Transmission hydraulic state evaluation method, controller, vehicle and storage medium

Abstract

The invention relates to the technical field of transmissions, and discloses a transmission hydraulic state evaluation method, a controller, a vehicle and a storage medium. According to the transmission hydraulic state evaluation method, real-time rotating speed data of a vehicle is obtained, prediction processing is conducted on the real-time rotating speed data through a hydraulic state prediction model, and a hydraulic state evaluation result of the vehicle is obtained; wherein the hydraulic state prediction model is obtained by training an initial differential neural network model according to rotation speed random data and hydraulic state random data; the hydraulic state random data is determined according to a preset hydraulic simulation model and the rotating speed random data. According to the hydraulic state prediction model, the rotating speed random data and the hydraulic state random data serve as training samples, the sample richness is increased, and the model generalization ability can be improved. Meanwhile, the dynamic physical process of the hydraulic system can be accurately simulated based on the initial differential neural network model, the generalization ability of the model is further improved, and the evaluation deviation is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of transmissions, and particularly to a method for evaluating the hydraulic state of a transmission, a controller, a vehicle, and a storage medium. Background Art

[0002] The hydraulic system of a hybrid vehicle transmission consists of a clutch, a piston, a return spring, a control oil pump, a hydraulic oil filter, an oil storage chamber, etc., and realizes the engagement and separation of the clutch by controlling the forward and reverse rotation of the oil pump. When the transmission is working, it is necessary to obtain the real-time values of the internal parameters of the hydraulic system in order to accurately control the clutch pressure by controlling the rotation speed of the oil pump. In the prior art, the vehicle monitors the hydraulic system parameters through virtual sensor software, and the virtual sensor software includes a neural network model established by using machine learning algorithms. The hydraulic system belongs to a dynamic physical system, and the hydraulic system parameters will fluctuate greatly under different transmission working conditions. The existing neural network model can only rely on the test data under specific working conditions during training, resulting in poor generalization ability of the neural network model, and different degrees of deviation will occur when monitoring the hydraulic system parameters under different transmission working conditions, and the accuracy of the hydraulic state evaluation result cannot be guaranteed. Summary of the Invention

[0003] Based on this, it is necessary to provide a method for evaluating the hydraulic state of a transmission, a controller, a vehicle, and a storage medium for the above technical problems, so as to solve the problems of poor generalization ability of the model and easy deviation of the evaluation result in the process of evaluating the hydraulic state of the transmission.

[0004] A method for evaluating the hydraulic state of a transmission includes: Obtain the real-time rotation speed data of the vehicle, and perform prediction processing on the real-time rotation speed data through a hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle; wherein, the hydraulic state prediction model is obtained by training an initial differential neural network model according to rotation speed random data and hydraulic state random data; the hydraulic state random data is determined according to a preset hydraulic simulation model and rotation speed random data.

[0005] A controller includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the method for evaluating the hydraulic state of the transmission as described above is implemented.

[0006] A vehicle includes the controller as described above.

[0007] A computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the method for evaluating the hydraulic state of the transmission as described above.

[0008] In the above transmission hydraulic state evaluation method, controller, vehicle, and storage medium, the transmission hydraulic state evaluation method obtains the real-time rotational speed data of the vehicle, and performs prediction processing on the real-time rotational speed data through a hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle. Among them, the hydraulic state prediction model is obtained by training an initial differential neural network model based on rotational speed random data and hydraulic state random data. The hydraulic state random data is determined according to a preset hydraulic simulation model and rotational speed random data. The hydraulic state prediction model of the present invention uses rotational speed random data and hydraulic state random data as training sample data, which can cover rotational speed data and hydraulic state data under different working conditions, increasing the sample richness and helping to improve the generalization ability of the neural network model. At the same time, the hydraulic state prediction model of the present invention is obtained by training on the basis of an initial differential neural network model. The initial differential neural network model helps to more accurately simulate the dynamic physical continuous change process of the hydraulic system, enabling the model to better adapt to different inputs and data distributions, further improving the generalization ability of the model, and thus accurately predicting the states of various parameters in the hydraulic system. In addition, the initial differential neural network model can also reduce the consumption of computing resources and time while ensuring the model performance, thereby improving the computing efficiency of the hydraulic state prediction model and realizing rapid evaluation of the transmission hydraulic state. Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0010] Figure 1 It is a schematic flowchart of a transmission hydraulic state evaluation method in an embodiment of the present invention; Figure 2 It is a schematic diagram of a neural network structure of a transmission hydraulic state evaluation method in an embodiment of the present invention; Figure 3 It is a schematic diagram of the random number distribution of rotational speed variables in a transmission hydraulic state evaluation method in an embodiment of the present invention; Figure 4 It is a schematic diagram of a rotational speed curve of a transmission hydraulic state evaluation method in an embodiment of the present invention; Figure 5 It is a schematic diagram of a hydraulic state curve of a transmission hydraulic state evaluation method in an embodiment of the present invention; Figure 6 It is a schematic diagram of a rotational speed curve under the working condition of Embodiment 1 in the transmission hydraulic state evaluation method of the present invention; Figure 7 It is a schematic diagram of the rotational speed curve under the working condition of Embodiment 2 in the transmission hydraulic state evaluation method of the present invention; Figure 8 It is a schematic diagram of the hydraulic state comparison curve under the working condition of Embodiment 1 in the transmission hydraulic state evaluation method of the present invention; Figure 9 It is a schematic diagram of the hydraulic state comparison curve under the working condition of Embodiment 2 in the transmission hydraulic state evaluation method of the present invention. Detailed implementation manners

[0011] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0012] In one embodiment, as Figure 1 shown, a transmission hydraulic state evaluation method is provided, including the following step S10: S10. Obtain the real-time rotational speed data of the vehicle, and perform prediction processing on the real-time rotational speed data through a hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle; wherein, the hydraulic state prediction model is trained based on rotational speed random data and hydraulic state random data for an initial differential neural network model; the hydraulic state random data is determined according to a preset hydraulic simulation model and the rotational speed random data.

[0013] It can be understood that the hydraulic system of a hybrid vehicle transmission consists of a clutch, a piston, a return spring, a control oil pump, a hydraulic oil filter, an oil storage chamber, etc., and realizes the separation and engagement of the clutch by controlling the forward and reverse rotation of the oil pump. During the operation of the transmission, the rotational speed of the control oil pump fluctuates with time, and the hydraulic state inside the hydraulic system also fluctuates with time. The real-time rotational speed data of the vehicle refers to the real-time rotational speed of the motor of the control oil pump. Input the real-time rotational speed data into the hydraulic state prediction model for prediction processing, and the hydraulic state prediction model can output the hydraulic state evaluation result corresponding to the real-time rotational speed data. The hydraulic state prediction model is a neural network model that is pre-trained to predict and evaluate the current operating state of the hydraulic system according to the input real-time rotational speed data. The hydraulic state evaluation result refers to the parameter result used to characterize the current operating state of the hydraulic system. The rotational speed of the control oil pump can be directly measured, and the evaluation parameters of the hydraulic state can be predicted by the neural network model based on the rotational speed.

[0014] The evaluation indexes of the hydraulic state include parameters such as the inlet pressure of the oil pump, the outlet pressure of the oil pump, the axial displacement of the clutch driving plate, and the axial displacement speed of the clutch driving plate. Controlling the rotational speed change of the oil pump will cause a significant change in the internal flow rate of the hydraulic system, which will indirectly affect the inlet pressure and outlet pressure of the oil pump. A clutch is a device that allows the driving part and the driven part to make axial displacement on the same axis to achieve separation and engagement. If the axial displacement is too large, the clutch may not work properly. If the axial displacement speed is not properly controlled, it may cause the engagement process to be uneven, resulting in impacts and vibrations, which will not only affect the driving comfort of the vehicle but may also damage the components of the clutch.

[0015] The performance of the hydraulic state prediction model is affected by various factors, including the quality of the training sample data, the selection of the neural network model, and the parameter configuration, etc. Therefore, when training the hydraulic state prediction model, these factors need to be comprehensively considered to ensure the accuracy and reliability of the model. In this embodiment, the rotational speed random data and the hydraulic state random data are used as the training sample data, and the initial differential neural network model is selected to train the model parameters. The hydraulic state prediction model is obtained by training the initial differential neural network model according to the rotational speed random data and the hydraulic state random data. Further, the initial differential neural network model is a neural network model pre-constructed based on the Ordinary Differential Equation Net (ODE).

[0016] The rotational speed random data is the rotational speed change amount data corresponding to different time points generated by the random number method, which can reflect the dynamic change of the rotational speed of the controlled oil pump fluctuating with time. The hydraulic state random data is the evaluation index data of the hydraulic state corresponding to the rotational speed random data generated by the simulation method. The same time point corresponds to a rotational speed random data and also corresponds to a hydraulic state random data. Among them, the hydraulic state random data is determined according to the preset hydraulic simulation model and the rotational speed random data. The preset hydraulic simulation model is a physical simulation model established in advance for analyzing the influence relationship between the rotational speed data and the hydraulic state indexes in the hydraulic system. Physical simulation is a calculation method that simulates the behavior of an actual physical system through mathematical models and algorithms. For example, when modeling the hydraulic system based on AMESim, AMESim will automatically solve the dynamic behavior of the hydraulic system and output the performance indexes of the hydraulic state, such as pressure, displacement, speed, etc. After inputting the rotational speed random data into the preset hydraulic simulation model, the preset hydraulic simulation model can output the hydraulic state random data.

[0017] In this embodiment, the real-time rotational speed data of the vehicle is obtained, and the real-time rotational speed data is predicted and processed through a hydraulic state prediction model to obtain a hydraulic state evaluation result of the vehicle. Among them, the hydraulic state prediction model is obtained by training an initial differential neural network model based on rotational speed random data and hydraulic state random data. The hydraulic state random data is determined according to a preset hydraulic simulation model and rotational speed random data. The hydraulic state prediction model of this embodiment uses rotational speed random data and hydraulic state random data as training sample data, which can cover rotational speed data and hydraulic state data under different working conditions, increasing the sample richness and helping to improve the generalization ability of the neural network model. At the same time, the hydraulic state prediction model of this embodiment is trained based on the initial differential neural network model. In the initial differential neural network model, ordinary differential equations are used to parameterize the continuous dynamics of the network layer, which helps to more accurately simulate the dynamic physical continuous change process of the hydraulic system, enabling the model to better adapt to different inputs and data distributions, further improving the generalization ability of the model, and thus accurately predicting the states of various parameters in the hydraulic system. In addition, combining the neural network with ordinary differential equations in the initial differential neural network model can reduce the consumption of computing resources and time while ensuring the model performance, thereby improving the computing efficiency of the hydraulic state prediction model and achieving a rapid evaluation of the transmission hydraulic state.

[0018] In one embodiment, before step S10, that is, before predicting and processing the real-time rotational speed data through the hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle, it includes: S101. Train the initial differential neural network model through the rotational speed random data and the hydraulic state random data to obtain the loss function value of the initial differential neural network model and the loss gradient corresponding to the loss function value; S102. Update the neural network parameters of the initial differential neural network model according to the loss function value and the loss gradient to obtain the neural network update parameters corresponding to minimizing the loss function value; S103. Generate a hydraulic state prediction model according to the neural network update parameters.

[0019] It can be understood that before training the initial differential neural network model, it is necessary to first establish the initial differential neural network model. By analyzing the state variables (hydraulic state evaluation indicators) and input variables (rotational speed) of the hydraulic system state space, a black-box state space model is established. In order to accurately describe the working states of the clutch and the hydraulic system, the state space expression should include 4 state variables, namely the inlet pressure of the oil pump, the outlet pressure of the oil pump, the axial displacement of the clutch driving plate, and the axial displacement speed of the clutch driving plate, and a system of differential equations is established. The formula of the system of differential equations is as follows: wherein, represents time, with the unit of s; is the rotational speed of the oil pump, with the unit of r / min; represents the inlet pressure of the oil pump, represents the outlet pressure of the oil pump, and the units are both bar; represents the axial displacement of the clutch driving disk, with the unit of m; represents the axial displacement speed of the clutch driving disk, with the unit of m / s; ~ is a function used to describe the relationship between the state variables of the hydraulic system and the corresponding derivatives, corresponds to the inlet pressure of the oil pump, corresponds to the outlet pressure of the oil pump, corresponds to the axial displacement of the clutch driving disk, corresponds to the axial displacement speed of the clutch driving disk.

[0020] Using the matrix function to replace ~ , the differential equation system can be written in matrix form: . Among them, , represents the state variables at time , represents the rotational speed at time , which is the input of the model. Therefore, the expression of the initial differential neural network model is , where Figure 2 represents the neural network parameters, that is, the weights and biases of the network layers. In a specific embodiment, the neural network structure of the initial differential neural network model is as shown in

[0021] After establishing the initial differential neural network model, the neural network parameters in are continuously updated using the gradient descent method. When the loss function is minimized, is made to approximate the state space derivative function . During the model training process of updating the neural network parameters once, represents the neural network parameters before the update, represents the neural network parameters after the update. Random rotational speed data is input into the initial differential neural network model Perform prediction processing to obtain a hydraulic state prediction result corresponding to the rotational speed random data. At this time, the loss function value of the initial differential neural network model represents the error rate between the hydraulic state prediction result and the hydraulic state random data. To minimize the loss function, it is also necessary to calculate the loss function value with respect to the neural network parameters of the loss gradient , and then update the neural network parameters along the opposite direction of the gradient to obtain the updated neural network parameters . Based on the initial differential neural network model perform iterative updates of the neural network parameters. When the loss gradient after iteration approaches zero or is less than or equal to the preset gradient threshold, stop the iteration to obtain the neural network update parameters corresponding to the minimized loss function value. The neural network update parameters refer to the neural network parameters when the final iteration stops. Substitute the neural network update parameters into the initial differential neural network model to obtain the hydraulic state prediction model.

[0022] In this embodiment, the initial differential neural network model is trained using the rotational speed random data and the hydraulic state random data, and the neural network parameters of the initial differential neural network model are iteratively updated according to the loss function value and the loss gradient, so as to construct an accurate and reliable hydraulic state prediction model, making the hydraulic state prediction model applicable to the dynamic changes of the hydraulic system.

[0023] In one embodiment, in step S101, that is, before training the initial differential neural network model with the rotational speed random data and the hydraulic state random data, it includes: S1011. Obtain the rotational speed variable value, the time variable value, and the random number value, and input the rotational speed variable value, the time variable value, and the random number value into a preset random number generator to obtain the rotational speed random data output by the preset random number generator.

[0024] Understandably, the preset random number generator is an algorithm model that is pre-set to generate random numbers. The user can input the minimum random number value, the maximum random number value, and the random number value of the search variable as needed, and the pseudo-random number generator will generate random numbers between the minimum random number value and the maximum random number value, and the number of random numbers is consistent with the random number value.

[0025] In one embodiment, the preset random number generator may be a pseudo random number generator (PRNG). The pseudo random number generator is used to calculate random numbers through a series of seed values when the system needs random numbers. For example, when the seed values include two search variables, a certain arithmetic combination (summation, ratio, or product) of the two search variables is directly used to generate random numbers. In this embodiment, since the rotational speed of the oil pump fluctuates with time, it is necessary to generate rotational speed random data based on two search variables that form the seed values. The two search variables are the rotational speed variable and the time variable. The rotational speed variable value is used to represent the maximum and minimum values of the rotational speed. There is a rotational speed stable time between two adjacent rotational speed fluctuations. The time variable value represents the maximum and minimum values of the rotational speed stable time, and the random number value is used to represent the total number of random numbers.

[0026] In a specific embodiment, to obtain comprehensively covered rotational speed data, the rotational speed change amount and the rotational speed stable time are used as search variables, and the rotational speed variable value, the time variable value, and the random number value are set. Among them, as Figure 3 shown in the schematic diagram of the rotational speed variable random number distribution, the rotational speed variable value includes the maximum rotational speed change amount of 4000 r / min, the minimum rotational speed change amount of 4000 r / min, and the rotational speed variable granularity of 50 r / min. The time variable value includes the maximum rotational speed stable time of 3 s, the minimum rotational speed stable time of 0.5 s, and the rotational speed stable time granularity of 0.5 s. The random number value is 500. The rotational speed variable value, the time variable value, and the random number value are input into the preset random number generator to obtain 500 rotational speed variable random numbers output by the preset random number generator. Further, the maximum rotational speed is limited to 6000 r / min, and the minimum rotational speed is 0 r / min. Since the initial rotational speed is 0 r / min (the rotational speed at the 0th second is 0 r / min), the rotational speed random data for the 1st to 500 s can be obtained according to the 500 rotational speed variable random numbers in the way of corresponding one rotational speed value per second.

[0027] Based on the preset random number generator, this embodiment generates rotational speed random data according to the rotational speed variable value, the time variable value, and the random number value, and can generate sufficient rotational speed random data as needed to improve the coverage of the rotational speed data.

[0028] In one embodiment, in step S101, that is, before training the initial differential neural network model with the rotational speed random data and the hydraulic state random data, it further includes: S1012. Generating a rotational speed curve according to the rotational speed random data; S1013. Performing simulation processing on the rotational speed curve through a preset hydraulic simulation model to obtain a hydraulic state curve; S1014. Obtain hydraulic state random data according to the hydraulic state curve.

[0029] Understandably, in one embodiment, after completing the physical modeling of the hydraulic system, when performing simulation analysis based on a preset hydraulic simulation model, it is necessary to set the time step of the simulation and the duration of the simulation. The preset hydraulic simulation model can monitor the dynamic response and performance of the hydraulic system through curve graphs, animation graphs, etc.

[0030] In this embodiment, curve fitting is performed on the rotational speed random data from 1 to 500 s, and a rotational speed curve schematic diagram as shown in Figure 4 can be obtained, that is, the dynamic curve between the rotational speed of the oil pump and time. When the time step of the simulation is set to 0.1 s and the duration of the simulation is 500 s, the rotational speed curve is simulated through the preset hydraulic simulation model, and a hydraulic state curve as shown in Figure 5 can be output, including the dynamic curve between the inlet pressure of the oil pump and time, the dynamic curve between the outlet pressure of the oil pump and time, the axial displacement of the clutch driving plate and time, and the axial displacement speed of the clutch driving plate and time. Analyze the hydraulic state curve, and 500 inlet pressure parameters of the oil pump, outlet pressure parameters of the oil pump, axial displacement parameters of the clutch driving plate, and axial displacement speed of the clutch driving plate can be extracted respectively in the way that each second corresponds to a hydraulic state index parameter, that is, the hydraulic state random data corresponding to the rotational speed random data can be obtained. At this time, each second corresponds to a random rotational speed value, and at the same time corresponds to an inlet pressure parameter value of the oil pump, an outlet pressure parameter value of the oil pump, an axial displacement parameter value of the clutch driving plate, and an axial displacement speed value of the clutch driving plate, and a set of training sample data is generated by combination. A total of 500 sets of training sample data can be generated. Divide the training data set and the test data set according to the preset test ratio. For example, when the preset test ratio is 8:2, 400 sets of training sample data are extracted as the training data set, and the remaining 100 sets of training sample data are used as the test data set.

[0031] This embodiment uses the preset hydraulic simulation model to simulate the rotational speed curve, and can obtain the hydraulic state random data corresponding to the rotational speed random data, ensuring the richness of the input samples during subsequent model training and helping to improve the generalization ability of the model.

[0032] In one embodiment, in step S1013, that is, before obtaining the hydraulic state curve by simulating the rotational speed curve through the preset hydraulic simulation model, it includes: S10131. Collect test rotational speed data and test hydraulic state data corresponding to the test rotational speed data under preset test conditions; S10132. Perform simulation processing on the test rotational speed data through an initial hydraulic simulation model to obtain simulation hydraulic state data corresponding to the test rotational speed data; S10133. Compare the test hydraulic state data with the simulation hydraulic state data to obtain a simulation test comparison result; S10134. Adjust the simulation model parameters of the initial hydraulic simulation model according to the simulation test comparison result, and obtain a preset hydraulic simulation model according to the adjusted simulation model parameters.

[0033] Understandably, the initial hydraulic simulation model is a physical simulation model of a hydraulic system established through a preset modeling tool (such as AMEsim). According to the hydraulic system principle of the hydraulic control part of the hybrid transmission for shifting gears, a physical simulation model of the hydraulic system is established to realize the oil filling and oil discharging processes of the clutch. For example, in AMEsim, a hydraulic system is established through a graphical interface, component models such as a clutch, a piston, a return spring, a control oil pump, a hydraulic oil filter, and an oil storage chamber are selected, and these components are connected together through connection lines to form a complete hydraulic system structure, and then the initial hydraulic simulation model can be obtained. The initial hydraulic simulation model includes simulation model parameters. After the physical modeling of the system is completed and before the simulation analysis, it is necessary to adjust the parameters of the initial hydraulic simulation model, and a preset hydraulic simulation model is obtained according to the adjusted simulation model parameters.

[0034] The preset test conditions are transmission working condition conditions preset for testing the changes in rotational speed and hydraulic state data in the laboratory, such as idle conditions, different speed conditions for shifting gears, etc. The test rotational speed data is the rotational speed value collected by a rotational speed sensor. The test hydraulic state data is the hydraulic state index value collected by an oil pressure sensor under the same preset test conditions as the test rotational speed data. The simulation hydraulic state data refers to the simulation value of the hydraulic state index obtained by performing simulation processing on the test rotational speed data through the initial hydraulic simulation model. The test hydraulic state data corresponding to the same test rotational speed data is compared with the simulation hydraulic state data to obtain a simulation test comparison result. The simulation test comparison result is data used to characterize the deviation between the hydraulic state index value and the simulation value of the hydraulic state index, such as the deviation rate. The simulation model parameters of the initial hydraulic simulation model are adjusted according to the simulation test comparison result, so that the deviation in the simulation test comparison result obtained from the initial hydraulic simulation model with the adjusted simulation model parameters is less than a preset deviation threshold (for example, the preset deviation threshold is that the deviation rate is less than 5%). At this time, the initial hydraulic simulation model with the adjusted simulation model parameters is determined as the preset hydraulic simulation model, that is, a preset hydraulic simulation model is obtained according to the adjusted simulation model parameters.

[0035] After obtaining the initial hydraulic simulation model through physical modeling in this embodiment, the test rotation speed data and test hydraulic state data during the test process are collected. The simulation hydraulic state data is calculated using the test rotation speed data as the simulation input, and the simulation model parameters are adjusted through the comparison between the test hydraulic state data and the simulation hydraulic state data, realizing the calibration and calibration of the initial hydraulic simulation model. The preset hydraulic simulation model obtained after calibration can accurately calculate the hydraulic state indicators of the dynamic process of the hydraulic system, ensuring the usability and accuracy of the preset hydraulic simulation model.

[0036] In one embodiment, the initial differential neural network model includes an ordinary differential equation network and a preset differential equation solver; in step S101, that is, training the initial differential neural network model with the rotation speed random data and the hydraulic state random data to obtain the loss function value of the initial differential neural network model and the loss gradient corresponding to the loss function value, includes: S1015. Perform equation solving processing on the ordinary differential equation network through the preset differential equation solver to obtain a differential solving result; S1016. Calculate the backpropagation error for the differential solving result to obtain the loss function value of the initial differential neural network model; S1017. Perform integration processing on the loss function value and the neural network parameters to determine the loss gradient corresponding to the loss function value.

[0037] Understandably, the initial differential neural network model includes an ordinary differential equation network and a preset differential equation solver. In the ordinary differential equation network, the residual connection of the neural network is replaced by an ordinary differential equation, and the preset differential equation solver is a tool preset for solving the ordinary differential equation in the ordinary differential equation network. The differential solving result refers to the solution of the ordinary differential equation in the ordinary differential equation network.

[0038] In one embodiment, the initial differential neural network model includes an ordinary differential equation network and a preset differential equation solver. After performing equation solving processing on the ordinary differential equation in the ordinary differential equation network through the preset differential equation solver, a differential solving result is obtained. The differential solving result is , where represents the differential start time, represents the differential end time, represents differentiating with respect to time. Calculate the backpropagation error for the differential solving result to obtain the loss function value of the initial differential neural network model. The loss function is specifically expressed as . Using the adjoint sensitivity method for the loss function value and the neural network parameters Perform integration processing to determine the loss function corresponding loss gradient , during one iteration, the loss gradient is specifically expressed as , where the adjoint state represents the loss function depending on the rate of change of the hydraulic state evaluation index .

[0039] In this embodiment, the loss function is determined according to the differential solution result of the ordinary differential equation network, and the adjoint sensitivity method is used to determine the loss gradient, which helps to continuously update the neural network parameters using the gradient descent method in the subsequent stage, so that approximate the state space derivative function , and further realize the accurate simulation of the dynamic physical process of the hydraulic system, and further improve the generalization ability of the model In one embodiment, in step S10, that is, before the real-time rotational speed data is predicted by the hydraulic state prediction model, it includes: S104. Compile the hydraulic state prediction model to obtain a prediction model code; S105. Write the prediction model code into the controller of the vehicle through a preset flashing tool.

[0040] Understandably, the prediction model code is software data that converts the hydraulic state prediction model into a readable code form for the controller. The preset flashing tool is a tool for writing the readable code form of the hydraulic state prediction model into the controller software. The controller of the vehicle can be a vehicle transmission controller, a vehicle integrated controller, or a combination of a vehicle transmission controller and a vehicle integrated controller. In order to apply the hydraulic state prediction model to an actual vehicle, the hydraulic state prediction model is compiled into a code language corresponding to the controller (such as C language code), and the compiled prediction model code is deployed to the controller of the vehicle using the preset flashing tool, so as to accurately predict the parameter results of the hydraulic state evaluation index in the hydraulic system during the operation of the vehicle transmission hydraulic system.

[0041] In this embodiment, the hydraulic state prediction model is compiled into a code language corresponding to the controller and deployed to the controller of the vehicle, ensuring the effective utilization of the hydraulic state prediction model by the controller of the vehicle.

[0042] In order to verify the accuracy and generalization ability of the hydraulic state prediction model, the real-time rotational speed data actually measured under two different working conditions are selected as the model input data for Embodiment 1 and Embodiment 2 respectively. Among them, the schematic diagram of the rotational speed curve of Embodiment 1 is as Figure 6 shown, and the schematic diagram of the rotational speed curve of Embodiment 2 is as Figure 7As shown. The real-time rotational speed data of Embodiment 1 is input into a preset hydraulic simulation model to obtain the simulation value of the hydraulic state evaluation result in Embodiment 1. The real-time rotational speed data of Embodiment 1 is input into the hydraulic state prediction model to obtain the predicted value of the hydraulic state evaluation result in Embodiment 1, and curve fitting is performed to generate as Figure 8 shown in the schematic diagram of the hydraulic state comparison curve. The real-time rotational speed data of Embodiment 2 is input into the preset hydraulic simulation model to obtain the simulation value of the hydraulic state evaluation result in Embodiment 2. The real-time rotational speed data of Embodiment 2 is input into the hydraulic state prediction model to obtain the predicted value of the hydraulic state evaluation result in Embodiment 2, and curve fitting is performed to generate as Figure 9 shown in the schematic diagram of the hydraulic state comparison curve. From Figure 8 and Figure 9 , it can be seen that for the same real-time rotational speed data, the change trends between the predicted values obtained by the hydraulic state prediction model and the simulation values in the preset hydraulic simulation model are almost completely coincident, indicating that the hydraulic state prediction model has extremely high accuracy and generalization ability.

[0043] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0044] In one embodiment, a transmission hydraulic state evaluation device is provided, and the transmission hydraulic state evaluation device corresponds one-to-one to the transmission hydraulic state evaluation method in the above embodiment. The transmission hydraulic state evaluation device includes a hydraulic state prediction module. The detailed description of each functional module is as follows: The hydraulic state prediction module is used to obtain the real-time rotational speed data of the vehicle, and perform prediction processing on the real-time rotational speed data through the hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle; wherein, the hydraulic state prediction model is trained based on rotational speed random data and hydraulic state random data for the initial differential neural network model; the hydraulic state random data is determined according to the preset hydraulic simulation model and the rotational speed random data.

[0045] In one embodiment, the hydraulic state prediction module includes: The loss calculation unit is used to perform model training on the initial differential neural network model through the rotational speed random data and the hydraulic state random data to obtain the loss function value of the initial differential neural network model and the loss gradient corresponding to the loss function value; The parameter update unit is used to update the neural network parameters of the initial differential neural network model according to the loss function value and the loss gradient to obtain the neural network update parameters corresponding to minimizing the loss function value; A hydraulic state prediction model generation unit, configured to generate a hydraulic state prediction model according to the updated parameters of the neural network.

[0046] In one embodiment, the hydraulic state prediction module further includes: A rotational speed random data generation unit, configured to obtain a rotational speed variable value, a time variable value, and a random number value, input the rotational speed variable value, the time variable value, and the random number value into a preset random number generator, and obtain rotational speed random data output by the preset random number generator.

[0047] In one embodiment, the hydraulic state prediction module further includes: A rotational speed curve generation unit, configured to generate a rotational speed curve according to the rotational speed random data; A simulation processing unit, configured to perform simulation processing on the rotational speed curve through a preset hydraulic simulation model to obtain a hydraulic state curve; A hydraulic state random data determination unit, configured to obtain hydraulic state random data according to the hydraulic state curve.

[0048] In one embodiment, the hydraulic state prediction module further includes: A test data acquisition unit, configured to acquire test rotational speed data and corresponding test hydraulic state data under preset test conditions; A test data simulation processing unit, configured to perform simulation processing on the test rotational speed data through an initial hydraulic simulation model to obtain simulated hydraulic state data corresponding to the test rotational speed data; A simulation test comparison unit, configured to compare the test hydraulic state data with the simulated hydraulic state data to obtain a simulation test comparison result; A preset hydraulic simulation model determination unit, configured to adjust the simulation model parameters of the initial hydraulic simulation model according to the simulation test comparison result, and obtain a preset hydraulic simulation model according to the adjusted simulation model parameters.

[0049] In one embodiment, the hydraulic state prediction module further includes: An equation solving processing unit, configured to perform equation solving processing on the ordinary differential equation network through a preset differential equation solver to obtain a differential solving result; A loss function value determination unit, configured to perform backpropagation error calculation on the differential solving result to obtain a loss function value of the initial differential neural network model; A loss gradient determination unit, configured to perform integral processing on the loss function value and the neural network parameters to determine a loss gradient corresponding to the loss function value.

[0050] In one embodiment, the hydraulic state prediction module further includes: A prediction model code generation unit, configured to compile the hydraulic state prediction model to obtain a prediction model code; A prediction model code writing unit, configured to write the prediction model code into the controller of the vehicle through a preset flashing tool.

[0051] For the specific limitations of the transmission hydraulic state evaluation device, reference can be made to the limitations of the transmission hydraulic state evaluation method in the foregoing text, which will not be elaborated herein. Each module in the above transmission hydraulic state evaluation device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0052] In one embodiment, a controller is provided. For the specific limitations of the controller, reference can be made to the limitations of the transmission hydraulic state evaluation method in the foregoing text, which will not be elaborated herein. Each module in the above controller can be implemented in whole or in part by software, hardware, and their combination. The controller includes a processor and a memory connected through a system bus. Among them, the processor of the controller is used to provide computing and control capabilities. The memory of the controller includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The database of the controller is used to store the data involved in the transmission hydraulic state evaluation method. The network interface of the controller is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a transmission hydraulic state evaluation method is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0053] In one embodiment, a controller is provided. The controller includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the transmission hydraulic state evaluation method is implemented.

[0054] A vehicle includes the above-mentioned controller. The controller can be a controller of a vehicle transmission, a vehicle controller, or an integration or combination of a controller of a vehicle transmission and a vehicle controller.

[0055] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. Computer-readable instructions are stored on the readable storage media. When the computer-readable instructions are executed by one or more processors, the following steps are implemented: Obtain the real-time rotational speed data of the vehicle, and perform prediction processing on the real-time rotational speed data through a hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle; wherein, the hydraulic state prediction model is obtained by training an initial differential neural network model based on rotational speed random data and hydraulic state random data; the hydraulic state random data is determined according to a preset hydraulic simulation model and rotational speed random data.

[0056] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through computer-readable instructions. The computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0057] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A transmission hydraulic state evaluation method, characterized in that: include: The real-time speed data of the vehicle is acquired, and the real-time speed data is predicted and processed by a hydraulic state prediction model to obtain a hydraulic state evaluation result of the vehicle; wherein the hydraulic state prediction model is obtained by training an initial differential neural network model according to the speed random data and the hydraulic state random data; the hydraulic state random data is determined according to a preset hydraulic simulation model and the speed random data.

2. The transmission hydraulic state evaluation method according to claim 1, characterized in that: Before the real-time speed data is predicted and processed by the hydraulic state prediction model to obtain the hydraulic state evaluation result of the vehicle, the method includes: Performing model training on the initial differential neural network model through the rotational speed random data and the hydraulic state random data to obtain a loss function value of the initial differential neural network model and a loss gradient corresponding to the loss function value; The neural network parameters of the initial differential neural network model are updated according to the loss function value and the loss gradient to obtain the neural network update parameters corresponding to the minimized loss function value; A hydraulic state prediction model is generated according to the neural network update parameters.

3. The transmission hydraulic state evaluation method according to claim 2, characterized in that: Before the initial differential neural network model is trained by the rotation speed random data and the hydraulic state random data, the method includes: The speed variable value, the time variable value and the random number value are obtained, and the speed variable value, the time variable value and the random number value are input into a preset random number generator to obtain the speed random data output by the preset random number generator.

4. The transmission hydraulic state evaluation method according to claim 2, characterized in that: Before the initial differential neural network model is trained by the rotation speed random data and the hydraulic state random data, the method includes: generating a speed curve according to the speed random data; The speed curve is simulated by a preset hydraulic simulation model to obtain a hydraulic state curve; Hydraulic state random data is obtained according to the hydraulic state curve.

5. The transmission hydraulic state evaluation method according to claim 4, characterized in that: Before simulating the speed curve by using a preset hydraulic simulation model to obtain a hydraulic state curve, the method includes: Collecting test speed data and test hydraulic state data corresponding to the test speed data under preset test conditions; Performing simulation processing on the test speed data through an initial hydraulic simulation model to obtain simulated hydraulic state data corresponding to the test speed data; Comparing the test hydraulic state data with the simulated hydraulic state data to obtain a simulation test comparison result; The simulation model parameters of the initial hydraulic simulation model are adjusted according to the simulation test comparison results, and the preset hydraulic simulation model is obtained according to the adjusted simulation model parameters.

6. The transmission hydraulic state evaluation method according to claim 2, characterized in that: The initial differential neural network model includes an ordinary differential equation network and a preset differential equation solver; The initial differential neural network model is trained by using the rotation speed random data and the hydraulic state random data to obtain a loss function value of the initial differential neural network model and a loss gradient corresponding to the loss function value, including: Performing equation solving processing on the ordinary differential equation network by using a preset differential equation solver to obtain a differential solution result; Performing back propagation error calculation on the differential solution result to obtain the loss function value of the initial differential neural network model; The loss function value and the neural network parameters are integrated to determine the loss gradient corresponding to the loss function value.

7. The transmission hydraulic state evaluation method according to claim 1, characterized in that: Before the real-time speed data is predicted by the hydraulic state prediction model, the method includes: Compiling the hydraulic state prediction model to obtain a prediction model code; The prediction model code is written into the controller of the vehicle through a preset flashing tool.

8. A controller comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that: When the processor executes the computer-readable instructions, the transmission hydraulic state evaluation method according to any one of claims 1 to 7 is implemented.

9. A vehicle, characterized in that: Comprising a controller as claimed in claim 8.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to perform the transmission hydraulic state evaluation method according to any one of claims 1 to 7.

Images