Method and system for identifying time-varying characteristics of heavy-duty vehicle suspension

Abstract

A method and system for identifying the time-varying characteristics of a heavy-duty vehicle suspension, wherein a sensor collects the sequential states of a mining truck operation, and a deep learning network is used to predict parameter-related factors based on the sequential states; at the same time, a linear dynamic model that considers the longitudinal-vertical coupling effect is used to roughly estimate the stiffness and damping coefficients as key suspension parameters, and then a nonlinear dynamic model that considers the longitudinal-vertical coupling effect is used to predict the state of the next time step based on the parameter-related factors and key suspension parameters. The present invention integrates a deep learning network into the physical model of a mining truck, establishes an accurate longitudinal-vertical dynamic model of a mining truck and identifies accurate suspension parameters, and gives the time-varying characteristics of the stiffness and damping of the mining truck suspension through a physical model-data driven method, and the model itself has certain interpretability and generalization; the four suspension stiffnesses and damping are obtained only through the information on the sprung part, and it has the advantages of simple architecture and easy implementation.

Description

Technical Field

[0001] The present invention relates to a technology in the field of mining truck control, in particular to a method and system for identifying the time-varying characteristics of a heavy-duty vehicle suspension. Background Art

[0002] The existing mining truck suspension simulation technology is unable to simulate and quantitatively calculate the time-varying stiffness and damping characteristics of the suspension, and most literature only presents statistical quantities. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention proposes a method and system for identifying the time-varying characteristics of the suspension of a heavy-duty vehicle, integrates a deep learning network into the physical model of a mining truck, establishes an accurate longitudinal-vertical dynamic model of the mining truck and identifies accurate suspension parameters, and gives the time-varying characteristics of the suspension stiffness and damping of the mining truck through a physical model-data-driven method. The model itself has certain interpretability and generalization. Four suspension stiffnesses and damping are obtained only through sprung information (one IMU), which has the advantages of simple architecture and easy implementation.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a method for identifying the time-varying characteristics of a heavy-duty vehicle suspension. A sensor collects the sequential states of a mining truck operation, and a deep learning network is used to predict parameter-related factors according to the sequential states. At the same time, a linear dynamics model considering the longitudinal-vertical coupling effect is used to roughly estimate the stiffness and damping coefficients as key suspension parameters, and a nonlinear dynamics model considering the longitudinal-vertical coupling effect is used to predict the state of the next time step according to the parameter-related factors and the key suspension parameters.

[0006] The state loss Lp between the predicted state and the target state detected by the sensor is used as the loss error term of the deep learning network to update the network parameters.

[0007] The linear dynamic model considering the longitudinal-vertical coupling effect includes: c The vehicle body with a rotational inertia of Ic and the unsprung mass arranged below it are m tf and m tr The front and rear axles of the vehicle body are: The suspension force transmitted from the front and rear axles to the vehicle body is F f and F r , the road excitation of the front and rear tires is z qf and z qr , the degrees of freedom (DoF) of the model include the center of gravity (CoG) z c The vertical displacement and pitch angle θ c and the front and rear unsprung masses z tf and ztr The vertical displacement, F c represents the inertial force acting on CoG due to the acceleration a and velocity v moving along the x-axis, M c represents the moment of inertia, where: the suspension stiffness and damping coefficient are k and c respectively, the subscripts f and r represent the front suspension and rear suspension respectively, and the subscript t represents the tire. The distance between CoG and the front / rear axle is a c and b c , the vertical distance between CoG and the pitch center PC is Δh.

[0008] The deep learning network includes: a three-layer LSTM network layer and a fully connected network layer, wherein: each layer of the LSTM network layer has 128 hidden units, and the fully connected network layer outputs the suspension stiffness damping correction coefficient.

[0009] Technical Effects

[0010] The present invention combines physics and learning methods to establish a hybrid modeling framework to separately identify the front and rear suspension stiffness and damping parameters of mining trucks that vary over time under different conditions. Compared with the prior art, the present invention facilitates the subsequent hybrid model to be expressed in a unified form, thereby enabling the identification of suspension parameters in a data-driven manner, accelerating the convergence of the neural network, and accurately predicting the vertical dynamic characteristics of the mining truck. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the vertical apparent dynamics identification method of the mining truck;

[0012] Figure 2 Schematic diagram of the nonlinear dynamic model considering the longitudinal-vertical coupling effect;

[0013] Figure 3 Schematic diagram of hybrid modeling architecture;

[0014] Figure 4 Schematic diagram of the raw signals collected by the sensor;

[0015] Figure 5 Schematic diagram of the minecart speed in each test set;

[0016] Figure 6 Schematic diagram of vertical acceleration comparison at CoG (Part 1);

[0017] Figure 7 Schematic diagram of vertical acceleration comparison at CoG (Part 5);

[0018] Figure 8 It is a schematic diagram of the front suspension vertical speed comparison;

[0019] Fig. 9It is a schematic diagram of the comparison of the vertical speed of the rear suspension;

[0020] Fig.10 Schematic diagram of the change of front and rear suspension stiffness;

[0021] Fig.11 Schematic diagram of the change law of front and rear suspension damping. DETAILED DESCRIPTION

[0022] like Figure 1 As shown, this embodiment involves a method for identifying the time-varying characteristics of a heavy-duty vehicle suspension, in which a sensor collects the sequential states of a mining truck operation, and a deep learning network is used to predict parameter-related factors based on the sequential states; at the same time, a linear dynamic model considering the longitudinal-vertical coupling effect is used to roughly estimate the suspension stiffness and damping coefficient as key suspension parameters, and then a nonlinear dynamic model considering the longitudinal-vertical coupling effect is used to predict the state of the next time step based on the parameter-related factors and key suspension parameters.

[0023] like Figure 2 As shown, the nonlinear dynamic model considering the longitudinal-vertical coupling effect has a positive direction definition, including: having a mass m c The vehicle body with a rotational inertia of Ic and the unsprung mass arranged below it are m tf and m tr The front and rear axles of the vehicle body are: The suspension force transmitted from the front and rear axles to the vehicle body is F f and F r , the road excitation of the front and rear tires is z qf and z qr The degrees of freedom (DoF) of the model include the center of gravity (CoG) z c The vertical displacement and pitch angle θ c and the front and rear unsprung masses z tf and z tr The vertical displacement, F c represents the inertial force acting on CoG due to the acceleration a and velocity v moving along the x-axis, M c represents the moment of inertia, where: the suspension stiffness and damping coefficient are k and c respectively, the subscripts f and r represent the front suspension and rear suspension respectively, and the subscript t represents the tire. The distance between CoG and the front / rear axle is a c and b c , the vertical distance between CoG and the pitch center PC is Δh.

[0024] According to the D'Alembert principle, the longitudinal-vertical dynamic model of a mining truck is:

[0025] Among them: The force exerted by the vehicle body on the suspension is

[0026] Preferably, the vertical acceleration and velocity of the center of gravity are calculated according to the nonlinear mapping relationship of the IMU Specifically include:

[0027] 1) According to Euler's theorem of rigid body kinematics, the motion of a point in the absolute coordinate system r is decomposed into the translation displacement R in the relative coordinate system plus the vector sum of the rotation around the base point: Where: A is the coordinate rotation matrix, u i is the projection on the coordinate axis after rotation. OXYZ is an inertial reference system (absolute reference system) fixed on the ground; i X m Y m Z m It is a coordinate system fixed on the IMU (relative coordinate system), and the three-axis directions are the same as the inertial reference system; i X i Y i Z i Indicates the coordinate system definition of the IMU itself. Point P is the point to be measured, and u is the vector from the coordinate axis reference point to the measured point.

[0028] 2) According to the actual situation, reasonably determine the rotation order of the coordinate axes and calculate the rotation matrix: According to the knowledge of vehicle kinematics, the roll angle and pitch angle are very small relative to the yaw angle. Therefore, the rotation order when selecting the coordinate axes is: first rotate around the Z axis (corresponding to the yaw angle α c ); then rotate around the Y axis (corresponding to the pitch angle θ c ); finally rotate around the X axis (corresponding to the rolling angle γ c ). We get the rotation matrix expression A(α c ,θ c ,γ c )=A γ A β A α , where: the positive direction of rotation is the right-handed spiral direction, and the expressions of the three rotation matrices are

[0029] 3) Calculate the velocity of point P and take the derivative of both sides of the expressions of the three rotation matrices with respect to time. The rotation matrix is ​​regular, and its derivative with time is calculated by linear algebra. Among them: Each axial angular velocity matrix is ​​an antisymmetric matrix Get the velocity expression of the measured point And the acceleration expression of the measured point (the angular acceleration of each axis) That is, calculate the acceleration (including angular acceleration) of any point rigidly connected to the IMU installation position.

[0030] The movement of the vehicle body is thus obtained, completing the longitudinal-vertical dynamic model of the mining truck.

[0031] The key suspension parameters, namely the suspension stiffness and damping coefficient, are obtained in the following way: Taking the front suspension as an example, under the force F f Under the input of , the vertical displacement z on the sprung spring is obtained according to the transmission relationship between the suspension force and the suspension displacement of the mining truck's dynamic system. f , the suspension stiffness is calculated by system identification method, and the unsprung position displacement is calculated, including:

[0032] 1) According to the Laplace transform relationship between force and displacement Among them: All are defined above, and the curly hairs are the Laplace transforms of the corresponding variables

[0033] 2) By eliminating the vertical displacement of the unsprung mass, the transmission relationship between the suspension force and the suspension displacement of the mining truck dynamic system Where: s is the time domain differential operator. According to the measured real-time mine car timing signal, the continuous time system simplified refined instrumental variable method (SRIVC) is used for system identification

[22] , rewrite the system transfer equation into the standard form of band-limited continuous-time white noise e(t) Among them: the parameter vector to be identified is θ = [a1, a2, a3, b0, b1, b2] T ;

[0034] 3) The entire equation is low-pass filtered through a continuous-time state variable filter to obtain the input and output pre-filtered time derivatives required for identification. It is related to the noise vector, so the recognition result is asymptotically biased. To eliminate the influence of the noise vector, define the tool vector at step j Calculate the parameter estimates for step j+1 Continue iterating until the error between the recognition result and the previous recognition result is small enough, that is,

[0035] 4) Based on the identified system transfer function and the original input-output relationship, the normalized root mean square error (NRMSE) and final prediction error (FPE) are used as evaluation indicators to evaluate the accuracy of the model.

[0036] In this embodiment, high-order terms are selected as the calculation method for parameter regression, and the calculation method of cf = a3 / b2, kf = a2 / b2 is adopted. Considering that this test is carried out under the same initial conditions, the suspension stiffness parameters should also be the same. Therefore, the least square method is used to calculate the suspension stiffness parameters under all working conditions, and the average value of the error evaluation index is given. Similarly, the rear suspension also uses the same calculation process. In addition, it also provides a basis for the establishment and training of subsequent network models, and is directly used for the calculation of physical models.

[0037] The state of the next time step is identified by combining data drive with a neural network to identify suspension parameters, specifically including:

[0038] 1) Taking the front suspension as an example, Perform backward differentiation to obtain the suspension discrete system expression at time k: Where: superscript k and k-1 represent the current state and the last state respectively. The sampling time difference of each state; since the unsprung mass displacement is obtained through parameter identification, its differential is not accurate and needs to be identified in combination with other equations. The speed of the unsprung mass at time k is Where: The subscript new represents the modified parameters of the neural network. When the suspension force at time k-1 is introduced, the entire algorithm will iterate, resulting in the accumulation of errors each time. The forward differential of the vehicle DoF is introduced to represent the unsprung mass velocity at time k-1. Get unsprung mass After the speed By solving the function, we can get the force Ff transmitted by the front suspension at time k:

[0039] 2) For the rear suspension, the equation is Calculate F at time k r Therefore, the motion of the vehicle is According to the forward differentiation method, the predicted state of the vehicle body is summarized as follows: An iterative method for obtaining the body parameters of mining trucks in discrete time.

[0040] 3) The front and rear suspension parameters of mining trucks show nonlinear characteristics under different working conditions. Based on the previous parameter identification, neural networks are used for refined modeling, that is, the updated iterative parameters are Neural network based on time series signal Get N out =[P kf ,P kr ,P cf ,P cr ].

[0041] like Figure 3As shown, the neural network adopts a recurrent neural network (RNN), including: a three-layer long short-term memory (LSTM) network and a fully connected layer, wherein: the hidden layer size is 128, and the fully connected layer calculates the suspension correction parameters and outputs them.

[0042] The loss function of the neural network is the mean square error (MSE), and the physical signals compared are the vertical acceleration and pitch rate of the sprung mass of the vehicle body, specifically:

[0043] In order to avoid problems such as gradient disappearance or gradient explosion during neural network training, the Xavier initialization method is used to initialize the RNN network. The linear output layer uses the normal distribution initialization method to enhance training stability and accelerate network convergence.

[0044] After specific practical experiments, the TR100A mining truck produced by TEREX was selected as the experimental platform. The test site is located in the Ruomaoshan mining area in Wuhu City, Anhui Province, China. In the experiment, the truck position information and all kinematic information are collected by the highly robust, high-precision positioning and attitude setting system IFS2100 designed by DAISCH. All data are collected at 100Hz, and the past 2-step state is input into the hybrid neural network model and physical model for calculation.

[0045] The sensor is installed on the cockpit floor, and its positive x-axis and z-axis are in the same direction as the vehicle coordinate system. However, the positive y-axis is opposite to the vehicle coordinate system, so the positive direction needs to be considered in the calculation.

[0046] The experiment was conducted on a flat paved road with the mine car in an unloaded state. For simplicity, the ground excitation was ignored compared to the tire size during transportation. The experiment included longitudinal acceleration and deceleration tests, and the specific tests are shown in Table 1.

[0047] Table 1. Mining truck test status

[0048]

[0049] The required mining car size parameters, stiffness characteristics, inertia information, and sensor installation positions in the calculation process are shown in Table 2, where Δx, Δy, and Δz represent the coordinates of the CoG relative to the IMU installation position. The suspension stiffness parameters are calculated, and the average fitting accuracy (1-NRMSE) under all working conditions is 84.76%, and the average FPE is 0.024. The fitting results are highly accurate and serve as reference inputs for subsequent model and network calculations.

[0050] Table 2. Mining card parameters

[0051] parameter Value (unit) parameter Value (unit) <![CDATA[a c ]]> 2.33(m) <![CDATA[m tf ]]> 8150.45(kg) <![CDATA[b c ]]> 2.24(m) <![CDATA[m tr ]]> 18423.55(kg) Δx -3.30(m) <![CDATA[k tf ]]> <![CDATA[5.25×10 7 (N / m)]]> Δy -1.24(m) <![CDATA[k tr ]]> <![CDATA[7.96×10 7 (N / m)]]> Δz 0.45(m) <![CDATA[k f ]]> <![CDATA[1.87×10 6 (N / m)]]> Δh 1.12(m) <![CDATA[k r ]]> <![CDATA[1.93×10 6 (N / m)]]> <![CDATA[m c ]]> 42046 / (kg) <![CDATA[c f ]]> <![CDATA[1.59×10 4 (Ns / m)]]> <![CDATA[I c ]]> <![CDATA[234333(kg·m 2 )]]> <![CDATA[c r ]]> 2.00×104(Ns / m)

[0052] After clarifying the parameters and working conditions of the mining truck, the above equations are used to train the network and build the model. The specific comparison results are given below. Before calculation and network training, all collected data are aligned on the same timeline and forward and dimensionally transformed according to the theory proposed in this invention.

[0053] like Figure 4 As shown in , these are the original values ​​of the longitudinal acceleration, vertical acceleration, and pitch angular velocity measured by the sensor. Figure 5 Shown are the truck speeds in each part of the test set.

[0054] During the network training process, training sets and test sets are distinguished for all working conditions in Table 1. In order to conduct reasonable training and comparison under different working conditions, the first 40% and the last 40% of all working condition data are used as training sets, and the remaining 20% ​​of the data are used as test sets.

[0055] In order to better verify the accuracy of the present invention, a physical model, a data-driven model and a hybrid model are selected for comparison. Among them, the physical model is a classic half-car model, and the data-driven model adopts an LSTM network. All training parameters participating in the present invention are consistent to verify the applicability of each model. The design of mining trucks generally prevents cargo from falling due to excessive vertical acceleration during transportation, thereby avoiding loss of transportation capacity and road interference. In addition, obtaining the suspension movement speed is also beneficial to subsequent suspension control and tire dynamic load calculation. Therefore, the present invention focuses on these two indicators and compares the calculation results of different models.

[0056] Select the first and fifth parts of the working condition to display to verify the model effect during acceleration and deceleration. Figure 6 and Figure 7 As shown in the figure, the comparison results of the vertical acceleration of the center of gravity of the two working condition test groups are shown respectively. In the physical model, the calculation of the unsprung position requires solving differential equations during the calculation process. This process will introduce errors and oscillations, resulting in violent oscillations in the final state calculation at the next moment. However, the calculation process of this method is simple and can reflect the motion law of the mine car. The accuracy of the data-driven model is higher than that of the pure physical model, but its generalization ability is limited by the limited samples and the large number of iterations (about 10-15 epochs, depending on the sample size). The present invention can effectively reflect the real state of motion and converge within a limited number of iterations (about 2-3 epochs). It can also be seen from the comparison results that under some harsh working conditions, the hybrid model still has strong robustness. Overall, due to the combination of the advantages of the physical model and the data-driven model, the hybrid model performs better.

[0057] The comparison results of the vertical velocities of the front and rear suspension spring positions calculated by the vertical velocity of the center of gravity (obtained by the integration of acceleration) and the pitch angular velocity in the fourth part of the working condition are as follows: Figure 8 and Fig. 9 As shown. It can be seen that after the integration process, the accuracy of the solution of the physical model is significantly improved and can represent the law of vehicle motion. However, like the data-driven model, the error is still large under specific or harsh working conditions. The hybrid model accurately reflects the sprung velocity change pattern in most cases, but in very few working conditions, the accuracy is not as good as the data-driven model. The reason for this phenomenon may be that only two continuous states are used in the hybrid model, and due to sensor noise, the physical model may amplify the predicted value, resulting in more significant prediction errors at individual points. But overall, the calculation accuracy of the hybrid model is higher than that of the other two models, and its prediction results are relatively more reliable.

[0058] In order to give a qualitative comparison of the three models, the present invention uses the MSE of all test sets as a measurement indicator. The smaller the value, the higher the accuracy of the model. c , front suspension spring speed v f and the rear suspension sprung velocity v r The MSE comparison results are shown in Table 3.

[0059] Table 3 Comparison of MSE in experiments

[0060] MSEvalue Physicsmodel Data-driven model Hybridmodel <![CDATA[a c (m / s 2 ) 2 ]]> 13.69 0.18 0.13 <![CDATA[v f (×10 -2 m / s) 2 ]]> 2.36 0.76 0.32 <![CDATA[v r (×10 -2 m / s 2 ) 2 ]]> 2.23 0.78 0.31

[0061] It can be seen that the present invention has good performance in predicting different motion parameters. It is worth mentioning that the hybrid model method of the present invention only relies on the body sensor signal and does not require unsprung information. This advantage makes it possible to directly predict the dynamic characteristics of the mine car without building a complex mine car sensor measurement system.

[0062] In addition, hydro-pneumatic suspension is particularly common on mining trucks due to its excellent durability. However, due to its design, the modeling of the nonlinear stiffness and damping characteristics of hydro-pneumatic suspension is challenging and difficult to directly apply to the calculation of truck dynamic characteristics.

[0063] The invention provides the suspension parameters of the mining truck, which is very important for the structural design and performance prediction of the mining truck. Fig.10 and Fig.11 As shown, the changing rules of front and rear suspension stiffness and damping coefficient.

[0064] When the mining truck accelerates, i.e., in conditions 1 and 2, the rear suspension compresses and the front suspension tensions. When the hydro-pneumatic suspension is compressed, the internal gas pressure increases and the stiffness increases, and vice versa. Therefore, if Fig.10As shown, the stiffness of the rear suspension should be higher than that of the front suspension. When the mining truck decelerates, the stiffness of the front suspension is significantly more significant. In addition, when observing the changes in suspension stiffness during the acceleration and deceleration switching of the mining truck, it is not difficult to find that the front suspension has undergone significant changes. This phenomenon is because the cross-section and internal gas mass of the front suspension are relatively small, so under the same volume and pressure change rate (isothermal assumption), its stiffness changes significantly. Since the rear suspension mainly bears the load-bearing function and its cross-section is larger, the suspension stiffness changes less when the working conditions are switched. Generally, the front and rear suspension stiffness change regulations should be opposite.

[0065] From the damping coefficient, as the speed of the mine car increases, its vertical movement will also increase. At this time, the speed at which the liquid in the oil-gas suspension flows through the throttle hole will also tend to increase. According to fluid mechanics theory, the damping will increase in this case, such as Fig.11 However, the factors affecting damping are very complex, so a more reasonable explanation is given here.

[0066] Compared with the existing technology, this method can solve the time-varying stiffness and damping values ​​of the front and rear suspensions of a mining truck running on an unstructured road only by using the IMU.

[0067] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A method for identifying time-varying characteristics of a heavy-duty vehicle suspension, characterized in that: The sensors collect the sequence states of mining truck operation, and the deep learning network predicts the parameter-related factors based on the sequence states. At the same time, the linear dynamics model considering the longitudinal-vertical coupling effect roughly estimates the stiffness and damping coefficients as key suspension parameters, and the nonlinear dynamics model considering the longitudinal-vertical coupling effect predicts the state of the next time step based on the parameter-related factors and key suspension parameters. The state loss Lp between the predicted state and the target state detected by the sensor is used as the loss error term of the deep learning network to update the network parameters; The linear dynamic model considering the longitudinal-vertical coupling effect includes: c The vehicle body with a rotational inertia of Ic and the unsprung mass arranged below it are m tf and m tr The front and rear axles of the vehicle body are: The suspension force transmitted from the front and rear axles to the vehicle body is F f and F r , the road excitation of the front and rear tires is z qf and z qr , the degrees of freedom (DoF) of the model include the center of gravity (CoG) z c The vertical displacement and pitch angle θ c and the front and rear unsprung masses z tf and z tr The vertical displacement, F c represents the inertial force acting on CoG due to the acceleration a and velocity v moving along the x-axis, M c represents the moment of inertia, where: the suspension stiffness and damping coefficient are k and c respectively, the subscripts f and r represent the front suspension and rear suspension respectively, the subscript t represents the tire, and the distance between CoG and the front / rear axle is a c and b c , the vertical distance between CoG and the pitch center PC is Δh.

2. The method for identifying the time-varying characteristics of a heavy-duty vehicle suspension according to claim 1, characterized in that: Calculate the vertical acceleration and velocity of the center of gravity based on the nonlinear mapping relationship of the inertial sensor Specifically include: 1) According to Euler's theorem of rigid body kinematics, the motion of a point in the absolute coordinate system r is decomposed into the translation displacement R in the relative coordinate system plus the vector sum of the rotation around the base point: Where: A is the coordinate rotation matrix, u i is the projection on the coordinate axis after rotation. OXYZ is an inertial reference system fixed on the ground, that is, an absolute reference system. i X m Y m Z m It is the coordinate system fixed on the inertial sensor, that is, the relative coordinate system, and the three-axis directions are the same as the inertial reference system; i X i Y i Z i Indicates the coordinate system definition of the inertial sensor itself, point P is the point to be measured, and u is the vector from the coordinate axis reference point to the measured point; 2) According to the actual situation, reasonably determine the rotation order of the coordinate axes and calculate the rotation matrix: According to the knowledge of vehicle kinematics, the roll angle and pitch angle are very small relative to the yaw angle. Therefore, the rotation order when selecting the coordinate axes is: first rotate around the Z axis corresponding to the yaw angle α c ; Then rotate around the Y axis to the corresponding pitch angle θ c ; Finally, rotate around the X axis to the corresponding rolling angle γ c , we get the rotation matrix expression A(α c ,θ c ,γ c )=A γ A β A α , where: the positive direction of rotation is the right-handed spiral direction, and the expressions of the three rotation matrices are 3) Calculate the velocity of point P and take the derivative of both sides of the expressions of the three rotation matrices with respect to time. The rotation matrix is ​​regular, and its derivative with time is calculated by linear algebra. Among them: Each axial angular velocity matrix is ​​an antisymmetric matrix Get the velocity expression of the measured point And the acceleration expression of the measured point, that is, the angular acceleration of each axis That is, the acceleration of any point rigidly connected to the installation position of the inertial sensor is calculated, thereby obtaining the movement of the vehicle body and completing the longitudinal-vertical dynamic model of the mining truck.

3. The method for identifying the time-varying characteristics of a heavy-duty vehicle suspension according to claim 1, wherein: The key suspension parameters, namely the suspension stiffness and damping coefficient, are obtained in the following way: Taking the front suspension as an example, under the force F f Under the input of , the vertical displacement z on the sprung spring is obtained according to the transmission relationship between the suspension force and the suspension displacement of the dynamic system of the heavy-load vehicle. f , the suspension stiffness is calculated by system identification method, and the unsprung position displacement is calculated, including: 1) According to the Laplace transform relationship between force and displacement Among them: All are defined above, the squiggles are the Laplace transforms of the corresponding variables; 2) By eliminating the vertical displacement of the unsprung mass, the transmission relationship between the suspension force and the suspension displacement of the dynamic system of the heavy-duty vehicle Where: s is the time domain differential operator. According to the measured real-time mine car timing signal, the continuous-time system simplified refined instrumental variable method (SRIVC) is used for system identification, and the system transfer equation is rewritten into the standard form of band-limited continuous-time white noise e(t): Among them: the parameter vector to be identified is θ = [a1, a2, a3, b0, b1, b2] T ; 3) The entire equation is low-pass filtered through a continuous-time state variable filter to obtain the input and output pre-filtered time derivatives required for identification. Since the regression vector after filtering It is related to the noise vector, so the recognition result is asymptotically biased. To eliminate the influence of the noise vector, the tool vector is iterated j times. Calculate the parameter estimates for iteration j+1 Continue iterating until the error between the recognition result and the previous recognition result is small enough, that is, 4) Based on the identified system transfer function and the original input-output relationship, the normalized root mean square error (NRMSE) and final prediction error (FPE) are used as evaluation indicators to evaluate the accuracy of the model.

4. The method for identifying the time-varying characteristics of a heavy-duty vehicle suspension according to claim 1, wherein: The state of the next time step is identified by combining data drive with a neural network to identify suspension parameters, specifically including: 1) Taking the front suspension as an example, Perform backward differentiation to obtain the suspension discrete system expression at time k: Where: superscript k and k-1 represent the current state and the last state, respectively, and the sampling time difference of each state; since the unsprung mass displacement is obtained through parameter identification, its differential is not accurate and needs to be identified in combination with other equations. The speed of the unsprung mass at time k is Where: The subscript new represents the modified parameters of the neural network. When the suspension force at time k-1 is introduced, the entire algorithm will iterate, resulting in the accumulation of errors each time. The forward differential of the vehicle DoF is introduced to represent the unsprung mass velocity at time k-1. Get unsprung mass After the speed By solving the function, we can get the force Ff transmitted by the front suspension at time k: 2) For the rear suspension, the equation is Calculate F at time k r , so the motion of the vehicle is According to the forward differentiation method, the predicted state of the vehicle body is summarized as follows: An iterative method to obtain the parameters of the mining truck body in discrete time; 3) The front and rear suspension parameters of mining trucks show nonlinear characteristics under different working conditions. Based on the previous parameter identification, neural networks are used for refined modeling, that is, the updated iterative parameters are Neural network based on time series signal Get N out =[P kf ,P kr ,P cf ,P cr ].

5. The method for identifying the time-varying characteristics of a heavy-duty vehicle suspension according to claim 1, characterized in that: The deep learning network includes: a three-layer LSTM network layer and a fully connected network layer, wherein: each layer of the LSTM network layer has 128 hidden units, and the fully connected network layer outputs the suspension stiffness damping correction coefficient.

Images