature prediction model P, and denormalize the prediction results to obtain a temperature prediction value; normalize the calculated internal clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and input them into the temperature prediction model P, and denormalize the prediction results to obtain another temperature prediction value.

[0011] Step D, combining the two sets of temperature prediction values ​​according to the transmission gear position signal of the clutch connection, to obtain the final temperature prediction values ​​of the inner and outer clutch friction plates, the dual steel plates, and the dual steel plate surface cooling oil of the wet dual clutch.

[0012] In the method of the present invention, it is critical to use the temperature prediction model Pouter and the temperature prediction model Pinner, which are constructed using the following method:

[0013] Step 1: Obtain clutch temperature prediction model training samples through a wet dual clutch heat transfer simulation model.

[0014] Since it is difficult to obtain the clutch friction plate temperature through experimental acquisition methods for rotating components such as clutches, the temperature prediction model training samples in the present invention are obtained by establishing a wet dual clutch heat transfer simulation model.

[0015] The heat transfer simulation model in the present invention is constructed offline based on the thermal network theory according to the heat transfer mode between the components of the wet dual clutch or between the clutch and the surrounding fluid, and the heat transfer process of the clutch. The reliability of the heat transfer model is verified by comparing with experimental data.

[0016] Sample acquisition includes: acquisition of inner and outer clutch friction plates, dual steel plates, dual steel plate surface cooling oil, and clutch outlet cooling oil temperature samples under different sliding friction work, different input speeds, different inlet cooling oil temperatures, and different inlet cooling oil flows in the two states of wet dual clutch outer clutch engagement, inner clutch disengagement, and inner clutch engagement and outer clutch disengagement; as well as acquisition of the transmission torque of the outer and inner clutches under different oil pressures, different oil temperatures, and different sliding differences, as well as acquisition of the gear signal at the transmission end.

[0017] Step 2, constructing a temperature prediction model P, including establishing a prediction model Pexternal for the temperatures of the inner and outer clutch friction plates, dual steel plates, and cooling oil on the steel plate surfaces when the outer clutch of the wet dual clutch is engaged and the inner clutch is disengaged; and a prediction model Pinner for the temperatures of the inner and outer clutch friction plates, dual steel plates, and cooling oil on the steel plate surfaces when the inner clutch is engaged and the outer clutch is disengaged.

[0018] Finally, the constructed model needs to be tested offline.

[0019] In one embodiment of the present invention, step 1 comprises:

[0020] Step 1.1, based on the thermal network method, a wet dual clutch heat transfer simulation model is established, and the temperature of the dual steel plate and the dual steel plate cooling oil collected in the experiment is compared with the simulation results of the heat transfer simulation model to verify the accuracy of the simulation results of the model. Of course, this step first needs to analyze the factors affecting the temperature of the wet dual clutch friction plate, dual steel plate, and clutch steel plate surface cooling oil, including the working state of the wet dual clutch, sliding friction work, clutch input speed, clutch inlet cooling oil flow, and clutch inlet cooling oil temperature.

[0021] Step 1.2, according to the wet dual-clutch heat transfer simulation model, obtain samples of the outer clutch and inner clutch friction plates, dual steel plates, and dual steel plate surface cooling oil temperatures under different sliding friction works, different input speeds, different inlet cooling oil temperatures, and different inlet cooling oil flow rates when the wet dual clutch is in the two states of outer clutch engaged, inner clutch disengaged, and inner clutch engaged, outer clutch disengaged.

[0022] Sample acquisition includes: acquisition of inner and outer clutch friction plates, dual steel plates, dual steel plate surface cooling oil, and clutch outlet cooling oil temperature samples under different sliding friction work, different input speeds, different inlet cooling oil temperatures, and different inlet cooling oil flows in the two states of wet dual clutch outer clutch engagement, inner clutch disengagement, and inner clutch engagement and outer clutch disengagement; as well as acquisition of the transmission torque of the outer and inner clutches under different oil pressures, different oil temperatures, and different sliding differences, as well as acquisition of the gear signal at the transmission end.

[0023] In one embodiment of the present invention, step 2 comprises:

[0024] Step 2.1, the temperature samples obtained by the simulation model are grouped according to the two working states of the clutch, the C outer temperature sample set when the outer clutch is engaged and the inner clutch is disengaged, and the C inner temperature sample set when the inner clutch is engaged and the outer clutch is disengaged.

[0025] Step 2.2, respectively, take the C outer temperature sample set and the C inner temperature sample set, with the sliding friction work, clutch input end speed, clutch inlet cooling oil flow, clutch inlet cooling oil temperature as the input variable group, to form the C outer input variable sample set and the C inner input variable sample set; the outer clutch and inner clutch friction plates, dual steel plates, and dual steel plate surface cooling oil temperatures are taken as the output variable group to form the C outer output variable sample set and the C inner output variable sample set.

[0026] Step 2.3, divide the C external input variable sample set and the corresponding C external output variable sample set into the training sample set Test_C external input variable sample set, Test_C external output variable sample set, Train_C external input variable sample set, and Train_C external output variable sample set outside the prediction model P; similarly, divide the C internal input variable sample set and the C internal output variable sample set into the training sample set Test_C internal input variable sample set, Test_C internal output variable sample set, Train_C internal input variable sample set, and Train_C internal output variable sample set within the prediction model P.

[0027] Step 2.4, conduct genetic neural network training and testing of more than three layers, and construct temperature prediction models Pouter when the outer clutch is engaged and the inner clutch is separated, and temperature prediction models Pinner when the inner clutch is engaged and the outer clutch is separated.

[0028] In one embodiment of the present invention, the step 2.4 further comprises:

[0029] Step 2.4.1, normalize the sample set of external input variables of Test_C, the sample set of external output variables of Test_C, the sample set of external input variables of Train_C, and the sample set of external output variables of Train_C, and normalize the sample set of internal input variables of Test_C, the sample set of internal output variables of Test_C, the sample set of internal input variables of Train_C, and the sample set of internal output variables of Train_C;

[0030] Step 2.4.2, respectively determine the number of layers and the number of hidden layer neuron nodes of the neural network outside the temperature prediction model P and inside the temperature prediction model P, establish a neural network weight parameter updating method based on a genetic algorithm, and construct a GA-BP genetic neural network model with more than three layers;

[0031] Step 2.4.3, use the normalized Test_C external input variable sample set, Test_C external output variable sample set, Train_C external input variable sample set, and Train_C external output variable sample set as the external input and output of the temperature prediction model P; use the normalized Test_C internal input variable sample set, Test_C internal output variable sample set, Train_C internal input variable sample set, and Train_C internal output variable sample set as the internal input and output of the temperature prediction model P;

[0032] Step 2.4.4, initialize the GA-BP genetic neural network model parameters, set the number of iterations, genetic algorithm population, mutation probability, crossover probability, evolutionary generations, and finally train the model;

[0033] Step 2.4.5, according to the error of the test set on the trained model, continuously iterate and update the model until the error meets the requirement, and then store the trained temperature prediction model Pouter and the temperature prediction model Pinner respectively;

[0034] Step 2.4.6, according to the transmission gear signal connected to the wet dual clutch, design the temperature prediction model Pexternal and temperature prediction model Pinternal calling logic, and construct the temperature prediction model P of the outer clutch, inner clutch friction plate, dual steel plate, and steel plate surface cooling oil of the wet clutch under different working states and different working environments.

[0035] In one embodiment of the present invention, the three-dimensional difference table of the transmission torque of the inner and outer clutches is formed by experimentally obtaining the transmission torque of the outer and inner clutches under different oil pressures, different oil temperatures, and different slip differences.

[0036] In another aspect, the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for rapidly predicting the temperature of a wet dual clutch friction plate when executing the computer program.

[0037] The beneficial effects of the present invention relative to the prior art are:

[0038] 1. The present invention uses a wet dual clutch heat transfer simulation model to obtain training samples of a temperature prediction model, and constructs a temperature prediction model Pouter and a temperature prediction model Pinner to predict the temperature. On the one hand, it greatly saves the cost and test time of collecting data through experiments, and on the other hand, it improves the accuracy of the prediction.

[0039] 2. The present invention can accurately predict the temperature of the dual steel plates and the surface cooling oil of the dual steel plates of the inner and outer clutches of a wet dual clutch without the need for an additional clutch outlet cooling oil temperature sensor, with a small error. At the same time, it provides a solution to the difficulty in monitoring the temperature of the clutch friction plates. It can be used for monitoring and predicting the temperature of the dual steel plates, friction plates, and the surface cooling oil of the dual steel plates of a wet dual clutch under actual operating conditions, and has broad application prospects.

[0040] The method of the present invention can be used to predict the wet dual-clutch friction plate, dual steel plate, and dual steel plate surface oil temperature during vehicle driving, and can be used to formulate a clutch cooling strategy to prevent the clutch friction plate from failing due to excessive temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the wet dual clutch structure;

[0042] Figure 2 It is the thermal network structure diagram of wet dual clutch;

[0043] Figure 3 It is the heat transfer simulation model verification process;

[0044] Figure 4 It is a flow chart of the clutch temperature prediction model constructed in the state of outer clutch engagement and inner clutch disengagement of the present invention;

[0045] Figure 5 It is a combination of the output results of the temperature prediction model when the outer clutch is engaged and the inner clutch is separated and the temperature prediction model when the inner clutch is engaged and the outer clutch is separated under different gear signals of the wet dual clutch described in the present invention;

[0046] Figure 6 It is a schematic diagram of the topological structure of a three-layer BP neural network for predicting temperature when the outer clutch of a wet dual clutch is engaged and the inner clutch is separated as described in the present invention;

[0047] Figure 7 It is a schematic diagram of the topological structure of a three-layer BP neural network for predicting temperature when the inner clutch of a wet dual clutch is engaged and the outer clutch is separated as described in the present invention;

[0048] Figure 8 It is a flow chart of the genetic algorithm of the present invention for optimizing the weights and threshold parameters of the BP neural network;

[0049] Fig. 9 It is the fitness curve of the GA-BP neural network evolution process;

[0050] Fig.10 The invention discloses a temperature prediction process of inner and outer clutch friction plates, dual steel plates and dual steel plate surface cooling oil. DETAILED DESCRIPTION

[0051] Hereinafter, the present technology will be described in detail in conjunction with exemplary embodiments with reference to the accompanying drawings.

[0052] See also Figure 1 , is a schematic diagram of the wet dual clutch structure mentioned in the present invention. Normally, the working process of this wet dual clutch is divided into two states: one state is that the outer clutch is engaged and the inner clutch is disengaged, and the other state is that the inner clutch is engaged and the outer clutch is disengaged.

[0053] See also Figure 4 , is a prediction model establishment method for the surface cooling oil temperature of wet dual clutch friction plate, dual steel plate, and clutch steel plate based on genetic neural network:

[0054] Step 1: Obtaining training samples of clutch temperature prediction model. The process is as follows:

[0055] Step 1.1, analyze the factors affecting the surface cooling oil temperature of the wet dual clutch friction plate, dual steel plate, and clutch steel plate, including the working state of the wet dual clutch, sliding friction work, clutch input end speed, clutch inlet cooling oil flow, and clutch inlet cooling oil temperature.

[0056] Step 1.2, establish a wet dual clutch heat transfer simulation model based on the thermal network method, and use the experimentally collected temperatures of the dual steel plates and the dual steel plate cooling oil to compare with the simulation results of the heat transfer simulation model to verify the accuracy of the simulation results of the model.

[0057] Figure 2 This is the thermal network structure diagram of the wet dual clutch heat transfer simulation model in the present invention. According to the structure of each clutch component and the heat exchange method with the cooling oil, the clutch is discretized into many mass blocks. Each mass block is connected with the calculated thermal resistance and heat transfer resistance, and finally forms the following Figure 2The heat transfer simulation model thermal network structure shown in the figure can be used to construct a heat transfer simulation model of a wet dual clutch based on the heat transfer simulation software according to the thermal network structure diagram.

[0058] Figure 3 This is a process for verifying the accuracy of simulation results of the wet dual clutch heat transfer simulation model established in the present invention. The process is, through a large number of experimental verifications, the simulation results of the constructed wet dual clutch heat transfer simulation model are compared with the bench test results under the conditions of inputting the same amount of sliding work, clutch input end speed, clutch inlet cooling oil temperature, and clutch inlet cooling oil, and the model is continuously optimized to make the relative error between the simulation results and the test results within 3%. The simulation model is considered reliable, and the wet dual clutch heat transfer simulation model is used to simulate the change process of the friction plate, dual steel plate, and dual steel plate surface cooling oil temperature under the actual working conditions of the wet dual clutch.

[0059] Step 1.3, based on the wet dual-clutch heat transfer simulation model established offline, obtain the temperature samples of the outer clutch and inner clutch friction plates, dual steel plates, and dual steel plate surface cooling oil under different sliding friction work, different input speeds, different inlet cooling oil temperatures, and different inlet cooling oil flow rates when the wet dual clutch is in the two states of outer clutch engaged, inner clutch disengaged and inner clutch engaged, outer clutch disengaged.

[0060] Step 2, offline respectively establish prediction models Pexternal for the temperature of inner and outer clutch friction plates, dual steel plates, and cooling oil on the steel plate surface when the outer clutch is engaged and the inner clutch is separated; and Pinner for the temperature of inner and outer clutch friction plates, dual steel plates, and cooling oil on the steel plate surface when the inner clutch is engaged and the outer clutch is separated, the process is as follows:

[0061] Step 2.1, the temperature samples obtained by the simulation model are grouped according to the two working states of the clutch, the C outer temperature sample set when the outer clutch is engaged and the inner clutch is separated, and the C inner temperature sample set when the inner clutch is engaged and the outer clutch is separated.

[0062] Step 2.2, respectively, the C outer temperature sample set and the C inner temperature sample set are used as the input variable group, with the sliding friction work, the clutch input end speed, the clutch inlet cooling oil flow rate, and the clutch inlet cooling oil temperature as the input variable group, to form the C outer input variable sample set and the C inner input variable sample set; the outer clutch and inner clutch friction plate, the dual steel plate, and the dual steel plate surface cooling oil temperature are used as the output variable group to form the C outer output variable sample set and the C inner output variable sample set.

[0063] Step 2.3, respectively divide the C external input variable sample set and the corresponding C external output variable sample set into the training sample set Test_C external input variable sample set, Test_C external output variable sample set, Train_C external input variable sample set, and Train_C external output variable sample set outside the prediction model P; similarly, respectively divide the C internal input variable sample set and the C internal output variable sample set into the training sample set Test_C internal input variable sample set, Test_C internal output variable sample set, Train_C internal input variable sample set, and Train_C internal output variable sample set inside the prediction model P.

[0064] Step 2.4, perform genetic neural network training and testing of more than three layers, and construct temperature prediction models Pouter and Pinner respectively when the outer clutch is engaged and the inner clutch is separated.

[0065] Step 2.4.1: normalize the sample set of input variables outside Test_C, the sample set of output variables outside Test_C, the sample set of input variables outside Train_C, and the sample set of output variables outside Train_C; normalize the sample set of input variables inside Test_C, the sample set of output variables inside Test_C, the sample set of input variables inside Train_C, and the sample set of output variables inside Train_C

[0066] Step 2.4.2, respectively determine the number of layers and the number of hidden layer neuron nodes of the neural network outside the temperature prediction model P and inside the temperature prediction model P, establish a neural network weight parameter update method based on the genetic algorithm, and construct a GA-BP genetic neural network model with more than three layers.

[0067] Step 2.4.3, use the normalized Test_C external input variable sample set, Test_C external output variable sample set, Train_C external input variable sample set, and Train_C external output variable sample set as the input and output of the temperature prediction model P external; use the normalized Test_C internal input variable sample set, Test_C internal output variable sample set, Train_C internal input variable sample set, and Train_C internal output variable sample set as the input and output of the temperature prediction model P internal

[0068] Step 2.4.4, initialize the GA-BP genetic neural network model parameters, set the number of iterations, genetic algorithm population, mutation probability, crossover probability, evolutionary generations, and finally train the model.

[0069] Step 2.4.5, according to the error of the test set on the trained model, continuously iterate and update the model until the error meets the requirements, and then store the trained temperature prediction model Pouter and temperature prediction model Pinner respectively.

[0070] Step 2.4.6, according to the transmission gear signal connected to the wet dual clutch, design the temperature prediction model P external and temperature prediction model P internal calling logic, and construct the temperature prediction model P of the outer clutch, inner clutch friction plate, dual steel plate, and steel plate surface cooling oil of the wet clutch under different working states and different working environments.

[0071] The above content can be combined Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Fig. 9 To further explain:

[0072] See also Figure 4 , is the process of establishing a temperature prediction model for a wet dual clutch when the outer clutch is engaged and the inner clutch is disengaged, which is the same as the process of establishing a temperature prediction model for the inner clutch is engaged and the outer clutch is disengaged, including collecting data samples using the constructed wet dual clutch heat generation simulation model, normalizing the sample data, sending it to the constructed GA-BP genetic neural network model for training, and saving the final model, wherein the data sample collection includes the following steps:

[0073] Firstly, the input and output of the wet dual clutch temperature prediction model are determined. The input is the clutch input end speed, cooling oil inlet temperature, cooling oil inlet flow rate, and sliding power. The output is the outer clutch friction plate temperature, the outer clutch dual steel plate temperature, the inner clutch friction plate temperature, the inner clutch dual steel plate temperature, the outer clutch dual steel plate surface oil temperature, and the inner clutch dual steel plate surface oil temperature. Secondly, based on the constructed wet dual clutch heat transfer model, data samples under the outer clutch engaged and inner clutch disengaged states, as well as data samples under the inner clutch engaged and outer clutch disengaged states are collected according to the input and output. Finally, the data samples are normalized. The normalization is performed according to the formula xi=2*(Xi-Xmin) / (Xmax-Xmin)-1. Taking the input speed as an example, xi is the normalized result of input speed sample i, Xi is the value of input speed sample i, Xmin is the minimum value in the input speed sample, and Xmax is the maximum value in the input speed sample.

[0074] After the temperature prediction model under the state of outer clutch engaged and inner clutch disengaged is constructed, and the temperature prediction model under the state of inner clutch engaged and outer clutch disengaged is constructed, the output results of the two temperature prediction models are merged according to the transmission gear position signal control connected to the clutch. The merging process flow chart is shown in Figure 5 As shown, the process is as follows

[0075] First, according to the transmission gear signal, determine whether the signal value L is an odd number or an even number. When the L value is an odd number, output K1=1, K2=0; when the L value is an even number, output K1=0, K2=1; finally, multiply the output result of the temperature prediction model when the outer clutch is engaged and the inner clutch is disengaged by K1, and then multiply the output result of the temperature prediction model when the inner clutch is engaged and the outer clutch is disengaged by K2, and finally output the temperature prediction result of the wet dual clutch.

[0076] Figure 6 , Figure 7 They are respectively a schematic diagram of the three-layer BP neural network topology structure for temperature prediction when the outer clutch is engaged and the inner clutch is disengaged, and a schematic diagram of the three-layer BP neural network topology structure for temperature prediction when the inner clutch is engaged and the outer clutch is disengaged.

[0077] First, the empirical formula is estimated based on the number of neurons In the formula, n and l are the number of neurons in the input layer and the output layer respectively, a is a constant ranging from 1 to 10, and m is the number of neurons in the hidden layer. Substituting the determined n and l into the above formula, we can obtain the value range of m of the three-layer BP neural network for temperature prediction when the outer clutch is engaged and the inner clutch is separated and the three-layer BP neural network for temperature prediction when the inner clutch is engaged and the outer clutch is separated, which are both 4-14. Then, different numbers of neurons are selected in turn to train 10 times respectively, and the average value of the mean square error and correlation coefficient in the 10 training results is taken for evaluation. Finally, it is determined that m of the three-layer BP neural network for temperature prediction when the outer clutch is engaged and the inner clutch is separated is 11, and m of the three-layer BP neural network for temperature prediction when the inner clutch is engaged and the outer clutch is separated is 14.

[0078] The three-layer BP neural network for predicting the temperature when the outer clutch is engaged and the inner clutch is separated and the three-layer BP neural network for predicting the temperature when the inner clutch is engaged and the outer clutch is separated are trained respectively, and the weights and thresholds of the BP neural network are updated using the genetic optimization algorithm. The weight and threshold update process of the two neural networks is the same, such as Figure 8 The genetic algorithm shown in the figure optimizes the BP neural network weights and threshold parameter flow chart. The process is as follows:

[0079] First, (1) set the initial weights and thresholds, encode the weights and thresholds to obtain the initial population, (2) decode to obtain the initial weights and thresholds, (3) assign the initial weights and thresholds to the established three-layer neural network, (4) use the training samples to train the network, use the test samples to test the network, and obtain the test error, (5) calculate the fitness, select chromosomes with high fitness to replicate, crossover, mutate, and obtain a new population, (6) determine whether the error condition or the set total number of iterations is met. If so, decode the weights and thresholds to obtain the final neural network weights and thresholds; if not, return to step (2).

[0080] Fig. 9 This is the fitness curve of the evolution process of the genetic algorithm. The average fitness of the samples decreases with the increase of the number of iterations and finally tends to be stable, indicating that the error of the training samples is small.

[0081] Fig.10 The above temperature prediction model is used to predict the temperature of the inner and outer clutch friction plates, dual steel plates, and the surface cooling oil of the dual steel plates. The process is as follows:

[0082] Step A, online acquisition of the outer clutch oil pressure, slip difference, inner clutch oil pressure, slip difference, clutch input speed, inlet cooling oil flow and inlet cooling oil temperature of the wet dual clutch, and a transmission gear position signal connected to the clutch.

[0083] Step B, according to the three-dimensional difference table of the inner and outer clutch transmission torque, as well as the inner and outer clutch oil pressure and slip difference, respectively calculate the inner and outer clutch transmission torque, and respectively calculate the inner and outer clutch slip work based on the transmission torque calculation results.

[0084] Here, the three-dimensional difference table of the inner and outer clutch transmission torques is formed by experimentally obtaining the transmission torques of the outer and inner clutches under different oil pressures, different oil temperatures, and different slip differences.

[0085] Step C: normalize the calculated external clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and input them to the temperature prediction model P, and denormalize the prediction results to obtain a temperature prediction value; normalize the calculated internal clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and input them to the temperature prediction model P, and denormalize the prediction results to obtain another temperature prediction value.

[0086] Step D, according to the transmission gear position signal of the clutch connection, the two sets of temperature prediction values ​​are combined to obtain the final temperature prediction values ​​of the inner and outer clutch friction plates, the dual steel plates, and the cooling oil on the surface of the dual steel plates of the wet dual clutch.

[0087] To explain more clearly, the above steps generally include online data collection of temperature prediction model input variables, sliding friction work calculation, data collection and sliding friction work normalization, input to the trained model, output result denormalization, and obtain clutch temperature prediction value.

[0088] The signal values ​​of the offline data collection are the outer clutch oil pressure, the outer clutch slip difference, the inner clutch oil pressure, the inner clutch slip difference, the clutch input speed, the clutch inlet cooling oil temperature, and the clutch inlet cooling oil temperature; the slip work calculation includes the slip work calculation of the outer clutch and the inner clutch, and the process is the same, as follows:

[0089] The transmission torque of the inner and outer clutches is calculated by utilizing the transmission torque of the outer and inner clutches collected offline at different oil pressures, different oil temperatures, and different slip differences. The transmission torque of the inner and outer clutches at different oil pressures, slip differences, and oil temperatures is calculated by difference. The oil temperature refers to the oil temperature on the surface of the outer clutch dual steel plates and the oil temperature on the surface of the inner clutch dual steel plates in the temperature prediction results of the wet dual clutch.

[0090] The slip work of the inner and outer clutches is calculated according to the theoretical formula P=T*w slip difference. Taking the outer clutch as an example, P is the slip work of the outer clutch, T is the torque transmitted by the outer clutch, and w slip difference is the slip difference of the outer clutch.

[0091] The output result is denormalized according to the formula Xi = (xi + 1) * (Xmax - Xmin) / 2 + Xmin. Taking the inner clutch friction plate temperature as an example, Xi is the actual predicted value of the inner clutch friction plate temperature after denormalization, xi is the predicted output value of the genetic neural network, and Xmax and Xmin are the maximum and minimum values ​​of the corresponding variables of the sample training data, respectively.

[0092] Through the above embodiments and figures, it can be proved that the present invention can directly and quickly predict the temperature of the clutch friction plate, the dual steel plate and the dual steel plate according to the working condition of the wet dual clutch only, without the need for an additional clutch outlet cooling oil temperature, thus saving the cost of the clutch outlet cooling oil temperature sensor; and by obtaining model training samples through simulation methods, a large number of samples can be obtained in a short time, thus saving experimental cost and time.

Claims

1. A method for rapid prediction of friction plate temperature of a wet dual clutch, characterized by: include: Step A, online acquisition of the outer clutch oil pressure, slip difference, inner clutch oil pressure, slip difference, clutch input speed, inlet cooling oil flow and inlet cooling oil temperature of the wet dual clutch, and a transmission gear position signal connected to the clutch; Step B, calculating the inner and outer clutch transmission torques according to the three-dimensional difference table of the inner and outer clutch transmission torques, as well as the inner and outer clutch oil pressures and slip differences, and calculating the inner and outer clutch slip work based on the transmission torque calculation results and the slip difference; Step C, normalizing the calculated external clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and inputting them into the temperature prediction model Pouter, denormalizing the prediction result to obtain a temperature prediction value; normalizing the calculated internal clutch friction work, clutch input speed, inlet cooling oil flow rate and inlet cooling oil temperature and inputting them into the temperature prediction model Pinner, denormalizing the prediction result to obtain another temperature prediction value; the temperature prediction model Pouter and the temperature prediction model Pinner are constructed by the following method: Step 1, obtaining clutch temperature prediction model training samples through a wet dual clutch heat transfer simulation model; The heat transfer simulation model is constructed offline based on the thermal network theory according to the heat transfer mode between the components of the wet dual clutch or between the components and the surrounding fluid and the heat transfer process of the clutch; Step 2, constructing a temperature prediction model P, including establishing a prediction model Pexternal for the temperature of the inner and outer clutch friction plates, the dual steel plates, and the cooling oil on the steel plate surface when the outer clutch of the wet dual clutch is engaged and the inner clutch is disengaged; and a prediction model Pinner for the temperature of the inner and outer clutch friction plates, the dual steel plates, and the cooling oil on the steel plate surface when the inner clutch is engaged and the outer clutch is disengaged; Step 3, performing an offline test on the temperature prediction model P; Step D, according to the transmission gear position signal of the clutch connection, the two sets of temperature prediction values ​​are combined to obtain the final temperature prediction values ​​of the inner and outer clutch friction plates, the dual steel plates, and the dual steel plate surface cooling oil of the wet dual clutch.

2. The method for rapid prediction of friction plate temperature of a wet dual clutch according to claim 1, characterized in that: The step 1 comprises: Step 1.1, establish a wet dual clutch heat transfer simulation model based on the thermal network method, compare the temperature of the dual steel plate and the dual steel plate cooling oil collected in the experiment with the simulation results of the heat transfer simulation model to verify the accuracy of the simulation results of the model; Step 1.2, according to the wet dual-clutch heat transfer simulation model, obtain samples of the outer clutch and inner clutch friction plates, dual steel plates, and dual steel plate surface cooling oil temperatures under different sliding friction works, different input speeds, different inlet cooling oil temperatures, and different inlet cooling oil flow rates when the wet dual clutch is in the two states of outer clutch engaged, inner clutch disengaged, and inner clutch engaged, outer clutch disengaged.

3. The method for rapid prediction of friction plate temperature of a wet dual clutch according to claim 1, characterized in that: The step 2 comprises: Step 2.1, grouping the temperature samples obtained by the simulation model according to the two working states of the clutch, the C outer temperature sample set when the outer clutch is engaged and the inner clutch is disengaged, and the C inner temperature sample set when the inner clutch is engaged and the outer clutch is disengaged; Step 2.2, respectively, the C outer temperature sample set and the C inner temperature sample set are used as the input variable group, with the sliding friction work, the clutch input end speed, the clutch inlet cooling oil flow rate, and the clutch inlet cooling oil temperature as the input variable group, to form the C outer input variable sample set and the C inner input variable sample set; the outer clutch and inner clutch friction plate, the dual steel plate, and the dual steel plate surface cooling oil temperature are used as the output variable group to form the C outer output variable sample set and the C inner output variable sample set; Step 2.3, respectively divide the C external input variable sample set and the corresponding C external output variable sample set into the training sample set Test_C external input variable sample set, Test_C external output variable sample set, Train_ C external input variable sample set, and Train_ C external output variable sample set outside the prediction model P; similarly, respectively divide the C internal input variable sample set and the C internal output variable sample set into the training sample set Test_C internal input variable sample set, Test_C internal output variable sample set, Train_ C internal input variable sample set, and Train_ C internal output variable sample set within the prediction model P; Step 2.4, conduct genetic neural network training and testing of more than three layers, and construct temperature prediction models Pouter when the outer clutch is engaged and the inner clutch is separated, and temperature prediction models Pinner when the inner clutch is engaged and the outer clutch is separated.

4. The method for rapid prediction of friction plate temperature of a wet dual clutch according to claim 3, characterized in that: The step 2.4 comprises: Step 2.4.1, normalize the sample set of external input variables of Test_C, the sample set of external output variables of Test_C, the sample set of external input variables of Train_C, and the sample set of external output variables of Train_C, and normalize the sample set of internal input variables of Test_C, the sample set of internal output variables of Test_C, the sample set of internal input variables of Train_C, and the sample set of internal output variables of Train_C; Step 2.4.2, respectively determine the number of layers and the number of hidden layer neuron nodes of the neural network outside the temperature prediction model P and inside the temperature prediction model P, establish a neural network weight parameter updating method based on a genetic algorithm, and construct a GA-BP genetic neural network model with more than three layers; Step 2.4.3, using the normalized Test_C external input variable sample set, Test_C external output variable sample set, Train_ C external input variable sample set, and Train_ C external output variable sample set as the input and output of the temperature prediction model P; using the normalized Test_C internal input variable sample set, Test_C internal output variable sample set, Train_ C internal input variable sample set, and Train_ C internal output variable sample set as the input and output of the temperature prediction model P; Step 2.4.4, initialize the GA-BP genetic neural network model parameters, set the number of iterations, genetic algorithm population, mutation probability, crossover probability, evolutionary generations, and finally train the model; Step 2.4.5, according to the error of the test set on the trained model, continuously iterate and update the model until the error meets the requirement, and then store the trained temperature prediction model Pouter and the temperature prediction model Pinner respectively; Step 2.4.6, according to the transmission gear signal connected to the wet dual clutch, design the temperature prediction model Pexternal and temperature prediction model Pinternal calling logic, and construct the temperature prediction model P of the outer clutch, inner clutch friction plate, dual steel plate, and steel plate surface cooling oil of the wet clutch under different working states and different working environments.

5. The method for rapid prediction of friction plate temperature of a wet dual clutch according to any one of claims 1 to 4, characterized in that: The three-dimensional difference table of the transmission torque of the inner and outer clutches is formed by experimentally obtaining the transmission torque of the outer clutch and the inner clutch under different oil pressures, different oil temperatures, and different slip differences.

6. The method for rapid prediction of friction plate temperature of a wet dual clutch according to any one of claims 1 to 4, characterized in that: The calculation formula of the slip work of the inner and outer clutches is: P=T*w slip difference, wherein P is the slip work of the outer clutch, T is the torque transmitted by the outer clutch, and w slip difference is the slip difference of the outer clutch.

7. The method for rapid prediction of friction plate temperature of a wet dual clutch according to claim 4, characterized in that: The output result is denormalized according to the formula Xi=(xi+1)*(Xmax-Xmin) / 2+ Xmin, where Xi is the actual predicted value of the inner clutch friction plate temperature after denormalization, xi is the predicted output value of the genetic neural network, and Xmax and Xmin are the maximum and minimum values ​​of the corresponding variables of the sample training data, respectively.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for quickly predicting the temperature of the friction plate of a wet dual clutch as described in any one of claims 1 to 7 is implemented.

Images