Method for constructing dynamical model of electromechanical composite transmission system driven by data and mechanism

Abstract

The invention discloses a method for constructing a dynamical model of an electromechanical composite transmission system driven by data and mechanisms. The method comprises the steps that transmission input data, transmission output data and multi-source working condition data of the electromechanical composite transmission system are collected; inputting the transmission input data into a mechanism dynamics model to obtain predicted output data; constructing a training data set by adopting the transmission input data, the transmission output data, the prediction output data and the multi-source working condition data, and training a neural network model by adopting the training data set to obtain an error correction network model; the mechanism dynamic model and the error correction network model are fused to obtain a dynamic model of the electromechanical composite transmission system, and the error correction network model carries out error correction on the prediction output data of the mechanism dynamic model. Through the combination of the model and the data, the prediction precision and generalization ability of the model under complex working conditions are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of dynamic model simulation, and particularly relates to a data and mechanism driven electromechanical composite transmission system dynamic model construction method and device, a readable storage medium, a terminal equipment and a program product. BACKGROUND

[0002] As a transmission device of tracked vehicles, the electromechanical composite transmission system is a new generation of technology after the hydro-mechanical comprehensive transmission device. Compared with the traditional mechanical transmission system, the electromechanical composite transmission system has significant advantages, especially in complex use environments, which can significantly improve the mobility, low noise and work efficiency of the vehicle. Therefore, its application prospect is extremely broad, and it will gradually become one of the key technologies of future vehicle equipment.

[0003] In order to fully exert the advantages of the electromechanical composite transmission system, especially the potential in improving the mobility, low noise and work efficiency of the vehicle, it is crucial to establish an accurate and efficient dynamic model. Existing electromechanical composite transmission system modeling has achieved some results: on the one hand, commercial simulation software has strong simulation analysis capability and is widely used in transmission mechanism dynamic modeling, which can preliminarily predict the dynamic behavior of the system; on the other hand, traditional physical modeling methods, such as Newton-Euler method and Lagrange method, are also used for transmission system modeling. These methods analyze the physical properties of each component in the transmission system and establish a dynamic differential equation model of the system, providing a basis for studying the dynamic characteristics of the vehicle under complex working conditions.

[0004] However, the existing transmission system modeling method still has limitations: the adaptability and precision under complex working conditions are often insufficient, and it is difficult to cope with the changing environmental factors and dynamic changes in actual application. SUMMARY

[0005] Therefore, the embodiments of the present application provide a data and mechanism driven electromechanical composite transmission system dynamic model construction method and device, aiming to solve the problem of insufficient modeling precision and adaptability of the electromechanical composite transmission mechanism in the prior art, and improve the prediction precision and generalization ability of the model under complex working conditions.

[0006] To achieve the above-mentioned purpose, according to the first aspect of the embodiments of the present application, a data and mechanism driven electromechanical composite transmission system dynamic model construction method is provided, comprising: Collecting transmission input data, transmission output data and multi-source working condition data of the electromechanical composite transmission system; Inputting the transmission input data into a mechanism dynamic model to obtain predicted output data, wherein the mechanism dynamic model is constructed according to the transmission structure of the electromechanical composite transmission system; The training data set is constructed by using the transmission input data, the transmission output data, the predicted output data and the multi-source working condition data, and a neural network model is trained by using the training data set to obtain an error correction network model; The error correction network model is fused with the mechanism dynamics model to obtain a dynamics model of the electromechanical composite transmission system, wherein the error correction network model corrects errors of the predicted output data of the mechanism dynamics model.

[0007] Further, the neural network model is a multi-layer feedforward neural network, including an input layer, a hidden layer and an output layer; the number of nodes of the input layer is equal to the sum of the number of categories of the transmission input data, the transmission output data and the multi-source working condition data; the output of the output layer is the error of the predicted output data, and the number of nodes of the output layer is equal to the number of categories of the predicted output data.

[0008] Further, the neural network model includes two hidden layers; the activation functions of the input layer and the hidden layer are Sigmoid functions, and the activation function of the output layer is a linear activation function.

[0009] Further, the electromechanical composite transmission system includes a driving motor and a transmission mechanism coupled with the driving motor; the transmission input data includes the speed and torque of the driving motor; the transmission output data includes the output speed, output torque and output angular acceleration of the transmission mechanism actually collected by a sensor; the multi-source working condition data includes the oil temperature and oil pressure of lubricating oil in the electromechanical composite transmission system; and the predicted output data includes the output speed, output torque and output angular acceleration of the transmission mechanism obtained by using the mechanism dynamics model.

[0010] Further, the error correction network model is obtained by training a neural network model by using the training data set, including: initializing the weights and thresholds of each node in the neural network model to obtain the initial values of the weights and thresholds of each node; optimizing the initial values of the weights and the initial values of the thresholds by using a genetic algorithm to obtain the optimal initial values of the weights and the optimal initial values of the thresholds corresponding to each node; based on the optimal initial values of the weights, the optimal initial values of the thresholds and the training data set, the output error of the neural network model is calculated by using a forward propagation algorithm, and the error is reversely transmitted by using a back propagation algorithm to adjust the weights and the thresholds of the nodes; in response to reaching a training termination condition, the last updated weights and thresholds are obtained, and the neural network model at the end of training is correspondingly taken as an error correction network model.

[0011] Further, the genetic algorithm is used to optimize the weight initial value and the threshold initial value, to obtain the optimal weight initial value and the optimal threshold initial value corresponding to each node, comprising: The weight and the threshold of all the nodes are encoded into chromosomes, and the chromosomes are initialized to obtain an initialized population; The fitness of each chromosome is calculated according to a preset fitness function, and genetic operations are performed according to the fitness of the chromosomes, and the population is iterated; the genetic operations include selection, crossover and mutation; In response to satisfying an iteration termination condition, the optimal chromosome of the last obtained population is decoded and output to obtain the optimal weight initial value and the optimal threshold initial value corresponding to each node.

[0012] According to a second aspect of the embodiment of the present application, a data and mechanism driven electromechanical composite transmission system dynamics model construction device is provided, comprising: A collection module is configured to collect transmission input data, transmission output data and multi-source working condition data of the electromechanical composite transmission system; A prediction module is configured to input the transmission input data into a mechanism dynamics model to obtain predicted output data, and the mechanism dynamics model is constructed according to the transmission structure of the electromechanical composite transmission system; A training module is configured to construct a training data set using the transmission input data, the transmission output data, the predicted output data and the multi-source working condition data, and train a neural network model using the training data set to obtain an error correction network model; A fusion module is configured to fuse the mechanism dynamics model and the error correction network model to obtain a dynamics model of the electromechanical composite transmission system, wherein the error correction network model corrects errors of the predicted output data of the mechanism dynamics model.

[0013] According to a third aspect of the embodiment of the present application, a computer readable storage medium is provided, which stores a program, and the program is executed by a processor to implement the steps of the data and mechanism driven electromechanical composite transmission system dynamics model construction method according to the first aspect of the present application.

[0014] According to a fourth aspect of the embodiment of the present application, a terminal device is provided, which comprises a memory, a processor and a program stored in the memory and executable on the processor, and the processor implements the steps of the data and mechanism driven electromechanical composite transmission system dynamics model construction method according to the first aspect of the present application when executing the program.

[0015] According to a fifth aspect of the embodiments of the present application, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps in the data and mechanism driven electromechanical composite transmission system dynamics model construction method according to the first aspect of the present application.

[0016] The embodiments of the present application have at least one of the following advantages or beneficial effects: The embodiments of the present application describe the basic framework of the physical mechanism and dynamic behavior of the electromechanical composite transmission system through the mechanism dynamics model, and correct and optimize the mechanism dynamics model based on the neural network model using data, which makes up for the shortcomings of relying solely on mechanism models or data-driven models, realizes the complementary advantages of the two, and effectively improves the accuracy of the dynamics model of the electromechanical composite transmission system.

[0017] When constructing the training data set of the neural network model, the embodiments of the present application not only use transmission input data, transmission output data and predicted output data, but also use multi-source working condition data to consider the nonlinear factors and multi-source coupling effects existing in the transmission system, so that the error correction network model trained can more comprehensively capture the dynamic characteristics of the system and adapt to different working conditions.

[0018] When training the neural network model, the embodiments of the present application optimize the initial weights and thresholds of the feedforward neural network through the genetic algorithm, avoid the problem of traditional feedforward neural network falling into local optimum, make the model have stronger adaptability and generalization ability, and can provide more accurate prediction in complex time-varying environment.

[0019] The dynamics model constructed by the embodiments of the present application provides a more reliable and accurate theoretical basis for the design, control and optimization of the electromechanical composite transmission system. Through accurate model prediction, the maneuverability, low noise and work efficiency of the vehicle can be better guided to improve, which is helpful to promote the development of related technical fields.

[0020] The above technical solutions can be combined with each other in the present application to realize more preferred combination schemes. Other features and advantages of the present application will be described in the subsequent specification, and some advantages will become apparent from the specification, or will be understood by implementing the present application. The purpose and other advantages of the present application can be achieved and obtained from the contents specifically pointed out in the specification and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification, illustrate embodiments of the present application and together with the description serve to explain the principles of the present application, and should not be necessarily construed as limiting the present application. Figure 1 is the main flowchart of the data and mechanism driven electromechanical composite transmission system dynamics model construction method of the embodiments of the present application. Figure 2 is a network structure schematic diagram of a feedforward neural network in an embodiment of the present application; Figure 3 is a main flow schematic diagram of training a neural network model in an embodiment of the present application; Figure 4 is a main flow schematic diagram of neural network model initial value optimization in an embodiment of the present application; Figure 5 is a main flow schematic diagram of a data and mechanism driven electromechanical composite transmission system dynamics model construction method in another embodiment of the present application; Figure 6 is a composition schematic diagram of a bilateral motor coupling driving device in another embodiment of the present application; Figure 7 is a transmission force analysis schematic diagram of a bilateral motor coupling driving device in another embodiment of the present application; Figure 8 is a network structure schematic diagram of a feedforward neural network in another embodiment of the present application; Figure 9 is a flow schematic diagram of neural network model initial value optimization in another embodiment of the present application; Figure 10 is a main composition module schematic diagram of a data and mechanism driven electromechanical composite transmission system dynamics model construction device in an embodiment of the present application; Figure 11 is a composition schematic diagram of a terminal device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The preferred embodiments of the present application will be described in detail below with reference to the drawings, wherein the drawings constitute a part of this application and serve to explain the principles of the embodiments of the present application, but are not used to limit the scope of the present application.

[0023] Embodiment one Figure 1 is a main flow schematic diagram of a data and mechanism driven electromechanical composite transmission system dynamics model construction method in an embodiment of the present application. As shown in Figure 1 , the data and mechanism driven electromechanical composite transmission system dynamics model construction method in this embodiment of the present application includes the following steps S101 to S104.

[0024] Step S101, collecting transmission input data, transmission output data and multi-source working condition data of the electromechanical composite transmission system; Step S102, inputting the transmission input data into a mechanism dynamics model to obtain predicted output data, and the mechanism dynamics model is constructed according to the transmission structure of the electromechanical composite transmission system; Step S103, using the transmission input data, the transmission output data, the predicted output data, and the multi-source working condition data to construct a training data set, and using the training data set to train a neural network model to obtain an error correction network model; Step S104, fusing the mechanism dynamics model and the error correction network model to obtain a dynamics model of the electromechanical composite transmission system, wherein the error correction network model corrects errors of the predicted output data of the mechanism dynamics model.

[0025] It can be understood that the method in the embodiment and some embodiments of the present application is based on the combination of data and models, and a more accurate and time-varying dynamics model can be constructed, which not only improves the adaptability of the model, but also helps to improve the prediction ability of the transmission system behavior under complex working conditions, so as to provide a reference for the control and research of the vehicle equipped with the electromechanical composite transmission mechanism under complex working conditions.

[0026] It can be understood that in the embodiment of the present application, the electromechanical composite transmission system includes a driving motor and a transmission mechanism coupled with the driving motor. The transmission input data refers to the excitation data acting on the input end of the electromechanical composite transmission system, which drives the system to run. The transmission output data refers to the response data of the output end of the electromechanical composite transmission system, which reflects the dynamic behavior of the system under the input excitation. The multi-source working condition data refers to various environmental and working condition data that affect the operation of the electromechanical composite transmission system, which reflects the actual operating environment and working conditions of the system, for example, the oil temperature and oil pressure of the lubricating oil, the external load and the environmental temperature, etc.

[0027] Specifically, in the embodiment and some embodiments of the present application, the transmission input data includes the speed and torque of the driving motor, the transmission output data includes the output speed, output torque and output angular acceleration of the transmission mechanism, and the multi-source working condition data includes the oil temperature and oil pressure of the lubricating oil in the electromechanical composite transmission system. Correspondingly, the predicted output data includes the output speed, output torque and output angular acceleration of the transmission mechanism obtained by using the mechanism dynamics model.

[0028] Figure 2 is a network structure schematic diagram of the feedforward neural network in the embodiment of the present application. Specifically, as shown in Figure 2As shown in this embodiment and some embodiments of the present invention, the neural network model is a multi-layer feedforward neural network, such as a feedforward neural network, including an input layer, an output layer, and n interconnected hidden layers between them, used to effectively model and correct errors in the mechanistic dynamics model. The number of nodes in the input layer is equal to the sum of the number of types of the transmission input data, the transmission output data, and the multi-source operating condition data, to comprehensively reflect the main factors affecting the dynamic behavior of the system. The output of the output layer is the error of the predicted output data, and the number of nodes in the output layer is equal to the number of types of the predicted output data, corresponding to the actually observed dynamic response value, such as physical quantities like the actual angular velocity measured by sensors, used to compare with the model output and guide the network to learn the error compensation relationship. The number of nodes in the hidden layer is determined through cross-validation to achieve a balance between fitting accuracy and network complexity, avoiding underfitting or overfitting. The activation functions of the input layer and hidden layers are selected from commonly used Sigmoid functions according to the fitting target to enhance the network's ability to express nonlinear relationships, while the output layer uses a linear activation function, enabling the network to output continuous values, adapting to the regression prediction task of actual dynamic quantities. The neural network model of this invention has a structure that takes into account both the diversity and physical meaning of the input variables and has strong nonlinear modeling capabilities, providing an efficient model foundation for subsequent error correction.

[0029] Specifically, in this embodiment and some embodiments of the present invention, the first Layer The input of each node (i.e., neuron) can be represented as: , In the formula: Indicates the first The first in the layer Input of each node; For the first Number of neurons in a layer; for Layer Each node to Layer The connection weights between nodes; For the first Layer The output of each node; For the first Layer Threshold for each node; ; This excludes the number of network layers, which is the input layer.

[0030] The activation function for the hidden layer is the Sigmoid function: , In the network Layer nodes The output can be represented as: .

[0031] Understandably, in this embodiment and some embodiments of the present invention, after the neural network structure design is completed, a training sample set needs to be constructed for training. The selected variables should be able to comprehensively reflect the time-varying factors affecting the dynamic behavior of the transmission system. In order to improve training efficiency and ensure that the neural network can converge quickly, while avoiding the phenomenon of gradient vanishing or exploding during training, all training sample data are normalized to give them a uniform numerical range.

[0032] Figure 3 This is a schematic diagram of the main process of training a neural network model in an embodiment of the present invention. Specifically, as shown below... Figure 3 As shown, in this embodiment and some embodiments of the present invention, step S103, which involves training a neural network model using the training dataset to obtain an error correction network model, includes steps S103a to S103d: Step S103a: Initialize the weights and thresholds of each node in the neural network model to obtain the initial weight values ​​and initial threshold values ​​of each node. Step S103b: Use a genetic algorithm to optimize the initial weight value and the initial threshold value to obtain the optimal initial weight value and the optimal initial threshold value for each node. Step S103c: Based on the initial value of the optimal weight, the initial value of the optimal threshold, and the training dataset, the output error of the neural network model is calculated using the forward propagation algorithm, and the error is propagated in the back direction using the back propagation algorithm to adjust the weight and threshold of the node. Step S103d: Determine whether the training termination condition is met. If the determination result is met, the training ends and the last updated weights and thresholds are obtained. The neural network model at the end of training is used as the error correction network model. If the determination result is not met, return to step S103c.

[0033] Understandably, in this embodiment and some embodiments of the present invention, the feedforward neural network is highly sensitive to the selection of initial weights and thresholds during training. Due to the randomness of the initial parameters, the network often gets trapped in local optima, preventing the model from obtaining the global optimum, thereby affecting the training effect and final performance of the model. To solve this problem, the embodiments of the present invention optimize the initial parameters of the feedforward neural network, thereby improving the performance of the neural network. Figure 4 This is a schematic diagram of the main process of initial value optimization for the neural network model in an embodiment of the present invention. Specifically, as shown below... Figure 4As shown, in this embodiment and some embodiments of the present invention, step S103b includes steps S103b1 to S103b3.

[0034] Step S103b1: Encode the weights and thresholds of all the nodes into chromosomes, and initialize the chromosomes to obtain an initialized population. It is understood that the population includes multiple different chromosomes.

[0035] Step S103b2: Calculate the fitness of each chromosome according to a preset fitness function, and perform genetic operations on the population according to the fitness of the chromosomes; the genetic operations include selection, crossover and mutation.

[0036] Understandably, selection, crossover, and mutation are the three key steps in optimization, working synergistically to drive the population towards the optimal solution. Specifically, in this embodiment and some embodiments of the present invention, selection is the first step in genetic operations, determining which individuals can continue to reproduce. More specifically, chromosomes are selected based on the fitness of each individual; chromosomes with higher fitness have a greater probability of being selected. Fitness is typically determined by calculating the prediction error of the neural network; individuals with smaller errors have higher fitness. During the selection process, chromosomes with higher fitness have a greater probability of being selected, and may be selected multiple times. Through this process, genetic operations can pass on superior individuals (i.e., better network parameters) to the next generation. Crossover is the second step in genetic operations, simulating the gene recombination process in nature. More specifically, two chromosomes (parents) are randomly selected from the current population and exchanged at a random position to produce new chromosomes (offspring). The goal of crossover is to combine the advantages of the parents to generate better offspring individuals. The selection of the crossover point is random, meaning that each crossover operation will produce different results, thereby helping genetic operations explore more solution spaces and avoid getting trapped in local optima. Mutation is the third step in genetic operations, designed to enhance population diversity and prevent genetic operations from getting trapped in local optima. Mutation generates new solutions by occasionally altering certain genes (i.e., weights or thresholds) in a chromosome. The probability of mutation is usually low, but it plays a crucial role throughout the optimization process because it introduces new gene variations, helping GA explore undiscovered regions of the search space and effectively avoid the trap of local optima. Mutation allows genetic operations to maintain greater flexibility and exploratory power during global searches.

[0037] Step S103b3: Determine whether the iteration termination condition is met. If the condition is met, decode and output the optimal chromosome of the population to obtain the initial optimal weight and initial optimal threshold values ​​for each node. If the condition is not met, return to step S103b2 to continue iteration. It is understood that the optimal chromosome is the chromosome with the highest fitness in the population at the time of iteration termination.

[0038] Specifically, in this embodiment and some embodiments of the present invention, selection, crossover, and mutation operations are performed alternately. The population is updated in each generation, and the performance of the current population is evaluated by calculating the fitness of each chromosome. After multiple generations of evolution, the genetic operations gradually find the optimal solution, i.e., the optimal combination of neural network weights and thresholds. Finally, the optimized solution is used as the initial weights and thresholds of the feedforward neural network, thus providing a better starting point, enabling the feedforward neural network to converge faster in subsequent training processes and improving the model's accuracy.

[0039] It is understood that the method provided in this invention overcomes the following shortcomings of existing commercial simulation software in the dynamic modeling of transmission mechanisms: First, the modeling process of commercial software is highly encapsulated, making it difficult for users to obtain or modify its underlying computational logic, thus limiting in-depth research on dynamic mechanisms and innovative applications of new methods; second, its closed architecture and limited secondary development interfaces restrict functional customization and system integration. Therefore, this invention has broad practical value.

[0040] Understandably, the method provided by the embodiments of the present invention has a high tolerance for modeling errors and unknown disturbances, and can effectively compensate for them. It breaks through many limitations of existing traditional modeling methods such as the Newton-Euler method and the Lagrange method, as well as commercial software modeling, when faced with complex working conditions and high-precision requirements. For example, it is difficult to effectively describe common nonlinear dynamic behaviors in transmission systems, such as gear meshing friction and changes in lubrication state. Furthermore, many physical parameters involved in the model may change with environmental and load variations during actual operation, leading to a decrease in the accuracy of the mechanistic model. Moreover, once a traditional model is built, its structure and parameters are basically fixed, lacking the ability to adaptively adjust based on new data, and thus failing to meet the real-time prediction and control requirements of the system under varying working conditions.

[0041] It is understood that the method of the present invention can not only improve the accuracy of the dynamic model of the transmission system, but also better reflect the influence of time-varying factors on the system performance.

[0042] As will be understood by those skilled in the art, all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium.

[0043] Example 2 Figure 5 This is a schematic diagram illustrating the main flow of the data and mechanism-driven dynamic model construction method for the electromechanical composite transmission system according to this embodiment. Figure 6 This is a schematic diagram of the dual-motor coupling drive device in this embodiment. Figure 5 and Figure 6 As shown, the data and mechanism-driven dynamic model construction method for electromechanical composite transmission system of the present invention in this embodiment is applied to the dual-motor coupled drive device as shown in Figure 6, and includes the following steps S201 to S204.

[0044] Step S201: Collect the transmission input data, transmission output data, and multi-source operating condition data of the dual-side motor coupling drive transmission device; Step S202: Input the transmission input data into the dual-motor coupling drive mechanism model to obtain the predicted output data. The dual-motor coupling drive mechanism model is constructed based on the transmission structure of the dual-motor coupling drive device. Step S203: Construct a training dataset using the transmission input data, the transmission output data, the predicted output data, and the multi-source working condition data; and train a neural network model using the training dataset to obtain an error correction network model. Step S204: The dual-motor coupling drive mechanism model and the error correction network model are fused to obtain the dynamic model of the electromechanical composite transmission system, wherein the error correction network model corrects the error in the predicted output data of the dual-motor coupling drive mechanism model.

[0045] Specifically, such as Figure 6 As shown, in this embodiment and some embodiments of the present invention, the dual-side motor coupling drive device consists of two drive motors 1 / 2 and their reduction gears 3 / 4, a power coupling mechanism 5, and two gearboxes 8 / 9 on both sides. The ratio of the number of teeth of the ring gear and the sun gear of the planetary gear set of the motor reduction gear 3 / 4 is [value missing]. The power coupling mechanism 5 consists of two symmetrical planetary gear sets 6 / 7, with the ratio of the number of teeth on the planetary gear set ring to the number of teeth on the sun gear being [value missing]. The 8th and 9th gears of the transmission have two gears. The ratio of the number of teeth on the planetary gear ring and the sun gear is [value missing]. .

[0046] Specifically, in this embodiment and some embodiments of the present invention, the two outermost gearboxes 8 / 9 are connected to the sun gears by clutches 10 / 11, and the gear ring is connected to the transmission housing by brakes 12 / 13. Simultaneously engaging the brakes 12 / 13 on both sides and disengaging the clutches 10 / 11, the two gearboxes are in first gear, and the gear ratio of gearboxes 8 / 9 is... This allows the vehicle to travel in low gears; simultaneously, clutches 10 / 11 on both sides are engaged, brakes 12 / 13 are disengaged, and the gearboxes on both sides are in second gear. Gearboxes 8 / 9 have a gear ratio of 1, enabling the vehicle to travel in high gears. When the output speeds of both motors are the same, the vehicle travels in a straight line; when the output speeds of the two motors are different, the vehicle performs differential steering. The power coupling mechanism 5 is key to achieving dual-motor coupled drive. It can transfer regenerative steering power from the low-speed side track to the high-speed side track, but the amount of power transferred depends on the planetary gear parameters of the power coupling mechanism 5.

[0047] Understandably, we take the transmission system in low-gear operating mode as the research object. In this mode, the outer 8 / 9 gear ring is fixed, and the transmission ratio is greater than 1. The overall transmission system is a two-degree-of-freedom system. The input end is the sun gear of the two motor reduction gears (3 / 4 gear), and the output end is the planetary carrier of the two 8 / 9 gear sets. We use the Newton-Euler method to analyze the forces acting on a typical planetary gear set. The analysis results are as follows... Figure 7 As shown. First, the force analysis of the sun gear is performed. The direction shown in the figure is defined as the positive direction. Assuming that the force on each planet gear is the same, the sun gear is currently only subjected to the input torque and the meshing force between the teeth of the planet gears, so equation (1) can be obtained. The same force analysis is performed on the planet gears, the ring gear, and the planet carrier to obtain equations (2)-(4).

[0048] ; ; ; ; In the formula: Let be the moment of inertia of the transmission components. For the sun wheel, For planetary gears, For gear rings, For planetary support, The angular acceleration of the transmission component. The number of planetary gears. The meshing force between gears, This represents the meshing relationship between the sun gear and the planetary gears. This represents the meshing relationship between the gear ring and the planetary gears. Let be the base circle radius of the transmission component. Let be the radius of rotation of the planet carrier. For the mass of the planetary gear, For gear pressure angle, This refers to the input torque of the transmission mechanism.

[0049] For the transmission mechanism as a whole, the same force analysis method is adopted. If the component being analyzed is composed of multiple components, the angular acceleration and angular velocity of the combined component are consistent. In the analysis, it is only necessary to superimpose the moment of inertia and the force on each part of the combined component. According to the simplified diagram of the transmission mechanism, the reduction gear 3 / 4, the power coupling mechanism 5 and the speed change gears 8 / 9 on both sides are analyzed respectively. First, the two reduction gears 3 / 4 directly connected to the motor 1 / 2 are analyzed. Assuming that the forces on each planetary gear are consistent and ignoring the influence of gear tooth stiffness and damping, the dynamic differential equations corresponding to each gear are derived by the Newton-Euler method, and equations (5)-(10) can be obtained. Here, the influence of bearing viscous friction is considered, and a negative friction torque is added after each formula. This means that the rotating component as a whole is subjected to a frictional torque that is opposite to the direction of rotation.

[0050] ; ; ; ; ; ; In the formula: Let be the moment of inertia of the transmission components. For the sun wheel, For planetary gears, For gear rings, For planetary support, It is the 3rd planet in the planetary system. It is the 4th planet in the planetary order. It is the 6th planet in the planetary order. It is the 7th planet in the planetary system. It is the 8th planet in the planetary system. It is the 9th planet. The angular acceleration of the transmission component. The angular velocity of the transmission component. The number of planetary gears. The meshing force between gears, This represents the meshing relationship between the sun gear and the planetary gears. This represents the meshing relationship between the gear ring and the planetary gears. Let be the base circle radius of the transmission component. The radius of rotation of the planet carrier. For the mass of the planetary gear, For gear pressure angle, The bearing viscous friction coefficient, and is the input torque of the transmission mechanism, and is half the output torque of the motor.

[0051] The same force analysis was performed on the intermediate power coupling mechanism 5, resulting in equations (11)-(14). The definitions of the variables in the equations are the same as before.

[0052] ; ; ; ; Performing the same force analysis on the two gearboxes 8 / 9 connected to the output shaft on the outer side, we obtain equations (15)-(18). The variable definitions in the equations are the same as before. and The load torque of the transmission mechanism is denoted as , and the load torque of the planetary carriers output from the 8th and 9th gearboxes on both sides is denoted as .

[0053] ; ; ; ; The relative meshing displacement of the meshing gears in each planetary gear set along the meshing line direction is calculated as shown in equation (19). This represents the relative meshing displacement between the sun gear and the planet gears. This represents the relative meshing displacement between the gear ring and the planetary gears. The definitions of the remaining variables are the same as before.

[0054] ; The meshing force of each gear can be expressed as: ; In the formula, For equivalent meshing stiffness, This is the equivalent meshing damping.

[0055] The torque of the forces exerted on the high-speed and low-speed tracks during the movement of a tracked vehicle, converted to the output end of the steering mechanism (i.e., the planetary carriers of the 8 / 9 gear sets on both sides). and It is obtained from equations (21)-(26). The first term in equations (21) and (22) is the resistance of the tracked vehicle in straight-line travel, the second term is the resistance torque during the turning process, the third term is the inertial torque of the tracked vehicle in straight-line travel, the fourth term is the air resistance torque, the fifth term is the inertial torque of the two tracks, and the sixth term is the inertial torque during the angular acceleration of the vehicle body when turning. and These are the speeds of the high-speed side track and the low-speed side track, respectively. The relative turning radius, This is the steering drag coefficient. For tracked vehicles at a specified radius The maximum steering resistance coefficient during steering.

[0056] ; ; ; ; ; ; In the formula, For the overall vehicle quality, It is the acceleration due to gravity. The ground deformation resistance coefficient, The radius of the driving wheel, This is the steering drag coefficient. The track ground contact length, The center distance between the two tracks. air density, The aerodynamic drag coefficient, For windward area, The equivalent rotational inertia of the track, Let be the moment of inertia of the vehicle about its vertical axis. For the side transmission ratio, For side transmission efficiency.

[0057] Combining equations (1)-(26), the transmission mechanism model of the overall dual-sided motor coupled drive device can be expressed as: ; in All of the included equations (5)-(18) And the corresponding first derivative, representing the system's state variables. This represents a control variable.

[0058] Integrating the state-space equations using the Runge-Kutta method yields... The changes of each variable over time are used to obtain the rotational speed, angular acceleration, and gear meshing force of each gear transmission component at each moment.

[0059] Specifically, such as Figure 8As shown in this embodiment and some embodiments of the present invention, the structure of the feedforward neural network is designed, and the neural network model includes two hidden layers. The neural network constructed in these embodiments of the present invention adopts a four-layer feedforward neural network structure, consisting of an input layer, two hidden layers, and an output layer, to achieve effective modeling and correction of errors in the physical dynamics model. The number of nodes in the input layer is determined according to the dimensions of the variables involved in the experiment, including the motor speed and torque at the input end of the electromechanical composite transmission mechanism, as well as multi-source environmental factors reflecting the system's operating conditions, such as lubricating oil temperature, oil pressure, and external loads, which can comprehensively reflect the main factors affecting the dynamic behavior of the system. The number of nodes in the output layer corresponds to the actual observed dynamic response values ​​of the system, such as physical quantities like the actual angular velocity measured by sensors, used to compare with the model output and guide the network to learn the error compensation relationship. The number of nodes in the hidden layer is determined through cross-validation to achieve a balance between fitting accuracy and network complexity, avoiding underfitting or overfitting. The network's activation function is the commonly used sigmoid function, chosen based on the fitting target, to enhance the network's ability to express nonlinear relationships. The output layer uses a linear activation function, enabling the network to output continuous values, thus adapting to regression prediction tasks of real-world dynamic quantities. This neural network architecture design balances the diversity and physical meaning of input variables while possessing strong nonlinear modeling capabilities, providing an efficient model foundation for subsequent error correction.

[0060] No. Layer The input of each node (i.e., neuron) can be represented as: ; In the formula: Indicates the first The first in the layer Input of each node; For the first Number of neurons in a layer; for Layer Each node to Layer The connection weights between nodes; For the first Layer The output of each node; For the first Layer Threshold for each node; ; This refers to the number of network layers (excluding the input layer).

[0061] The activation function for the hidden layer is the Sigmoid function: ; In the network Layer nodes The output can be represented as: ; It is understood that in this embodiment and some embodiments of the present invention, the sample input includes the left and right motor speeds at the input end of the transmission mechanism. and and the corresponding input torque and The output rotational speed calculated from the physical dynamics model before optimization and During the experiment, the lubricating oil temperature (Temp) and oil pressure (pres), and the output load (output torque) were measured. and These variables comprehensively reflect the time-varying factors affecting the dynamic behavior of the transmission system. The sample output is the output angular acceleration calculated from the physical dynamics model before optimization. and Compared with actual measurements and The difference in data allows the neural network to learn the gap between the model and actual behavior.

[0062] Specifically, such as Figure 9 As described in this embodiment and some embodiments of the present invention, by introducing a genetic algorithm to optimize the initial weights and thresholds, not only is the drawback of neural networks getting trapped in local optima effectively avoided, but the training process of the network can also be accelerated. The global optimization capability of genetic operations ensures that the network has stronger adaptability under various operating conditions, enabling the modified dynamic model to provide more accurate predictions under complex time-varying conditions. Through this method, the combination of genetic algorithms and neural networks can significantly improve the predictive ability of the model, especially when dealing with complex nonlinear systems, exhibiting extremely high accuracy and stability.

[0063] In the training and validation process of feedforward neural networks, the Levenberg-Marquardt (LM) backpropagation training algorithm is executed based on the initial parameters obtained through genetic algorithm optimization. Specifically, the collected experimental dataset is first divided into training, validation, and test sets according to a certain ratio, typically 70% for the training set, 15% for the validation set, and 15% for the test set. Then, forward propagation is performed using the training set data to calculate the error between the network output and the actual target value. This error is then propagated backward using the backpropagation algorithm to adjust the network weights and thresholds. The Levenberg-Marquardt algorithm combines the advantages of gradient descent and Newton's method, automatically adjusting the update step size during training to accelerate convergence while avoiding the learning rate setting problems that may occur in traditional gradient descent.

[0064] After each training round, the network performs error evaluation on the validation set. Error metrics (such as mean squared error and mean absolute error) are calculated to check the model's generalization ability, and the training process is adjusted accordingly to prevent overfitting. The test set is used to finally evaluate the model's performance, ensuring its predictive ability on unseen data. During training, the network iteratively updates weights and biases until the error converges and reaches the preset performance standard. Finally, after training, the network structure and parameters are fixed, resulting in an optimized neural network model. This optimized model is then used as a corrector to compensate for errors in the initial dynamics model, significantly improving the prediction accuracy and adaptability of the transmission system under complex operating conditions.

[0065] Because the model only outputs the output end and The difference from the actual value also needs to be derived from the transmission ratio between gears to obtain the angular acceleration differences of other gears and shafts. The overall model is then encapsulated as follows: function, and The definition is consistent with the previous text.

[0066] Specifically, in this embodiment and some embodiments of the present invention, after completing the physical modeling and training of the neural network correction model, the mechanistic model is combined with the neural network model to obtain the corrected dynamic model: ; The output of the mechanistic model and key operating parameters (such as lubricating oil temperature, oil pressure, and load) are used as input features and fed into a trained neural network to predict the error correction amount under actual operating conditions. This correction amount is then superimposed on the output of the mechanistic model to form the final dynamic response result. Integrating this result yields the changes of each corrected variable over time. This fusion method combines the interpretability of the mechanistic model with the nonlinear fitting capability of the data-driven model, effectively improving the accuracy and generalization ability of the dynamic model under complex time-varying environments.

[0067] Example 3 Figure 10 This is a schematic diagram of the main components of the data- and mechanism-driven electromechanical composite transmission system dynamic model construction device according to an embodiment of the present invention. Figure 10 As shown, the data- and mechanism-driven dynamic model construction device for electromechanical composite transmission system of this invention in this embodiment includes: The data acquisition module is used to acquire the transmission input data, transmission output data, and multi-source operating condition data of the electromechanical composite transmission system. The prediction module is used to input the transmission input data into the mechanism dynamics model to obtain the predicted output data. The mechanism dynamics model is constructed based on the transmission structure of the electromechanical composite transmission system. The training module is used to construct a training dataset using the transmission input data, the transmission output data, the predicted output data, and the multi-source operating condition data, and to train a neural network model using the training dataset to obtain an error correction network model. The fusion module is used to fuse the mechanistic dynamics model and the error correction network model to obtain the dynamics model of the electromechanical composite transmission system, wherein the error correction network model corrects the error in the predicted output data of the mechanistic dynamics model.

[0068] Example 4 like Figure 11 As shown, Embodiment 4 of the present invention provides a terminal device, including at least one processor and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps in the data and mechanism-driven electromechanical composite transmission system dynamic model construction method as described in the first aspect of the present invention.

[0069] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripherals, voltage regulators, and power management circuits, via interfaces, as is well known in the art. Interfaces provide a connection between the bus and the transceiver, such as communication interfaces or user interfaces. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0070] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0071] Example 5 Embodiment 5 of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the data and mechanism-driven method for constructing a dynamic model of an electromechanical composite transmission system as described in the first aspect of the present invention.

[0072] Those skilled in the art will understand from the foregoing description that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes, but is not limited to, various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic storage devices, and optical storage devices.

[0073] Example 6 Embodiment 6 of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the data and mechanism-driven method for constructing a dynamic model of an electromechanical composite transmission system as described in the first aspect of the present invention.

[0074] Based on the above detailed description of the embodiments of the present invention, it can be seen that the embodiments of the present invention have the following advantages: The embodiments of the present invention obtain an error correction network model based on data and neural network models, and use the error correction network model to correct the prediction error of the mechanism dynamics model, thereby effectively improving the accuracy of the dynamics model of the electromechanical composite transmission system.

[0075] The embodiments of the present invention describe the basic framework of the physical mechanism and dynamic behavior of the electromechanical composite transmission system through a mechanistic dynamics model. The neural network model is used to modify and optimize the mechanistic dynamics model with data, which makes up for the shortcomings of relying solely on the mechanistic model or the data-driven model and achieves the complementary advantages of the two.

[0076] In constructing the training dataset for the neural network model, this invention not only uses transmission input data, transmission output data, and predicted output data, but also uses multi-source operating condition data to consider the nonlinear factors and multi-source coupling effects present in the transmission system. This enables the trained error correction network model to more comprehensively capture the dynamic characteristics of the system and adapt to different operating conditions.

[0077] In the embodiments of the present invention, when training a neural network model, the initial weights and thresholds of the feedforward neural network are optimized by a genetic algorithm, which avoids the problem of traditional feedforward neural networks getting trapped in local optima, making the model more adaptable and generalizable, and able to provide more accurate predictions in complex time-varying environments.

[0078] The dynamic model constructed in this invention provides a more reliable and accurate theoretical basis for the design, control, and optimization of electromechanical hybrid transmission systems. Through precise model predictions, it can better guide the improvement of vehicle mobility, noise reduction, and operational efficiency, thus contributing to the development of related technological fields.

[0079] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing a dynamic model of an electromechanical composite transmission system driven by data and mechanisms, characterized in that, include: Collect transmission input data, transmission output data, and multi-source operating condition data of the electromechanical composite transmission system; The transmission input data is input into the mechanism dynamics model to obtain the predicted output data. The mechanism dynamics model is constructed based on the transmission structure of the electromechanical composite transmission system. A training dataset is constructed using the transmission input data, the transmission output data, the predicted output data, and the multi-source operating condition data. An error correction network model is then trained using the training dataset. By integrating the mechanistic dynamics model and the error correction network model, a dynamic model of the electromechanical composite transmission system is obtained, wherein the error correction network model corrects the error in the predicted output data of the mechanistic dynamics model.

2. The method according to claim 1, characterized in that, The neural network model is a multi-layer feedforward neural network, including an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is equal to the sum of the number of types of the transmission input data, the transmission output data, and the multi-source operating condition data; the output of the output layer is the error of the predicted output data, and the number of nodes in the output layer is equal to the number of types of the predicted output data.

3. The method according to claim 2, characterized in that, The neural network model includes two hidden layers; the activation function of the input layer and the hidden layer is the sigmoid function, and the activation function of the output layer is a linear activation function.

4. The method according to any one of claims 1-3, characterized in that, The electromechanical composite transmission system includes a drive motor and a transmission mechanism coupled to the drive motor. The transmission input data includes the speed and torque of the drive motor. The transmission output data includes the output speed, output torque, and output angular acceleration of the transmission mechanism collected by actual sensors. The multi-source operating condition data includes the oil temperature and oil pressure of the lubricating oil in the electromechanical composite transmission system. The predicted output data includes the output speed, output torque, and output angular acceleration of the transmission mechanism obtained using the mechanistic dynamics model.

5. The method according to any one of claims 1-3, characterized in that, The error correction network model is obtained by training a neural network model using the aforementioned training dataset, including: The weights and thresholds of each node in the neural network model are initialized to obtain the initial weight values ​​and initial threshold values ​​of each node. A genetic algorithm is used to optimize the initial weight values ​​and the initial threshold values ​​to obtain the optimal initial weight values ​​and the optimal initial threshold values ​​for each node. Based on the initial values ​​of the optimal weights, the initial values ​​of the optimal thresholds, and the training dataset, the output error of the neural network model is calculated using the forward propagation algorithm, and the error is propagated backward using the backpropagation algorithm to adjust the weights and thresholds of the nodes. In response to the completion of the training termination condition, the last updated weights and thresholds are obtained, and the neural network model at the end of training is used as the error correction network model.

6. The method according to claim 1, characterized in that, A genetic algorithm is used to optimize the initial weight values ​​and the initial threshold values ​​to obtain the optimal initial weight values ​​and the optimal initial threshold values ​​for each node, including: The weights and thresholds of all the nodes are encoded into chromosomes, and the chromosomes are initialized to obtain an initialized population. The fitness of each chromosome is calculated according to a preset fitness function, and genetic operations are performed on the population based on the fitness of the chromosomes; the genetic operations include selection, crossover, and mutation. In response to the fulfillment of the iteration termination condition, the optimal chromosome of the population obtained at the end is decoded and output to obtain the optimal initial weight value and the optimal initial threshold value corresponding to each node.

7. A data- and mechanism-driven device for constructing a dynamic model of an electromechanical composite transmission system, characterized in that, include: The data acquisition module is used to acquire the transmission input data, transmission output data, and multi-source operating condition data of the electromechanical composite transmission system. The prediction module is used to input the transmission input data into the mechanism dynamics model to obtain the predicted output data. The mechanism dynamics model is constructed based on the transmission structure of the electromechanical composite transmission system. The training module is used to construct a training dataset using the transmission input data, the transmission output data, the predicted output data, and the multi-source operating condition data, and to train a neural network model using the training dataset to obtain an error correction network model. The fusion module is used to fuse the mechanistic dynamics model and the error correction network model to obtain the dynamics model of the electromechanical composite transmission system, wherein the error correction network model corrects the error in the predicted output data of the mechanistic dynamics model.

8. A computer-readable storage medium, characterized in that, It stores a program that, when executed by a processor, implements a data- and mechanism-driven method for constructing a dynamic model of an electromechanical composite transmission system as described in any one of claims 1-6.

9. A terminal device, characterized in that, It includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the data and mechanism-driven method for constructing a dynamic model of an electromechanical composite transmission system as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the data and mechanism-driven method for constructing a dynamic model of an electromechanical composite transmission system as described in any one of claims 1-6.

Images