Vibration Suppression Method and System for EPS Bumpy Road Surface Based on Multi-Signal Fusion and Parameter Identification

The EPS system, through multi-signal fusion and parameter identification, can identify the characteristics of bumpy roads in real time and calculate the compensation torque, thus solving the stability and adaptability problems of steering wheel vibration suppression on bumpy roads and improving driver comfort.

CN122852692BActive Publication Date: 2026-11-10TIANJIN DECO INTELLIGENT CONTROL CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611339897.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-09-01
Publication Date
2026-11-10
Estimated Expiration
2046-09-01

AI Technical Summary

Technical Problem

When a vehicle is driving on a bumpy road, the vibration of the steering wheel is difficult to suppress effectively. Existing EPS systems are prone to misjudgment and frequent switching when identifying bumpy road conditions and compensation torque, which leads to driver discomfort.

Method used

By employing a multi-signal fusion and parameter identification method, and through the comprehensive processing of torque sensor, rotor position sensor, lateral acceleration sensor and vehicle speed signals, combined with Mamdani-type fuzzy inference and recursive least squares method with adaptive forgetting factor, road excitation characteristics and steering system parameters are identified in real time, and compensation torque is calculated and adjusted to suppress steering wheel vibration.

Benefits of technology

It improves the stability of bumpy road condition recognition and the adaptability of compensation torque, reduces driver discomfort, avoids frequent switching of compensation functions, and uses existing signals to achieve vibration suppression without additional sensors or mechanical structures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122852692B_ABST
    Figure CN122852692B_ABST
Patent Text Reader

Abstract

This invention relates to the field of vehicle EPS system control technology, and discloses a method and system for EPS bumpy road vibration suppression based on multi-signal fusion and parameter identification. The method involves acquiring torque, rotor position, lateral acceleration, and vehicle speed signals, extracting torque, speed, and lateral acceleration fluctuation characteristics, and determining the bumpy road condition through fuzzy inference. After meeting the identification conditions, a recursive least squares method with an adaptive forgetting factor is used to identify steering system parameters, and a fast Fourier transform is combined to obtain the dominant frequency and amplitude of the road excitation. The compensation torque is calculated based on the road excitation and system dynamic parameters, and after amplitude limiting, ramp entry, and exponential decay exit, it is converted into a Q-axis current compensation amount and superimposed on the assist current, thereby suppressing steering wheel vibration caused by bumpy roads.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of vehicle EPS system control technology, specifically to an EPS bumpy road vibration suppression method and system based on multi-signal fusion and parameter identification. Background Technology

[0002] When a vehicle travels on a bumpy road, the unevenness of the road surface is transmitted in the opposite direction to the steering wheel through the tires, steering tie rods, steering gear, and steering shaft, causing circumferential vibration and axial movement of the steering wheel. This manifests as a tingling or throbbing sensation in the driver's hands. The frequency and amplitude of steering wheel vibration are related to the level of road unevenness, vehicle speed, tire characteristics, the mechanical structure of the steering system, and the driver's steering operation state, exhibiting time-varying and random characteristics.

[0003] The EPS system provides assist torque to the vehicle's steering mechanism via a steering motor. When using the steering motor to suppress vibrations from bumpy roads, it is necessary to distinguish between normal changes in hand force generated by the driver's active steering and vibrations caused by the reverse transmission of road excitation. Relying solely on a single torque sensor signal or a fixed threshold can easily lead to misjudgments of bumpy road conditions due to vehicle lateral movement, speed changes, driver's active steering, and sensor noise.

[0004] The main frequency of road excitation, the amplitude of road excitation, and the dynamic parameters of the EPS system vary with road type, vehicle speed, and steering system status. When using fixed system parameters and fixed compensation gain, the compensation torque is difficult to adapt to different bumpy road conditions such as gravel roads, washboard roads, Belgian roads, and speed bumps. If the compensation torque is engaged or disengaged too quickly, it may also cause sudden changes in steering wheel torque. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for suppressing vibrations on bumpy roads in EPS systems using multi-signal fusion and parameter identification. The technical problem to be solved is to improve the stability of identifying the bumpy road conditions and the degree of bumpiness, to obtain the main frequency of road excitation, the amplitude of road excitation, the natural frequency of the system, and the equivalent damping ratio of the system in real time, to calculate the compensation torque based on the road excitation characteristics and the dynamic parameters of the steering system, to reduce steering wheel vibration caused by bumpy roads, and to avoid frequent switching of the compensation function.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] This invention provides a method for suppressing vibrations on EPS bumpy road surfaces based on multi-signal fusion and parameter identification, including:

[0008] The system acquires signals from a torque sensor, a rotor position sensor, a lateral acceleration signal, and a vehicle speed signal. The torque sensor signal is filtered to obtain the DC component and torque fluctuation component. The motor speed is calculated based on the rotor position sensor signal, and the average speed and speed fluctuation component are extracted from the motor speed. The lateral acceleration signal is filtered to obtain the steady-state component and lateral acceleration fluctuation component.

[0009] Within a sliding time window, calculate the effective value of torque fluctuation, peak value of torque fluctuation, effective value of speed fluctuation, peak value of speed fluctuation, effective value of lateral acceleration fluctuation, and peak value of lateral acceleration fluctuation. Calculate the feel fluctuation based on the effective value of torque fluctuation, peak value of torque fluctuation, and number of zero crossings per unit time.

[0010] The Mamdani-type fuzzy inference method is used to determine the continuous road surface grade value based on the effective values ​​of torque fluctuation, speed fluctuation, lateral acceleration fluctuation, and vehicle speed signal. The fuzzy rule base also corrects the continuous road surface grade value or outputs a compensation suppression state based on the number of zero crossings per unit time, the ratio of peak torque fluctuation to effective torque fluctuation, steering wheel angle rate, and torque change rate.

[0011] Once the continuous road surface grade reaches the activation threshold, the vehicle speed exceeds the minimum compensation speed threshold, and the identification activation duration is maintained, a recursive least squares method with an adaptive forgetting factor is used to identify discrete system parameters online. Based on the nominal steering system model, feature extraction sampling period, and second-order vibration system discretization method, a parameter conversion relationship is established between the discrete system parameters and the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system.

[0012] A fast Fourier transform is performed on the torque fluctuation signal to obtain the dominant frequency of the road excitation. The system's natural frequency and equivalent damping ratio are calculated based on the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system. The road excitation amplitude is calculated based on the system's amplitude-frequency gain and the peak value of the torque fluctuation.

[0013] The compensation torque amplitude is calculated based on the road excitation amplitude, the main frequency of the road excitation, the system's natural frequency, the system's equivalent damping ratio, the road grade, and the vehicle speed signal. The direction of the compensation torque is determined based on the direction of the speed fluctuation component and the phase correction of the compensation torque.

[0014] A maximum limit is applied to the compensation torque. When compensation begins, the compensation torque increases from zero to the target compensation torque in a ramp-like manner. When compensation ends, the compensation torque decreases to zero exponentially. The compensation torque is divided by the steering motor torque coefficient to form the Q-axis current compensation amount. This Q-axis current compensation amount is then superimposed with the Q-axis current generated by the power steering control to form the target Q-axis current.

[0015] In addition, the present invention also provides an EPS bumpy road vibration suppression system with multi-signal fusion and parameter identification, including: a torque sensor for acquiring torque sensor signals;

[0016] Rotor position sensor, used to acquire rotor position sensor signals;

[0017] Lateral acceleration sensor, used to collect lateral acceleration signals;

[0018] Vehicle speed signal interface, used to acquire vehicle speed signals;

[0019] EPS controller, including:

[0020] The road condition recognition unit is used to filter and extract features from torque sensor signals, rotor position sensor signals and lateral acceleration signals through sliding time windows to obtain torque fluctuation features, speed fluctuation features and lateral acceleration fluctuation features. Based on the torque fluctuation features, speed fluctuation features, lateral acceleration fluctuation features and vehicle speed signals, it performs fuzzy inference to obtain continuous road level values ​​and determines the bumpy road condition through hysteresis comparison.

[0021] The parameter identification unit is used to update the steering system parameters using an online parameter identification method with an adaptive forgetting factor when the identification start condition is met under bumpy road conditions, thereby obtaining the main frequency of road excitation, the amplitude of road excitation, the natural frequency of the system, and the equivalent damping ratio of the system.

[0022] The compensation torque calculation unit is used to calculate the compensation torque based on the main frequency of road excitation, the amplitude of road excitation, the natural frequency of the system, the equivalent damping ratio of the system, the road grade and the vehicle speed signal. It performs amplitude limiting, slope entry and exponential decay exit processing on the compensation torque, converts the processed compensation torque into the Q-axis current compensation amount and adds it to the Q-axis current generated by the power assist control to form the Q-axis target current.

[0023] The steering motor, connected to the EPS controller, is used to output compensating torque based on the target current of the Q-axis to suppress steering wheel vibration caused by bumpy roads.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] This invention comprehensively utilizes torque fluctuation characteristics, speed fluctuation characteristics, lateral acceleration fluctuation characteristics, and vehicle speed signals for multi-signal fusion judgment, which can reduce the interference of driver active steering, vehicle speed changes, and sensor noise on the identification results of bumpy road conditions. The Mamdani-type fuzzy inference method outputs continuous road level values ​​based on different signal feature combinations, and handles high-frequency continuous vibration, low-frequency large impact, and driver active steering conditions based on the number of zero crossings per unit time, the ratio of peak torque fluctuation to effective torque fluctuation, steering wheel angle rate, and torque change rate. A recursive least squares method with an adaptive forgetting factor adjusts parameters to track speed based on the continuous road level value. The forgetting factor decreases when the continuous road level value increases and increases when the continuous road level value decreases or the feel fluctuation reaches a steady state. The feel fluctuation comprehensively reflects the torque fluctuation intensity, peak torque fluctuation, and torque fluctuation frequency, and can serve as a calibration basis for evaluating the effect of forgetting factor adjustment and compensation. The compensation weight is determined based on the main frequency of road excitation and the natural frequency of the system. Low-frequency driver-driven active steering components are limited by a pre-stage bandpass filter and compensation suppression rules, while high-frequency noise is limited by a pre-stage low-pass filter and bandpass filter. Road surface excitation near the system's natural frequency participates in the compensation torque calculation according to the compensation weight. Slope entry, exponential decay exit, and two-stage hysteresis control can reduce sudden changes in compensation torque during the start-up and shutdown of the compensation function. This invention utilizes existing signals from the EPS controller and the steering motor to achieve vibration suppression without adding dedicated road surface detection sensors or mechanical structures. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 This is the main flowchart of the method of the present invention.

[0028] Figure 2 This is a flowchart of the branch process for determining the condition of bumpy roads in this invention.

[0029] Figure 3 This is a flowchart of the parameter identification sub-process with covariance matrix correction of the present invention.

[0030] Figure 4 This is the hysteresis time-domain response diagram for EPS bumpy road surface level identification in this invention.

[0031] Figure 5 This is the hysteresis loop characteristic curve for the fourth-level road surface identification of the present invention. Detailed Implementation

[0032] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0033] The following is in conjunction with the appendix Figures 1-5 The embodiments of the present invention will be described in detail below.

[0034] This invention discloses an EPS (Electric Power Suppressor) bumpy road vibration suppression system based on multi-signal fusion and parameter identification, including a torque sensor, a rotor position sensor, a lateral acceleration sensor, a vehicle speed signal interface, an EPS controller, and a steering motor. The EPS controller includes a road condition recognition unit, a parameter identification unit, and a compensation torque calculation unit.

[0035] The road condition recognition unit receives signals from the torque sensor, rotor position sensor, lateral acceleration, and vehicle speed. It outputs torque fluctuation characteristics, speed fluctuation characteristics, lateral acceleration fluctuation characteristics, feel fluctuation, continuous road surface grade, bumpy road surface condition, and compensation suppression condition. The parameter identification unit receives torque fluctuation components, feel fluctuation, continuous road surface grade, bumpy road surface condition, compensation suppression condition, and motor output torque. It outputs the main road excitation frequency, road excitation amplitude, system natural frequency, and system equivalent damping ratio. The compensation torque calculation unit receives speed fluctuation components, vehicle speed signal, continuous road surface grade, compensation suppression condition, main road excitation frequency, road excitation amplitude, system natural frequency, and system equivalent damping ratio. It outputs the compensation torque and Q-axis current compensation amount. The steering motor control unit receives the Q-axis current compensation amount and the Q-axis current generated by the power steering control, forms the target Q-axis current, and drives the steering motor.

[0036] The torque sensor sampling frequency is 1kHz. The torque sensor signal is filtered by a first-order low-pass filter with a cutoff frequency of 50Hz to remove high-frequency noise, and then filtered by a second-order Butterworth bandpass filter with a passband range of 5Hz to 50Hz to extract the torque fluctuation component. The rotor position sensor uses a 12-bit resolver. The rotor position sensor sampling frequency is 1kHz. The motor speed is obtained using a combination of discrete differentiation and filtering. The lateral acceleration sensor sampling frequency is 500Hz, with a measurement range of ±3g. The lateral acceleration signal is converted to a 1kHz feature extraction time base using a sample-and-hold method. The lateral acceleration signal is filtered by a second-order Butterworth bandpass filter with a passband range of 5Hz to 50Hz to extract the lateral acceleration fluctuation component. The vehicle speed signal comes from the CAN bus, with an update frequency of 100Hz. The vehicle speed signal that has most recently passed the validity check is maintained between adjacent vehicle speed messages. The sliding time window length is 50ms. The torque sensor signal, rotor position sensor signal, and lateral acceleration signal after time base conversion are all processed at a 1kHz feature extraction frequency, and the sliding time window contains 50 sampling points. The second-order Butterworth bandpass filter is discretized using the bilinear transform method and implemented in the EPS controller.

[0037] The torque sensor signal is decomposed according to the following formula: ;

[0038] In the formula, for The torque sensor signal at any given time, in N·m; for The DC component of torque that constantly reflects the driver's steady-state steering force, expressed in N·m; for The torque fluctuation component that constantly reflects the dynamic changes in road surface excitation, expressed in N·m; It represents continuous time, measured in seconds (s).

[0039] The effective value of torque ripple is calculated according to the following formula:

[0040] ;

[0041] In the formula, For The effective value of torque fluctuation within the sliding time window at the end time, in N·m; The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. For the first time window within the sliding time window The torque fluctuation value at each sampling point is expressed in N·m. For summation operations; This refers to the arithmetic square root operation.

[0042] The peak torque fluctuation is calculated using the following formula:

[0043] ;

[0044] In the formula, For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; This is for calculating the maximum value. This is for absolute value operations.

[0045] The motor speed in the discrete time domain is calculated according to the following formula:

[0046] ;

[0047] In the formula, For the first The motor speed at each sampling time, in rad / s; For the first The rotor position at each sampling time, in rad; For the first The rotor position at each sampling time, in rad; The feature extraction sampling period is expressed in seconds (s). This represents the discrete sampling time sequence number, which is dimensionless.

[0048] Motor speed can be decomposed according to the following formula:

[0049] ;

[0050] In the formula, For the first The average rotational speed at each sampling moment reflects the driver's steering speed, and the unit is rad / s; For the first Each sampling time reflects the rotational speed fluctuation component that reflects the road surface excitation speed fluctuation, with the unit being rad / s.

[0051] The effective value of speed fluctuation is calculated according to the following formula:

[0052] ;

[0053] In the formula, For The effective value of the rotational speed fluctuation within the sliding time window at the end time, in rad / s; For the first time window within the sliding time window The rotational speed fluctuation value at each sampling point, in rad / s; The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. It is a continuous time interval, measured in seconds (s). For summation operations; This refers to the arithmetic square root operation.

[0054] The peak value of the speed fluctuation is calculated according to the following formula:

[0055] ;

[0056] In the formula, For The peak value of the rotational speed fluctuation within the sliding time window at the end time, in rad / s; This is for calculating the maximum value. This is for absolute value operations.

[0057] The lateral acceleration signal is decomposed according to the following formula:

[0058] ;

[0059] In the formula, For the first The transverse acceleration signal at each sampling time, in m / s²; For the first Each sampling moment reflects the steady-state component of the lateral acceleration of the vehicle's steady-state steering motion, with units of m / s². For the first Each sampling moment reflects the lateral acceleration fluctuation component of the road surface bump excitation lateral vibration, with the unit being m / s².

[0060] The effective value of lateral acceleration fluctuation is calculated according to the following formula:

[0061] ;

[0062] In the formula, For The effective value of the lateral acceleration fluctuation within the sliding time window at the end time, in m / s². For the first time window within the sliding time window The lateral acceleration fluctuation value of each sampling point is expressed in m / s². The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. It is a continuous time interval, measured in seconds (s). For summation operations; This refers to the arithmetic square root operation.

[0063] The peak value of lateral acceleration fluctuation is calculated according to the following formula:

[0064] ;

[0065] In the formula, For The peak value of the lateral acceleration fluctuation within the sliding time window at the end time, in m / s². This is for calculating the maximum value. This is for absolute value operations.

[0066] The fluctuation in feel is calculated using the following formula:

[0067] ;

[0068] In the formula, for The tactile fluctuation at any time is a calibration evaluation value formed by weighting different engineering dimension characteristics, without being assigned a unified physical dimension. The weighting coefficient for the effective value of torque fluctuation is determined based on the calibration range of the effective value of torque fluctuation. This is the weighting factor for the peak torque fluctuation; the engineering quantity scale is determined based on the calibration range of the peak torque fluctuation. This is the weighting factor for the number of zero crossings per unit time. The scale of the project quantity is determined based on the calibration range of the number of zero crossings per unit time. For The effective value of torque fluctuation within the sliding time window at the end time, in N·m; For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; for The number of times the torque fluctuation signal crosses the zero line per unit time, expressed in Hz; , and The values ​​are 0.4, 0.3 and 0.3 respectively.

[0069] The sign function value of the torque fluctuation signal is determined according to the following formula:

[0070] ;

[0071] In the formula, For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. For the first time window within the sliding time window The torque fluctuation value at each sampling point is expressed in N·m. This is the index of the sampling points within the sliding time window, and it is dimensionless.

[0072] The number of zero crossings per unit time is calculated using the following formula:

[0073] ;

[0074] In the formula, for The number of times the torque fluctuation signal crosses the zero line per unit time, expressed in Hz; The length of the sliding time window, in seconds; For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. The time is continuous, measured in seconds; a value of 2 indicates that a complete oscillation contains two zero crossings. For summation operations; This is for absolute value operations.

[0075] Required for calculating hand feel fluctuation , and All three input parameters use a fixed signal engineering scale. The same signal engineering scale is used for the same vehicle model and the same controller software version to ensure repeatable comparability of feel fluctuations.

[0076] The fuzzy determination of road surface levels employs the Mamdani-type fuzzy inference method. This method includes input fuzzification, fuzzy rule inference, and output defuzzification. Bumpy roads are classified into four levels: Level 1, Level 2, Level 3, and Level 4. Level 1 corresponds to slight bumps, Level 2 to moderate bumps, Level 3 to severe bumps, and Level 4 to extreme bumps. The fuzzy inference inputs include the effective values ​​of torque fluctuation, engine speed fluctuation, lateral acceleration fluctuation, and vehicle speed signal.

[0077] The domain of discourse for the effective value of torque fluctuation is [0, 6] N·m. A small fuzzy subset of the effective value of torque fluctuation is assigned a descending half-trapezoidal membership function, with a membership of 1 in the interval from 0 N·m to 0.2 N·m, and a linear decrease to 0 in the interval from 0.2 N·m to 0.5 N·m. A small fuzzy subset of the effective value of torque fluctuation is assigned a triangular membership function, with a membership of 0 at 0.3 N·m, 1 at 0.8 N·m, and 0 at 1.5 N·m.

[0078] The middle fuzzy subset of the effective torque fluctuation value adopts a triangular membership function, with a membership degree of 0 at 1.0 N·m, 1 at 2.0 N·m, and 0 at 3.0 N·m. The large fuzzy subset of the effective torque fluctuation value adopts a triangular membership function, with a membership degree of 0 at 2.0 N·m, 1 at 3.0 N·m, and 0 at 4.5 N·m. The very large fuzzy subset of the effective torque fluctuation value adopts an ascending semi-trapezoidal membership function, with a membership degree of 0 at 3.5 N·m, linearly increasing to 1 in the range of 3.5 N·m to 5.0 N·m, and remaining at 1 in the range of 5.0 N·m to 6.0 N·m.

[0079] After converting the effective values ​​of speed fluctuations to r / min, the universe of discourse is [0, 150] r / min. The very small fuzzy subset of the effective values ​​of speed fluctuations uses a descending half-trapezoidal membership function, with a membership of 1 in the range of 0 r / min to 5 r / min, and a linear decrease to 0 in the range of 5 r / min to 10 r / min. The small fuzzy subset of the effective values ​​of speed fluctuations uses a triangular membership function, with a membership of 0 at 5 r / min, 1 at 15 r / min, and 0 at 30 r / min. The medium fuzzy subset of the effective values ​​of speed fluctuations uses a triangular membership function, with a membership of 0 at 20 r / min, 1 at 40 r / min, and 0 at 60 r / min. The large fuzzy subset of the effective values ​​of speed fluctuations uses a triangular membership function, with a membership degree of 0 at 50 r / min, 1 at 75 r / min, and 0 at 100 r / min. The very large fuzzy subset of the effective values ​​of speed fluctuations uses an ascending semi-trapezoidal membership function, with a membership degree of 0 at 80 r / min, linearly increasing to 1 in the range of 80 r / min to 120 r / min, and remaining at 1 in the range of 120 r / min to 150 r / min.

[0080] After converting the effective value of lateral acceleration fluctuations into a multiple of gravitational acceleration, the universe of discourse is [0, 0.8]g. A small fuzzy subset of the effective value of lateral acceleration fluctuations uses a descending half-trapezoidal membership function, with a membership degree of 1 in the interval from 0g to 0.02g, and a linear decrease to 0 in the interval from 0.02g to 0.05g. A small fuzzy subset of the effective value of lateral acceleration fluctuations uses a triangular membership function, with a membership degree of 0 at 0.03g, 1 at 0.08g, and 0 at 0.15g. A medium fuzzy subset of the effective value of lateral acceleration fluctuations uses a triangular membership function, with a membership degree of 0 at 0.1g, 1 at 0.2g, and 0 at 0.3g. The large fuzzy subset of the effective values ​​of lateral acceleration fluctuations adopts a triangular membership function, with a membership degree of 0 at 0.25g, 1 at 0.4g, and 0 at 0.55g. The very large fuzzy subset of the effective values ​​of lateral acceleration fluctuations adopts an ascending semi-trapezoidal membership function, with a membership degree of 0 at 0.45g, linearly increasing to 1 in the interval from 0.45g to 0.65g, and remaining at 1 in the interval from 0.65g to 0.8g.

[0081] The universe of discourse for the vehicle speed signal is [0, 120] km / h. The low-speed fuzzy subset of the vehicle speed signal uses a descending semi-trapezoidal membership function, with a membership of 1 in the range of 0 km / h to 20 km / h, and a linear decrease to 0 in the range of 20 km / h to 40 km / h. The medium-speed fuzzy subset of the vehicle speed signal uses a triangular membership function, with a membership of 0 at 30 km / h, 1 at 60 km / h, and 0 at 90 km / h. The high-speed fuzzy subset of the vehicle speed signal uses an ascending semi-trapezoidal membership function, with a membership of 0 at 70 km / h, a linear increase to 1 in the range of 70 km / h to 100 km / h, and a membership of 1 in the range of 100 km / h to 120 km / h.

[0082] The universe of discourse for outputting continuous pavement grade values ​​is [1,4]. The membership function for grade 1 has a membership of 1 when the continuous pavement grade value is 1, and linearly decreases to 0 as the continuous pavement grade value increases from 1 to 2. The membership function for grade 2 uses a triangular membership function, with a membership of 0 when the continuous pavement grade value is 1.5, 1 when the continuous pavement grade value is 2.0, and 0 when the continuous pavement grade value is 2.5. The membership function for grade 3 also uses a triangular membership function, with a membership of 0 when the continuous pavement grade value is 2.5, 1 when the continuous pavement grade value is 3.0, and 0 when the continuous pavement grade value is 3.5. The membership function for grade 4 has a membership of 0 when the continuous pavement grade value is 3, and linearly increases to 1 as the continuous pavement grade value increases from 3 to 4.

[0083] For each fuzzy rule, the minimum value among all input membership degrees is taken as the activation strength of the fuzzy rule. For the same road surface grade, the maximum value among the activation strengths of the corresponding fuzzy rules is taken as the road surface grade membership degree.

[0084] The grade value of a continuous pavement is calculated according to the following formula:

[0085] ;

[0086] In the formula, for The continuous road surface grade value after defuzzification at each time step is dimensionless and ranges from [1,4]. The number of fuzzy rules involved in defuzzification is dimensionless. This is a fuzzy rule index, dimensionless; for Time of the first The activation intensity of a fuzzy rule is dimensionless. For the first The fuzzy rule outputs the center value of the level, which is dimensionless; It is a continuous time interval, measured in seconds (s). For summation operations.

[0087] The center values ​​for levels 1, 2, 3, and 4 are 1.0, 2.0, 3.0, and 4.0, respectively.

[0088] The fuzzy rule base includes the following rules: Rule 1: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all very small, output level 1. Rule 2: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all small, output level 2. Rule 3: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all medium, output level 2. Rule 4: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all medium, output level 3. Rule 5: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all large, output level 3. Rule 6: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all large, output level 3. Rule 7: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all large, output level 4. Rule 8: When the effective values ​​of torque fluctuation, speed fluctuation, and lateral acceleration fluctuation are all very large, output level 4. Rule 9: When the effective value of torque fluctuation is medium, the effective value of speed fluctuation is large, and the effective value of lateral acceleration fluctuation is small, output level 3. Rule 10: When the effective value of torque fluctuation is small, the effective value of speed fluctuation is large, and the effective value of lateral acceleration fluctuation is medium, output level 2. Rule 11: When the vehicle speed signal is low and the effective value of torque fluctuation is medium, decrease the continuous road surface level by 1. Rule 12: When the vehicle speed signal is high and the effective value of torque fluctuation is medium, decrease the continuous road surface level by 1. Rule 13: When the number of zero crossings per unit time is greater than 30Hz and the ratio of the peak torque fluctuation to the effective value of torque fluctuation is less than 1.5, decrease the continuous road surface level by 1. Rule 14: When the number of zero crossings per unit time is less than 10Hz and the ratio of the peak torque fluctuation to the effective value of torque fluctuation is greater than 2.5, increase the continuous road surface level by 1. Rule 15: When the steering wheel angle rate is greater than 50° / s, output compensation suppression mode. Rule 16: When the torque change rate is greater than 10 N·m / ms, output compensation suppression mode.

[0089] Rules 1 to 10 constitute the basic rules. When the current input membership combination is not completely consistent with rules 1 to 10, the nearest basic rule is selected according to the minimum difference between the current input membership and the corresponding predecessor membership of rules 1 to 10; if there are two or more identical minimum differences, the weighted average of the corresponding output level center values ​​is used. The nearest basic rule processing is used to ensure that all input combinations generate continuous road surface level values.

[0090] The continuous road surface grade value uses a first-level hysteresis band with a width of 0.6. When the continuous road surface grade value reaches 2.0 and the vehicle speed is greater than the minimum compensation speed threshold for 100ms, the road surface condition recognition unit outputs a bumpy road surface state. When the continuous road surface grade value is lower than 1.4 and the exit delay time is reached, the road surface condition recognition unit exits the bumpy road surface state. The minimum compensation speed threshold is calibrated based on historical road data of the vehicle under low-speed steering and low-speed bumpy road surface conditions. The exit delay time is calibrated based on the torque fluctuation decay time after the vehicle leaves the bumpy road surface. The compensation mode (i.e., the working mode in which the EPS controller executes compensation torque output and superposition) is set with a second-level hysteresis. The compensation mode is activated when the bumpy road surface state is valid and the compensation suppression state is invalid. The compensation mode exits when the bumpy road surface state is invalid or the compensation suppression state is valid. The compensation mode is maintained for at least 100ms after activation and at least 100ms after exit to avoid continuous switching of the compensation mode.

[0091] The steering wheel angle is obtained based on the rotor position sensor signal and the nominal steering ratio. The nominal steering ratio is derived from the mechanical parameters of the EPS system.

[0092] The EPS system under bumpy road conditions is represented by the following formula:

[0093] ;

[0094] In the formula, The equivalent rotational inertia of the steering system is expressed in kg·m². for The steering wheel angle at any given moment, measured in rad. for The angular velocity of the steering wheel at any given moment, in rad / s; for The steering wheel angular acceleration at a given moment, in rad / s². The equivalent damping coefficient of the steering system is expressed in N·m·s / rad. The equivalent torsional stiffness of the steering system is expressed in N·m / rad. for The motor output torque at any given time, expressed in N·m; for The road surface excitation torque at time t, expressed in N·m; For continuous time, the unit is seconds; a dot above the letter indicates the first derivative with respect to continuous time; two dots above the letter indicate the second derivative with respect to continuous time.

[0095] Within a single identification time window, the road surface excitation torque is expressed by the following formula:

[0096] ;

[0097] In the formula, The amplitude of the road surface excitation within a single identification time window is expressed in N·m. The dominant frequency of road surface excitation within a single identification time window, in Hz; The initial phase of the road surface excitation within a single identification time window, in rad; for The random perturbation torque at time t, expressed in N·m; It is a sine function; Pi; It represents continuous time, measured in seconds (s).

[0098] After discretizing the second-order vibration system, the input-output model is established according to the following formula:

[0099] ;

[0100] In the formula, For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; , , and The parameters of the discrete system to be identified are dimensionless. This represents the discrete sampling time sequence number, which is dimensionless.

[0101] The parameter estimation vector is established according to the following formula:

[0102] ;

[0103] In the formula, For the first The parameter estimation vector at each sampling time point; , , and The first The discrete system parameter estimates at each sampling time point are dimensionless; superscript. Indicates matrix transpose; This represents the discrete sampling time sequence number, which is dimensionless.

[0104] The observation vector is established according to the following formula:

[0105] ;

[0106] In the formula, For the first The observation vector at each sampling time; For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; For the first Motor output torque input at each sampling time, in N·m; superscript Indicates matrix transpose; This represents the discrete sampling time sequence number, which is dimensionless.

[0107] The system input-output relationship is expressed by the following formula:

[0108] ;

[0109] In the formula, For the first Torque fluctuation output at each sampling time, in N·m; For the first The transpose of the observation vector at each sampling time; For the first The parameter estimation vector at each sampling time point; This represents the discrete sampling time sequence number, which is dimensionless.

[0110] The recursive gain vector is updated according to the following formula:

[0111] ;

[0112] In the formula, For the first The recursive gain vector at each sampling time; For the first The covariance matrix at each sampling time; For the first The observation vector at each sampling time; For the first The transpose of the observation vector at each sampling time; For the first The forgetting factor at each sampling time is dimensionless. This represents the discrete sampling time sequence number, which is dimensionless.

[0113] The parameter estimation vector is updated according to the following formula:

[0114] ;

[0115] In the formula, For the first The parameter estimation vector at each sampling time point; For the first The parameter estimation vector at each sampling time point; For the first The recursive gain vector at each sampling time; For the first Torque fluctuation output at each sampling time, in N·m; For the first The transpose of the observation vector at each sampling time; This represents the discrete sampling time sequence number, which is dimensionless.

[0116] The covariance matrix is ​​updated according to the following formula:

[0117] ;

[0118] In the formula, For the first The covariance matrix at each sampling time; It is the identity matrix; For the first The recursive gain vector at each sampling time; For the first The transpose of the observation vector at each sampling time; For the first The covariance matrix at each sampling time; For the first The forgetting factor at each sampling time is dimensionless. This represents the discrete sampling time sequence number, which is dimensionless.

[0119] The forgetting factor is adaptively adjusted according to the following formula:

[0120] ;

[0121] In the formula, For the first The forgetting factor at each sampling time is dimensionless. The maximum forgetting factor is dimensionless. It is the smallest forgetting factor, dimensionless; For the first The normalized pavement grade is obtained by linearly mapping the continuous pavement grade values ​​to the [0,1] interval at each sampling time, and is dimensionless. This represents the discrete sampling time sequence number, which is dimensionless. The maximum forgetting factor is 0.995, the minimum forgetting factor is 0.950, and the initial forgetting factor is 0.985.

[0122] The initial covariance matrix is The values ​​100, 100, 10, and 10 correspond to the initial covariances of the four discrete system parameters in the parameter estimation vector, respectively. The initial values ​​of the parameter estimation vector are set according to the nominal steering system model.

[0123] Parameter identification begins after the continuous road surface grade value reaches 2.0 and remains so for 100ms. Parameter updates are paused when the steering wheel angle rate exceeds 50° / s or the torque change rate exceeds 10 N·m / ms. When the rate of change of steering feel fluctuation is lower than the steady-state judgment threshold, the forgetting factor is increased, but the increased forgetting factor does not exceed the maximum forgetting factor. When the increase in steering feel fluctuation during the road surface abrupt change judgment time exceeds the road surface abrupt change threshold, the forgetting factor is decreased, but the decreased forgetting factor is not lower than the minimum forgetting factor. The steady-state judgment threshold is calibrated based on historical data of the rate of change of steering feel fluctuation corresponding to normal paved roads and stable bumpy roads. The road surface abrupt change judgment time and the road surface abrupt change threshold are calibrated based on the increase in steering feel fluctuation when a vehicle enters a gravel road, washboard road, Belgian road, and speed bump from a normal paved road.

[0124] The parameter estimation vector is subjected to amplitude limiting and first-order low-pass filtering. Parameter validity is checked every 100ms. The valid range of parameters is determined based on the nominal steering system model and bench identification results. When the parameter estimate exceeds the valid range, the parameter identification unit restores the initial value of the parameter estimation vector and the initial covariance matrix.

[0125] Based on the nominal steering system model, feature extraction sampling period, and discretization method, a parameter conversion relationship is pre-established between the discrete system parameters and the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system. This parameter conversion relationship is stored in the EPS controller using a discrete parameter lookup table. The discrete parameter lookup table is obtained by discretizing the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system from the nominal steering system model using a parameter mesh, and calculating the corresponding discrete system parameters using the same feature extraction sampling period and discretization method. The parameter identification unit performs interpolation in the discrete parameter lookup table based on the parameter estimation vector to obtain the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system. The interpolation results are limited by the effective range of the parameters.

[0126] A Fast Fourier Transform (FFT) is performed on the torque fluctuation signal. The FFT uses a 256-point transform with a sliding step size of 64 points. When the torque sensor sampling frequency is 1kHz, the frequency resolution is 3.90625Hz. When the ratio of the amplitude of the main peak to the amplitude of the secondary peak is greater than 1.5, the frequency corresponding to the main peak is determined as the effective road excitation main frequency. When the ratio is not greater than 1.5, the parameter identification unit retains the previous effective road excitation main frequency and sets the current frequency's effective state to invalid. The compensation torque calculation unit uses the previous effective road excitation main frequency when the current frequency's effective state is invalid.

[0127] The system amplitude-frequency gain is calculated according to the following formula:

[0128] ;

[0129] In the formula, For the steering system at frequency The amplitude-frequency gain at that point is dimensionless. The frequency to be calculated is in Hz. The system's inherent frequency, measured in Hz; The system's equivalent damping ratio is dimensionless. This refers to the arithmetic square root operation.

[0130] The road surface excitation amplitude is calculated according to the following formula:

[0131] ;

[0132] In the formula, The amplitude of the road surface excitation within a single identification time window is expressed in N·m. For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; The amplitude-frequency gain of the steering system at the main frequency of road excitation is dimensionless. The dominant frequency of road surface excitation within a single identification time window, in Hz; It represents continuous time, measured in seconds (s).

[0133] The system's equivalent damping ratio is calculated using the following formula:

[0134] ;

[0135] In the formula, The system's equivalent damping ratio is dimensionless. The equivalent damping coefficient of the steering system is expressed in N·m·s / rad. The equivalent rotational inertia of the steering system is expressed in kg·m². The equivalent torsional stiffness of the steering system is expressed in N·m / rad. This refers to the arithmetic square root operation.

[0136] The system's natural angular frequency is calculated using the following formula:

[0137] ;

[0138] In the formula, The system's natural angular frequency is expressed in rad / s. The equivalent torsional stiffness of the steering system is expressed in N·m / rad. The equivalent rotational inertia of the steering system is expressed in kg·m². This refers to the arithmetic square root operation.

[0139] The system's natural frequency is calculated using the following formula:

[0140] ;

[0141] In the formula, The system's inherent frequency, measured in Hz; The system's natural angular frequency is expressed in rad / s. Pi; The equivalent torsional stiffness of the steering system is expressed in N·m / rad. The equivalent rotational inertia of the steering system is expressed in kg·m². This refers to the arithmetic square root operation.

[0142] The basic amplitude of the compensating torque is calculated according to the following formula:

[0143] ;

[0144] In the formula, The compensation torque is the basic amplitude, expressed in N·m. The amplitude of the road surface excitation within a single identification time window is expressed in N·m. It is the basic compensation gain coefficient, which is dimensionless. The compensation weight at the dominant frequency of the road excitation is dimensionless. The dominant frequency of the road surface excitation within a single identification time window is expressed in Hz.

[0145] The compensation weight is calculated according to the following formula:

[0146] ;

[0147] In the formula, For frequency The compensation weight at the location is dimensionless. The frequency to be calculated is in Hz. The system's inherent frequency, measured in Hz; The bandwidth adjustment coefficient for compensating for the frequency range is dimensionless. This refers to the arithmetic square root operation.

[0148] The base compensation gain coefficient is 0.6. The bandwidth adjustment coefficient is 0.3. When the bandwidth adjustment coefficient is 0.3, the compensation weight reaches a relatively high value in the vicinity of the system's natural frequency. When it is necessary to expand the compensation frequency range, the bandwidth adjustment coefficient can be adjusted to 0.7 through road calibration. The low-frequency driver active steering component is limited by a 5Hz to 50Hz bandpass filter and compensation suppression rules. High-frequency sensor noise is limited by a first-order low-pass filter with a cutoff frequency of 50Hz and a bandpass filter.

[0149] The final compensation torque amplitude is calculated according to the following formula:

[0150] ;

[0151] In the formula, The final compensation torque amplitude is expressed in N·m. The compensation torque is the basic amplitude, expressed in N·m. The pavement grade gain coefficient is determined based on the pavement grade and is dimensionless. The vehicle speed correlation coefficient is determined based on the vehicle speed signal and is dimensionless.

[0152] The road surface grade gain coefficient and vehicle speed correlation coefficient are stored in the EPS controller using a lookup table. The road surface grade gain coefficient increases with the continuous increase of the road surface grade value. The vehicle speed correlation coefficient gradually decreases below the minimum compensation speed threshold, maintains the calibrated value within the compensation working speed range, and gradually decreases in areas with increased high-speed interference. The road surface grade gain coefficient and vehicle speed correlation coefficient are calibrated based on data from gravel roads, washboard roads, Belgian roads, and speed bump roads. The calibration target is to retain some torque fluctuations after compensation, with the basic compensation gain coefficient corresponding to a calibration target of retaining 40% of the road surface feedback. The final compensation torque amplitude is limited to a maximum value, with a maximum absolute value of 2.5 N·m.

[0153] The direction of the compensation torque is determined according to the damping control direction, which is opposite to the direction of the speed fluctuation component. The damping compensation torque is calculated according to the following formula:

[0154] ;

[0155] In the formula, for The damping compensation torque at time t, expressed in N·m; This represents the damping gain, measured in N·m·s / rad; the negative sign indicates that the direction of the damping compensation torque is opposite to the direction of the speed fluctuation component. for The rotational speed fluctuation component that constantly reflects the road surface excitation speed fluctuation, with the unit being rad / s; It represents continuous time, measured in seconds (s).

[0156] The damping gain is determined based on the continuous pavement grade value. The damping gain corresponding to grade 2 is 0.05 N·m·s / rad, the damping gain corresponding to grade 3 is 0.08 N·m·s / rad, and the damping gain corresponding to grade 4 is 0.12 N·m·s / rad. When the continuous pavement grade value is between the center values ​​of adjacent grades, the damping gain is a linear interpolation of the adjacent damping gains.

[0157] The phase correction of the compensating torque is calculated according to the following formula:

[0158] ;

[0159] In the formula, This is the phase correction amount for the compensating torque, expressed in rad. It is the arctangent function; The system's equivalent damping ratio is dimensionless. The dominant frequency of road surface excitation within a single identification time window, in Hz; This is the system's inherent frequency, measured in Hz.

[0160] The phase lead time constant is 15ms. The compensation torque calculation unit determines the target compensation torque in the direction of the band based on the sign of the damping compensation torque, the phase correction amount of the compensation torque, and the phase lead time constant. The absolute value of the target compensation torque in the direction of the band does not exceed the amplitude of the final compensation torque.

[0161] When the compensation start-up conditions are met, the compensation torque increases according to the following formula:

[0162] ;

[0163] In the formula, for The compensation torque at any given time, expressed in N·m; for The target compensation torque is determined at all times based on the final compensation torque amplitude and the compensation torque direction, and the unit is N·m; The cumulative time since entering compensation mode, in seconds; The time for the slope to ascend is measured in seconds. It represents continuous time, measured in seconds (s).

[0164] The ramp-up time is 200ms. After the cumulative time after entering compensation mode reaches the ramp-up time, the compensation torque is updated according to the target compensation torque, and the maximum value limit continues to be executed.

[0165] When the compensation exit condition is met, the compensation torque decays exponentially according to the following formula:

[0166] ;

[0167] In the formula, For the first The compensation torque at each compensation torque update moment, in N·m; For the first The compensation torque at each compensation torque update moment, in N·m; The compensation torque update period is expressed in seconds (s). is the exponentially decaying time constant, in seconds; It is a natural exponential function; This represents the discrete sampling time sequence number, which is dimensionless.

[0168] The exponential decay time constant is 300ms. The compensation torque is set to 0 once its absolute value decreases to the value corresponding to the minimum controllable torque of the steering motor. The minimum controllable torque of the steering motor is obtained based on the Q-axis current detection resolution and the steering motor torque coefficient calibration.

[0169] The target current for the Q-axis is calculated using the following formula:

[0170] ;

[0171] In the formula, For the first The target Q-axis current at each compensation torque update moment, in A; For the first The Q-axis current generated by the assist control at each compensation torque update moment is expressed in A. For the first The compensation torque at each compensation torque update moment, in N·m; This is the torque coefficient of the steering motor, expressed in N·m / A; This represents the discrete sampling time sequence number, which is dimensionless.

[0172] The target current for the Q-axis is limited by the allowable current of the steering motor, the bus voltage, the safety status of the EPS system, and the fault status of the EPS system.

[0173] To facilitate a better understanding of the present invention by those skilled in the art, the present invention will be further illustrated below with reference to specific implementation examples.

[0174] Example 1: This example uses an EPS controller with an NXP S32K344 as the main microcontroller. The steering motor is a three-pole non-polar permanent magnet synchronous motor with a peak torque of 8.5 N·m and a rated speed of 1100 r / min. The steering motor current control loop executes at a frequency of 16 kHz, and the steering motor current control cycle is 62.5 μs. The update frequency of the road condition recognition unit, parameter identification unit, and compensation torque calculation unit is 1 kHz, with a corresponding update cycle of 1 ms. The Q-axis current compensation output by the road condition recognition unit, parameter identification unit, and compensation torque calculation unit is executed by the steering motor current control loop in subsequent steering motor current control cycles. The torque sensor sampling frequency is 1 kHz. The rotor position sensor sampling frequency is 1 kHz. The lateral acceleration sensor sampling frequency is 500 Hz. The vehicle speed signal update frequency is 100 Hz. The sliding time window length is 50 ms, and the sliding time window contains 50 feature extraction sampling points.

[0175] The road condition recognition unit calculates the effective value of torque fluctuation, peak torque fluctuation, effective value of speed fluctuation, peak speed fluctuation, effective value of lateral acceleration fluctuation, peak lateral acceleration fluctuation, number of zero crossings per unit time, and feel fluctuation. The road condition recognition unit obtains the continuous road surface grade value based on a fuzzy rule base. When the continuous road surface grade value reaches 2.0 and the vehicle speed exceeds the minimum compensated vehicle speed threshold for 100 ms, the parameter identification unit initiates parameter updates. The parameter identification unit performs recursive least squares parameter updates at an update frequency of 1 kHz. The initial forgetting factor is 0.985, and the forgetting factor is adjusted between 0.950 and 0.995 based on the continuous road surface grade value. The parameter identification unit performs parameter validity checks every 100 ms. When the parameter estimate exceeds the valid parameter range, the parameter identification unit restores the initial value of the parameter estimate vector and the initial covariance matrix. The Fast Fourier Transform uses a 256-point transform and a 64-point sliding step size. The frequency resolution is 3.90625 Hz. When the ratio of the amplitude of the main peak to the amplitude of the secondary peak in the spectrum is greater than 1.5, the parameter identification unit outputs the effective road excitation main frequency. The basic compensation gain coefficient is 0.6. The bandwidth adjustment coefficient is 0.3. The damping gains corresponding to levels 2, 3, and 4 are 0.05 N·m·s / rad, 0.08 N·m·s / rad, and 0.12 N·m·s / rad, respectively. The phase lead time constant is 15 ms. The maximum absolute value of the compensation torque is 2.5 N·m. The ramp-up time is 200 ms. The exponential decay time constant is 300 ms.

[0176] When a vehicle transitions from a paved road to a gravel road, washboard road, Belgian road, or speed bump, the road condition recognition unit outputs a continuous road surface grade value and a bumpy road surface condition. The parameter identification unit outputs the road excitation main frequency, road excitation amplitude, system natural frequency, and system equivalent damping ratio. The compensation torque calculation unit outputs the compensation torque and Q-axis current compensation amount. The steering motor control unit drives the steering motor to output the compensation torque based on the Q-axis target current. After the vehicle leaves the bumpy road surface, if the continuous road surface grade value drops below 1.4 and remains below the exit delay time, the compensation torque decreases to 0 in an exponential decay manner, and the Q-axis target current returns to the Q-axis current generated by the power assist control.

[0177] This invention was validated on a real vehicle using the R-EPS system. Test surfaces included gravel roads, washboard roads, Belgian roads, and speed bumps. Under Level 2 gravel road conditions, the effective value of steering wheel torque fluctuation decreased by 35% to 45%, and the peak value decreased by 30% to 40%. Under Level 3 washboard road conditions, the effective value of steering wheel torque fluctuation decreased by 40% to 55%, and the peak value decreased by 45% to 55%. Under Level 4 Belgian road conditions, the steering wheel torque fluctuation decreased by 30% to 40%. After the vehicle entered a normally paved road surface, the compensation torque decreased to 0 via exponential decay, and the Q-axis target current returned to the Q-axis current generated by the power steering control. The delay from road surface excitation to compensation torque output was less than 20ms. Slope entry and exponential decay exit reduced the rate of change of compensation torque during the start and stop of the compensation function.

[0178] Example 2: Based on the hardware and software architecture of Example 1, this example addresses the covariance matrix saturation and covariance matrix storm problems that are prone to occur in the online parameter identification process with adaptive forgetting factor, as well as the conflict between road vibration compensation torque and driver feel feedback under high vehicle speed and high steering force conditions. It introduces a covariance matrix correction process and a dynamic amplitude limiting correction mechanism.

[0179] When a real vehicle is driving for a long time or continuously traversing a flat road, due to insufficient excitation of the input signal, the covariance matrix in the recursive least squares algorithm with adaptive forgetting factor... The elements may continue to grow monotonically, triggering a covariance matrix storm, which will cause severe disturbances in parameter estimation when the road surface changes suddenly. Conversely, under severe and continuous bumpy conditions, the adaptive forgetting factor causes the covariance matrix to decay too quickly, which may lead to covariance matrix saturation and loss of the ability to track changes in system parameters.

[0180] To eliminate covariance matrix storms and covariance matrix saturation, the parameter identification unit in the first... The covariance matrix is ​​updated at each sampling time. Then, the trace of the covariance matrix is ​​calculated in real time:

[0181] ;

[0182] In the formula, For the first The trace of the covariance matrix at each sampling time point, dimensionless; For the first The covariance matrix at each sampling time; For the first The covariance matrix at the nth sampling time is the first... Line number The diagonal elements of the column head are dimensionless. Index of diagonal elements, dimensionless; These are discrete sampling time sequence numbers, dimensionless; For summation operations.

[0183] Set the lower limit threshold for the trace of the covariance matrix. Upper limit threshold of covariance matrix trace The parameter identification unit performs renormalization correction on the covariance matrix according to the following formula:

[0184] ;

[0185] In the formula, For the first The covariance matrix corrected at each sampling time; For the first The covariance matrix at each sampling time; For the first The trace of the covariance matrix at each sampling time point, dimensionless; The lower bound threshold of the covariance matrix trace is dimensionless. The upper limit threshold of the trace of the covariance matrix is ​​dimensionless. for The identity matrix; This represents the discrete sampling time sequence number, which is dimensionless.

[0186] When the trace of the covariance matrix is ​​below 0.1, it indicates that the matrix is ​​showing signs of saturation. This can be addressed by injecting a small positive definite perturbation along the upper diagonal. To maintain the algorithm's sensitivity to parameter mutations; when the trace of the covariance matrix exceeds 500.0, the matrix is ​​scaled proportionally to prevent numerical overflow and severe oscillations in parameter estimation caused by excessive covariance. In the... Calculate the recursive gain vector at each sampling time. At that time, the parameter identification unit automatically uses Replace the original Participate in matrix multiplication calculations.

[0187] In the compensation torque calculation unit, to prevent excessive bump compensation torque from weakening the driver's road feel or even interfering with the main steering intention when driving on high-speed curves or when the driver applies large hand force for emergency obstacle avoidance, the final compensation torque amplitude is adjusted. Perform dynamic amplitude limiting correction.

[0188] The dynamic amplitude limiting correction is calculated using the following formula to determine the final compensation torque amplitude after limiting:

[0189] ;

[0190] In the formula, for The final compensation torque amplitude after time-limiting, in N·m; The final compensation torque amplitude is expressed in N·m. for The upper limit of the dynamic allowable torque at any given time, in N·m; This is for calculating the minimum value; It represents continuous time, measured in seconds (s).

[0191] Dynamic allowable torque limit Based on the DC component of torque that reflects the driver's steady-state steering force and real-time vehicle speed signal The comprehensive construction is as follows:

[0192] ;

[0193] In the formula, for The upper limit of the dynamic allowable torque at any given time, in N·m; The maximum allowable compensation torque is calibrated to be 2.5 N·m. The force-weighted suppression coefficient is calibrated to 0.6. for The DC component of torque that constantly reflects the driver's steady-state steering force, expressed in N·m; The maximum steady-state hand force calibrated for the driver is set to 8.0 N·m. The high-speed attenuation factor is calibrated to 0.4; for The vehicle speed signal at any given time, in km / h; The benchmark maximum speed is set at 120 km / h. It is a continuous time interval, measured in seconds (s). This is for absolute value operations.

[0194] When the driver's steady-state steering force reaches 8.0 N·m and the vehicle speed reaches 120 km / h, the dynamic allowable torque limit automatically decreases to:

[0195] ;

[0196] It effectively constrains the physical boundaries of steering motor intervention under extreme conditions, returning active safety control entirely to the driver.

[0197] The EPS system incorporating a covariance matrix correction process and a dynamic amplitude limiting correction mechanism was validated in real vehicle bench and road tests. The vehicle cruised at 80 km / h on a composite test track including a smooth paved road and a continuous jagged bumpy road. In the transient response test where the vehicle suddenly transitioned from a paved road to a bumpy road, the traditional algorithm without covariance matrix correction experienced a parameter oscillation convergence period of up to 180 ms due to the covariance matrix exceeding its limit, resulting in a brief torque noise in the steering wheel. However, after adopting the covariance matrix correction process in this embodiment, the covariance matrix trace was stably limited to between 0.1 and 500.0, and parameter identification converged rapidly to the true physical value within 35 ms. The entire process delay from the occurrence of the bump to the smooth output of the compensation torque was shortened to 14 ms. In the high-speed, high-angle cornering bump test, the dynamic amplitude limiting correction mechanism smoothly reduced the upper limit of the compensation torque to 0.8 N·m, resulting in a clear and natural steering feel with no human intervention, and the system operated stably and reliably.

[0198] Example 3: This example further optimizes Example 2. This method is applicable to EPS systems using permanent magnet synchronous motors or brushless DC motors equipped with rotor position sensors, including R-EPS, P-EPS, DP-EPS, and some C-EPS systems using the aforementioned motor types. While the different EPS types differ in mechanical structure, they all possess the torque sensor signal, rotor position sensor signal, and controllable motor output torque upon which this invention relies. Therefore, they can all achieve vibration suppression on bumpy roads through the same signal processing and parameter identification framework.

[0199] Regarding the forgetting factor adjustment mechanism in the aforementioned recursive least squares parameter identification process with adaptive forgetting factor, it can be further clarified that it includes two layers of logic: The first layer is a basic adjustment mechanism based on the linear mapping of continuous road surface grade values ​​to the [0,1] interval, that is, when the continuous road surface grade value increases, the forgetting factor is linearly decreased to accelerate the parameter tracking speed, and when the continuous road surface grade value decreases, the forgetting factor is linearly increased to improve the stability of parameter estimation; the second layer is an auxiliary adjustment mechanism based on the rate of change of tactile fluctuation and the detection of road surface abrupt changes, that is, under steady-state conditions, the forgetting factor is further increased to suppress the amplification of parameter estimation noise, and under road surface abrupt changes, the forgetting factor is further decreased to capture transient parameter changes. The second auxiliary adjustment mechanism is superimposed on the first basic adjustment mechanism, and the two together determine the value of the forgetting factor at the current moment, and under any circumstances, the forgetting factor is limited to between the maximum forgetting factor of 0.995 and the minimum forgetting factor of 0.950.

[0200] Preferably, the typical range of the steady-state determination threshold is 0.5 N·m / s to 2.0 N·m / s; the typical range of the pavement change determination time is 50 ms to 200 ms; and the typical range of the pavement change threshold is 1.0 N·m to 3.0 N·m, in engineering units of tactile fluctuation.

[0201] In other preferred embodiments, regarding the aforementioned system amplitude-frequency gain With compensation weight This further clarifies that they have different physical meanings and control functions: The passive frequency response gain of a steering system under road surface excitation describes the amplification or attenuation effect of the system's inherent characteristics (equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness) on the transmission of road vibrations. The damping ratio in its mathematical expression is the system's equivalent damping ratio. It is used to infer the actual road excitation amplitude from the measured peak torque fluctuation. ; This describes the frequency-weighted strategy for active compensation torque, which adjusts the amplitude distribution weights of the compensation torque based on the position of the main road excitation frequency relative to the system's natural frequency. Its mathematical expression employs an independent bandwidth adjustment coefficient. (rather than) The two parameters are used to control the range of compensation frequencies. They are adjusted independently through different parameters, thereby achieving decoupled control of "road excitation back-pushing" and "compensation torque weighting", avoiding confusion between the passive response characteristics of the system and the active compensation strategy.

[0202] Furthermore, the preferred value ranges for some of the aforementioned key calibration parameters are as follows:

[0203] The typical range for the minimum compensated vehicle speed threshold is 5 km / h to 15 km / h; the typical range for the exit delay time is 200 ms to 1000 ms.

[0204] The above ranges are determined as follows: When the minimum compensation speed threshold is too low, the vibration suppression function is prone to accidental triggering under conditions such as low-speed parking, interfering with the driver's normal steering feel; when the minimum compensation speed threshold is too high, the vibration suppression function does not intervene in time when the vehicle is driving on a low-speed bumpy road (such as a washboard road). When the exit delay time is too short, the compensation function may frequently start and stop when the vehicle is driving on a continuous bumpy road section; when the exit delay time is too long, the compensation torque continues to act after the vehicle leaves the bumpy road surface, which may affect normal road feel feedback.

[0205] In some other preferred embodiments, the aforementioned maximum allowable compensation torque for the foundation is... The calibrated value of 2.5 N·m is consistent with the maximum absolute value of the compensation torque (static limit value) of 2.5 N·m. The two form a cascaded complementary relationship in the design intent: when the driver's steady-state steering force and vehicle speed are both zero (i.e., the vehicle is stationary or traveling in a straight line at a constant speed with no steering input), the upper limit of the dynamic allowable torque is... At this point, the dynamic limiting degenerates into the static limiting, and the maximum output of the compensation torque is limited to 2.5 N·m. When the driver applies steering force or the vehicle speed increases, the upper limit of the dynamic allowable torque decreases as the absolute value of the DC component of the torque increases, and also decreases as the vehicle speed signal increases, automatically shrinking to below 2.5 N·m. This effectively constrains the intervention boundary of the steering motor compensation torque under high vehicle speed and large steering conditions, returning the active safety control to the driver and achieving a cascaded complementary effect of "static limiting for safety and dynamic limiting for feel".

[0206] Preferably, the specific formula for the upper limit of the dynamic allowable torque can be in the following form:

[0207] ;

[0208] in This represents the upper limit of the dynamic allowable torque at time t, in N·m. The maximum allowable compensation torque is based on the given values, expressed in N·m. is the force-weighted inhibition coefficient, which is dimensionless; The DC component of the torque at time t reflects the driver's steady-state steering force, and its unit is N·m; The threshold for calibrating the driver's maximum steady-state hand force is given in N·m. is the high-speed attenuation factor, which is dimensionless; The vehicle speed signal at time t is expressed in km / h. The reference maximum speed is expressed in km / h; t represents continuous time, expressed in seconds. For absolute value calculations. When the driver's steady-state steering force reaches 8.0 N·m and the vehicle speed reaches 120 km / h, the upper limit of the dynamic allowable torque automatically drops to 0.6 N·m, effectively constraining the physical boundary of steering motor intervention under extreme conditions.

[0209] To facilitate a more intuitive understanding of the aforementioned fuzzy rule base, it is summarized in the following table: Table 1:

[0210] Table 1:

[0211]

[0212] Note: In rules 1 to 12, "—" indicates that the input quantity does not participate in the determination of this rule; rules 13 to 16 are correction / suppression rules based on feature conditions and do not depend on fuzzy membership degree calculation.

[0213] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0214] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for suppressing vibrations on EPS bumpy road surfaces using multi-signal fusion and parameter identification, characterized in that, include: Collect torque sensor signals, rotor position sensor signals, lateral acceleration signals, and vehicle speed signals; The torque sensor signal, rotor position sensor signal, and lateral acceleration signal are filtered and feature extracted using a sliding time window to obtain torque fluctuation characteristics, speed fluctuation characteristics, and lateral acceleration fluctuation characteristics. Fuzzy inference is performed based on torque fluctuation characteristics, speed fluctuation characteristics, lateral acceleration fluctuation characteristics and vehicle speed signal to obtain continuous road surface grade values ​​and determine the bumpy road surface state by hysteresis comparison. Under bumpy road conditions, after the continuous road surface grade value reaches 2.0 and remains so for 100ms, the discrete system parameters are identified online using a recursive least squares method with an adaptive forgetting factor; parameter updates are paused when the steering wheel angle rate is greater than 50° / s or the torque change rate is greater than 10N·m / ms. The forgetting factor is adjusted based on the following formula: ; In the formula, λ(k) is the forgetting factor at the k-th sampling time, which is dimensionless; λmax is the maximum forgetting factor, which is dimensionless; and λmin is the minimum forgetting factor, which is dimensionless. The normalized pavement grade is obtained by linearly mapping the continuous pavement grade values ​​to the interval [0,1] at the k-th sampling time, and is dimensionless; The forgetting factor is reduced when the continuous road surface grade value increases and increased when the continuous road surface grade value decreases. The feel fluctuation is formed by weighting the effective value of torque fluctuation, the peak value of torque fluctuation, and the number of zero crossings per unit time. The forgetting factor is further increased when the rate of change of feel fluctuation is lower than the steady-state judgment threshold. The forgetting factor is further reduced when the increase of feel fluctuation in the road surface change judgment time exceeds the road surface change threshold. The forgetting factor is limited to the maximum forgetting factor and the minimum forgetting factor. The parameter estimation vector is subjected to amplitude limiting and first-order low-pass filtering. The parameter validity is checked every 100ms. The effective range of the parameters is determined based on the nominal steering system model and bench identification results. When the parameter estimation value exceeds the effective range, the initial value of the parameter estimation vector and the initial covariance matrix are restored. Based on the nominal steering system model, feature extraction sampling period, and second-order vibration system discretization method, a discrete parameter lookup table is pre-established. This table is obtained by discretizing the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system using a parameter mesh, and calculating the corresponding discrete system parameters using the same feature extraction sampling period and discretization method. Interpolation is then performed on the discrete parameter lookup table based on the parameter estimation vector to obtain the equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system. Perform a fast Fourier transform on the torque fluctuation signal, determine the frequency corresponding to the main peak of the spectrum as the main frequency of road excitation, calculate the system natural frequency and system equivalent damping ratio based on the equivalent rotational inertia of the steering system, the equivalent damping coefficient of the steering system and the equivalent torsional stiffness of the steering system, and calculate the road excitation amplitude based on the system amplitude-frequency gain and the peak value of torque fluctuation. The compensation torque is calculated based on the main frequency of the road excitation, the amplitude of the road excitation, the natural frequency of the system, the equivalent damping ratio of the system, the road grade, and the vehicle speed signal. The compensation torque is subjected to amplitude limiting, ramp entry, and exponential decay exit processing. The processed compensation torque is converted into Q-axis current compensation and superimposed on the Q-axis current generated by the assist control to form the Q-axis target current.

2. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 1, characterized in that, The torque sensor signal is decomposed according to the following formula: ; In the formula, for The torque sensor signal at any given time, in N·m; for The DC component of torque that constantly reflects the driver's steady-state steering force, expressed in N·m; for The torque fluctuation component that constantly reflects the dynamic changes in road surface excitation, expressed in N·m; It is a continuous time interval, measured in seconds (s). The effective value of torque fluctuation is calculated within the sliding time window using the following formula: ; In the formula, For The effective value of torque fluctuation within the sliding time window at the end time, in N·m; The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. For the first time window within the sliding time window The torque fluctuation value at each sampling point is expressed in N·m. For summation operations; For the arithmetic square root operation; The peak torque fluctuation is calculated within the sliding time window using the following formula: ; In the formula, For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; This is for calculating the maximum value. This is for absolute value operations.

3. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 2, characterized in that, Motor speed is calculated using the following formula: ; In the formula, For the first The motor speed at each sampling time, in rad / s; For the first The rotor position at each sampling time, in rad; For the first The rotor position at each sampling time, in rad; The feature extraction sampling period is expressed in seconds (s). These are discrete sampling time sequence numbers, dimensionless; Motor speed can be decomposed according to the following formula: ; In the formula, For the first The average rotational speed at each sampling moment reflects the driver's steering speed, and the unit is rad / s; For the first Each sampling time reflects the rotational speed fluctuation component that reflects the road surface excitation speed fluctuation, with the unit being rad / s; The effective value of speed fluctuation is calculated according to the following formula: ; In the formula, For The effective value of the rotational speed fluctuation within the sliding time window at the end time, in rad / s; For the first time window within the sliding time window The rotational speed fluctuation value at each sampling point, in rad / s; The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. It is a continuous time interval, measured in seconds (s). For summation operations; For the arithmetic square root operation; The peak value of the speed fluctuation is calculated according to the following formula: ; In the formula, For The peak value of the rotational speed fluctuation within the sliding time window at the end time, in rad / s; This is for calculating the maximum value. For absolute value operations; The lateral acceleration signal is decomposed according to the following formula: ; In the formula, For the first The transverse acceleration signal at each sampling time, in m / s²; For the first Each sampling moment reflects the steady-state component of the lateral acceleration of the vehicle's steady-state steering motion, with units of m / s². For the first Each sampling moment reflects the lateral acceleration fluctuation component of the road surface bump excitation lateral vibration, with the unit being m / s²; The effective value of lateral acceleration fluctuation is calculated according to the following formula: ; In the formula, For The effective value of the lateral acceleration fluctuation within the sliding time window at the end time, in m / s². For the first time window within the sliding time window The lateral acceleration fluctuation value of each sampling point is expressed in m / s². The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. It is a continuous time interval, measured in seconds (s). For summation operations; For the arithmetic square root operation; The peak value of lateral acceleration fluctuation is calculated according to the following formula: ; In the formula, For The peak value of the lateral acceleration fluctuation within the sliding time window at the end time, in m / s². This is for calculating the maximum value. This is for absolute value operations.

4. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 3, characterized in that, The fluctuation in feel is calculated using the following formula: ; In the formula, for The tactile fluctuation at any time is a calibration evaluation value formed by weighting different engineering dimension characteristics, without being assigned a unified physical dimension. The weighting coefficient for the effective value of torque fluctuation is determined based on the calibration range of the effective value of torque fluctuation. This is the weighting factor for the peak torque fluctuation; the engineering quantity scale is determined based on the calibration range of the peak torque fluctuation. This is the weighting factor for the number of zero crossings per unit time. The scale of the project quantity is determined based on the calibration range of the number of zero crossings per unit time. For The effective value of torque fluctuation within the sliding time window at the end time, in N·m; For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; for The number of times the torque fluctuation signal crosses the zero line per unit time, expressed in Hz; , and The values ​​are 0.4, 0.3, and 0.3 respectively; The sign function value of the torque fluctuation signal is determined according to the following formula: ; In the formula, For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. For the first time window within the sliding time window The torque fluctuation value at each sampling point is expressed in N·m. This is a dimensionless index for the sampling points within the sliding time window. The number of zero crossings per unit time is calculated using the following formula: ; In the formula, for The number of times the torque fluctuation signal crosses the zero line per unit time, expressed in Hz; The length of the sliding time window, in seconds; For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. For the first time window within the sliding time window The sign function value at each sampling point is dimensionless. The number of sampling points within the sliding time window, dimensionless; This is a dimensionless index for the sampling points within the sliding time window. The time is continuous, measured in seconds; a value of 2 indicates that a complete oscillation contains two zero crossings. For summation operations; This is for absolute value operations.

5. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 4, characterized in that, The Mamdani-type min-max inference method is adopted to determine the road surface grade based on the effective values ​​of torque fluctuation, speed fluctuation, lateral acceleration fluctuation, and vehicle speed signal. For each fuzzy rule, the minimum value among all input membership degrees is taken as the fuzzy rule activation strength, and for the same road surface grade, the maximum value among the corresponding fuzzy rule activation strengths is taken as the road surface grade membership degree. Defuzzification is then performed according to the following formula: ; In the formula, for The continuous road surface grade value after defuzzification at each time step is dimensionless and ranges from [1,4]. The number of fuzzy rules involved in defuzzification is dimensionless. This is a fuzzy rule index, dimensionless; for Time of the first The activation intensity of a fuzzy rule is dimensionless. For the first The fuzzy rule outputs the center value of the level, which is dimensionless; It is a continuous time interval, measured in seconds (s). For summation operations; The center values ​​for Level 1, Level 2, Level 3, and Level 4 are 1.0, 2.0, 3.0, and 4.0, respectively. When the number of zero crossings per unit time is greater than 30Hz and the ratio of the peak torque fluctuation to the effective torque fluctuation is less than 1.5, the continuous road surface level value is reduced by one level. When the number of zero crossings per unit time is less than 10Hz and the ratio of the peak torque fluctuation to the effective torque fluctuation is greater than 2.5, the continuous road surface level value is increased by one level. When the steering wheel angle rate is greater than 50° / s or the torque change rate is greater than 10N·m / ms, the output compensation suppression state is reached. When the continuous road surface level value reaches 2.0 and the vehicle speed is greater than the minimum compensation speed threshold for 100ms, the bumpy road surface state is determined. When the continuous road surface level value is lower than 1.4 and the exit delay time is reached, the bumpy road surface state is exited.

6. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 5, characterized in that, The steering wheel angle is obtained based on the rotor position sensor signal and the nominal steering gear ratio, and the EPS system under bumpy road conditions is represented as a second-order vibration system: ; In the formula, The equivalent moment of inertia of the steering system is expressed in kg·m². for The steering wheel angle at any given moment, measured in rad. for The angular velocity of the steering wheel at any given moment, in rad / s; for The steering wheel angular acceleration at any given moment, in rad / s². The equivalent damping coefficient of the steering system is expressed in N·m·s / rad. The equivalent torsional stiffness of the steering system is expressed in N·m / rad. for The motor output torque at any given time, expressed in N·m; for The road surface excitation torque at time t, expressed in N·m; It is a continuous time interval, measured in seconds (s). A dot above a letter indicates the first derivative with respect to continuous time; two dots above a letter indicate the second derivative with respect to continuous time. Within a single identification time window, the road surface excitation torque is expressed by the following formula: ; In the formula, The amplitude of the road surface excitation within a single identification time window is expressed in N·m. The dominant frequency of road surface excitation within a single identification time window, in Hz; The initial phase of the road surface excitation within a single identification time window, in rad; for The random perturbation torque at time t, expressed in N·m; It is a sine function; Pi; It is a continuous time interval, measured in seconds (s). After discretizing the second-order vibration system, the input-output model is established according to the following formula: ; In the formula, For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; , , and The parameters of the discrete system to be identified are dimensionless. This represents the discrete sampling time sequence number, which is dimensionless.

7. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 6, characterized in that, The parameter estimation vector is established according to the following formula: ; In the formula, For the first The parameter estimation vector at each sampling time point; , , and The first The discrete system parameter estimates at each sampling time point are dimensionless; superscript. Indicates matrix transpose; These are discrete sampling time sequence numbers, dimensionless; The observation vector is established according to the following formula: ; In the formula, For the first The observation vector at each sampling time; For the first Torque fluctuation output at each sampling time, in N·m; For the first Torque fluctuation output at each sampling time, in N·m; For the first The motor output torque input at each sampling time, in N·m; For the first Motor output torque input at each sampling time, in N·m; superscript Indicates matrix transpose; These are discrete sampling time sequence numbers, dimensionless; The system input-output relationship is expressed by the following formula: ; In the formula, For the first Torque fluctuation output at each sampling time, in N·m; For the first The transpose of the observation vector at each sampling time; For the first The parameter estimation vector at each sampling time point; These are discrete sampling time sequence numbers, dimensionless; The recursive gain vector is updated according to the following formula: ; In the formula, For the first The recursive gain vector at each sampling time; For the first The covariance matrix at each sampling time; For the first The observation vector at each sampling time; For the first The transpose of the observation vector at each sampling time; For the first The forgetting factor at each sampling time is dimensionless. These are discrete sampling time sequence numbers, dimensionless; The parameter estimation vector is updated according to the following formula: ; In the formula, For the first The parameter estimation vector at each sampling time point; For the first The parameter estimation vector at each sampling time point; For the first The recursive gain vector at each sampling time; For the first Torque fluctuation output at each sampling time, in N·m; For the first The transpose of the observation vector at each sampling time; These are discrete sampling time sequence numbers, dimensionless; The covariance matrix is ​​updated according to the following formula: ; In the formula, For the first The covariance matrix at each sampling time; It is the identity matrix; For the first The recursive gain vector at each sampling time; For the first The transpose of the observation vector at each sampling time; For the first The covariance matrix at each sampling time; For the first The forgetting factor at each sampling time is dimensionless. These are discrete sampling time sequence numbers, dimensionless; The maximum forgetting factor is 0.995, the minimum forgetting factor is 0.950, and the initial forgetting factor is 0.

985.

8. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 7, characterized in that, The discrete parameters are stored in the EPS controller via a lookup table. The equivalent moment of inertia, equivalent damping coefficient, and equivalent torsional stiffness of the steering system obtained by interpolation based on the parameter estimation vector in the discrete parameter lookup table are limited by the effective range of the parameters. A fast Fourier transform is performed on the torque fluctuation signal, and the frequency corresponding to the main peak of the spectrum is determined as the main frequency of the road excitation. The system amplitude-frequency gain is calculated according to the following formula: ; In the formula, For the steering system at frequency The amplitude-frequency gain at that point is dimensionless. For the steering system at frequency Frequency response at that location; The frequency to be calculated is in Hz. The system's inherent frequency, measured in Hz; The system's equivalent damping ratio is dimensionless. For amplitude calculation; For the arithmetic square root operation; The road surface excitation amplitude is calculated according to the following formula: ; In the formula, The amplitude of the road surface excitation within a single identification time window is expressed in N·m. For The peak value of torque fluctuation within the sliding time window at the end time, in N·m; The amplitude-frequency gain of the steering system at the dominant frequency of road excitation is dimensionless. The dominant frequency of road surface excitation within a single identification time window, in Hz; It is a continuous time interval, measured in seconds (s). The system's equivalent damping ratio is calculated using the following formula: ; In the formula, The system's equivalent damping ratio is dimensionless. The equivalent damping coefficient of the steering system is expressed in N·m·s / rad. The equivalent moment of inertia of the steering system is expressed in kg·m². The equivalent torsional stiffness of the steering system is expressed in N·m / rad. For the arithmetic square root operation; The system's natural angular frequency is calculated using the following formula: ; In the formula, The system's natural angular frequency is expressed in rad / s. The equivalent torsional stiffness of the steering system is expressed in N·m / rad. The equivalent moment of inertia of the steering system is expressed in kg·m². For the arithmetic square root operation; The system's natural frequency is calculated using the following formula: ; In the formula, The system's inherent frequency, measured in Hz; The system's natural angular frequency is expressed in rad / s. Pi; The equivalent torsional stiffness of the steering system is expressed in N·m / rad. The equivalent moment of inertia of the steering system is expressed in kg·m². The arithmetic square root operation is performed; the fast Fourier transform uses a 256-point transform and a 64-point sliding step size. The road excitation main frequency is confirmed to be effective when the ratio of the amplitude of the main peak of the spectrum to the amplitude of the secondary peak of the spectrum is greater than 1.

5.

9. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 8, characterized in that, The basic amplitude of the compensating torque is calculated according to the following formula: ; In the formula, The compensation torque is the basic amplitude, expressed in N·m. The amplitude of the road surface excitation within a single identification time window is expressed in N·m. It is the basic compensation gain coefficient, which is dimensionless. The compensation weight at the dominant frequency of the road excitation is dimensionless. The dominant frequency of road surface excitation within a single identification time window, in Hz; The compensation weight is calculated according to the following formula: ; In the formula, For frequency The compensation weight at the location is dimensionless. The frequency to be calculated is in Hz. The system's inherent frequency, measured in Hz; The bandwidth adjustment coefficient for compensating for the frequency range is dimensionless. For the arithmetic square root operation; The final compensation torque amplitude is calculated according to the following formula: ; In the formula, The final compensation torque amplitude is expressed in N·m. The compensation torque is the basic amplitude, expressed in N·m. The pavement grade gain coefficient is determined based on the pavement grade and is dimensionless. The vehicle speed correlation coefficient is determined based on the vehicle speed signal and is dimensionless. The phase correction of the compensating torque is calculated according to the following formula: ; In the formula, The compensation torque phase correction is expressed in rad. It is the arctangent function; The system's equivalent damping ratio is dimensionless. The dominant frequency of road surface excitation within a single identification time window, in Hz; The system's inherent frequency, measured in Hz; The direction of the compensation torque is determined according to the damping control direction opposite to the direction of the speed fluctuation component, and is corrected according to the phase correction amount of the compensation torque; the basic compensation gain coefficient is 0.6, and the maximum absolute value of the compensation torque is 2.5 N·m.

10. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 7, characterized in that, In the process of identifying discrete system parameters online using the recursive least squares method with an adaptive forgetting factor, covariance matrix renormalization correction is also performed: The covariance matrix is ​​updated at the k-th sampling time. Then, the trace of the covariance matrix is ​​calculated in real time: ; In the formula, Up to 4, For the first The main diagonal element in the j-th row and j-th column of the covariance matrix at each sampling time; Set the lower limit threshold for the trace of the covariance matrix. Upper limit threshold of trace of covariance matrix ; When the trace of the covariance matrix is ​​lower than the lower limit threshold, it is corrected by injecting a positive definite perturbation into the main diagonal of the covariance matrix. The corrected covariance matrix is: ; in Let I be the dimension of the covariance matrix, and let I be the identity matrix. When the trace of the covariance matrix exceeds the upper limit threshold of the trace, it is corrected by proportional scaling: ; No correction is performed when the trace of the covariance matrix is ​​between the lower limit threshold and the upper limit threshold. In the When calculating the recursive gain vector at each sampling time, the corrected covariance matrix is ​​used. Replace the original covariance matrix Participate in the calculation.

11. The EPS bumpy road vibration suppression method based on multi-signal fusion and parameter identification according to claim 9, characterized in that, During the process of applying the maximum limit to the compensation torque, dynamic amplitude limiting correction is also performed: A dynamic allowable torque upper limit is constructed based on the DC component of torque reflecting the driver's steady-state steering force and the vehicle speed signal. The dynamic allowable torque upper limit decreases as the absolute value of the DC component of torque increases and decreases as the vehicle speed signal increases. The dynamic allowable torque upper limit is not greater than the preset maximum absolute value of the compensation torque. The final compensation torque amplitude is limited to within the upper limit of the dynamic allowable torque: ; in This is the final compensation torque amplitude after limiting. For the final compensation torque amplitude, This is the upper limit of the dynamic allowable torque.

12. A vibration suppression system for EPS bumpy road surfaces based on multi-signal fusion and parameter identification, characterized in that, The EPS bumpy road vibration suppression method for implementing multi-signal fusion and parameter identification as described in claim 1 includes: a torque sensor for acquiring torque sensor signals; Rotor position sensor, used to acquire rotor position sensor signals; Lateral acceleration sensor, used to collect lateral acceleration signals; Vehicle speed signal interface, used to acquire vehicle speed signals; EPS controller, including: The road surface condition recognition unit is used to filter and extract features from torque sensor signals, rotor position sensor signals and lateral acceleration signals through sliding time windows to obtain torque fluctuation features, speed fluctuation features and lateral acceleration fluctuation features. Based on the torque fluctuation features, speed fluctuation features, lateral acceleration fluctuation features and vehicle speed signals, it performs fuzzy inference to obtain continuous road surface grade values ​​and determines the bumpy road surface condition through hysteresis comparison. The parameter identification unit is used to update the steering system parameters using an online parameter identification method with an adaptive forgetting factor when the identification start condition is met under bumpy road conditions, thereby obtaining the main frequency of road excitation, the amplitude of road excitation, the natural frequency of the system, and the equivalent damping ratio of the system. The compensation torque calculation unit is used to calculate the compensation torque based on the main frequency of road excitation, the amplitude of road excitation, the natural frequency of the system, the equivalent damping ratio of the system, the road grade and the vehicle speed signal. It performs amplitude limiting, slope entry and exponential decay exit processing on the compensation torque, converts the processed compensation torque into the Q-axis current compensation amount and adds it to the Q-axis current generated by the power assist control to form the Q-axis target current. The steering motor, connected to the EPS controller, is used to output compensating torque based on the target current of the Q-axis to suppress steering wheel vibration caused by bumpy roads.

Citation Information

Patent Citations

  • Electric power steering device and its controller

    JP2007118794A

  • Method and system for fault isolation in an electric power steering system

    US9751556B1