Method and apparatus for hybrid control of a regenerative suspension

Abstract

This application relates to a hybrid control method and apparatus for energy-regenerative suspension. The method includes: acquiring the current speed of a tracked vehicle and the smoothness of the road surface ahead; establishing an active mode system model and an energy-regenerative mode system model of the tracked vehicle in the ECU based on a NARX neural network; generating a first predicted system response of the active mode system model and a second predicted system response of the energy-regenerative mode system model; generating a first comprehensive evaluation index and a second comprehensive evaluation index that reflect the driving state of the tracked vehicle based on the filtered first and second predicted system responses as index parameters; and determining the optimal operating mode of the suspension system under different driving states by judging the first and second comprehensive evaluation indices based on preset judgment criteria. The apparatus using this method enables energy recovery from the suspension to compensate for the energy consumption of the control system while meeting performance requirements.

Description

[Technical Field]

[0001] This application relates to the field of hybrid control technology for energy-feeding suspension, and in particular to a hybrid control method and device for energy-feeding suspension. [Background Technology]

[0002] Tracked vehicles often travel on off-road terrain with highly complex conditions, resulting in significant vibration amplitudes. Furthermore, tracked vehicles are heavy and have multiple points of contact with the road surface, representing a substantial potential for energy recovery. However, vehicle handling stability, ride comfort, and energy recovery characteristics are mutually restrictive. A single-mode semi-active control system cannot reconcile the conflict between ride comfort and handling stability, while active control suffers from excessive energy consumption. A single-mode control strategy is ill-suited to the complex and varied operating environment of the suspension system.

[0003] In tracked vehicles, the suspension system is a crucial component, and its performance often determines the overall combat performance. However, current research on suspension systems often overlooks their impact on the firepower and mobility of tracked vehicles. This means a lack of comprehensive consideration of the suspension's influence on firing accuracy, comfort, constraints, and reliability. Consequently, it's difficult to measure the corresponding performance indicators of the tracked vehicle's suspension system by selecting an appropriate response, leading to the suspension system operating outside its optimal mode and severely impacting the overall combat performance. [Summary of the Invention]

[0004] To address the problem that the impact of suspension systems on tracked vehicle performance is often overlooked during suspension system research, making it difficult to measure the corresponding performance indicators of tracked vehicle suspension systems and resulting in the suspension system not operating in its optimal mode, which seriously affects the overall vehicle's combat performance indicators, this invention obtains the current speed of the tracked vehicle and the smoothness of the road surface ahead. Based on the NARX neural network and ECU, it establishes an active mode system model and a regenerative mode system model for the tracked vehicle. Then, it generates the first predicted system response of the active mode system model and the second predicted system response of the regenerative mode system model and filters them to generate a first comprehensive evaluation index and a second comprehensive evaluation index that reflect the driving state of the tracked vehicle, thereby determining the optimal operating mode of the suspension system under different driving states.

[0005] The present invention proposes the following solution:

[0006] A hybrid control method for energy-feeding suspension includes:

[0007] Obtain the current speed of the tracked vehicle and the smoothness of the road surface ahead;

[0008] Based on the NARX neural network, an active mode system model and a regenerative mode system model of the tracked vehicle are established in the ECU.

[0009] Based on the current speed of the tracked vehicle and the smoothness of the road surface ahead, the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model are generated.

[0010] Based on the filtered responses of the first and second prediction systems, a first comprehensive evaluation index and a second comprehensive evaluation index are generated to reflect the driving status of the tracked vehicle.

[0011] Based on preset judgment criteria, the first comprehensive evaluation index and the second comprehensive evaluation index are judged to determine the optimal working mode of the suspension system under different driving conditions.

[0012] The above-described energy-recharge suspension hybrid control method, specifically the step of establishing the active mode system model and the energy-recharge mode system model of the tracked vehicle in the ECU based on the NARX neural network, includes:

[0013] The system input of the tracked vehicle is simulated by controlling the vibrator when the tracked vehicle is traveling at different speeds and on different roads, and a sample library of active mode and energy feeding mode is established.

[0014] Based on the NARX neural network, the active mode sample library, and the power feeding mode sample library, a first gray box model of the active mode prediction response and a second gray box model of the power feeding mode prediction response are established.

[0015] Based on the first gray box model and the second gray box model, determine the error of the NARX neural network model on the validation set and the error on the test set;

[0016] Based on the errors in the validation set and the errors in the test set, a first error correction model and a second error correction model are established.

[0017] Based on the first gray box model and the first error correction model, an active mode system model is generated;

[0018] Based on the second gray box model and the second error correction model, a power feeding mode system model is generated.

[0019] The above-described energy-feeding suspension hybrid control method, comprising the steps of establishing a first gray-box model of the active mode predictive response and a second gray-box model of the energy-feeding mode predictive response based on a NARX neural network, an active mode sample library, and an energy-feeding mode sample library, includes:

[0020] Train an open-loop NARX model based on the training set, and generate a closed-loop NARX model.

[0021] The closed-loop NARX model is validated based on the validation set to determine its predictive performance.

[0022] When the prediction performance reaches a satisfactory level, the accuracy of the closed-loop NARX model is tested based on the test set.

[0023] If the accuracy does not meet the requirements, retrain the open-loop NARX model.

[0024] The energy-feeding suspension hybrid control method described above, wherein the step of generating a first comprehensive evaluation index and a second comprehensive evaluation index of the tracked vehicle's driving state based on the filtered first and second predicted system responses as index parameters, includes:

[0025] Obtain the pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber;

[0026] By integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber, a first comprehensive evaluation index and a second comprehensive evaluation index are generated to provide feedback on the driving status of the tracked vehicle.

[0027] The energy-feeding suspension hybrid control method described above, in the step of integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber to generate a first comprehensive evaluation index and a second comprehensive evaluation index for the driving state of the tracked vehicle, includes:

[0028] The method for determining the pitch vibration angular velocity is as follows:

[0029] ;

[0030] ;

[0031] In the formula, D 1l The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the n sets of vibration angular velocities collected over a time period T. The root mean square value, This refers to the value that the vibration angular velocity should satisfy during the time delay Δt between the gunner firing and the projectile leaving the barrel, when a moving tank is at distance D and aims at a target of height H, so that the deviation of the projectile's height at the point of impact does not exceed the target's range.

[0032] The method for determining the vertical acceleration of the driver's seat is as follows:

[0033]

[0034] In the formula, D 2lIt is an evaluation of the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents n sets of driver seat vertical accelerations collected over a period of time T. The root mean square value after frequency f weighting This represents the effective value of the vertical acceleration of the tank above the first wheel when the tank is traveling on a Class A road surface.

[0035] The method for determining the duration of driver's seat vibration is as follows:

[0036]

[0037] In the formula, D 3l It is an evaluation index with standardized dimensions of cumulative driving time;

[0038] The method for determining the average damping power generated by the shock absorber is as follows:

[0039]

[0040] In the formula D 4l The maximum average damping power N of the hydraulic shock absorber Dmax Evaluation index after dimensionless measurement, N s This refers to the heat dissipation power of the hydraulic shock absorber;

[0041] The method for integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber is as follows:

[0042] h l =(ω1D 1l +ω2D 2l +ω3D 3l +ω4D 4l ) / 4

[0043] In the formula, ω1~ω4 are D 1l ~D 4l The weighting coefficients. Let h be the comprehensive evaluation index in the active mode. l The comprehensive evaluation index under the power feeding mode is h' l It can be known that h l <h' l .

[0044] The energy-rechargeable suspension hybrid control method described above, wherein the step of determining the optimal operating mode of the suspension system under different driving conditions based on the first comprehensive evaluation index and the second comprehensive evaluation index according to preset judgment criteria includes:

[0045] Based on the binding indicators, the feed voltage and feed efficiency of the energy recovery unit, a preset judgment criterion is generated;

[0046] Based on preset judgment criteria, the first comprehensive evaluation index, and the second comprehensive evaluation index, the optimal working mode of the suspension system under different driving conditions is determined.

[0047] The energy-feeding suspension hybrid control method described above, wherein the step of generating preset judgment criteria based on constraint indicators, energy-feeding voltage and energy-feeding efficiency of the energy recovery unit includes:

[0048] Obtain the dynamic deflection of the suspension;

[0049] Based on the dynamic deflection of the suspension, in the energy-gathering mode, it is determined whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the dynamic stroke of the road wheel. If so, it is determined that the active suspension is applicable to the driving state.

[0050] If the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel, compare the energy supply voltage with the voltage threshold of the energy storage device. If the energy supply voltage is less than the voltage threshold of the energy storage device, determine that the active suspension is suitable for the driving state.

[0051] If the energy supply voltage is greater than the voltage threshold of the energy storage device, the energy supply efficiency is compared with the preset energy supply efficiency threshold. If the energy supply efficiency is less than the preset energy supply efficiency threshold, it is determined that the driving state is suitable for active suspension. Otherwise, the energy supply suspension is used to put the tracked vehicle in a semi-active working condition.

[0052] A power-feeding suspension hybrid control device, comprising:

[0053] The acquisition module is used to acquire the current speed of the tracked vehicle and the smoothness of the road surface ahead;

[0054] A module is established to build active mode system model and energy feeding mode system model of tracked vehicle in ECU based on NARX neural network;

[0055] The generation module is used to generate the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model based on the current speed of the tracked vehicle and the smoothness of the road surface ahead.

[0056] The filtering module is used to generate a first comprehensive evaluation index and a second comprehensive evaluation index that provide feedback on the driving status of the tracked vehicle, based on the filtered responses of the first and second prediction systems as index parameters.

[0057] The determination module is used to determine the first comprehensive evaluation index and the second comprehensive evaluation index based on preset judgment criteria, and to determine the optimal working mode of the suspension system under different driving conditions.

[0058] The energy-feeding suspension hybrid control device described above, wherein the establishment module includes:

[0059] The first establishment unit is used to control the system input of the exciter to simulate the tracked vehicle traveling at different speeds and on different roads, and to establish a sample library of active mode and a sample library of energy feeding mode.

[0060] The second establishment unit is used to establish a first gray box model of the active mode prediction response and a second gray box model of the energy-feeding mode prediction response based on the NARX neural network, the active mode sample library and the energy-feeding mode sample library.

[0061] The first determining unit is used to determine the error of the NARX neural network model on the validation set and the error on the test set based on the first gray box model and the second gray box model.

[0062] The third establishment unit is used to establish a first error correction model and a second error correction model based on the error in the validation set and the error in the test set.

[0063] The first generation unit is used to generate an active mode system model based on the first gray box model and the first error correction model.

[0064] The second generation unit is used to generate a power feeding mode system model based on the second gray box model and the second error correction model.

[0065] The second establishment unit includes:

[0066] Generate a sub-unit, used to train an open-loop NARX model based on the training set, and generate a closed-loop NARX model;

[0067] A subunit is determined for verifying the closed-loop NARX model based on a validation set, and for determining the predictive performance of the closed-loop NARX model.

[0068] The prediction subunit is used to test the accuracy of the closed-loop NARX model based on the test set when the prediction performance reaches a satisfactory level.

[0069] The training subunit is used to retrain the open-loop NARX model if the accuracy does not meet the requirements.

[0070] The filtering module includes:

[0071] The acquisition unit is used to acquire pitch vibration angular velocity, driver seat vertical acceleration, driver seat vibration frequency, driver seat vibration duration, and average damping power generated by the shock absorber.

[0072] The third generation unit is used to integrate pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration and average damping power generated by shock absorber to generate the first comprehensive evaluation index and the second comprehensive evaluation index of the tracked vehicle's driving status.

[0073] The third generation unit includes:

[0074] The method for determining the pitch vibration angular velocity is as follows:

[0075] ;

[0076] ;

[0077] In the formula, D 1l The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the n sets of vibration angular velocities collected over a time period T. The root mean square value, This refers to the value that the vibration angular velocity should satisfy during the time delay Δt between the gunner firing and the projectile leaving the barrel, when a moving tank is at distance D and aims at a target of height H, so that the deviation of the projectile's height at the point of impact does not exceed the target's range.

[0078] The method for determining the vertical acceleration of the driver's seat is as follows:

[0079]

[0080] In the formula, D 2l It is an evaluation of the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents n sets of driver seat vertical accelerations collected over a period of time T. The root mean square value after frequency f weighting This represents the effective value of the vertical acceleration of the tank above the first wheel when the tank is traveling on a Class A road surface.

[0081] The method for determining the duration of driver's seat vibration is as follows:

[0082]

[0083] In the formula, D 3l It is an evaluation index with standardized dimensions of cumulative driving time;

[0084] The method for determining the average damping power generated by the shock absorber is as follows:

[0085]

[0086] In the formula D 4l The maximum average damping power N of the hydraulic shock absorber Dmax Evaluation index after dimensionless measurement, N s This refers to the heat dissipation power of the hydraulic shock absorber;

[0087] The method for integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber is as follows:

[0088] h l =(ω1D 1l +ω2D 2l +ω3D 3l +ω4D 4l ) / 4

[0089] In the formula, ω1~ω4 are D 1l ~D 4l The weighting coefficients. Let h be the comprehensive evaluation index in the active mode. l The comprehensive evaluation index under the power feeding mode is h' l It can be known that h l <h' l ;

[0090] The determining module includes:

[0091] The fourth generation unit is used to generate preset judgment criteria based on the constraint indicators, the feed voltage and feed efficiency of the energy recovery unit;

[0092] The second determining unit is used to determine the optimal working mode of the suspension system under different driving conditions based on preset judgment criteria, the first comprehensive evaluation index and the second comprehensive evaluation index.

[0093] The fourth generation unit includes:

[0094] Obtain the sub-unit, used to obtain the dynamic deflection of the suspension;

[0095] The judgment subunit is used to determine, in the energy feeding mode, whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the road wheel dynamic stroke based on the dynamic deflection of the suspension. If so, it determines that the driving state is suitable for active suspension.

[0096] The first comparison subunit is used to compare the energy supply voltage with the voltage threshold of the energy storage device if the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel. If the energy supply voltage is less than the voltage threshold of the energy storage device, the active suspension is determined to be suitable for the driving state.

[0097] The second comparison subunit is used to compare the energy feeding efficiency with a preset energy feeding efficiency threshold if the energy feeding voltage is greater than the voltage threshold of the energy storage device. If the energy feeding efficiency is less than the preset energy feeding efficiency threshold, it determines that the driving state is suitable for active suspension. Otherwise, it applies energy feeding suspension so that the tracked vehicle is in a semi-active working condition.

[0098] A computer-readable storage medium storing a computer program that, when executed by a power supply suspension hybrid control device, implements the power supply suspension hybrid control method as described above.

[0099] This invention, through obtaining the current speed of the tracked vehicle and the smoothness of the road surface ahead, establishes an active mode system model and an energy-replenishing mode system model of the tracked vehicle based on the NARX neural network and ECU. Then, it generates a first predicted system response of the active mode system model and a second predicted system response of the energy-replenishing mode system model and filters them to generate a first comprehensive evaluation index and a second comprehensive evaluation index that reflect the driving state of the tracked vehicle. This determines the optimal working mode of the suspension system under different driving states and switches the optimal working mode, so that while meeting performance requirements, energy recovery of the suspension can also be achieved to compensate for the energy consumption of the control system. [Attached Image Description]

[0100] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0101] Figure 1 This is a flowchart of the energy-feeding suspension hybrid control method according to the first embodiment of the present invention;

[0102] Figure 2 yes Figure 1 Detailed flowchart of step S12;

[0103] Figure 3 yes Figure 2 Detailed flowchart of step S122;

[0104] Figure 4 yes Figure 1 Detailed flowchart of step S14;

[0105] Figure 5 yes Figure 1 Detailed flowchart of step S15;

[0106] Figure 6 yes Figure 5Detailed flowchart of step S151;

[0107] Figure 7 This is a block diagram of the mode switching control system of the energy-feeding suspension hybrid control method according to the first embodiment of the present invention;

[0108] Figure 8 This is a structural block diagram of the energy-feeding suspension hybrid control device according to the second embodiment of the present invention;

[0109] Figure 9 yes Figure 8 Create a detailed structural diagram of the module;

[0110] Figure 10 yes Figure 9 Detailed structural block diagram of the second building unit;

[0111] Figure 11 yes Figure 8 Detailed structural diagram of the filtering module;

[0112] Figure 12 yes Figure 8 The detailed structural block diagram of the module is determined in the middle;

[0113] Figure 13 yes Figure 12 Detailed structural block diagram of the fourth generation unit.

Detailed Implementation Methods

[0114] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Well-known modules, units, and their connections, links, communications, or operations are not shown or described in detail. Furthermore, the described features, architectures, or functions can be combined in any way in one or more embodiments. Those skilled in the art should understand that the various embodiments described below are only for illustrative purposes and not for limiting the scope of protection of the present invention. It is also readily understood that the modules, units, or processing methods in the various embodiments described herein and shown in the accompanying drawings can be combined and designed in various different configurations. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0115] The definitions of various terms or methods used in the following embodiments are, except where logically impossible, generally defined as broad concepts that can be implemented under the premise of the content disclosed in the embodiments. Under this understanding, all specific subordinate limitations of the terms or methods should be considered as part of the invention and should not be narrowly interpreted or biased simply because the specification does not disclose such a specific limitation. Similarly, provided that it is logically feasible, the order of the steps in the method is flexible and varied, and all specific subordinate limitations in the broad concepts of various terms or methods fall within the scope of protection of this invention.

[0116] First embodiment:

[0117] Please refer to Figures 1 to 7 As shown, this embodiment proposes a hybrid control method for energy-feeding suspension, including S11-S15, wherein:

[0118] S11. Obtain the current speed of the tracked vehicle and the smoothness of the road surface ahead.

[0119] In this embodiment, the current speed of the tracked vehicle and the smoothness of the road surface ahead are obtained through a sensor module. The sensor module includes: a speed sensor, a high-definition camera 1, a high-definition camera 2, a gyroscope, an accelerometer, a velocity sensor, and a displacement sensor. The speed sensor is installed on the first pair of road wheels to detect the travel speed; the two high-definition cameras are respectively installed on the left and right sides of the front composite deck of the tracked vehicle to obtain images of the road surface ahead; the gyroscope is installed at the vehicle's center of gravity to obtain the pitch vibration angular velocity of the vehicle; the accelerometer is installed below the surface of the driver's seat to obtain the vertical acceleration and vibration frequency at the driver's location; the velocity sensor is installed on each road wheel to obtain the dynamic velocity of the road wheel relative to the vehicle body; and the displacement sensor is installed on each road wheel to obtain the dynamic stroke of the road wheel.

[0120] In this embodiment, two high-definition cameras and a vehicle speed sensor are connected to the ECU via signal lines. A method based on visual sensing technology is used to obtain road surface unevenness. To obtain sufficient road information even under adverse driving conditions, a high-definition camera is installed on each side of the front composite deck of the tracked vehicle. The left camera is offset to the right at a certain angle, and the right camera is offset to the left at a certain angle. The angle between the left camera and the horizontal plane is negative, while the angle between the right camera and the horizontal plane is positive. This camera arrangement ensures sufficient road surface information regardless of the tracked vehicle's posture. This information is further processed in the ECU to identify road surface unevenness, and the current vehicle speed is obtained through the vehicle speed sensor.

[0121] S12. Based on the NARX neural network, establish the active mode system model and the energy feeding mode system model of the tracked vehicle in the ECU.

[0122] This embodiment is based on the NARX neural network to establish sample libraries for active mode and energy recovery mode. Based on these, a first gray box model and a second gray box model are established. Based on the error of the validation set and the error in the test set, a first error correction model and a second error correction model are established. In this way, the active mode system model and the energy recovery mode system model of the tracked vehicle are established in the ECU. The simulation reliability is stronger, which enables the energy recovery of the suspension to compensate for the energy consumption of the control system while meeting the performance requirements.

[0123] As a preferred option rather than a specific limitation, step S12 includes S121-S126, wherein:

[0124] S121. Control the exciter to simulate the system input of the tracked vehicle when it travels at different speeds and on different roads, and establish a sample library of active mode and a sample library of energy feeding mode.

[0125] This embodiment uses sensors mounted on the tracked vehicle to obtain pitch vibration angular velocity, vertical acceleration and vibration frequency at the driver's location, the speed of the road wheels relative to the vehicle body, and the travel distance of the road wheels. Finally, it establishes sample libraries for both active and regenerative braking modes. The simulated test scenarios include the tracked vehicle traveling at speeds of 5, 15, 25, and 35 km / h on random road surfaces and A- to H grade road surfaces, with each road segment travel time being 15 seconds. The data characteristics include both random road spectrum and standard road surface data, resulting in comprehensive sample libraries for both active and regenerative braking modes.

[0126] S122. Based on the NARX neural network, the sample library of active mode and the sample library of power feeding mode, establish the first gray box model of active mode prediction response and the second gray box model of power feeding mode prediction response.

[0127] In this embodiment, 70% of the time-series sample library is used as the training set, 15% as the validation set, and 15% as the test set. First, an open-loop NARX model is trained on the training set. After training, the open-loop NARX is converted to a closed-loop model. Then, the prediction performance of this closed-loop NARX on the validation set is examined, and the decision to use the NARX network is based on the prediction results. If the prediction results are satisfactory, its accuracy on the test set is further examined to prevent overfitting. Otherwise, the NARX is retrained or its topology is optimized until it meets the requirements on both the validation and test sets.

[0128] In this embodiment, road surface unevenness and vehicle speed are selected as inputs to the NARX neural network. Pitch vibration angular velocity, vertical acceleration and vibration frequency at the driver's location, speed of the road wheel relative to the vehicle body, and road wheel travel are selected as outputs. The maximum order of input and output delay is set to 2, the number of hidden layers is 1, and the number of nodes in the hidden layer is:

[0129]

[0130] In the formula, m is the number of output layer nodes, which is set to 4; n is the number of input layer nodes, which is set to 2; and a is an integer between 0 and 10. The number of nodes in the hidden layer is set to 6.

[0131] The training steps for the NARX neural network are as follows:

[0132] a. Network initialization

[0133] b. Calculate the output of the hidden layer

[0134] c. Calculate the output of the output layer

[0135] d. Error Calculation

[0136] e. Weighting Update

[0137] f. Threshold update

[0138] g. Determine if the iteration has ended; otherwise, return to step b.

[0139] For step g, the root mean square error (RMSE) is used to determine whether the iteration has ended. RMSE is the square root of the ratio of the sum of squares of the deviations between the predicted and true values ​​to the number of observations n. If the output RMSE is less than a preset value δ, training stops after the prediction accuracy on the training set reaches the termination condition. After the termination condition is reached, the current weights are the final weights, and training is complete.

[0140] As a preferred option rather than a specific limitation, step S122 includes S1221-S1224, wherein:

[0141] S1221. Train an open-loop NARX model based on the training set to generate a closed-loop NARX model.

[0142] In this embodiment, the open-loop NARX model is continuously trained using a training set. After training is complete, the open-loop NARX model is transformed into a closed-loop model to make the NARX model more stable.

[0143] S1222. Validate the closed-loop NARX model based on the validation set, and determine the predictive performance of the closed-loop NARX model.

[0144] This embodiment validates the closed-loop NARX model using a validation set, performs performance prediction, and decides whether to adopt the NARX network based on the prediction results, effectively filtering out NARX models that do not meet the requirements.

[0145] S1223. When the prediction performance reaches a satisfactory level, test the accuracy of the closed-loop NARX model based on the test set.

[0146] If the prediction results are satisfactory, the accuracy on the test set is checked to ensure that the model's accuracy meets the requirements without overfitting, thus making the model more stable.

[0147] S1224. If the accuracy does not meet the requirements, retrain the open-loop NARX model.

[0148] In this embodiment, when the accuracy requirement is not met, the NARX model is repeatedly trained and verified, which can better establish the corresponding NARX model and achieve better simulation results.

[0149] S123. Based on the first gray box model and the second gray box model, determine the error of the NARX neural network model on the validation set and the error on the test set.

[0150] After the NARX neural network trained in this embodiment meets the requirements on the validation set and the test set, the error of its output on these two sample sets is calculated to determine whether the NARX neural network model is the required model, so as to improve the simulation effect and enable the energy recovery of the suspension to compensate for the energy consumption of the control system while meeting the performance requirements.

[0151] S124. Based on the error in the validation set and the error in the test set, establish a first error correction model and a second error correction model.

[0152] In this embodiment, the error and the sample input are Fourier transformed to the frequency domain. The ratio of the error to the sample input in the frequency domain yields an equation about the error, thus establishing an error correction model to ensure high accuracy and reliability.

[0153] S125. Generate an active mode system model based on the first gray box model and the first error correction model.

[0154] In this embodiment, the first gray box model obtained based on the NARX neural network is superimposed with the first error correction model to obtain the final active mode system model. By inputting the road surface unevenness and vehicle speed into the final model, the response can be predicted relatively accurately.

[0155] S126. Generate the energy feeding mode system model based on the second gray box model and the second error correction model.

[0156] In this embodiment, the second gray box model obtained based on the NARX neural network is superimposed with the second error correction model to obtain the final energy feeding mode system model. By inputting the road surface unevenness and vehicle speed into the final model, the response can be predicted relatively accurately.

[0157] S13. Based on the current speed of the tracked vehicle and the smoothness of the road surface ahead, generate the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model.

[0158] This embodiment predicts and calculates the active mode system model and the energy feeding mode system model based on the current speed of the tracked vehicle and the smoothness of the road surface ahead, and obtains the first predicted system response and the second predicted system response as the selection index, which can be more accurate and reliable.

[0159] S14. Based on the filtered first prediction system response and second prediction system response as indicator parameters, generate a first comprehensive evaluation index and a second comprehensive evaluation index to reflect the driving status of the tracked vehicle.

[0160] In this embodiment, the first and second prediction system responses, after being screened, are used as indicator parameters. After evaluating the indicators, the first and second comprehensive evaluation indicators are determined, which can better determine the optimal working mode of the suspension system under different driving conditions.

[0161] As a preferred option rather than a specific limitation, step S14 includes S141-S142, wherein:

[0162] S141. Obtain the pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber.

[0163] This embodiment takes into account the requirements for tank gun firing accuracy, comfort, and constraints, and selects the pitch vibration angular velocity. To reflect the impact of vibration on artillery firing accuracy; select the vertical acceleration of the driver's seat. Comfort indicators are measured by the driver's seat vibration frequency f and the driver's seat vibration duration t; for constraint indicators, considering whether the heat capacity of the hydraulic shock absorber can meet the requirements of the suspension system, the average damping power generated by the shock absorber is selected as the measure.

[0164] S142. By integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber, a first comprehensive evaluation index and a second comprehensive evaluation index are generated to provide feedback on the driving status of the tracked vehicle.

[0165] In this embodiment, the five parameters in step S142 are fused and processed to obtain their respective comprehensive evaluation indicators in different working modes. Within a time period T, the system model that inputs the collected inputs into the ECU can obtain n sets of estimated response values. In subsequent steps, these n sets of data are processed, and each additional set of data is processed once.

[0166] In order to facilitate the direct fusion of parameters, this embodiment requires the dimension unification of each parameter.

[0167] As a preferred option rather than a specific limitation, the method for determining the pitch vibration angular velocity is specifically as follows:

[0168] ;

[0169] ;

[0170] In the formula, D 1l The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the n sets of vibration angular velocities collected over a time period T. The root mean square value, This refers to the value that the vibration angular velocity must satisfy during the time delay Δt between the gunner firing the gun and the projectile leaving the barrel, when a moving tank is at a distance D and aims at a target of height H, so that the deviation of the projectile at the point of impact does not exceed the target's range.

[0171] As a preferred option rather than a specific limitation, the determination of the driver's seat vertical acceleration And the method for determining the vibration frequency f, specifically:

[0172]

[0173] In the formula, D 2l This is an evaluation based on the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents n sets of driver seat vertical accelerations collected over a period of time T. The root mean square value after frequency f weighting This represents the effective value of the vertical acceleration of the tank hull above the first wheel when the tank is traveling on a Class A road surface.

[0174] The driver's seat vertical acceleration is collected in the ECU, and its time history f(t) is obtained. A Fourier transform is performed to obtain F(ω), which is then multiplied by the frequency weighting function W(f) to obtain W(f)F(ω). Finally, an inverse Fourier transform is performed on this product to obtain the time history f of the frequency-weighted acceleration. w (t), which can be obtained through further calculation.

[0175]

[0176] As a preferred option rather than a specific limitation, the method for determining the duration of driver's seat vibration is specifically as follows:

[0177]

[0178] In the formula, D 3l It is an evaluation index after the cumulative driving time is standardized in terms of dimensions.

[0179] As a preferred option and not a specific limitation, the method for determining the average damping power generated by the shock absorber is as follows:

[0180]

[0181] In the formula D 4l The maximum average damping power N of the hydraulic shock absorber Dmax Evaluation index after dimensionless measurement, N s This refers to the heat dissipation power of the hydraulic shock absorber.

[0182] For hydraulic shock absorbers, the damping force F generated on the suspension xd It is the speed of the road wheels relative to the vehicle body. Therefore, the damping force can be approximately expressed as a function of , and thus the damping force can be approximated as .

[0183]

[0184] In the formula, c is the damping coefficient of the suspension system. This refers to the speed of the road wheel relative to the vehicle body when the hydraulic shock absorber valve is opened.

[0185] Since each load-bearing wheel in the vibration system model obtains n sets of data within time T, the average damping power of each hydraulic damper can be obtained. The one with the largest average damping power can be expressed as:

[0186]

[0187] The heat dissipation power of the hydraulic shock absorber can be obtained as follows:

[0188] N s =k T AΔT;

[0189] In the formula, k T Let ΔT be the heat dissipation coefficient of the vibration damper, A be the heat dissipation area, and ΔT be the difference between the surface temperature of the vibration damper and the ambient temperature.

[0190] As a preferred embodiment and not a specific limitation, the method for integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber is specifically as follows:

[0191] Each evaluation indicator is assigned a weight coefficient, and the indicators are then fused together after dimension normalization to obtain a comprehensive evaluation indicator.

[0192] h l =(ω1D 1l +ω2D 2l +ω3D 3l +ω4D 4l ) / 4;

[0193] In the formula, ω1~ω4 are D 1l ~D 4l The weighting coefficients. Let h be the comprehensive evaluation index in the active mode. l The comprehensive evaluation index under the power feeding mode is h' l It can be known that h l <h' l .

[0194] S15. Based on preset judgment criteria, determine the first comprehensive evaluation index and the second comprehensive evaluation index to determine the optimal working mode of the suspension system under different driving conditions.

[0195] In this embodiment, the obtained comprehensive evaluation index is input into the preset judgment criteria for multiple judgments. First, the comprehensive evaluation index h' is... l With a threshold h ε Compare, if h' l <h ε If the driving state is not suitable for passive suspension, the system will determine that passive suspension is appropriate. Otherwise, it will compare the difference between the comprehensive evaluation index of energy recovery mode and active mode with a preset value. If h' l -h l If the value is greater than Δh, the system determines that active suspension is suitable for this driving state; otherwise, it proceeds to the next step.

[0196] As a preferred option rather than a specific limitation, step S15 includes S151-S153, wherein:

[0197] S151. Generate preset judgment criteria based on the binding indicators, the feed voltage and feed efficiency of the energy recovery unit.

[0198] The constraint indicators described in this embodiment include ensuring that the dynamic deflection of the suspension does not exceed the allowable suspension travel, which is measured using the travel of the load-bearing wheels. Furthermore, a preset judgment criterion is generated by combining the feed voltage and feed efficiency of the energy recovery unit, ensuring accuracy and reliability.

[0199] As a preferred option rather than a specific limitation, step S151 includes S1511-S1514, wherein:

[0200] S1511. Obtain the dynamic deflection of the suspension.

[0201] This embodiment measures the dynamic deflection of the suspension by the travel of the load-bearing wheel, which is more accurate and reliable.

[0202] S1512. Based on the dynamic deflection of the suspension, in the energy-recharge mode, determine whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the dynamic stroke of the road wheel. If so, determine that the driving state is suitable for active suspension.

[0203] In this embodiment, to ensure that the probability of suspension breakdown is extremely low when the vehicle is traveling at speed u on a road surface of a specified grade, [x] should be present. d ≥3x rms , where [x d The travel distance for the first road wheel is designed to be x. rms This represents the root mean square value of the load-bearing wheel travel. In the regenerative braking mode, [x] d ]<3x rms If so, the system determines that active suspension is applicable to this driving state.

[0204] S1513. If the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel, compare the energy supply voltage with the voltage threshold of the energy storage device. If the energy supply voltage is less than the voltage threshold of the energy storage device, determine that the active suspension is suitable for the driving state.

[0205] In this embodiment, when the designed dynamic stroke of the first load-bearing wheel is less than three times the root mean square value of the dynamic stroke of the load-bearing wheel, the energy feeding voltage U is compared with the voltage threshold U of the energy storage device. ε Compare, if U ε If so, the system determines that active suspension is applicable to this driving state.

[0206] S1514. If the energy supply voltage is greater than the voltage threshold of the energy storage device, compare the energy supply efficiency with the preset energy supply efficiency threshold. If the energy supply efficiency is less than the preset energy supply efficiency threshold, determine that the driving state is suitable for active suspension. Otherwise, apply energy supply suspension to put the tracked vehicle in a semi-active working condition.

[0207] In this embodiment, when the feed voltage is greater than the voltage threshold of the energy storage device, the feed efficiency η is compared with a preset feed efficiency threshold η. ε Compare, if η < η ε If the system determines that the driving state is suitable for active suspension, then it will use energy-regenerating suspension. In this case, the tracked vehicle is in a semi-active operating condition.

[0208] ​S152. Based on preset judgment criteria, the first comprehensive evaluation index, and the second comprehensive evaluation index, determine the optimal working mode of the suspension system under different driving conditions.

[0209] This embodiment uses preset judgment criteria to determine the first comprehensive evaluation index and the second comprehensive evaluation index. By adopting active suspension, it can achieve energy recovery of the suspension while meeting performance requirements, so as to compensate for the energy consumption of the control system. This allows for better switching to the optimal working mode, ensuring that energy recovery of the suspension can compensate for the energy consumption of the control system while meeting performance requirements.

[0210] This embodiment obtains the current speed of the tracked vehicle and the smoothness of the road surface ahead. Based on the NARX neural network and ECU, it establishes an active mode system model and an energy recovery mode system model for the tracked vehicle. Then, it generates the first predicted system response of the active mode system model and the second predicted system response of the energy recovery mode system model and filters them to generate a first comprehensive evaluation index and a second comprehensive evaluation index that reflect the driving state of the tracked vehicle. This determines the optimal working mode of the suspension system under different driving states and switches the optimal working mode so that, while meeting performance requirements, energy recovery of the suspension can also be achieved to compensate for the energy consumption of the control system.

[0211] Second embodiment:

[0212] Please refer to Figures 8 to 13 As shown, this embodiment proposes a power-feeding suspension hybrid control device 100, including an acquisition module 110, an establishment module 120, a generation module 130, a filtering module 140, and a determination module 150, wherein:

[0213] The acquisition module 110, connected to the establishment module 120, is used to acquire the current speed of the tracked vehicle and the smoothness of the road surface ahead.

[0214] The establishment module 120, connected to the generation module 130, is used to establish the active mode system model and the energy feeding mode system model of the tracked vehicle in the ECU based on the NARX neural network.

[0215] As a preferred embodiment rather than a specific limitation, the establishment module 120 includes a first establishment unit 121, a second establishment unit 122, a first determination unit 123, a third establishment unit 124, a first generation unit 125, and a second generation unit 126, wherein:

[0216] The first establishment unit 121, connected to the second establishment unit 122, is used to control the system input of the exciter to simulate the tracked vehicle traveling at different speeds and on different roads, and to establish a sample library of active mode and a sample library of energy feeding mode.

[0217] The second establishment unit 122 is connected to the first determination unit 123 and is used to establish a first gray box model of the active mode prediction response and a second gray box model of the energy-feeding mode prediction response based on the NARX neural network, the active mode sample library and the energy-feeding mode sample library.

[0218] As a preferred embodiment rather than a specific limitation, the second establishment unit 122 includes a generation subunit 1221, a determination subunit 1222, a prediction subunit 1223, and a training subunit 1224, wherein:

[0219] The generation subunit 1221 is connected to the determination subunit 1222 and is used to train an open-loop NARX model based on the training set to generate a closed-loop NARX model.

[0220] Determine subunit 1222, which is connected to prediction subunit 1223, for verifying the closed-loop NARX model based on the validation set and determining the prediction performance of the closed-loop NARX model.

[0221] The prediction subunit 1223, connected to the training subunit 1224, is used to test the accuracy of the closed-loop NARX model based on the test set when the prediction performance reaches a satisfactory level.

[0222] Training subunit 1224 is used to retrain the open-loop NARX model if the accuracy does not meet the requirements.

[0223] The first determining unit 123 is connected to the third establishing unit 124 and is used to determine the error of the NARX neural network model on the validation set and the error on the test set based on the first gray box model and the second gray box model.

[0224] The third establishment unit 124, connected to the first generation unit 125, is used to establish a first error correction model and a second error correction model based on the error in the validation set and the error in the test set.

[0225] The first generation unit 125 is connected to the second generation unit 126 and is used to generate an active mode system model based on the first gray box model and the first error correction model.

[0226] The second generation unit 126 is used to generate a power feeding mode system model based on the second gray box model and the second error correction model.

[0227] The generation module 130, connected to the filtering module 140, is used to generate the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model based on the current speed of the tracked vehicle and the smoothness of the road surface ahead.

[0228] The filtering module 140, connected to the determining module 150, is used to generate a first comprehensive evaluation index and a second comprehensive evaluation index that provide feedback on the driving status of the tracked vehicle, based on the filtered first prediction system response and second prediction system response as indicator parameters.

[0229] As a preferred embodiment rather than a specific limitation, the filtering module 140 includes an acquisition unit 141 and a third generation unit 142, wherein:

[0230] The acquisition unit 141, connected to the third generation unit 142, is used to acquire pitch vibration angular velocity, driver seat vertical acceleration, driver seat vibration frequency, driver seat vibration duration, and average damping power generated by the shock absorber.

[0231] The third generation unit 142 is used to integrate pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration and average damping power generated by shock absorber to generate a first comprehensive evaluation index and a second comprehensive evaluation index for feedback on the driving status of tracked vehicle.

[0232] As a preferred option rather than a specific limitation, the third generation unit 142 includes:

[0233] The method for determining the pitch vibration angular velocity is as follows:

[0234] ;

[0235] ;

[0236] In the formula, D 1l The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the n sets of vibration angular velocities collected over a time period T. The root mean square value, This refers to the value that the vibration angular velocity must satisfy during the time delay Δt between the gunner firing the gun and the projectile leaving the barrel, when a moving tank is at a distance D and aims at a target of height H, so that the deviation of the projectile at the point of impact does not exceed the target's range.

[0237] The method for determining the vertical acceleration of the driver's seat is as follows:

[0238]

[0239] In the formula, D 2l It is an evaluation of the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents n sets of driver seat vertical accelerations collected over a period of time T. The root mean square value after frequency f weighting This represents the effective value of the vertical acceleration of the tank above the first wheel when the tank is traveling on a Class A road surface.

[0240] The method for determining the duration of driver's seat vibration is as follows:

[0241]

[0242] In the formula, D 3l It is an evaluation index after the cumulative driving time is standardized in terms of dimensions.

[0243] The method for determining the average damping power generated by the shock absorber is as follows:

[0244]

[0245] In the formula D 4l The maximum average damping power N of the hydraulic shock absorber Dmax Evaluation index after dimensionless measurement, N s This refers to the heat dissipation power of the hydraulic shock absorber.

[0246] The method for integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber is as follows:

[0247] h l =(ω1D 1l +ω2D 2l +ω3D 3l +ω4D 4l ) / 4;

[0248] In the formula, ω1~ω4 are D 1l ~D 4l The weighting coefficients. Let h be the comprehensive evaluation index in the active mode. l The comprehensive evaluation index under the power feeding mode is h' l It can be known that h l <h' l .

[0249] The determination module 150 is used to determine the first comprehensive evaluation index and the second comprehensive evaluation index based on preset judgment criteria, and to determine the optimal working mode of the suspension system under different driving conditions.

[0250] As a preferred embodiment rather than a specific limitation, the determining module 150 includes a fourth generating unit 151 and a second determining unit 152, wherein:

[0251] The fourth generation unit 151, connected to the second determination unit 152, is used to generate preset judgment criteria based on the constraint indicators, the energy recovery unit's feed voltage and feed efficiency.

[0252] As a preferred embodiment rather than a specific limitation, the fourth generation unit 151 includes an acquisition subunit 1511, a judgment subunit 1512, a first comparison subunit 1513, and a second comparison subunit 1514, wherein:

[0253] Acquisition subunit 1511, connected to judgment subunit 1512, is used to acquire the dynamic deflection of the suspension.

[0254] The judgment subunit 1512 is connected to the first comparison subunit 1513. It is used to determine, in the energy feeding mode, whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the dynamic stroke of the road wheel, based on the dynamic deflection of the suspension. If so, it is determined that the driving state is suitable for active suspension.

[0255] The first comparison subunit 1513 is connected to the second comparison subunit 1514. It is used to compare the energy supply voltage with the voltage threshold of the energy storage device if the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel. If the energy supply voltage is less than the voltage threshold of the energy storage device, it is determined that the driving state is suitable for active suspension.

[0256] The second comparison subunit 1514 is used to compare the energy feeding efficiency with a preset energy feeding efficiency threshold if the energy feeding voltage is greater than the voltage threshold of the energy storage device. If the energy feeding efficiency is less than the preset energy feeding efficiency threshold, it determines that the driving state is suitable for active suspension. Otherwise, it applies energy feeding suspension so that the tracked vehicle is in a semi-active working condition.

[0257] The second determining unit 152 is used to determine the optimal working mode of the suspension system under different driving conditions based on preset judgment criteria, the first comprehensive evaluation index and the second comprehensive evaluation index.

[0258] This embodiment obtains the current speed of the tracked vehicle and the smoothness of the road surface ahead. Based on the NARX neural network and ECU, it establishes an active mode system model and an energy recovery mode system model for the tracked vehicle. Then, it generates the first predicted system response of the active mode system model and the second predicted system response of the energy recovery mode system model and filters them to generate a first comprehensive evaluation index and a second comprehensive evaluation index that reflect the driving state of the tracked vehicle. This determines the optimal working mode of the suspension system under different driving states and switches the optimal working mode so that, while meeting performance requirements, energy recovery of the suspension can also be achieved to compensate for the energy consumption of the control system.

[0259] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0260] This invention also provides a computer storage medium storing a computer program that, when executed by a processor, implements a power supply suspension hybrid control method as described in the above embodiments.

[0261] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of each of the above-described embodiments of the energy-feeding suspension hybrid control method. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0262] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, terminal, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, RAM, ROM, magnetic disks, or optical disks.

[0263] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0264] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A hybrid control method for energy-feeding suspension, characterized in that, include: Obtain the current speed of the tracked vehicle and the smoothness of the road surface ahead; Based on the NARX neural network, an active mode system model and a regenerative mode system model of the tracked vehicle are established in the ECU. Based on the current speed of the tracked vehicle and the smoothness of the road surface ahead, the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model are generated. Based on the filtered responses of the first and second prediction systems, a first comprehensive evaluation index and a second comprehensive evaluation index are generated to reflect the driving status of the tracked vehicle. Based on preset judgment criteria, the first comprehensive evaluation index and the second comprehensive evaluation index are judged to determine the optimal working mode of the suspension system under different driving conditions.

2. The energy-feeding suspension hybrid control method according to claim 1, characterized in that, The steps of establishing the active mode system model and the energy-feeding mode system model of the tracked vehicle in the ECU based on the NARX neural network include: The system input of the tracked vehicle is simulated by controlling the vibrator when the tracked vehicle is traveling at different speeds and on different roads, and a sample library of active mode and energy feeding mode is established. Based on the NARX neural network, the active mode sample library, and the power feeding mode sample library, a first gray box model for predicting the active mode response and a second gray box model for predicting the power feeding mode response are established. Based on the first gray box model and the second gray box model, determine the error of the NARX neural network model on the validation set and the error on the test set; Based on the errors in the validation set and the errors in the test set, a first error correction model and a second error correction model are established. Based on the first gray box model and the first error correction model, an active mode system model is generated; Based on the second gray box model and the second error correction model, a power feeding mode system model is generated.

3. The energy-feeding suspension hybrid control method according to claim 2, characterized in that, The steps of establishing a first gray-box model for the active mode prediction response and a second gray-box model for the feed mode prediction response based on the NARX neural network, the active mode sample library, and the feed mode sample library include: Train an open-loop NARX model based on the training set, and generate a closed-loop NARX model. The closed-loop NARX model is validated based on the validation set to determine its predictive performance. When the prediction performance reaches a satisfactory level, the accuracy of the closed-loop NARX model is tested based on the test set. If the accuracy does not meet the requirements, retrain the open-loop NARX model.

4. The energy-feeding suspension hybrid control method according to claim 1, characterized in that, The step of generating a first comprehensive evaluation index and a second comprehensive evaluation index to reflect the driving status of the tracked vehicle based on the filtered first and second prediction system responses as index parameters includes: Obtain the pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber; By integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber, a first comprehensive evaluation index and a second comprehensive evaluation index are generated to provide feedback on the driving status of the tracked vehicle.

5. The energy-feeding suspension hybrid control method according to claim 4, characterized in that, The steps of integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber to generate a first comprehensive evaluation index and a second comprehensive evaluation index for the tracked vehicle's driving status include: The method for determining the pitch vibration angular velocity is as follows: In the formula, The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the root mean square value of n sets of vibration angular velocities collected over a time period T. This refers to the time delay between the gunner firing and the bullet leaving the barrel when a moving tank is at a distance D and aims at a target of height H. Within t, in order to ensure that the deviation of the projectile in the height of the impact point does not exceed the range of the target, the vibration angular velocity should meet the following value; The method for determining the vertical acceleration of the driver's seat is as follows: In the formula, It is an evaluation of the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents the root mean square value of the vertical acceleration Z of the driver's seat collected over a period of time T, weighted by frequency f. This represents the effective value of the vertical acceleration of the tank hull above the first wheel when the tank is traveling on a Class A road surface. The method for determining the duration of driver's seat vibration is as follows: In the formula, It is an evaluation index with standardized dimensions of cumulative driving time; The method for determining the average damping power generated by the shock absorber is as follows: In the formula It is the maximum average damping power of the hydraulic shock absorber. Evaluation indicators after dimensionless measurement This refers to the heat dissipation power of the hydraulic shock absorber; The method for integrating pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration, and average damping power generated by the shock absorber is as follows: In the formula, They are respectively The weighting coefficients are used to make the comprehensive evaluation index under the active mode as follows: The comprehensive evaluation index under the power feeding mode is It can be known that .

6. The energy-feeding suspension hybrid control method according to claim 1, characterized in that, The step of determining the optimal operating mode of the suspension system under different driving conditions by judging the first comprehensive evaluation index and the second comprehensive evaluation index based on preset judgment criteria includes: Based on the binding indicators, the feed voltage and feed efficiency of the energy recovery unit, a preset judgment criterion is generated; Based on preset judgment criteria, the first comprehensive evaluation index, and the second comprehensive evaluation index, the optimal working mode of the suspension system under different driving conditions is determined.

7. The energy-feeding suspension hybrid control method according to claim 6, characterized in that, The step of generating preset judgment criteria based on constraint indicators, the feed voltage and feed efficiency of the energy recovery unit includes: Obtain the dynamic deflection of the suspension; based on the dynamic deflection of the suspension, in the energy feeding mode, determine whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the dynamic stroke of the road wheel. If so, determine that the driving state is suitable for active suspension. If the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel, compare the energy supply voltage with the voltage threshold of the energy storage device. If the energy supply voltage is less than the voltage threshold of the energy storage device, determine that the active suspension is suitable for the driving state. If the energy supply voltage is greater than the voltage threshold of the energy storage device, the energy supply efficiency is compared with the preset energy supply efficiency threshold. If the energy supply efficiency is less than the preset energy supply efficiency threshold, it is determined that the driving state is suitable for active suspension. Otherwise, the energy supply suspension is used to put the tracked vehicle in a semi-active working condition.

8. A hybrid control device for energy-feeding suspension, characterized in that, include: The acquisition module is used to acquire the current speed of the tracked vehicle and the smoothness of the road surface ahead; A module is established to build active mode system model and energy feeding mode system model of tracked vehicle in ECU based on NARX neural network; The generation module is used to generate the first predicted system response of the active mode system model and the second predicted system response of the energy-feeding mode system model based on the current speed of the tracked vehicle and the smoothness of the road surface ahead. The filtering module is used to generate a first comprehensive evaluation index and a second comprehensive evaluation index that provide feedback on the driving status of the tracked vehicle, based on the filtered responses of the first and second prediction systems as index parameters. The determination module is used to determine the first comprehensive evaluation index and the second comprehensive evaluation index based on preset judgment criteria, and to determine the optimal working mode of the suspension system under different driving conditions.

9. The energy-feeding suspension hybrid control device according to claim 8, characterized in that, The establishment module includes: The first establishment unit is used to control the system input of the exciter to simulate the tracked vehicle traveling at different speeds and on different roads, and to establish a sample library of active mode and a sample library of energy feeding mode. The second establishment unit is used to establish a first gray box model of the active mode prediction response and a second gray box model of the energy-feeding mode prediction response based on the NARX neural network, the active mode sample library, and the energy-feeding mode sample library. The first determining unit is used to determine the error of the NARX neural network model on the validation set and the error on the test set based on the first gray box model and the second gray box model. The third establishment unit is used to establish a first error correction model and a second error correction model based on the error in the validation set and the error in the test set. The first generation unit is used to generate an active mode system model based on the first gray box model and the first error correction model. The second generation unit is used to generate a power feeding mode system model based on the second gray box model and the second error correction model. The second establishment unit includes: Generate sub-units to train open-loop NARX models based on the training set and generate closed-loop NARX models; A sub-unit is determined for verifying the closed-loop NARX model based on a validation set, and for determining the predictive performance of the closed-loop NARX model. The prediction subunit is used to test the accuracy of the closed-loop NARX model based on the test set when the prediction performance reaches a satisfactory level. Training subunits are used to retrain the open-loop NARX model if the accuracy requirements are not met. The filtering module includes: The acquisition unit is used to acquire pitch vibration angular velocity, driver seat vertical acceleration, driver seat vibration frequency, driver seat vibration duration, and average damping power generated by the shock absorber. The third generation unit is used to integrate pitch vibration angular velocity, driver's seat vertical acceleration, driver's seat vibration frequency, driver's seat vibration duration and average damping power generated by shock absorber to generate the first comprehensive evaluation index and the second comprehensive evaluation index of the tracked vehicle's driving status. The third generation unit includes: The method for determining the pitch vibration angular velocity is as follows: In the formula, The evaluation index represents the dimensionless value of the root mean square value of the vibration angular velocity. This represents the n sets of vibration angular velocities collected within a time period T. The root mean square value, This refers to the time delay between the gunner firing and the bullet leaving the barrel when a moving tank is at a distance D and aims at a target of height H. t inside, To ensure that the deviation of the projectile in the height of the impact point does not exceed the target's range, the vibration angular velocity should meet the following value; The method for determining the vertical acceleration of the driver's seat is as follows: In the formula, It is an evaluation of the root mean square value of the driver's seat vertical acceleration after frequency weighting and dimension normalization. This represents the root mean square value of the vertical acceleration Z of the driver's seat collected over a period of time T, weighted by frequency f. This represents the effective value of the vertical acceleration of the tank hull above the first wheel when the tank is traveling on a Class A road surface. The method for determining the duration of driver's seat vibration is as follows: In the formula, It is an evaluation index with standardized dimensions of cumulative driving time; The method for determining the average damping power generated by the shock absorber is as follows: In the formula It is the maximum average damping power of the hydraulic shock absorber. Evaluation indicators after dimensionless measurement This refers to the heat dissipation power of the hydraulic shock absorber; The data includes pitch vibration angular velocity, driver's seat vertical acceleration, and driver's seat vibration frequency. The method for determining the duration of driver's seat vibration and the average damping power generated by the shock absorber has In the formula, They are respectively The weighting coefficients are used to make the comprehensive evaluation index under the active mode as follows: The comprehensive evaluation index under the power feeding mode is It can be known that ; The determining module includes: The fourth generation unit is used to generate preset judgment criteria based on the constraint indicators, the feed voltage and feed efficiency of the energy recovery unit; The second determining unit is used to determine the optimal working mode of the suspension system under different driving conditions based on preset judgment criteria, the first comprehensive evaluation index and the second comprehensive evaluation index. The fourth generation unit includes: an acquisition subunit, used to acquire the dynamic deflection of the suspension; The judgment subunit is used to determine, in the energy feeding mode, whether the design dynamic stroke of the first road wheel is less than three times the root mean square value of the road wheel dynamic stroke based on the dynamic deflection of the suspension. If so, it determines that the driving state is suitable for active suspension. The first comparison subunit is used to compare the energy supply voltage with the voltage threshold of the energy storage device if the design travel of the first road wheel is less than three times the root mean square value of the road wheel travel. If the energy supply voltage is less than the voltage threshold of the energy storage device, the active suspension is determined to be suitable for the driving state. The second comparison subunit is used to compare the energy feeding efficiency with a preset energy feeding efficiency threshold if the energy feeding voltage is greater than the voltage threshold of the energy storage device. If the energy feeding efficiency is less than the preset energy feeding efficiency threshold, it determines that the driving state is suitable for active suspension. Otherwise, it applies energy feeding suspension so that the tracked vehicle is in a semi-active working condition.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by the power supply suspension hybrid control device, implements the power supply suspension hybrid control method as described in any one of claims 1-7.

Images