A vehicle semi-active suspension adaptive control and state estimation method

Abstract

The application discloses a kind of vehicle semi-active suspension adaptive control and state estimation method, by calculating stability cross frequency, determine the control strategy of hybrid VBGH-ABGH, improve the steering stability in whole road excitation band;Comfort-oriented SH-ADD algorithm based on comfort cross frequency, can guarantee the real-time switching between SH algorithm and ADD algorithm in control process, so that SH-ADD algorithm has good adaptability and optimal control effect;The online real-time estimation method of vehicle suspension state variable based on FFT guarantees the engineering realizability of the above control method;A kind of vehicle semi-active suspension adaptive control method, comfort-oriented SH-ADD algorithm and steering stability-oriented VBGH-ABGH algorithm are used as bottom controller, and upper controller is used, based on the calculation of vehicle lateral acceleration, the two algorithms in bottom controller are switched in real time, driving conditions are considered, without sacrificing the performance of one aspect, fully taking into account the optimality of the two strategies in bottom layer.

Description

Technical Field

[0001] This invention belongs to the field of intelligent suspension technology for vehicles, specifically relating to a method for adaptive control and state estimation of semi-active suspension for vehicles. Background Technology

[0002] Semi-active suspension systems have become a research hotspot in the field of intelligent suspension engineering in recent years. With the increasing maturity of controllable damping shock absorbers (CDC and MRC) in terms of structure, intelligent vehicle suspension systems based on semi-active suspension technology are gradually being engineered and mass-produced.

[0003] The semi-active suspension system consists of controllable damping shock absorbers and a control system with control algorithms as the key technology. Regarding the structure of the controllable damping shock absorbers, the main manufacturers for CDC are Sachs (Germany), Bristol (USA), and Mono (USA); the main manufacturer for MRC is Beijing West Industries. However, these companies have not disclosed their core control algorithms.

[0004] In recent years, some domestic manufacturers have also begun mass production of controllable damping shock absorber structures. However, currently mass-produced vehicles still mainly focus on control under typical operating conditions, such as controlling pitch during braking and starting, and controlling roll during steering. For fully automatic control algorithms, due to difficulties in algorithm and engineering implementation methods, all OEMs are currently accelerating research and development.

[0005] The challenge of the control algorithm lies in the contradiction between comfort and handling stability, as well as the contradiction between the two in different road surface excitation frequency bands. Another challenge is the engineering feasibility.

[0006] The most widely known algorithm in the engineering community is the sky-hook (SH) damping algorithm proposed by Professor Karnopp in the United States, which is a comfort-oriented control strategy. Its characteristic is that it significantly improves the vibration acceleration of the sprung mass in the low-frequency band around the resonance point, but worsens the vibration acceleration in the mid- and high-frequency bands. Another algorithm, acceleration-driven damping (ADD), is exactly the opposite of SH; it worsens vibration acceleration in the low-frequency band but significantly improves it in the mid- and high-frequency bands. In 2007, Professor Savaresi of Imperial College London in Italy proposed the mixed SH-ADD algorithm, which combines the advantages of both algorithms by switching between them through the calculation of crossover frequencies.

[0007] However, both SH and SH-ADD are geared towards improving comfort, aiming to reduce the vibration acceleration of sprung mass. To improve vehicle handling stability, Cole et al. proposed the ground-hook (GH) algorithm in 1994. Similar to SH, this invention refers to it as VBGH (velocity-based ground-hook). VBGH is geared towards vehicle stability, aiming to reduce the dynamic load on the wheels. In 2020, Kopylovet, referencing ADD, proposed the ABGH (acceleration-based ground-hook) algorithm.

[0008] The engineering implementation of the SH-ADD, VBGH, and ABGH algorithms mentioned above all require absolute and relative velocity signals of the sprung and unsprung masses of the vehicle, using these state variable signals for damping switching. However, measuring these signals on a vehicle is extremely difficult. Using the Kalman filtering method results in high computational cost, is sensitive to model accuracy, and is difficult to implement in non-static environments. Summary of the Invention

[0009] In view of this, the present invention provides a method for adaptive control and state estimation of semi-active vehicle suspension, aiming to provide a technical solution for the engineering and production of semi-active intelligent vehicle suspension.

[0010] A semi-active adaptive control method for vehicle suspension includes: in operation stability control, when the road excitation dominant frequency f excitation When the frequency is less than the set crossover frequency Ω, the ABGH control algorithm is used to control the vehicle; when the road excitation dominant frequency is greater than the set crossover frequency Ω, the VBGH control algorithm is used to control the vehicle.

[0011] A preferred method for calculating the set crossover frequency Ω includes:

[0012] Maintain the stiffness K of each tire w Keeping the mass constant, within the range of the masses on each spring, the mass m on the spring is... s The value is increased from the minimum to the maximum in a set step size, yielding the mass values ​​on each spring at each increment order: m s =[m s,1 m s,2 …m s,N ]; where N is the increment order;

[0013] Within the range of values, the tire stiffness of each wheel is increased from its minimum value to its maximum value in a set step size, yielding the tire stiffness values ​​at each increment order: K w =[K w,1 K w,2 …Kw,N ].

[0014] For the i-th reel, with spring mass m s Tire stiffness k w Below, frequency response analysis of the VBGH control algorithm and ABGH algorithm is performed to obtain the stability-oriented crossover frequency Ω. n,h By iterating through n = 1, ..., N, h = 1, ..., N, we obtain the stability-oriented cross frequencies for pairwise combinations of sprung mass and tire stiffness, resulting in an N×N dimensional stability-oriented cross frequency matrix.

[0015]

[0016] When switching between the ABGH and VBGH control algorithms, the cross frequency corresponding to the current sprung mass and tire stiffness data is found in the stability-oriented cross frequency matrix based on the vehicle's current sprung mass and tire stiffness data. This frequency is then compared with the dominant frequency of the road excitation to complete the switching between the two algorithms.

[0017] Ideally, the tire stiffness k w Calculations based on tire pressure sensors:

[0018]

[0019] Where P is tire pressure, W is tire width, and D is tire diameter;

[0020] sprung mass m s Calculated using the following formula:

[0021]

[0022] In the formula, g represents the acceleration due to gravity, and k b Δs represents the suspension stiffness; Δs is the suspension deformation at static equilibrium, obtained through dynamic displacement sensors or steering angle sensors installed in intelligent suspension vehicles.

[0023] Preferably, the dominant frequency f of the road surface excitation excitation The following method is used to obtain it:

[0024] Collect vertical acceleration sensor signals at the wheel axle center. After performing an FFT on the signal, first divide by the number of points N, then remove the periodicity, and finally take the modulus of the signal. The largest modulus value is the dominant frequency f of the road excitation. excitation .

[0025] Ideally, the crossover frequency should be updated each time a vehicle starts.

[0026] A semi-active adaptive control method for vehicle suspension, when aiming at driving comfort control, when the road excitation dominant frequency fexcitation When the crossover frequency is less than the set crossover frequency α, the SH control algorithm is used to control the vehicle; when the road excitation dominant frequency is greater than the set crossover frequency α, the ADD control algorithm is used to control the vehicle; the calculation method for the crossover frequency α includes:

[0027] Maintain the stiffness K of each tire w Keeping the mass constant, within the range of the masses on each spring, the mass m on the spring is... s The value is increased from the minimum to the maximum in a set step size, yielding the mass values ​​on each spring at each increment order: m s =[m s,1 m s,2 …m s,N ]; where N is the increment order;

[0028] Within the range of values, the tire stiffness of each wheel is increased from its minimum value to its maximum value in a set step size, yielding the tire stiffness values ​​at each increment order: K w =[K w,1 K w,2 …K w,N ];

[0029] For the i-th reel, with spring mass m s Tire stiffness k w Next, frequency response analysis using the SH algorithm and ADD algorithm is performed to obtain the comfort-oriented crossover frequency α. n,h By iterating through n = 1, ..., N, h = 1, ..., N, we obtain the comfort-oriented cross-frequency combinations of sprung mass and tire stiffness, resulting in an N×N dimensional comfort-oriented cross-frequency matrix.

[0030]

[0031] Based on the vehicle's current sprung mass and tire stiffness data, the cross frequency corresponding to the current sprung mass and tire stiffness data is found in the comfort-oriented cross frequency matrix, and the SH algorithm and ADD algorithm are switched according to the cross frequency.

[0032] Preferred methods for estimating state variables required for vehicle control include:

[0033] (1) When the vehicle is started, a sliding window is used to sample the sprung mass acceleration signal and the unsprung mass acceleration signal; wherein, the interval between adjacent sliding windows is ΔN sampling points; the width of each sliding window is N, and the sequence number of the sliding window is k; sampling is performed once every 1ms, and the width of the sliding window is N = 1024 sampling points; the acceleration signal Stored in a buffer, where n = 0, 1, ..., N-1;

[0034] (2) After completing a sliding window sampling, the acceleration signal is low-pass filtered;

[0035] (3) The acceleration signal after low-pass filtering The time-domain acceleration signal A is obtained by performing an FFT transform. k [ω]; where ω represents frequency; A k [ω] represents the acceleration signal corresponding to the k-th sliding window;

[0036] (4) The frequency domain signal A after FFT transformation k [ω] Frequency components below 1Hz have their amplitudes set to zero;

[0037] (5) Next, apply frequency domain integration theory to the signal obtained in step (4) to perform frequency domain integration processing to obtain the frequency domain velocity signal:

[0038]

[0039] (6) Obtain the time-domain signal of velocity from the frequency-domain signal of velocity using IFFT:

[0040]

[0041] (7) When k≥1, during the frequency domain integration process, the initial value of this integration is set to the integration value at ΔN obtained in the previous sliding window. This is equivalent to adding this value as an offset increment to the current integration data. The corrected velocity is obtained by the following formula:

[0042]

[0043] (8) After every ΔN sampling points, the sliding window moves once, repeating the process from step 2 (2) to step 8 (8), thereby realizing the online estimation process of velocity state.

[0044] A semi-active adaptive control method for vehicle suspension utilizes an upper-level controller to switch between a hybrid SH-ADD control algorithm and a hybrid VBGH-ABGH control algorithm based on lateral acceleration, thereby controlling the vehicle. Specifically, when the lateral acceleration is less than a set threshold, the hybrid SH-ADD control algorithm is used; otherwise, the hybrid VBGH-ABGH control algorithm is used.

[0045] In the hybrid VBGH-ABGH control algorithm, when the road excitation dominant frequency f excitation When the frequency is less than the set crossover frequency Ω, the ABGH control algorithm is used to control the vehicle; when the road excitation dominant frequency is greater than the set crossover frequency Ω, the VBGH control algorithm is used to control the vehicle.

[0046] In the hybrid SH-ADD control algorithm, when the road excitation dominant frequency f excitation When the frequency is less than the set crossover frequency α, the SH control algorithm is used to control the vehicle; when the road excitation dominant frequency is greater than the set crossover frequency α, the ADD control algorithm is used to control the vehicle.

[0047] A preferred formula for calculating lateral acceleration is as follows:

[0048]

[0049] Where L is the vehicle wheelbase; K us δ is the understeer coefficient, v is the vehicle speed, and δ is the steering wheel angle.

[0050] The present invention has the following beneficial effects:

[0051] The intelligent suspension technology described in this invention represents a development direction in vehicle engineering and even the field of intelligent vibration reduction technology, and it has urgent demand and broad prospects for engineering production in the current field of intelligent vehicle technology. The beneficial effect of this invention is that the proposed semi-active intelligent suspension control algorithm can be implemented in engineering, solving the bottleneck problem in control algorithms that restricts the practical production of semi-active intelligent suspension technology. The algorithm proposed in this invention also has transfer and application value in other intelligent vibration reduction technology fields. Attached Figure Description

[0052] Figure 1 A dynamic model of single-wheel vibration of the vehicle suspension;

[0053] Figure 2 The frequency response between the hybrid SH-ADD vehicle body acceleration and road surface excitation;

[0054] Figure 3 This represents the frequency response of tire deformation relative to road surface excitation.

[0055] Figure 4 The process of identifying the dominant frequency of the road surface;

[0056] Figure 5 shows the simulation results of the identification of the dominant frequency of the road surface; among them, Figure 5(a) is the identification interval 2s-3s, and Figure 5(b) is the identification interval 6s-7s.

[0057] Figure 6 Frequency response of the hybrid VBGH-ABGH algorithm;

[0058] Figure 7 For the tire deformation response, where Figure 7 (a) is 1.5Hz. Figure 7 (b) is 10Hz;

[0059] Figure 8This refers to tire deformation under Class D road conditions.

[0060] Figure 9 Schematic diagram of the adaptive crossover frequency update mechanism;

[0061] Figure 10 2D mapping diagrams are established for cross-frequency calibration to address comfort and stability considerations;

[0062] Figure 11 This is the implementation process of the online state estimation algorithm of the present invention;

[0063] Figure 12 A comparison chart of online velocity estimation simulation and theoretical values;

[0064] Figure 13 This is a schematic diagram of the multi-objective control principle of the present invention;

[0065] Figure 14 A comparison of the model-calculated values ​​of lateral acceleration in this invention with the CARSIM simulation values;

[0066] Figure 15 This is a schematic diagram of the CarSim and Simulink co-simulation structure.

[0067] Figure 16 This is the lane-changing test scenario for the present invention;

[0068] Figure 17 This is a simulation diagram of scenario 1 of the present invention;

[0069] Figure 18 This invention provides a high-speed curved track; wherein, (a) is a race area scene, (b) is a circular curved track, (c) is a Class B road surface outline, and (d) is the road surface height;

[0070] Figure 19 These are comparison images of the curved track at different moments.

[0071] Figure 20 This is a performance comparison chart for curved tracks. Detailed Implementation

[0072] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0073] Comparative studies have revealed that, compared to passive suspension, ABGH significantly improves handling stability in the low-frequency range, while VBGH exhibits deterioration in performance in the low-frequency range. However, VBGH can reduce dynamic load in the range of wheel resonance frequencies, meaning that ABGH and VBGH exhibit different performance in different frequency bands.

[0074] Based on this finding, this invention proposes a hybrid VBGH-ABGH control strategy oriented towards operational stability. The aim is to improve handling stability across the entire road excitation frequency band. The VBGH-ABGH algorithm relies on the accurate acquisition of the stability cross-over frequency, but this frequency varies with occupant and tire stiffness. Therefore, this invention proposes an adaptive calculation method for the stability cross-over frequency.

[0075] Based on the hybrid SH-ADD algorithm, this invention proposes a comfort-oriented SH-ADD algorithm based on comfort crossover frequency, which can ensure real-time switching between the SH algorithm and the ADD algorithm during the control process, thereby enabling the SH-ADD algorithm to have good adaptability and optimal control effect.

[0076] This invention also proposes an online real-time estimation method for vehicle suspension state variables based on FFT, thereby ensuring the engineering feasibility of the control method proposed in this invention.

[0077] Based on the above-described invention, this invention proposes a semi-active adaptive control method for vehicle suspension. This method uses the comfort-oriented SH-ADD algorithm and the handling-stability-oriented VBGH-ABGH algorithm as the bottom-level controller, and employs an upper-level controller to switch between the two algorithms in real time based on the calculation of the vehicle's lateral acceleration. When the lateral acceleration exceeds a threshold value, indicating a risk of sideslip, the upper-level controller switches the bottom-level controller to the VBGH-ABGH algorithm to control the vehicle. Conversely, when the lateral acceleration is below the threshold value, typically during straight-line driving, the bottom-level controller operates using the comfort-oriented SH-ADD algorithm. The algorithm proposed in this invention considers driving conditions without sacrificing any aspect of performance, fully balancing the optimality of the two bottom-level strategies.

[0078] The proposed method has been verified by Simulink simulation or CarSim and Simulink co-simulation. The proposed adaptive control method has been tested by simulating a real vehicle. Through simulation tests in various typical driving scenarios, it has been proven that it has good adaptive performance under various real driving conditions.

[0079] First, let's introduce the background algorithm related to this invention:

[0080] 1.1. Single-wheel two-degree-of-freedom model of semi-active vehicle suspension

[0081] The single-wheel vibration dynamic model of a semi-active vehicle suspension is as follows: Figure 1 As shown.

[0082] according to Figure 1 The dynamic equations are established as follows:

[0083]

[0084] Where z b ,z w ,z r These represent the absolute vertical displacements of the vehicle body, wheels, and road surface elevations, respectively; m b m w , which are the sprung mass and the unsprung mass, respectively. b k w These represent the suspension and tire stiffness, respectively; Δs and Δw represent the compression of the suspension spring and tire stiffness at static equilibrium, respectively; the first-order dynamics reflects the damping at c in The delay relationship between the input and the actual output c(t), where β is the modulation bandwidth of the semi-active suspension damper.

[0085] The simulation verification model parameters of this invention are shown in Table 1.

[0086] Table 1. Simulation Model Parameters

[0087]

[0088] 1.2. Hybrid SH-ADD Algorithm

[0089] Professor Karnopp proposed the following ceiling damping algorithm for comfort:

[0090]

[0091] in, It is the absolute displacement of the sprung mass; It is the suspension dynamic speed.

[0092] Professor Savaresi proposed the following ADD control algorithm:

[0093]

[0094] in, It is the acceleration of the sprung mass.

[0095] The hybrid SH-ADD algorithm proposed by Professor Savaresi of Italy is as follows:

[0096]

[0097] in, It is the acceleration of the mass on the spring; It is the absolute velocity of the sprung mass. α is the dynamic velocity; α is the cross-over frequency of the hybrid SH-ADD algorithm.

[0098] The SH-ADD algorithm determines the dominant road excitation frequency f based on the following formula. excitation Determining the magnitude of the crossover frequency:

[0099]

[0100] This judgment method does not require real-time estimation of the dominant road excitation frequency f. excitation This is an indirect calculation estimation method in the time domain. Its advantage is good real-time performance. However, the condition is obtained indirectly based on the simple harmonic excitation condition, and its specific effect on random road surfaces needs to be evaluated.

[0101] As can be seen from the hybrid SH-ADD algorithm proposed by Professor Savaresi of Italy, based on a specific frequency point α, which this invention calls the comfort-oriented crossover frequency, the switching between the SH and ADD algorithms is determined. This allows the two algorithms to compensate for each other's shortcomings in different frequency bands, fully utilizing their advantages across different frequency bands. The frequency responses obtained from simulations of the passive suspension, SH, ADD, and hybrid SH-ADD algorithms are as follows: Figure 2 .

[0102] from Figure 2 It can be seen that the vibration acceleration obtained by the hybrid SH-ADD by switching between SH and ADD according to the dominant frequency of road excitation is the smallest in the entire frequency band.

[0103] 1.3. VBGH and ABGH Algorithms

[0104] Both VBGH and ABGH algorithms are heuristic algorithms, proposed with reference to SH and ADD algorithms.

[0105] The heuristic VBGH algorithm is as follows:

[0106]

[0107] in, It is the absolute velocity of the unsprung mass.

[0108] The acceleration-based heuristic algorithm ABGH proposed by Kopylov is as follows:

[0109]

[0110] in, It is the acceleration of the unsprung mass.

[0111] The following are the four main inventive points of this invention.

[0112] Example 1: Hybrid VBGH-ABGH Algorithm

[0113] 2.1. Evaluation of handling stability and comparative analysis of VBGH and ABGH performance

[0114] The improvement in vehicle vibration acceleration achieved by the hybrid SH-ADD method comes at the cost of wheel vibration. Wheel vibration affects dynamic load, and when the dynamic load exceeds the static load, wheel hop occurs, which affects handling stability. Reducing dynamic load can improve handling stability and road friendliness.

[0115] When the tire deformation (z) w -z r A wheel jump will occur when the following equation is satisfied:

[0116]

[0117] The frequency response of the dynamic load (tire deformation) of this invention is evaluated and calculated using the following formula, and the simulation results are shown below. Figure 3 .

[0118]

[0119] In the above formula, N is the number of wheels;

[0120] from Figure 3 It can be seen that VBGH minimizes the dynamic load near the unsprung mass resonance point, but deteriorates it in the low-frequency range. Conversely, ABGH deteriorates the dynamic load near the unsprung mass resonance point, but improves it near the sprung mass resonance point. Note that when the road excitation dominant frequency is greater than the crossover frequency, from Figure 3 The superiority of VBGH over ABGH is not very obvious. This is because the vibration near the unsprung mass resonance point is a high-frequency vibration, and the tire deformation itself is relatively small. The vibration of both is similar to that of passive suspension. Since the tire deformation itself is relatively small, it is meaningful for VBGH to be superior to ABGH. The simulation of the VBGH-ABGH algorithm proposed later also illustrates this point.

[0121] 2.2. The VBGH-ABGH algorithm proposed in this invention

[0122] First, define and determine the crossover frequency Ω for operational stability, based on... Figure 3As shown in Table 1, the frequency is around 7Hz. In the low-frequency range (0.2-2Hz) and before the crossover frequency Ω, the ABGH algorithm performs similarly to the minimum-damped passive suspension; however, the VBGH algorithm performs worse than the high-damped passive suspension. In the mid-frequency range (2-7Hz), both algorithms perform similarly to the low-damped passive suspension. However, in the mid-frequency band, the increase in wheel dynamic load has a negligible impact on handling performance because the resonant frequencies of sprung and unsprung masses are not set within this range. When the dominant road excitation frequency is greater than the crossover frequency Ω, VBGH outperforms ABGH. Since the vertical axis uses a logarithmic coordinate system, from... Figure 3 The magnified view clearly shows this.

[0123] Based on the above findings, this invention proposes for the first time a stability-oriented hybrid VBGH-ABGH algorithm. This algorithm identifies the dominant frequency of the road surface online and adaptively calculates the stability-oriented crossover frequency Ω. By comparing these two, it switches between VBGH and ABGH, thus fully utilizing their performance advantages in different excitation frequency bands. Specifically, the hybrid VBGH-ABGH algorithm switches to ABGH when the dominant frequency of the road surface is in the low-frequency range and switches to VBGH when the dominant frequency is in the high-frequency range.

[0124] The expression for the hybrid VBGH-ABGH algorithm is as follows:

[0125]

[0126] Where Ω is the crossover frequency for operational stability, and f excitation It is the dominant frequency of road surface excitation.

[0127] Unlike the SH-ADD algorithm, this invention uses the FFT method to identify the dominant frequency f of the road surface excitation. excitation The specific method is as follows:

[0128] 1) Utilizing the vertical acceleration sensor signal installed at the wheel axle As the input signal for FFT.

[0129] 2) such as Figure 4 As shown, the dominant frequency f of road surface excitation using a sliding window is... excitation The dynamic recognition is as follows:

[0130] First, divide the vertical acceleration sensing signal processed by FFT by the number of points N. Then, remove the periodicity. Finally, take the modulus of the signal. The largest modulus is the dominant frequency f. excitation After extensive simulation testing, the FFT parameter N = 1024 and the sampling frequency was selected as 1kHz. The simulation input signal was:

[0131]

[0132] Where w(t) is a white noise random signal, the simulation verification results are shown in Figure 5.

[0133] As shown in Figure 5, the sliding window FFT method proposed in this invention can effectively identify the active frequency signals of the road surface.

[0134] Simulation of the frequency response of the hybrid VBGH-ABGH algorithm is shown below. Figure 6 .

[0135] from Figure 6 It can be seen that the hybrid VBGH-ABGH algorithm has optimal performance in both the low-frequency and high-frequency regions. That is, it is almost superior to the ABGH algorithm in the low-frequency region before the crossover frequency point, and almost close to the VBGH algorithm in the high-frequency region.

[0136] To further verify the hybrid VBGH-ABGH algorithm proposed in this embodiment, simulations were performed in the time domain to verify two typical frequency harmonic excitations and random excitations.

[0137] First, two typical frequency simple harmonic sinusoidal signals Asin(w) are used. i t) is used as the road surface excitation, where w1 = 1.5 Hz (sprung mass resonant frequency) and w2 = 10 Hz (unsprung mass resonant frequency). The tire deformation response is as follows: Figure 7 As shown.

[0138] (A = 0.05)

[0139] from Figure 7 As can be seen, when the road excitation frequency is 1.5Hz, the performance of the VBGH algorithm deteriorates further than that of the nominal passive suspension, while the ABGH algorithm exhibits excellent performance; that is, in the low-frequency region, ABGH outperforms both VBGH and the passive suspension. Notably, as expected, the performance of the hybrid VBGH-ABGH algorithm is almost close to that of ABGH. Compared to the passive suspension, the stability performance of the proposed hybrid VBGH-ABGH algorithm is improved by 38% and 31% respectively, using the root mean square error (RMS) and maximum error (MaxAE) as evaluation metrics.

[0140] from Figure 7 As can be seen from the data, VBGH exhibits the best performance under 10Hz excitation, while the hybrid VBGH-ABGH, as expected, shows the same performance as VBGH. The hybrid VBGH-ABGH algorithm even improves the root mean square value by 5.8% compared to the ABGH algorithm. This may not seem very important at first glance, but it is still a valuable improvement because it increases the root mean square value of tire deformation by 2.6mm relative to the unsprung mass resonance frequency.

[0141] For simulations of random conditions, ISO-8608D level random road surface input was used, with a vehicle speed of 60 km / h. The time-domain response, root mean square (RMS) value, and maximum error (MaxAE) of tire deformation are as follows: Figure 8 As shown in Table 2.

[0142] Table 2 ISO-8608 D-level Road Input Tire Deformation

[0143]

[0144] from Figure 8 As shown in Table 2, the hybrid VBGH-ABGH algorithm exhibits a significant performance improvement, outperforming both the VBGH and ABGH algorithms. The hybrid VBGH-ABGH algorithm improves the root mean square value of the passive suspension by 29.4%.

[0145] Example 2: Stability-Oriented Adaptive Cross-Frequency Calibration Method

[0146] Maintain the stiffness K of each tire w Keeping the mass constant, within the range of the masses on each spring, the mass m on the spring is... s The value is increased from the minimum to the maximum in a set step size, yielding the mass values ​​on each spring at each increment order: m s =[m s,1 m s,2 …m s,N ]; where N is the increment order. Similarly, by increasing the tire stiffness of each wheel from its minimum value to its maximum value within the range of values, the values ​​of the tire stiffness of each wheel at each increment order are obtained:

[0147] K w =[K w,1 K w,2 …K w,N ]

[0148] For the i-th reel, the mass on the spring is m s,n Tire stiffness is K w,h Below, frequency response analysis of the VBGH control algorithm and ABGH algorithm is performed to obtain the stability-oriented crossover frequency Ω. n,h By iterating through n = 1, ..., N, and h = 1, ..., N, we obtain the stability-oriented cross frequencies for pairwise combinations of sprung mass and tire stiffness, resulting in an N×N dimensional stability-oriented cross frequency matrix.

[0149]

[0150] The input signal k of this invention w and m s No additional sensors are required; existing equipment in commercial vehicles can be used. Among these, tire stiffness k...w The following formula can be used to calculate the pressure from the tire pressure sensor:

[0151]

[0152] Where P is tire pressure (Pa), W is tire width, and D is tire diameter (m). The sprung mass of this invention is m. s This can be obtained by solving the dynamic equations when the vehicle is stationary. m s It is a function of the deformation when the suspension is in static equilibrium, and the formula is as follows:

[0153]

[0154] k b Δs represents the suspension stiffness; Δs is the static equilibrium value, which is the suspension deformation, obtainable through dynamic displacement sensors or steering angle sensors installed in intelligent suspension vehicles. Based on the above offline calculations, and based on the current (k) w ,m s For data pairs, this invention proposes an adaptive update mechanism to optimize the crossover frequency in the update algorithm for both comfort and stability. Since the data pairs (k...) w ,m s The crossover frequency does not change significantly during vehicle operation, therefore it is not necessary to update it in every control cycle. This invention only updates the crossover frequency when the vehicle starts, and continues until the next time the driver starts the vehicle; when switching between the ABGH and VBGH control algorithms, the crossover frequency is updated according to the vehicle's current (k) w ,m s The data is used to find (k) in the stability-oriented cross-frequency matrix. w ,m s The cross-frequency corresponding to the data is compared with the dominant frequency of the road surface excitation in that frequency domain to complete the switching between the two algorithms. See the diagram for the principle of the adaptive cross-frequency update mechanism. Figure 9 .

[0155] Example 3: Comfort-oriented SH-ADD algorithm based on comfort cross frequency.

[0156] In determining the crossover frequency of the hybrid SH-ADD algorithm, this invention follows the method of Example 2, for the i-th round, with a spring mass of m... s,n Tire stiffness is K w,h Next, frequency response analysis using the SH algorithm and ADD algorithm is performed to obtain the comfort-oriented crossover frequency α. n,h By iterating through n = 1, ..., N, and h = 1, ..., N, we obtain the comfort-oriented cross-frequency combinations of sprung mass and tire stiffness, resulting in an N×N dimensional comfort-oriented cross-frequency matrix.

[0157]

[0158] Similar to Example 2, this invention only updates the crossover frequency when the vehicle starts, and continues until the next time the driver starts the vehicle; when switching between the SH control algorithm and the ADD control algorithm, the frequency is updated according to the vehicle's current (k) frequency. w ,m s The data is used to find (k) in the comfort-oriented cross-frequency matrix. w ,m s The cross-frequency corresponding to the data is compared with the dominant frequency of the road surface excitation in this frequency domain to complete the switching between the two algorithms.

[0159] Figure 10 A 2D mapping diagram is calibrated for the cross-frequency alignment of comfort and stability. From... Figure 10 As can be seen, the comfort-oriented crossover frequency increases with increasing tire stiffness and decreases with increasing sprung mass; it depends on both variables, with the effect of sprung mass changes being more significant than that of tire stiffness changes. Simulation observations are consistent with theoretical expectations. The comfort index is sprung mass acceleration, which is mainly affected by sprung mass. The stability-oriented crossover frequency is directly proportional to both sprung mass and tire stiffness, and it depends on both variables. The effect of tire stiffness is more significant than that of sprung mass. Notably, this relationship is exactly the opposite of the comfort-oriented crossover frequency.

[0160] Example 4: Estimation Methods for State Variables

[0161] Estimating state variables is a key issue in the engineering implementation of semi-active suspension. The SH, ADD, and VBGH-ABG algorithms all require feedback states such as the absolute and relative velocities of the sprung and unsprung masses. These states are difficult to measure in a vehicle. To address this problem, this invention proposes a real-time state estimation calculation method based on FFT frequency domain integration. The proposed method requires only two accelerometers, installed on the sprung and unsprung masses respectively. The specific implementation method and simulation verification demonstrate the effectiveness of the method.

[0162] The implementation method based on FFT frequency domain integration is as follows:

[0163] The acceleration signal a(t) with a certain frequency ω component is:

[0164] a(t)=Ae jωt

[0165] A is the amplitude, the initial velocity is zero, and the velocity signal obtained by integration is:

[0166]

[0167] Where v(t) is the Fourier velocity component with frequency ω; V is the velocity amplitude, V = A / jω. When the initial velocity and displacement are zero, the displacement x(t) can be obtained by integrating the following equation:

[0168]

[0169] Where x(t) is the displacement Fourier component with frequency ω; X is the coefficient corresponding to x(t), X = -A / ω 2 .

[0170] After Fourier transforming all the different frequency components, the time-domain velocity and displacement signals can be obtained through inverse Fourier transform. Considering the measurement accuracy of the accelerometer in the low-frequency region, in practical engineering, the low-frequency components of the response can be set to 0 according to the actual situation, for example, the frequency components below 1Hz can be set to zero.

[0171] In the control algorithm proposed in this invention, only the velocity of the vibration response needs to be calculated in real time. Based on the above theory, the implementation process of the online real-time estimation algorithm of this invention is as follows: Figure 11 The online state estimation algorithm is implemented as follows:

[0172] (1) When the vehicle is started, in order to improve timeliness, the algorithm runs on a buffer window of length N, number of slides k (k=0,1,2,…) and slide distance ΔN, sampling once every 1ms. The buffer window has 1024 (N) sampling points. The acceleration signal is then sent to the buffer window. (Depending on the sensor location, the acceleration signal is divided into sprung mass acceleration signal) and unsprung mass acceleration signal (But the processing method is the same) stored in the buffer, where n = 0, 1, ..., N-1;

[0173] (2) After completing a window period of sampling, perform low-pass filtering and select a cutoff frequency of 50Hz.

[0174] (3) The time-domain acceleration signal sampled within the window period The time-domain acceleration signal A is obtained by performing an FFT transform. k [ω];

[0175] (4) The frequency domain signal A after FFT transformation k [ω] Frequency components below 1Hz have their amplitudes set to zero;

[0176] (5) Next, apply frequency domain integration theory to the signal obtained in step (4) to perform frequency domain integration processing to obtain the frequency domain velocity signal:

[0177]

[0178] (6) Obtain the time-domain signal of velocity from the frequency-domain signal of velocity using IFFT:

[0179]

[0180] (7) Since each window is processed independently, the integration process does not consider the cumulative nature of the velocity signal between windows, that is, the initial value of each integration is zero. This leads to discontinuities at the boundaries of continuous windows. To eliminate this error, an offset correction mechanism is applied:

[0181] When k=0, the data within the window is integrated with zero initial value, and no correction is needed;

[0182] When k≥1, the initial value of this integration is set to the integral value at ΔN obtained in the previous window (the (k-1)·ΔNth data point in total). This is equivalent to adding this value as an offset increment to the current integration data. The corrected velocity can be obtained by the following formula:

[0183]

[0184] This ensures that the first point in the new window is aligned with the incremental point of the previous window, satisfying:

[0185]

[0186] This offset correction method maintains signal continuity between consecutive sliding windows, which is crucial for achieving accurate, real-time state variable estimation in semi-active suspension control algorithms.

[0187] (8) After sampling ΔN increments, slide the window once and repeat the process from step 2 (2) to step 8 (8) to achieve online speed state estimation. The vehicle simulation parameters and the relevant parameters of the algorithm obtained after extensive optimization are shown in Table 3.

[0188] Table 3 Simulation parameters based on online FFT calculation of state variables

[0189]

[0190] Among them, z road Represents road surface grade; v longtitudinal t represents the vehicle's speed. initialization f is the initialization time; accelerometer f is the acceleration sampling frequency; estimation The velocity estimation frequency is the frequency at which the algorithm acquires the velocity. N is the number of sampling data points in the acceleration window, and ΔN is the increment of the number of sampling points in the sliding window.

[0191] The simulation diagram of the proposed online velocity estimation method is shown in Figure 12.

[0192] from Figure 12 As can be seen, compared with theoretical data, the algorithm proposed in this invention can estimate the speed of the sprung mass (the speed of the unsprung mass can also be obtained) with relatively high accuracy, meeting the accuracy requirements for algorithm engineering implementation, with a maximum MaxAE of less than 0.005 m / s. Applying this method removes the obstacles to the engineering implementation of the semi-active control algorithm proposed in this invention.

[0193] Example 5: A semi-active suspension adaptive control method

[0194] 5.1. The algorithm proposed in this invention

[0195] Based on the above-mentioned points of invention, this invention proposes an adaptive method for semi-active suspension. The idea behind this method is:

[0196] (1) The above-mentioned hybrid SH-ADD and hybrid VBGH-ABGH are combined, and an upper-level controller is used to switch between the two based on the collected typical operating condition vehicle information and driver operation signals, thereby determining in real time whether to prioritize comfort or handling stability. The schematic diagram of the control method proposed in this invention is shown below. Figure 13 .

[0197] (2) Figure 13 As shown, the switching logic proposed in this invention calculates lateral acceleration online based on vehicle speed and steering angle signals. Since excessive lateral acceleration can lead to wheel load imbalance and thus cause vehicle skidding safety risks, when the lateral acceleration exceeds a set threshold, the upper-level controller switches to stability as the control objective; otherwise, it switches to comfort as the control objective. That is:

[0198]

[0199] Among them, c SH-ADD The damping is determined by the hybrid SH-ADD, c VBGH-ABGH The damping is determined by the hybrid VBGH-ABGH. To calculate the obtained lateral acceleration, α treshold This represents the lateral acceleration threshold value.

[0200] (3) Lateral acceleration can be obtained using an IMU sensor. Considering the implementation cost and the lag problem of direct measurement, this invention proposes a method for calculating lateral acceleration based on vehicle speed and steering angle. The calculation formula is as follows:

[0201]

[0202] Where L is the wheelbase; K usδ is the understeer coefficient, v is the vehicle speed, and δ is the steering wheel angle. The values ​​of these parameters can be determined based on the specific vehicle. Real-time calculated values ​​can be filtered using a high-pass filter to eliminate DC components and low-frequency noise. Figure 14 This is a comparison between the calculated lateral acceleration values ​​and the CARSIM simulation.

[0203] 5.2. Simulation Verification of the Algorithm Proposed in This Invention

[0204] The proposed algorithm's effectiveness was verified using co-simulation with CarSim and Simulink. The co-simulation structure diagram is shown below. Figure 15 .

[0205] Scenario 1: Sudden lane change to avoid obstacles. The lane change test scenario is as follows: Figure 16 Provided.

[0206] Simulation results are shown below Figure 17 .

[0207] from Figure 17 It can be seen that the strategy conversion during lane change (a), track tracking effect (b), and tire dynamic load (c) (dynamic load is minimized when switching to stability mode) prove the effectiveness of the strategy proposed in scenario 1.

[0208] Scenario 2: High-speed curved track, the test scenario consists of... Figure 18 Provided.

[0209] Simulation results are shown below Figure 19 , Figure 20 .

[0210] from Figure 20 Statistical calculations show that the passive suspension experienced 12 wheel bounce events, while the control method of this invention reduced this to 2. Statistically, the root mean square value of vertical acceleration is reduced by 32% compared to the passive suspension. Figure 19 Two instantaneous vehicle state images were captured from the simulation. Figure 18 High-speed curved track (b) - At the sharp turn around 150m, from... Figure 19 (a) It can be seen that the passive suspension vehicle has slid off the track, while the vehicle equipped with the semi-active control strategy of the present invention has better handling stability than the passive suspension vehicle. Figure 19 (b) is a screenshot of the finish line, showing the blue vehicle, equipped with the control strategy of this invention, leading the red vehicle to the finish line. This demonstrates the effectiveness of scenario 2 of the proposed strategy. Scenario 1 and Scenario 2 represent typical driving conditions and comprehensively reflect the effectiveness of the control method proposed in this invention.

[0211] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A semi-active adaptive control method for vehicle suspension, characterized in that, include: When aiming at operational stability control, when the road excitation dominant frequency Less than the set cross frequency At that time, the ABGH control algorithm is used to control the vehicle; when the road excitation dominant frequency is greater than the set crossover frequency... At that time, the VBGH control algorithm is used to control the vehicle; Set cross frequency The calculation methods include: Maintain the stiffness of each tire Keeping the mass constant, within the range of the mass on each wheel spring, the mass on the spring is... The values ​​are increased from the minimum to the maximum in a set step size, yielding the mass values ​​on each spring at each increment order: ;in, The order of increment; Within the range of values, the tire stiffness of each wheel is increased from its minimum value to its maximum value in a set step size, resulting in the tire stiffness values ​​at each increment order: ; Regarding the first Wheel, mass on spring Tire stiffness Next, frequency response analysis of the VBGH control algorithm and ABGH algorithm is performed to obtain the stability-oriented crossover frequency. ; Traversal , The cross frequencies for stability under pairwise combinations of sprung mass and tire stiffness are obtained. Stability-oriented cross-frequency matrix of dimension: ； When switching between the ABGH control algorithm and the VBGH control algorithm, the cross frequency corresponding to the current sprung mass and tire stiffness data is found in the stability-oriented cross frequency matrix based on the current sprung mass and tire stiffness data. This frequency is then compared with the dominant frequency of road excitation to complete the switching between the two algorithms. Tire stiffness Calculations based on tire pressure sensors: ； in, It's tire pressure. It refers to the tire width. It is the tire diameter; Sprout mass Calculated using the following formula: In the formula, Represents gravitational acceleration. For suspension stiffness; It is the amount of suspension deformation when statically balanced, obtained through dynamic displacement sensors or steering angle sensors installed in intelligent suspension vehicles.

2. The vehicle semi-active suspension adaptive control method as described in claim 1, characterized in that, Road surface excitation dominant frequency The following method is used to obtain it: Collect vertical acceleration sensor signals at the wheel axle center. After performing an FFT on the signal, first divide it by the number of points N, then remove the periodicity, and finally take the modulus of the signal. The frequency with the largest modulus is the dominant frequency of the road excitation. .

3. The vehicle semi-active suspension adaptive control method as described in claim 2, characterized in that, The crossover frequency is updated each time a vehicle starts.

4. The vehicle semi-active suspension adaptive control method as described in claim 1, characterized in that, The state variable estimation methods required for vehicle control include: (1) When the vehicle is started, a sliding window is used to sample the sprung mass acceleration signal and the unsprung mass acceleration signal; wherein, the interval between adjacent sliding windows is set. There are N sampling points; the width of each sliding window is N, and the window number is set to N. Samples are taken once every 1 ms, with a sliding window width of N = 1024 sampling points; the acceleration signal... Stored in the buffer, where ; (2) After completing a sliding window sampling, the acceleration signal is low-pass filtered; (3) Acceleration signal after low-pass filtering The time-domain acceleration signal is obtained by performing an FFT transform. ;in, Indicates frequency; This represents the acceleration signal corresponding to the k-th sliding window; (4) For the frequency domain signal after FFT transformation Frequency components below 1Hz are processed to have their amplitude set to zero. (5) Next, apply frequency domain integration theory to the signal obtained in step (4) to perform frequency domain integration processing to obtain the frequency domain velocity signal: ； (6) Obtain the time-domain signal of velocity from the frequency-domain signal of velocity using IFFT: ； (7) When During frequency domain integration, the initial value for this integration is set to the value obtained in the previous sliding window. The integral value is then added as an offset increment to the current integral data. The corrected velocity is obtained using the following formula: (8) Every After sampling points, the sliding window moves once, repeating the process from step 2 (2) to step 8 (8), thereby realizing the online estimation process of velocity state.

Images