A Torque Reconfiguration and Allocation Method for Multi-Axle Special Vehicles Based on Hub Motor Fault Estimation

Abstract

This invention discloses a torque reconfiguration allocation method for multi-axle special vehicles based on hub motor fault estimation, belonging to the field of automotive intelligent control technology. The method includes: acquiring vehicle state information through onboard sensors; estimating the hub motor fault torque using a neural network observer; establishing a total driving torque fault model based on this estimate and considering the characteristics of electric wheel-driven multi-axle special vehicles; and finally, establishing an optimization objective function for torque reconfiguration allocation with the goal of maximizing the vehicle's adhesion margin. Under the actual constraints of multi-axle vehicles, the optimal torque reconfiguration allocation is achieved using a wolf pack algorithm. This invention can effectively estimate the additional fault torque caused by hub motor failure, and by introducing a real-time estimate of the hub motor fault torque in the torque reconfiguration allocation, it can effectively improve the vehicle's stability control performance under motor failure conditions, thereby enhancing the vehicle's driving safety.

Description

Technical Field

[0001] This invention belongs to the field of automotive intelligent control technology, and relates to multi-axle vehicle control technology, specifically to a torque reconfiguration and allocation method for multi-axle special vehicles based on hub motor fault estimation. Background Technology

[0002] Multi-axle special vehicles play a crucial role in national defense industry construction, especially in the transportation of rockets and large missiles. With increasing demands for vehicle safety in recent years, safety research has become an important part of current vehicle research, particularly for multi-axle special vehicles used for transport and launch. Compared to traditional vehicles, electric wheel drive multi-axle special vehicles save energy consumed in transmission mechanisms such as gearboxes, clutches, and drive shafts, improving efficiency and flexibility. However, electric wheels are very close to the ground during operation, making them susceptible to dust, mud, gravel, and even water, resulting in a harsh working environment. If a motor fails during operation, it can lead to a loss of driving torque, or in severe cases, a sudden change in yaw torque, causing the vehicle to veer, skid, or even fishtail, resulting in serious traffic accidents. To improve overall vehicle safety, torque reconfiguration and distribution control under hub motor failure has become an important research direction for electric wheel multi-axle vehicles.

[0003] The invention patent application CN112886905A, published on June 1, 2021, addresses the problem that 8×8 multi-axle electric wheel drive vehicles, due to the increased probability of drive system failure caused by the large number of drive motors, leading to the vehicle's inability to enter the desired driving state. It discloses a rule-based fault-tolerant control method for electric wheel drive eight-wheel vehicles, capable of simultaneously handling both single-sided and double-sided drive failures, thus improving the active safety of multi-axle vehicles. The invention patent application CN113968205A, published on January 25, 2022, discloses a composite braking fault-tolerant control method for multi-axle electric drive vehicles. This method effectively solves the problem that most current fault-tolerant control strategies for multi-axle vehicles are designed for all-wheel drive vehicles and are not very applicable to non-all-wheel drive vehicles, leading to safety issues caused by motor failure during braking. This method has significant research value for improving vehicle driving safety.

[0004] Current torque reconfiguration fault-tolerant control strategies applied to multi-axle special vehicles are rule-based. However, due to the large number of drive units in this study, resulting in numerous failure modes, rule-based fault-tolerant control strategies require developing corresponding control measures for each failure mode, leading to a significant workload. Furthermore, rule-based control strategies do not consider the optimization effect on vehicle stability margin. In addition, these methods assume that the fault torque or failure coefficient of the hub motor is known, which is impossible in practical applications. Therefore, to fully consider the optimization effect on vehicle stability margin and fit the actual application scenario, there is an urgent need for a torque reconfiguration allocation method for multi-axle special vehicles based on hub motor fault estimation to improve the driving safety of multi-axle vehicles under hub motor failure. Summary of the Invention

[0005] To address the driving safety issues of multi-axle special vehicles under hub motor failure, this invention proposes a torque reconfiguration distribution method for multi-axle special vehicles based on hub motor failure estimation. A fault model of the vehicle's total driving torque is established by combining the characteristics of electric wheel-driven multi-axle special vehicles. An optimization objective function for torque reconfiguration distribution is established with the goal of maximizing the vehicle's adhesion margin, thereby achieving optimal torque reconfiguration distribution and improving the driving safety of multi-axle special vehicles.

[0006] The present invention provides a torque reconfiguration and allocation method for multi-axle special vehicles based on hub motor fault estimation, comprising the following steps:

[0007] S1: Based on vehicle motion state information, a neural network observer is used to estimate the fault torque of the hub motor;

[0008] S2: Based on the hub motor fault torque estimated in step S1, and combined with the dynamic characteristics of electric wheel multi-axle special vehicles, establish a total driving torque fault model for the vehicle.

[0009] S3: Based on the vehicle total driving torque fault model established in step S2, establish an optimization objective function for torque reconfiguration distribution with the goal of maximizing the vehicle's adhesion margin, and establish constraint conditions based on the dynamic constraints and actual actuator constraints of multi-axle special vehicles.

[0010] S4: Based on the wolf pack algorithm, the objective function of torque reconfiguration allocation is optimized under the constraints established in step S3 to obtain the optimal torque reconfiguration allocation result under hub motor fault.

[0011] In step S1, vehicle driving status information, total driving torque, and wheel hub motor fault torque are collected through onboard sensors. The data undergoes preprocessing such as upsampling and standardization. A fully connected neural network is used to extract sample features of the vehicle status information, and a neural network observer is trained to estimate the wheel hub motor fault torque. The neural network observer comprises an input layer, n hidden layers, and a fully connected output layer. The vehicle status is input to the input layer, and the output layer outputs the estimated fault torque values ​​for all wheels of the vehicle; n is a positive integer.

[0012] In step S2, the established fault model for the vehicle's total driving torque is as follows:

[0013]

[0014] Where v is the total driving torque of the vehicle, v = [F x ΔM z ] T F x Let ΔM be the total longitudinal force of the vehicle. z Additional yaw moment for the vehicle; u d This is the expected output torque of the vehicle's wheel hub motor; B is the estimated fault torque of the vehicle's wheel hub motor; B is the parameter matrix, as follows:

[0015]

[0016] r is the wheel radius, d is the distance between wheels on the same axle, and δ il (i=1,2) represents the steering angle of the two front left wheels of the vehicle, δ ir (i = 1, 2) represents the steering angle of the first two right wheels of the vehicle.

[0017] In step S3, the established objective function for torque reconfiguration allocation is as follows:

[0018]

[0019] Where W represents the set of wheels for a multi-axle special vehicle, F xi F represents the longitudinal force of tire i. zi μ represents the vertical load on each wheel. i This represents the road adhesion coefficient of tire i.

[0020] In step S3, the constraints of the objective function for torque reconfiguration allocation include: inequality constraints of the longitudinal forces of each tire, equality constraints of the vehicle's total driving torque fault model, and constraints of the actual torque provided by the two hub motors of each axle.

[0021] The inequality constraints for the longitudinal forces of each tire are expressed as follows:

[0022]

[0023] Among them, T max_brake This represents the maximum braking torque, which includes both the maximum braking torque of the motor and the maximum braking torque of the hydraulic braking system; T max_drive This represents the maximum driving torque of the motor; r is the wheel radius; F yi μ represents the lateral force of tire i; x μ represents the longitudinal adhesion coefficient of the road surface. y This represents the lateral adhesion coefficient of the road surface.

[0024] The constraints on the actual torque provided by the two hub motors on each axle are as follows:

[0025] T ild +T ilf =T ird +T irf , i∈W1;

[0026] Where W1 represents the set of wheels on one side of a multi-axle special vehicle, T ild T represents the desired output torque of the hub motor for the left wheel i. ilf To estimate the fault torque of the hub motor of the left wheel i, T ird T represents the desired output torque of the hub motor for the right wheel i. irf The estimated fault torque of the hub motor of the right wheel i.

[0027] The beneficial effects of this invention are as follows:

[0028] (1) The method of the present invention can effectively estimate the additional fault torque caused by hub motor failure.

[0029] (2) The method of the present invention introduces a real-time estimate of the fault torque of the hub motor in the torque reconstruction distribution, which can effectively improve the stability control effect of the vehicle under motor failure.

[0030] (3) The method of the present invention can effectively reduce the incidence of vehicle instability accidents caused by hub motor failure and improve the safety performance of automobile driving. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below. Referring to the accompanying drawings will provide a clearer understanding of the features and advantages of the present invention. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort. Wherein:

[0032] Figure 1 This is a flowchart illustrating the torque reconfiguration and allocation method for multi-axle special vehicles based on hub motor fault estimation according to the present invention.

[0033] Figure 2 This is a schematic diagram of the implementation structure of torque reconfiguration and allocation for multi-axle special vehicles based on hub motor fault estimation according to the present invention.

[0034] Figure 3 This is a neural network structure diagram of the hub motor fault observer of the present invention;

[0035] Figure 4 This is a schematic diagram illustrating the realization of the total driving torque of the vehicle in terms of tire force according to the present invention;

[0036] Figure 5 This is a flowchart of the optimization solution of the wolf pack algorithm of the present invention. Detailed Implementation

[0037] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0038] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0039] like Figure 1 and 2As shown, this invention provides a torque reconfiguration distribution method for multi-axle special vehicles based on hub motor fault estimation, applicable to torque reconfiguration distribution control under hub motor faults in multi-axle vehicles. A neural network fault observer is trained using onboard sensor signals to estimate the fault torque of the hub motor. Based on this estimate, and combined with the dynamic characteristics of the electric wheel multi-axle special vehicle, a total driving torque fault model is established. Finally, an optimization objective function for torque reconfiguration distribution is established with the goal of maximizing the vehicle's adhesion margin. Under the actual constraints of the multi-axle vehicle, the optimal torque reconfiguration distribution is achieved based on a wolf pack algorithm. The method involves acquiring vehicle state information through onboard sensors, estimating hub motor faults using a neural network observer, and then establishing a total driving torque fault model based on this estimate, combined with the characteristics of the electric wheel-driven multi-axle special vehicle. Finally, an optimization objective function for torque reconfiguration distribution is established with the goal of maximizing the vehicle's adhesion margin. Under the actual constraints of the multi-axle vehicle, the optimal torque reconfiguration distribution is achieved based on a wolf pack algorithm, improving the driving safety of multi-axle special vehicles.

[0040] The torque reconstruction and allocation method for multi-axle special vehicles based on hub motor fault estimation of the present invention will be described in four steps below.

[0041] Step 1: Obtain vehicle status information through onboard sensors and use a neural network observer to estimate the fault torque of the hub motor.

[0042] Step 1-1: Use onboard sensors to collect vehicle driving status information, total driving torque and wheel hub motor fault torque. Preprocess the data based on data augmentation and standardization methods to enrich the quantity and quality of the original data and obtain training samples.

[0043] Steps 1-2: Use a fully connected neural network to extract data features of vehicle state information, train a neural network-based hub motor fault observer, and realize the observation of motor fault torque.

[0044] Motor fault torque estimation is a crucial aspect of this invention, as its accuracy determines the performance of the torque reconfiguration and distribution controller, directly impacting the vehicle's kinetic safety. This invention utilizes the nonlinear characteristics of a fully connected neural network to fit the mapping relationship between multiple state information measured by onboard sensors and the wheel hub motor fault torque, thereby improving the estimation accuracy of the wheel hub motor fault torque.

[0045] The embodiments of the present invention are as follows Figure 3 The fully connected neural network shown is used to fit the mapping relationship between multiple state information measured by onboard sensors and the fault torque of the wheel hub motor, thereby estimating the fault torque of the wheel hub motor. From Figure 3It can be seen that the neural network contains n=5 hidden layers, each with m=50 neurons, an input layer, and a fully connected output layer. The input layer has 6 neurons, which input the vehicle state, and the output layer has 8 neurons, representing the estimated fault torque values ​​of the 8 wheels of the multi-axle vehicle in this embodiment of the invention. Each vehicle in this embodiment of the invention is equipped with one hub motor. For example... Figure 2 As shown, the vehicle state input to the neural network fault observer of the present invention includes motor torque, front wheel steering angle, motor current, yaw rate, longitudinal vehicle speed and lateral vehicle speed, and outputs the estimated value of motor fault torque to the hub motor torque reconstruction and distribution controller.

[0046] If the input to the neural network is represented as z, then the neural network fault observer can be represented as...

[0047]

[0048] in, This is an estimated value for the fault torque of the hub motor. Represents the composition of functions, g i (i = 1, ..., n) represents the activation function of the i-th hidden layer, used to introduce non-linearity into the neural network, f i (i = 1, ..., n) represents the output of the i-th hidden layer, which can be expressed by the following formula:

[0049] f i (ξ i-1 ) = W i ξ i-1 +b i (2)

[0050] Where, ξ i-1 W represents the output vector of the previous hidden layer. i Let b represent the weight coefficient matrix of the i-th hidden layer. i This represents the bias vector of the i-th hidden layer.

[0051] To train the neural network fault observer, the loss function is defined as the mean squared error between the estimated value and the actual value of the hub motor fault torque:

[0052]

[0053] Where, N j u represents the total number of samples in the training data. f This is the actual torque of a faulty wheel hub motor.

[0054] By selecting ReLU as the activation function, the design of a neural network fault observer can be completed through training.

[0055] By using a trained neural network fault observer to fit the vehicle driving status information collected in real time by on-board sensors, the fault torque of the wheel hub motor is estimated.

[0056] Step 2: Based on the estimated fault torque of the hub motor obtained in Step 1, and combined with the dynamic characteristics of the electric wheel multi-axle special vehicle, establish the vehicle's total driving torque fault model.

[0057] Before designing the lower-level hub motor torque reconfiguration distribution controller, it is necessary to establish a fault model of the vehicle's total drive torque. This is based on the hub motor fault estimates. and Figure 4 The diagram shown illustrates the realization of the total driving torque of the vehicle in the tire force. The fault model of the total driving torque of the vehicle based on the estimated torque value of the hub motor fault can be expressed as follows:

[0058]

[0059] Where v is the total driving torque of the vehicle, v = [F x ΔM z ] T F x Let ΔM be the total longitudinal force of the vehicle. z This refers to the additional yaw moment of the vehicle; the x-axis is parallel to the vehicle's axis, and the y-axis is perpendicular to the vehicle's axis; u d This refers to the expected output torque of each wheel hub motor in the vehicle, as described in the embodiment of this invention. d =[T 1ld T 2ld T 3ld T 4ld T 1rd T 2rd T 3rd T 4rd ] T T represents the expected output torque of each hub motor. ild (i = 1, 2, 3, 4) represents the desired output torque of the four hub motors on the left, T ird (i = 1, 2, 3, 4) represents the expected output torque of the four hub motors on the right side; This is the estimated value of the fault torque caused by the hub motor failure calculated in step 1, T. ilf (i = 1, 2, 3, 4) represents the fault torque of the four hub motors on the left side, T irf (i = 1, 2, 3, 4) represents the fault torque of the four hub motors on the right side; u d and u f The sum of these values ​​represents the true output torque of the multi-axle special vehicle; parameter matrix B is as follows:

[0060]

[0061] r is the wheel radius, d is the distance between wheels on the same axle, and δ il (i=1,2) represents the steering angle of the first two wheels on the left, δ ir (i = 1, 2) represents the steering angle of the first two wheels on the right.

[0062] Step 3: Based on the vehicle total driving torque fault model, establish an optimization objective function for torque reconfiguration allocation with the goal of maximizing vehicle adhesion margin, and establish constraint conditions based on the dynamic constraints and actual actuator constraints of multi-axle special vehicles.

[0063] To avoid tire force saturation as much as possible, it is necessary to maximize the vehicle's grip margin. Tire utilization rate can accurately reflect the vehicle's grip margin. When the tire utilization rate reaches 100%, it indicates that the vehicle's grip margin is zero, meaning the vehicle is close to instability. As the tire utilization rate decreases, the vehicle's grip margin increases, and the vehicle's stability improves. The objective function chosen is to minimize the sum of squares of the tire utilization rates, as shown below:

[0064]

[0065] Where, η i F represents the utilization rate of tire i. xi F represents the longitudinal force of tire i. yi Let represent the lateral force of tire i. Sum of the squares of the utilization rates of the eight tires, the objective function can be expressed as:

[0066]

[0067] Where, μ i F represents the coefficient of adhesion of wheel i to the road surface. zi This represents the vertical load on wheel i.

[0068] Since this invention focuses only on controlling longitudinal force, the objective function is rewritten as follows:

[0069]

[0070] According to the theory of tire friction circle, the constraint caused by the friction circle is as follows:

[0071]

[0072] Among them, F xmax =μ x F zi ,F ymax =μ y F zi μ x and μ yThese are the longitudinal adhesion coefficient and the lateral adhesion coefficient of the road surface, respectively. F can be derived from equation (11). xi The constraint conditions under road surface limitations are expressed as follows:

[0073]

[0074] Furthermore, based on the external characteristics of the drive motor and the limitations of the output capacity of actuators such as the hydraulic braking system, F can be obtained. xi The constraints are as follows:

[0075]

[0076] Among them, T max_drive T represents the maximum driving torque of the motor. max_brake This represents the maximum braking torque, which includes both the maximum braking torque of the motor and the maximum braking torque of the hydraulic braking system, i.e., T. max_brake =T m_max_brake +T b_max_brake According to equations (12) and (13), the inequality constraint conditions for the longitudinal forces of each tire are expressed as follows:

[0077]

[0078] To ensure yaw stability in multi-axle vehicles, the actual torque provided by the two hub motors on each axle should be the same.

[0079] T ild +T ilf =T ird +T irf (i=1,2,3,4) (12)

[0080] Meanwhile, the equality constraint conditions are shown in equation (4). Combining equation (4), equation (11) and equation (12), we can obtain the final constraint conditions for the longitudinal forces of each tire.

[0081] Step 4: Solve the optimization problem of torque reconfiguration allocation under the final constraint condition based on the wolf pack algorithm, and realize the optimal torque reconfiguration allocation of multi-axle special vehicles under hub motor failure.

[0082] When using the wolf pack algorithm to solve the problem, under the constraint of satisfying the objective function of torque reconstruction allocation, the hub motor torque combinations are randomly generated as artificial wolves. The objective function value of the artificial wolves is obtained by calculating the objective function value J of the corresponding hub motor torque combination. The optimization solution process of the wolf pack algorithm in this embodiment of the invention is as follows: Figure 5 As shown, it includes the following:

[0083] (1) First, under the constraints of equations (4), (11) and (12), N sets of hub motor torque combinations are randomly generated as artificial wolf packs to determine the state of the artificial wolves and the maximum number of walks T. max .

[0084] (2) Then, calculate the objective function value of N sets of hub motor torque combinations, select the optimal value as the alpha wolf of the artificial wolves, and select a certain proportion of artificial wolves other than the alpha wolf as scout wolves to perform roaming behavior. When a scout wolf calculates the prey odor concentration J tan The objective function value is less than the alpha wolf J. tou If the number of roaming attempts T reaches its limit, the scout wolf's roaming behavior ends. Afterwards, the wolves initiate an attack. During their pursuit in the designated direction, the wolves continuously calculate their objective function value and compare it with the alpha wolf's objective function value. When the alpha wolf's objective function value J... meng Less than the wolf J tou At that time, the roles of the wolf and the alpha wolf are reversed; when the wolf's objective function value J... meng Greater than the alpha wolf or the distance d between the two is Satisfy d is When <d, d is a preset value, and the roles of both remain unchanged. It can be seen that whether the roles of the alpha wolf and the wolf switch are determined by which wolf perceives the concentration of the prey's scent.

[0085] (3) Finally, the spatial position of the alpha wolf is the required optimal hub motor torque distribution result.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A torque reconfiguration and allocation method for multi-axle special vehicles based on hub motor fault estimation, characterized in that, Includes the following steps: Step 1) Establish a neural network fault observer to estimate the fault torque of the hub motor; obtain the state of the multi-axle special vehicle through on-board sensors and input it into the neural network fault observer to estimate the fault torque of the hub motor; Step 2) Establish a fault model for the total driving torque of electric wheel-driven multi-axle special vehicles; The vehicle's total driving torque fault model is established based on the estimated fault torque value of the in-wheel motor as follows: ； Where v is the total driving torque of the vehicle. , The total longitudinal force of the vehicle, Additional yaw moment for the vehicle; This is the expected output torque of each wheel hub motor in the vehicle; B is the estimated fault torque of each wheel hub motor in the vehicle; B is the parameter matrix, as shown below: ； r is the wheel radius, and d is the distance between wheels on the same axle. This refers to the steering angle of the two front left wheels of the vehicle. This refers to the steering angle of the two front right wheels; Step 3) Establish an optimization objective function for torque reconfiguration allocation with the goal of maximizing vehicle adhesion margin, and establish constraints based on the dynamic limitations and actual actuator limitations of multi-axle special vehicles; The objective function for torque reconfiguration allocation is established as follows: ； Where W represents the set of all wheels of a multi-axle special vehicle. This represents the longitudinal force of tire i. This indicates the vertical load on each wheel. This represents the road adhesion coefficient of tire i; Step 4) Solve the objective function of torque reconfiguration allocation based on the wolf pack algorithm and output the optimal torque reconfiguration allocation result.

2. The method according to claim 1, characterized in that, In step 1, the neural network fault observer used includes an input layer, n hidden layers, and a fully connected output layer. The vehicle state is input into the input layer, and the output layer outputs the estimated fault torque values ​​for all wheels of the vehicle; n is a positive integer. The neural network observer is pre-trained, and the loss function is set as the mean square error between the estimated value of the hub motor fault torque and the actual hub motor fault torque. After training, a neural network observer is obtained that fits the mapping relationship between the vehicle state measured by the on-board sensor and the fault torque of the wheel hub motor.

3. The method according to claim 1 or 2, characterized in that, In step 1, the vehicle state data input to the neural network fault observer includes motor torque, front wheel angle, motor current, yaw rate, longitudinal speed, and lateral speed.

4. The method according to claim 1, characterized in that, In step 3, the constraints of the objective function for torque reconfiguration allocation include the inequality constraints of the longitudinal forces of each tire, the equality constraints of the vehicle's total driving torque fault model, and the constraints of the actual torque provided by the two hub motors of each axle. The inequality constraints for the longitudinal forces of each tire are as follows: ； in, This indicates the maximum braking torque, which includes both the maximum braking torque of the motor and the maximum braking torque of the hydraulic braking system. This represents the maximum driving torque of the motor; r is the wheel radius. This represents the lateral force on tire i; This represents the longitudinal adhesion coefficient of the road surface. This represents the lateral adhesion coefficient of the road surface; The constraints on the actual torque provided by the two hub motors on each axle are as follows: ,i W1; Where W1 represents the set of wheels on one side of a multi-axle special vehicle, Let be the desired output torque of the hub motor of the left wheel i. To estimate the fault torque of the hub motor of the left wheel i, Let be the desired output torque of the hub motor of the right wheel i. The estimated fault torque of the hub motor of the right wheel i.

5. The method according to claim 4, characterized in that, In step 4, when using the wolf pack algorithm to solve the problem, the hub motor torque combination is randomly generated as an artificial wolf under the constraint of satisfying the objective function of torque reconstruction allocation. The objective function value of the artificial wolf is obtained by calculating the objective function value J of the corresponding hub motor torque combination.

Images