Road surface roughness identification method based on trapezoidal load shape

By adopting a trapezoidal load form and an improved Kalman filter algorithm, the problem of inaccurate description of the continuous variation characteristics of road excitation is solved, the estimation accuracy of unknown inputs and the accuracy of road roughness identification are improved, and the robustness of the identification method is enhanced.

CN123104658APending Publication Date: 2026-11-17CHINA CONSTR EIGHT ENG DIV CORP LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611277477.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing road surface roughness identification methods based on vehicle dynamic response, the rectangular load approximation method leads to inaccurate description of the continuous change characteristics of road surface excitation and low estimation accuracy of unknown inputs, which affects the accuracy and robustness of road surface roughness identification.

Method used

Discrete state equations are established using trapezoidal load form, a system observation matrix is ​​constructed, and the Kalman filter algorithm is improved. Road surface unevenness is identified through the simplified form of trapezoidal load, thereby improving the estimation accuracy and identification accuracy of unknown inputs.

Benefits of technology

It can more accurately describe the continuous variation characteristics of road surface excitation, improve the estimation accuracy of unknown inputs, enhance the accuracy and robustness of road surface roughness identification, and reduce identification errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104658A_ABST
    Figure CN123104658A_ABST
Patent Text Reader

Abstract

This invention discloses a method for identifying road surface roughness based on trapezoidal load shape, comprising the following steps: Step 1: Establishing discrete state equations using trapezoidal load form; Step 2: Constructing a system observation matrix; Step 3: Identifying road surface roughness using an improved Kalman filter algorithm based on a simplified form of trapezoidal load. This invention relates to the field of traffic structure health monitoring technology and can solve the problems of inaccurate description of continuous road surface excitation characteristics, low estimation accuracy of unknown inputs, and low accuracy in identifying road surface roughness in existing technologies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of traffic structure health detection technology, and in particular to a method for identifying road surface unevenness based on trapezoidal load shape. Background Technology

[0002] Road surface roughness information is a crucial indicator characterizing road smoothness and service performance. Its identification results are widely used in vehicle dynamics analysis, road condition assessment, driving safety evaluation, vehicle comfort analysis, and road maintenance decision-making. Accurately obtaining road surface roughness information is of great significance for improving road inspection efficiency, ensuring vehicle operational safety, and optimizing road maintenance management.

[0003] Currently, road surface roughness identification methods based on vehicle dynamic response only require the use of conventional measurement equipment such as acceleration sensors and displacement sensors installed on the vehicle to obtain the vehicle vibration response, and then combine the vehicle dynamics model to invert the road surface roughness information. It has the advantages of low equipment cost, convenient implementation and wide applicability, and has received widespread attention in recent years.

[0004] In existing methods for identifying road surface roughness based on vehicle dynamic response, Kalman filtering and its improved algorithms are widely used in the fields of road surface roughness and identification due to their good state estimation capabilities and noise resistance. These methods typically treat road surface excitation as unknown input, establishing a vehicle dynamic state-space model and combining it with the vehicle vibration response to achieve inverse identification of road surface roughness. However, existing Kalman filtering methods often use a rectangular load approximation to handle unknown inputs between adjacent sampling times when establishing discrete state equations, assuming that the load remains constant within a single sampling time interval.

[0005] However, in reality, the road excitation experienced by vehicles varies continuously. Using a rectangular load approximation leads to abrupt changes in load between adjacent sampling times, resulting in a significant deviation from the actual road excitation. This is particularly problematic in identifying localized unevenness such as speed bumps, potholes, and protrusions, as well as in identifying high-frequency road surface irregularities. Such approximations easily introduce large discretization errors, reducing the accuracy of unknown input estimation and road surface irregularity identification. Furthermore, existing methods do not adequately utilize unknown input information at the current moment during state prediction, further impacting the stability and robustness of the identification results.

[0006] Therefore, there is a need to provide a road surface roughness identification method based on trapezoidal load shape, which can solve the problems of inaccurate description of continuous change characteristics of road surface excitation, low estimation accuracy of unknown input, and low accuracy of road surface roughness identification in the existing technology. Summary of the Invention

[0007] The purpose of this invention is to provide a road surface roughness identification method based on trapezoidal load shape, which can solve the problems of inaccurate description of continuous change characteristics of road surface excitation, low estimation accuracy of unknown input, and low accuracy of road surface roughness identification in the prior art.

[0008] This invention is implemented as follows:

[0009] A method for identifying road surface roughness based on trapezoidal load shape includes the following steps:

[0010] Step 1: Establish discrete state equations using trapezoidal load form;

[0011] Step 2: Construct the system observation matrix ;

[0012] Step 3: Identify road surface unevenness using an improved Kalman filter algorithm based on a simplified form of trapezoidal load.

[0013] In step 1, the state-space equation of the vehicle system under trapezoidal load can be expressed as:

[0014] (1)

[0015] Wherein, the state vector Displacement of the vehicle and speed Combining these elements, its expression is: Then the state vector The differential expression is: ;

[0016] Road surface unevenness Represented as: ,

[0017] This represents the road surface roughness experienced by the i-th (i=1,…,n) wheel;

[0018] System noise vector It is Gaussian white noise with a mean of 0;

[0019] , and The system transformation matrix is ​​expressed by the following formula:

[0020] (2)

[0021] The specific forms of coefficient matrices A and B are as follows:

[0022] (3)

[0023] in, For the vehicle's mass matrix, Here is the damping matrix of the vehicle. Let I be the stiffness matrix of the vehicle, and let I be a diagonal matrix.

[0024] When performing vibration analysis on vehicles, road surface unevenness is considered. Represented as:

[0025] ;

[0026] This is the wheel load position matrix. , For the j-th wheel spring Degrees of freedom regarding the point of contact with the vehicle;

[0027] The contact stiffness matrix is ​​composed of the arrangement of wheel springs in contact with the road surface, namely: .

[0028] In step 2, the system observation matrix The expression form is:

[0029] (4);

[0030] in, The measurement noise vector is Gaussian white noise with a mean of 0; H is the measurement matrix, and D is the input transformation matrix. The forms of H and D are as follows:

[0031] (5);

[0032] The system noise covariance matrix Q and the measurement noise covariance matrix T are expressed as follows:

[0033] (6).

[0034] In step 3, during the initial estimation stage, based on the assumptions of the improved Kalman filter algorithm, the state vector is... Road surface unevenness And its covariance matrix is ​​initialized;

[0035] (7)

[0036] (8);

[0037] Among them, the symbol " " represents the estimated value of the vector, It is the state vector at step i. The estimated value, The road surface unevenness at step i. The estimated value;

[0038] Vehicle state vector and road surface unevenness The initial covariance matrices are as follows:

[0039] (9)

[0040] (10);

[0041] During the Kalman gain calculation stage, the initial covariance matrix is ​​passed to obtain the predicted covariance matrix. Meanwhile, the Kalman gain matrix is ​​obtained based on the minimum mean square error estimation. ;

[0042] Predicting covariance matrix The calculation formula is:

[0043] (11);

[0044] Kalman gain matrix The calculation formula is:

[0045] (12);

[0046] In the unknown input estimation stage: the unknown input estimation stage mainly involves the external excitation covariance matrix. Update The expression is: Calculate the covariance matrix at the current time. Meanwhile, based on the covariance matrix at the current time... The road surface roughness at the current moment is estimated by combining the collected vibration response at the current moment. ;

[0047] Covariance matrix at the current time The expression is:

[0048] (13);

[0049] Road surface unevenness at the current moment The expression is:

[0050] (14);

[0051] In the state prediction phase: The goal of the state prediction phase is to predict the future state vector of the system based on the system state model and the estimated values ​​of external stimuli.

[0052] Future state vector The expression is:

[0053] (15);

[0054] In the correction phase: the predicted future state vector is corrected using the observed data and the Kalman gain matrix to obtain the estimated state vector value. Simultaneously, the estimated covariance matrix is ​​obtained through the Kalman gain matrix and the external excitation covariance matrix. ;

[0055] State covariance matrix The expression is:

[0056] (16);

[0057] Estimating the covariance matrix The expression is:

[0058]

[0059] (17).

[0060] Compared with the prior art, the present invention has the following advantages:

[0061] This invention is based on the vehicle dynamic response generated by road surface irregularities during vehicle operation. It can more accurately describe the continuous change characteristics of road surface excitation, improve the estimation accuracy of unknown inputs and enhance the accuracy of road surface irregularity identification. It solves the problems of existing road surface irregularity identification methods, such as the inability of rectangular load approximation to accurately represent the continuous change characteristics of road surface excitation, insufficient utilization of unknown input information at the current moment, and low accuracy and robustness of road surface irregularity identification. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of a seven-degree-of-freedom vehicle according to Embodiment 1 of the present invention;

[0063] Figure 2 This is a schematic diagram of the accelerometer sensor arrangement position in Embodiment 1 of the present invention;

[0064] Figure 3 This is a schematic diagram of the road surface unevenness experienced by the vehicle in Embodiment 1 of the present invention;

[0065] Figure 4 This is a schematic diagram of the vehicle acceleration response according to Embodiment 1 of the present invention;

[0066] Figure 5 This is a simplified schematic diagram of rectangular load and trapezoidal load according to Embodiment 1 of the present invention;

[0067] Figure 6This is a schematic diagram of the road surface roughness estimation results in Embodiment 1 of the present invention. Detailed Implementation

[0068] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0069] A method for identifying road surface roughness based on trapezoidal load shape includes the following steps:

[0070] Step 1: Establish discrete state equations using trapezoidal load form.

[0071] In step 1, the state-space equation of the vehicle system under trapezoidal load can be expressed as:

[0072] (1)

[0073] Wherein, the state vector Displacement of the vehicle and speed Combining these elements, its expression is: Then the state vector The differential can be expressed as: .

[0074] Road surface unevenness It can be represented as: ,

[0075] This represents the road surface unevenness experienced by the i-th (i=1,…,n) wheel.

[0076] System noise vector It is Gaussian white noise with a mean of 0.

[0077] , and The system transformation matrix is ​​expressed by the following formula:

[0078] (2)

[0079] The specific forms of coefficient matrices A and B are as follows:

[0080] (3)

[0081] in, For the vehicle's mass matrix, Here is the damping matrix of the vehicle. Let I be the stiffness matrix of the vehicle, and let I be a diagonal matrix.

[0082] When performing vibration analysis on vehicles, road surface unevenness is considered. It can be represented as:

[0083] .

[0084] This is the wheel load position matrix. , For the j-th wheel spring The degree of freedom of the contact position with the vehicle can be freely set according to the vehicle system.

[0085] The contact stiffness matrix is ​​an n-dimensional diagonal matrix composed of the arrangement of wheel springs in contact with the road surface, namely: .

[0086] Step 2: Construct the system observation matrix .

[0087] In step 2, the system observation matrix The expression form is:

[0088] (4).

[0089] in, The measurement noise vector is Gaussian white noise with a mean of 0. H is the measurement matrix, and D is the input transformation matrix. The forms of H and D are as follows:

[0090] (5).

[0091] The system noise covariance matrix Q and the measurement noise covariance matrix T are expressed as follows:

[0092] (6).

[0093] Step 3: Identify road surface unevenness using an improved Kalman filter algorithm based on a simplified form of trapezoidal load.

[0094] In step 3, during the initial estimation stage, based on the assumptions of the improved Kalman filter algorithm, the state vector is... Road surface unevenness And its covariance matrix is ​​initialized.

[0095] The goal of the initial estimation phase is to provide the necessary initial conditions for the subsequent prediction and correction processes, ensuring that the algorithm can recursively proceed from the initial moment.

[0096] (7)

[0097] (8).

[0098] Among them, the symbol " " represents the estimated value of the vector, It is the state vector at step i. The estimated value, The road surface unevenness at step i. The estimated value.

[0099] Vehicle state vector and road surface unevenness The initial covariance matrices are as follows:

[0100] (9)

[0101] (10).

[0102] In the Kalman gain calculation stage, the calculation of the Kalman gain matrix is ​​the core step of the Kalman filtering algorithm, as it determines the accuracy of the state estimation. This stage primarily involves transferring the initial covariance matrix to obtain the predicted covariance matrix. Meanwhile, the Kalman gain matrix is ​​obtained based on the minimum mean square error estimation. .

[0103] Predicting covariance matrix The calculation formula is:

[0104] (11).

[0105] Kalman gain matrix The calculation formula is:

[0106] (12).

[0107] The calculation of the Kalman gain matrix is ​​essentially about finding the future state vector. The optimal trade-off between the observed data (which can be the vehicle's acceleration, velocity, and displacement response) is used to correct the state estimate.

[0108] In the unknown input estimation stage: the unknown input estimation stage mainly involves the external excitation covariance matrix. Update The expression is: Calculate the covariance matrix at the current time. Meanwhile, based on the covariance matrix at the current time... The road surface roughness at the current moment is estimated by combining the collected vibration response at the current moment. .

[0109] Covariance matrix at the current time The expression is:

[0110] (13).

[0111] Road surface unevenness at the current moment The expression is:

[0112] (14).

[0113] In the state prediction phase: The goal of the state prediction phase is to determine the road surface roughness based on the system state model, i.e., formula (1) and the estimated value of the external excitation, i.e., the road surface roughness at the current moment. It predicts the future state vector of the system.

[0114] Future state vector The expression is:

[0115] (15).

[0116] In the correction phase: During the correction phase, the predicted future state vector is corrected using observed data (such as the vehicle's acceleration, velocity, and displacement response) and the Kalman gain matrix, resulting in a more accurate state vector estimate and covariance matrix. This state vector correction step significantly reduces the error in system state vector prediction and improves the algorithm's adaptability to complex systems.

[0117] The state vector is estimated and corrected based on the predicted state vector, the acquired vibration response at the current moment, and the Kalman gain matrix, to obtain the estimated state vector value. Simultaneously, through the Kalman gain matrix and the external excitation covariance matrix, which is the covariance matrix at the current time. Obtain the estimated covariance matrix .

[0118] State vector estimate The expression is:

[0119] (16).

[0120] Estimating the covariance matrix The expression is:

[0121]

[0122] (17).

[0123] Example 1: A schematic diagram of a seven-degree-of-freedom vehicle model is shown below. Figure 1 As shown. The vehicle's mass is... =1085kg, the body roll and pitch moments of inertia are respectively = and = The masses of the four wheels are respectively , , , Both are 40kg. The suspension stiffnesses are respectively , , , All of them The suspension damping is respectively , , , All of them The tire stiffnesses are respectively , , , All of them The tire damping is respectively , , , All of them The vertical distances from the front and rear wheels to the vehicle's center of gravity are respectively... =1.4m and =1.47m, the vertical distances from the left and right wheels to the measuring center of mass are respectively =0.7m and =0.75m.

[0124] Accelerometers are placed at the four wheels of the vehicle to monitor vibration response. A schematic diagram of the vehicle sensor layout is shown below. Figure 2 As shown, the accelerometers are respectively used , , and This indicates the unevenness of the road surface the vehicle traversed, such as... Figure 3 As shown. From Figure 3 As can be seen, the road surface unevenness experienced by the right front wheel and the right rear wheel and The same, only the time difference. The right wheel experienced different road surface unevenness. and It also has this characteristic.

[0125] First, based on the vehicle's parameter information and the Lagrange equation, the vehicle's system matrix is ​​constructed: the vehicle's mass matrix. Damping matrix of the vehicle and vehicle stiffness matrix And construct the wheel load position matrix. and contact stiffness matrix .

[0126] Then, acceleration responses during vehicle operation are collected using acceleration sensors installed on the vehicle. These collected acceleration responses are used as the system's observed response (i.e., the measured vehicle vibration response, which can be vehicle displacement, velocity, or acceleration) and input into the subsequent recognition algorithm. The observed sensor data are as follows: Figure 4 As shown. From Figure 4 It can be seen that the vibration range of the acceleration response of the four wheels is relatively consistent, mainly concentrated in .

[0127] Next, the discrete state equations and system observation matrix under trapezoidal load form are established, and the measured vehicle acceleration response is used as the observation. As the observed parameters serve as the system's observed response, they are progressively input into the improved Kalman filter algorithm according to the road surface roughness identification process. Traditional rectangular load discretization simplifies the load to a rectangle, resulting in simple calculations; the trapezoidal load discretization method of this invention simplifies the load to a trapezoid, considering the current load's input information, providing complete information and high approximation accuracy. Schematic diagrams of rectangular and trapezoidal load discretization are shown below. Figure 5 (a) and Figure 5 As shown in (b).

[0128] Finally, the unknown input improved Kalman filter algorithm is used for recursive calculation. The calculation process is as follows: First, the state vector, road surface roughness, and covariance matrix are initialized; then, the predicted state vector and predicted covariance matrix are calculated; further, the Kalman gain matrix is ​​calculated based on the system observation response; the unknown input, i.e., road surface roughness, is estimated using the system observation response at the current moment; finally, the state vector is corrected based on the estimated road surface roughness, and the state covariance matrix is ​​updated.

[0129] Repeat the recursive calculation process to obtain the road surface unevenness corresponding to each sampling time during vehicle travel, such as... Figure 6 As shown. Figure 6 (a) indicates the road surface roughness experienced by the right wheel of the vehicle. Figure 6 (b) indicates the road surface unevenness experienced by the left wheel of the vehicle.

[0130] from Figure 6 As can be seen, the pavement roughness estimation results using the trapezoidal discrete load form are very close to the actual pavement roughness profile, proving the effectiveness and applicability of the method of the present invention. By comparing the estimation results with the traditional simplified rectangular load form and the actual profile (using a frequency of 60Hz), the accuracy of the estimation results of the present invention is about 6 times that of the traditional rectangular load form. The calculation error ε of the simplified trapezoidal load is 0.09, and the calculation error ε of the simplified rectangular load is 0.52. The error ε is calculated using formula (18):

[0131] (18)

[0132] in, This represents the estimation result of the simplified algorithm for trapezoidal (or rectangular) loads. This represents the actual road surface unevenness profile. This demonstrates that the estimation effect of this method is significantly better than that of traditional Kalman filtering under conditions with low sampling frequency.

[0133] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying road surface unevenness based on a trapezoidal load shape, characterized by: Includes the following steps: Step 1: Establish discrete state equations using trapezoidal load form; Step 2: Constructing the system observation matrix ; Step 3: Identify road surface unevenness using an improved Kalman filter algorithm based on a simplified form of trapezoidal load.

2. The road surface roughness identification method based on trapezoidal load shape according to claim 1, characterized by: In step 1, the state-space equation of the vehicle system under trapezoidal load can be expressed as: (1) Wherein, the state vector Displacement of the vehicle and speed Combining these elements, its expression is: Then the state vector The differential expression is: ; Road surface unevenness Represented as: , This represents the road surface roughness experienced by the i-th (i=1,…,n) wheel; System noise vector It is Gaussian white noise with a mean of 0; , and The system transformation matrix is ​​expressed by the following formula: (2) The specific forms of coefficient matrices A and B are as follows: (3) in, For the vehicle's mass matrix, Here is the damping matrix of the vehicle. Let I be the stiffness matrix of the vehicle, and let I be a diagonal matrix. When performing vibration analysis on vehicles, road surface unevenness is considered. Represented as: ; This is the wheel load position matrix. , For the j-th wheel spring Degrees of freedom regarding the point of contact with the vehicle; The contact stiffness matrix is ​​composed of the arrangement of wheel springs in contact with the road surface, namely: .

3. The method for identifying road surface roughness based on trapezoidal load shape according to claim 1, characterized in that: In step 2, the system observation matrix The expression form is: (4); in, The measurement noise vector is Gaussian white noise with a mean of 0; H is the measurement matrix, and D is the input transformation matrix. The forms of H and D are as follows: (5); The system noise covariance matrix Q and the measurement noise covariance matrix T are expressed as follows: (6)。 4. The method for identifying road surface roughness based on trapezoidal load shape according to claim 1, characterized in that: In step 3, during the initial estimation stage, based on the assumptions of the improved Kalman filter algorithm, the state vector is... Road surface unevenness And its covariance matrix is ​​initialized; (7) (8); Among them, the symbol " " represents the estimated value of the vector, It is the state vector at step i. The estimated value, The road surface unevenness at step i. The estimated value; Vehicle state vector and road surface unevenness The initial covariance matrices are as follows: (9) (10); During the Kalman gain calculation stage, the initial covariance matrix is ​​passed to obtain the predicted covariance matrix. Meanwhile, the Kalman gain matrix is ​​obtained based on the minimum mean square error estimation. ; Predicting covariance matrix The calculation formula is: (11); Kalman gain matrix The calculation formula is: (12); In the unknown input estimation stage: the unknown input estimation stage mainly involves the external excitation covariance matrix. Update The expression is: Calculate the covariance matrix at the current time. Meanwhile, based on the covariance matrix at the current time... The road surface roughness at the current moment is estimated by combining the collected vibration response at the current moment. ; Covariance matrix at the current time The expression is: (13); Road surface unevenness at the current moment The expression is: ;(14); In the state prediction phase: The goal of the state prediction phase is to predict the future state vector of the system based on the system state model and the estimated values ​​of external stimuli. Future state vector The expression is: (15); In the correction phase: the predicted future state vector is corrected using the observed data and the Kalman gain matrix to obtain the estimated state vector value. Simultaneously, the estimated covariance matrix is ​​obtained through the Kalman gain matrix and the external excitation covariance matrix. ; State covariance matrix The expression is: (16); Estimating the covariance matrix The expression is: ;(17)。