Data-driven EMB system brake clearance online estimation method and system

Abstract

The invention provides a data-driven EMB system brake clearance online estimation method and system, and the method comprises the steps: building a nonlinear polynomial mapping model based on a Legendre orthogonal polynomial through mining a relation between a brake motor current and a rotation angle signal in a brake loading stage of an electronic mechanical brake system; and an enhanced recursive least square algorithm is designed to realize coefficient online identification of a current and corner polynomial, and a brake clearance value is analyzed from an orthogonal fitting function through a zero current extrapolation mechanism. According to the method, a force sensor and off-line calibration are not needed, low-cost on-line estimation of the brake clearance of the electronic mechanical brake system can be achieved, the high-precision brake clearance is provided for the drive-by-wire electronic mechanical brake system, and a theoretical basis is provided for self-adaptive adjustment of the brake clearance of the electronic mechanical brake system and design of a control algorithm.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving and intelligent chassis technology, specifically relating to a data-driven online estimation method and system for braking clearance in EMB systems. Background Technology

[0002] Currently, the main methods for estimating brake clearance in electromechanical braking (EMB) systems include:

[0003] 1. Threshold-based detection methods: Current and torque-based methods are significantly affected by temperature, motor efficiency, and noise. Under the same operating conditions, the signal fluctuates greatly and there are abnormal jumps, which leads to detection errors and reduced reliability. At the same time, the threshold depends on the recognition logic and needs to be repeatedly calibrated for different vehicle models, which lacks adaptability and robustness.

[0004] 2. Offline fitting model method: This method relies on the quality and quantity of the fitted data, places high demands on the storage and computing resources of the real-time control system, and is very costly. At the same time, it cannot be updated online and is difficult to cope with changes in brake clearance caused by brake pad wear.

[0005] 3. Based on the RLS recursive algorithm: It is extremely sensitive to noise and cannot meet the requirements of stable online estimation; at the same time, the regression vector is highly coupled internally, and the differences in various values ​​are too large, which leads to a sharp decline in the reliability of the estimated value.

[0006] Therefore, existing braking clearance estimation methods suffer from problems such as low accuracy, reliance on force sensors and offline calibration, lack of online adaptive capability, poor anti-interference ability, and numerical instability.

[0007] I. A search revealed that Chinese invention patent CN120277811A discloses a method for identifying brake gaps and estimating adaptive clamping force in an EMB system. This application differs from this one in the following ways: 1. Patent CN120277811A classifies brake gap types based on EMB contact time. It uses the motor torque balance equation and establishes a clamping force dynamic model using the motor's electromagnetic torque, angle, and angular velocity. Clamping force estimation is achieved through different types of extended state observers, with its core focus on the linkage between "gap classification + clamping force estimation." In contrast, this patent focuses on the online estimation of the brake gap itself. By exploring the correlation between motor current and rotation angle signals during the braking loading phase, it constructs a Legendre orthogonal polynomial nonlinear mapping model. Combined with an enhanced recursive least squares algorithm, it achieves online parameter identification and then directly analyzes the gap value through a zero-current extrapolation mechanism. The core technology emphasizes the accuracy and independence of gap estimation. 2. While patent CN120277811A achieves sensorless gap identification, it does not specify the accuracy index of gap estimation. Furthermore, its technical logic relies on the targeted design of the observer to match different gap types and is based on threshold judgment for identification, resulting in poor reliability and adaptability. In contrast, this patent eliminates the need for force sensors and offline calibration, explicitly achieving a steady-state estimation error of less than 0.6 rad (0.39%) for braking gap, and a relative root mean square error of less than 3.478 × 10⁻³ under large initial deviation conditions, demonstrating a greater advantage in balancing low cost and high accuracy. II. A search reveals that Chinese invention patent CN121062678A discloses a sensorless clamping force estimation method for EMB based on multi-signal fusion and displacement compensation, which differs from this application in the following ways:

[0008] 1. Patent CN121062678A, based on the mechanical model of the EMB system, achieves contact point detection through current-position information fusion, and then improves the clamping force estimation accuracy through position error compensation and backlash compensation. The core technology is the clamping force optimization strategy of "multi-signal fusion + multi-stage compensation". In contrast, this patent does not rely on contact point detection and subsequent compensation mechanisms. It can directly estimate the gap by using the polynomial mapping relationship between current and angle signals, combined with a dedicated identification algorithm. The technical path is simpler and avoids the cumulative errors that may be introduced by multiple compensation stages.

[0009] 2. The core objective of patent CN121062678A is to improve the clamping force estimation effect, with brake clearance related processing serving only as an auxiliary step in clamping force estimation. This patent, however, focuses on online brake clearance estimation as its core objective, building its technological advantage through clear accuracy indicators, while providing direct high-precision data support for clamping force control and clearance adaptive adjustment. Its technological positioning is more focused on the clearance parameter itself. III. A search reveals that Chinese invention patent CN118876927A discloses a multi-stage closed-loop control method and device for an EMB system based on a nonlinear disturbance observer, which differs from this application in the following ways:

[0010] 1. The core of patent CN118876927A is the design of differentiated closed-loop control strategies for different operating stages of EMB. In the braking gap elimination stage, it quickly eliminates the gap through a dual closed-loop system of speed loop and current loop, focusing on solving the braking response speed problem, with the goal of "rapid elimination" in gap handling. The core of this patent, however, is the "online accurate estimation" of the gap value. By establishing a precise mapping model of current-rotation angle, it achieves quantitative detection of the gap, rather than simply qualitative elimination, providing data for fine-tuning the gap.

[0011] 2. Patent CN118876927A achieves adaptive clearance adjustment through multi-closed-loop control, using contact angle and reference angle to control and achieve a fixed clearance. However, its adjustment basis does not explicitly include high-precision clearance estimation data. The high-precision clearance estimation method proposed in this patent can directly provide accurate clearance parameter input for this type of multi-stage control strategy, improving the adaptability and control accuracy of the control strategy from the bottom up. The two technologies are in a supporting and being supported relationship. IV. A search reveals that Chinese invention patent CN117341644A discloses an EMB system control method based on a three-closed-loop system and a load observer, which differs from this application in the following ways:

[0012] 1. Patent CN117341644A relies on a pressure sensor to detect the actual braking force, thereby determining the braking stage and switching the dual closed-loop control strategy. Its core is "sensor-assisted phased control," and its gap handling is implicit in the braking force judgment logic, without a separate gap estimation mechanism. In contrast, this patent eliminates the need for force or pressure sensors, constructing a separate online gap estimation model. High-precision gap detection can be achieved using only current signals, significantly reducing system costs.

[0013] 2. The technical advantages of patent CN117341644A are reflected in control aspects such as low torque jitter and smooth speed regulation. This patent not only achieves high-precision gap estimation but also provides a theoretical basis for the design of adaptive gap adjustment and control algorithms for EMB systems. Its technical value covers the entire chain of "parameter estimation - control support," and its application scope is wider. V. A search reveals CN118288958A, "An Automatic Brake Clearance Adjustment System and Method for Automotive EMB," which differs from this application in the following ways:

[0014] I. Patent CN118288958A monitors the braking clearance in real time by acquiring the displacement change of the roller screw nut. Its core relies on "mechanical structure displacement signal" for clearance detection, employing a cumulative sum of forward and reverse rotations and motor control based on a displacement threshold to maintain the braking clearance within a fixed range. However, its detection accuracy may be affected by mechanical transmission errors. In contrast, this patent constructs an estimation model based on the correlation between the motor current and rotation angle electrical signals. It eliminates the need for mechanical displacement detection components, avoiding mechanical error interference at the signal source level, and achieves higher estimation accuracy (steady-state error less than 0.6 rad).

[0015] Second, the key technology of patent CN118288958A is the closed-loop regulation mechanism of "gap monitoring + motor revolution adjustment". Gap monitoring is an auxiliary means of regulation. This patent takes online gap estimation technology as the core innovation point, which can not only support gap adjustment, but also provide basic data for the optimization of control algorithm of EMB system. The technological innovation focuses more on the estimation method itself, and the application scenarios are more flexible. Summary of the Invention

[0016] To address the aforementioned problems, this invention provides a data-driven online estimation method and system for brake clearance in EMB systems. This overcomes issues such as low accuracy, reliance on force sensors and offline calibration, lack of online adaptive capability, poor anti-interference performance, and numerical instability. It eliminates the need for force sensors and offline calibration, enabling low-cost online estimation of brake clearance in electromechanical braking systems. This provides high-precision brake clearance for drive-by-wire electromechanical braking systems and offers a theoretical basis for adaptive adjustment and control algorithm design for brake clearance in electromechanical braking systems.

[0017] To achieve the above objectives, the specific solution adopted by the present invention is as follows:

[0018] A data-driven online method for estimating braking clearance in an EMB system includes the following steps:

[0019] S1. Data anomaly preprocessing: Acquiring motor current during braking loading phase. Motor rotation angle during loading phase And perform outlier preprocessing on it;

[0020] S2. Construction of the Legendre Orthogonal Basis Regression Model: Based on the preprocessed current signal, a current-rotation angle regression model is constructed using Legendre orthogonal polynomials as basis functions. ;

[0021] S3, Enhanced RLS Parameter Recognition: Employing a forgetting factor Enhanced recursive least squares method for parameter vectors in regression models Perform online identification and updates;

[0022] S4. Zero-current extrapolation estimation of braking gap: This involves estimating the parameter vector online. Substituting the values ​​into the zero-current extrapolation calculation yields the estimated value of the braking gap. .

[0023] As a preferred embodiment, in step S1, the motor current during the loading stage is... Motor rotation angle during loading phase Outlier preprocessing is performed based on physical boundaries and single-point recursive statistics.

[0024] As a preferred embodiment, the construction of the regression model in step S2 specifically includes:

[0025] According to the braking motion law of the electromechanical braking system, during the entire loading process from the initial position, there is a definite mathematical relationship between its current-angle global mapping relationship and the current braking gap, as follows:

[0026] ;

[0027] in, This indicates the motor current during the loading phase. Indicates the motor rotation angle during the loading phase. express and The nonlinear function, where A, B, C, and D are constants;

[0028] Motor current during loading phase Perform linear normalization to interval :

[0029] ;

[0030] in, This indicates the preset maximum current value under actual measurement conditions. The auxiliary variable representing linear normalization, This indicates the motor current during the loading phase;

[0031] Legendre orthogonal base in the interval It has orthogonality, and the weight function The orthogonal polynomial form is as follows:

[0032] ;

[0033] in, and Represents the Legendre base. Auxiliary variables representing linear normalization;

[0034] For the current-rotation angle mapping relationship, the first four Legendre bases are used, namely:

[0035] ;

[0036] in, Represents the first Legendre base. Indicates the second Legendre base. Indicates the third Legendre base, This represents the fourth Legendre base. Auxiliary variables representing linear normalization;

[0037] At the same time, the Legendre basis satisfies the following orthogonality:

[0038] , ;

[0039] in, and Represents the Legendre base. Auxiliary variables representing linear normalization;

[0040] Constructing a regression vector Define parameter vector The current-rotation angle regression model based on the Legendre orthogonal basis is obtained as follows:

[0041] ;

[0042] in, The current-rotation regression model is based on a Legendre orthogonal basis. , , , Based on Legendre , , , The coefficients are those corresponding to the Legendre basis;

[0043] Current-rotation regression model based on Legendre orthogonal basis Convert to matrix form:

[0044] ;

[0045] in, The current-rotation regression model is based on a Legendre orthogonal basis. For the regression vector, For parameter vectors.

[0046] As a preferred embodiment, step S3 employs a forgetting factor. Furthermore, an enhanced RLS algorithm is established by combining the Legendre orthogonal basis for online identification and updating, and a current-rotation angle regression model based on the Legendre orthogonal basis is constructed. The predictor and cost function are defined as follows:

[0047] ;

[0048] ;

[0049] in, This indicates the predicted motor rotation angle. For the regression vector, Represents the parameter vector estimated online. This represents the weighted sum of squares of the prediction error. Forgetting factor, Current-rotation regression model based on Legendre orthogonal basis;

[0050] Constructing an enhanced RLS with a forgetting factor:

[0051] ;

[0052] in, This represents the gain matrix for parameter estimation. Represents the covariance matrix. Forgetting factor, For the regression vector, Represents the parameter vector estimated online. The current-rotation regression model is based on the Legendre orthogonal basis.

[0053] As a preferred embodiment, the current parameter estimate is obtained in step S4. Then, normalize the auxiliary variables. Extrapolating to the theoretical zero current point, i.e. At that time, the estimated value of the braking clearance is obtained. :

[0054] ;

[0055] in, This is the estimated value for the braking clearance. , , , These are the estimated Legendre basis coefficients.

[0056] As a preferred embodiment, step S5 is included after step S4:

[0057] Update and store historical data: After each braking cycle, based on the estimated braking clearance... Trigger a recursive iteration to update the initial value and store historical results to improve the cold start speed of the electromechanical braking system.

[0058] This invention also provides a data-driven online braking clearance estimation system for EMB systems, comprising:

[0059] A data preprocessing module is configured to perform step S1 in the method to preprocess outliers in the acquired current and angle signals.

[0060] A model building module, configured to execute step S2 in the method, for building a Legendre orthogonal basis regression model;

[0061] A parameter identification module is configured to execute step S3 in the method to achieve online adaptive identification of parameters.

[0062] A clearance calculation module, configured to perform step S4 in the method, for calculating an estimated braking clearance value;

[0063] The storage module is updated and configured to perform step S5 in the method to update the initial state and manage historical data.

[0064] As a preferred embodiment, the system is integrated into the vehicle's electromechanical braking system to provide real-time estimates for adaptive adjustment of the braking clearance.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] (1) Braking gap estimation can be completed using only motor current and angle signals, eliminating the dependence on high-cost force sensors and the calibration of complex stiffness models, and is suitable for mass-produced vehicles;

[0067] (2) Based on the Legendre enhanced RLS recursive algorithm, the condition number of the regression matrix is ​​significantly reduced, the problem of multicollinearity of parameters is solved, and the convergence speed of the RLS algorithm is accelerated. At the same time, through linear normalization, the numerical explosion of higher-order terms is suppressed, the influence of measurement noise is effectively suppressed, and it has strong robustness to non-ideal initial bias.

[0068] (3) The braking gap estimation maps the mathematical fitting results to the actual braking gap through the "zero current extrapolation" mechanism, which has clear physical meaning and interpretability, while being computationally efficient and meeting the functional safety verification and vehicle deployment requirements of the electromechanical braking system. Attached Figure Description

[0069] Figure 1 This is the technical approach of the present invention;

[0070] Figure 2 These are the results of data preprocessing and braking clearance estimation;

[0071] Figure 3 These are the results of the Legendre-RLS parameter estimation;

[0072] Figure 4 This is the result of the algorithm's robustness verification. Detailed Implementation

[0073] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The specific implementation methods of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0074] Example:

[0075] like Figure 1 As shown, this invention provides a data-driven online estimation method for braking clearance in an EMB system, comprising the following steps:

[0076] S1. Data anomaly preprocessing: Acquiring motor current during braking loading phase. Motor rotation angle during loading phase And perform outlier preprocessing on it;

[0077] S2. Construction of the Legendre Orthogonal Basis Regression Model: Based on the preprocessed current signal, a current-rotation angle regression model is constructed using Legendre orthogonal polynomials as basis functions. ;

[0078] S3, Enhanced RLS Parameter Recognition: Employing a forgetting factor Enhanced recursive least squares method for parameter vectors in regression models Perform online identification and updates;

[0079] S4. Zero-current extrapolation estimation of braking gap: This involves estimating the parameter vector online. Substituting the values ​​into the zero-current extrapolation calculation yields the estimated value of the braking gap. .

[0080] Preferably, in step S1, the motor current during the loading stage is... Motor rotation angle during loading phase Outlier preprocessing is performed based on physical boundaries and single-point recursive statistics.

[0081] Preferably, the construction of the regression model in step S2 specifically includes:

[0082] According to the braking motion law of the electromechanical braking system, during the entire loading process from the initial position, there is a definite mathematical relationship between its current-angle global mapping relationship and the current braking gap, as follows:

[0083] ;

[0084] in, This indicates the motor current during the loading phase. Indicates the motor rotation angle during the loading phase. express and The nonlinear function, where A, B, C, and D are constants;

[0085] Motor current during loading phase Perform linear normalization to interval :

[0086] ;

[0087] in, This indicates the preset maximum current value under actual measurement conditions. The auxiliary variable representing linear normalization, This indicates the motor current during the loading phase;

[0088] Legendre orthogonal base in the interval It has orthogonality, and the weight function The orthogonal polynomial form is as follows:

[0089] ;

[0090] in, and Represents the Legendre base. Auxiliary variables representing linear normalization;

[0091] For the current-rotation angle mapping relationship, the first four Legendre bases are used, namely:

[0092] ;

[0093] in, Represents the first Legendre base. Indicates the second Legendre base. Indicates the third Legendre base, This represents the fourth Legendre base. Auxiliary variables representing linear normalization;

[0094] At the same time, the Legendre basis satisfies the following orthogonality:

[0095] , ;

[0096] in, and Represents the Legendre base. Auxiliary variables representing linear normalization;

[0097] Constructing a regression vector Define parameter vector The current-rotation angle regression model based on the Legendre orthogonal basis is obtained as follows:

[0098] ;

[0099] in, The current-rotation regression model is based on a Legendre orthogonal basis. , , , Based on Legendre , , , The coefficients are those corresponding to the Legendre basis;

[0100] Current-rotation regression model based on Legendre orthogonal basis Convert to matrix form:

[0101] ;

[0102] in, The current-rotation regression model is based on a Legendre orthogonal basis. For the regression vector, For parameter vectors.

[0103] Preferably, step S3 uses a forgetting factor. Furthermore, an enhanced RLS algorithm is established by combining the Legendre orthogonal basis for online identification and updating, and a current-rotation angle regression model based on the Legendre orthogonal basis is constructed. The predictor and cost function are defined as follows:

[0104] ;

[0105] ;

[0106] in, This indicates the predicted motor rotation angle. For the regression vector, Represents the parameter vector estimated online. This represents the weighted sum of squares of the prediction error. Forgetting factor, Current-rotation regression model based on Legendre orthogonal basis;

[0107] Constructing an enhanced RLS with a forgetting factor:

[0108] ;

[0109] in, This represents the gain matrix for parameter estimation. Represents the covariance matrix. Forgetting factor, For the regression vector, Represents the parameter vector estimated online. The current-rotation regression model is based on the Legendre orthogonal basis.

[0110] Preferably, the current parameter estimate is obtained in step S4. Then, normalize the auxiliary variables. Extrapolating to the theoretical zero current point, i.e. At that time, the estimated value of the braking clearance is obtained. :

[0111] ;

[0112] in, This is the estimated value for the braking clearance. , , , These are the estimated Legendre basis coefficients.

[0113] Preferably, step S5 is included after step S4:

[0114] Update and store historical data: After each braking cycle, based on the estimated braking clearance... Trigger a recursive iteration to update the initial value and store historical results to improve the cold start speed of the electromechanical braking system.

[0115] This invention also provides a data-driven online braking clearance estimation system for EMB systems, comprising:

[0116] A data preprocessing module is configured to perform step S1 in the method to preprocess outlier values ​​in the current and angle signals acquired by the loaded data acquisition module.

[0117] A model building module, configured to execute step S2 in the method, for building a Legendre orthogonal basis regression model;

[0118] A parameter identification module is configured to execute step S3 in the method to achieve online adaptive identification of parameters.

[0119] A clearance calculation module, configured to perform step S4 in the method, for calculating an estimated braking clearance value;

[0120] The storage module is updated and configured to perform step S5 in the method to update the initial state and manage historical data.

[0121] Preferably, the system is integrated into the vehicle's electromechanical braking system to provide a real-time estimate for adaptive adjustment of the braking clearance.

[0122] Furthermore, to verify the effectiveness of the braking gap estimation method based on motor current fitting of this invention, and to verify the performance advantages of Legendre Enhanced RLS (LE-RLS) compared to traditional RLS, experimental verification was conducted, and the results are as follows: Figure 2 As shown, Figure 2 As shown in (a), based on the measured clamping force-rotation angle data, offline polynomial fitting yields the measured braking gap as follows: , Figure 2 (b) shows the offline fitting results of the motor current-rotation angle relationship before and after data processing. The fitting results of the two are as follows: and This demonstrates the feasibility of decoupling modeling relying solely on current signals, and that preprocessing does not affect braking gap estimation, thus further improving the data quality of recursive fitting. Figure 2(c) The online estimation performance of LP-RLS and traditional RLS was compared. LP-RLS converged quickly to the vicinity of the reference value within about 5 braking cycles (cumulative loading of 10 s), and the final error was within 0.6 rad (0.39%), showing excellent convergence and stability. In contrast, traditional RLS has an ill-conditioned regression matrix due to the non-orthogonalization of the regression vector, strong coupling between parameters and large differences in magnitude. It is highly sensitive to measurement noise and the parameters continue to oscillate.

[0123] like Figure 3 As shown, the parameter recursion process of Legendre-RLS is illustrated. The orthogonal coefficients of each order converge rapidly within 5 braking cycles. The specific errors are shown in Table 1.

[0124] Experimental results show that the introduction of the Legendre orthogonal basis significantly improves the numerical conditions of the regression model, achieves parameter decoupling, and greatly enhances the robustness and convergence reliability of the algorithm under dynamic conditions.

[0125] To further evaluate the convergence robustness of the proposed Legendre-enhanced RLS algorithm under non-ideal initial conditions, a series of extreme initial deviations were set up for verification, based on measured initial rotation angles. Apply respectively , , and The offset was used to construct a scene far exceeding the actual initial zero point, and the result was as follows: Figure 4 As shown, although the initial deviation is as high as The algorithm converged successfully under all test conditions, and the steady-state estimation error was within [a certain range]. Within, the relative error is lower than This fully demonstrates that the algorithm is highly insensitive to initial deviations, and the relative root mean square error of the iteration results is less than 3.478×10⁻³ under the condition of large initial deviations.

[0126] In summary, this invention enables reliable estimation without precise initial values, significantly improving the adaptive capability and application flexibility of electromechanical braking systems. Furthermore, based on existing signal data, it achieves low-cost, online, and high-precision estimation of braking gaps in electromechanical braking systems.

[0127] This invention can be widely applied to online estimation of braking clearance in electromechanical braking systems of passenger cars, commercial vehicles, motorcycles, and all-electric flying cars.

[0128] The above embodiments are for illustrative purposes only and are not intended to limit the scope of this invention. Although this invention has been described in detail with reference to the embodiments, those skilled in the art should understand that various combinations, modifications, or equivalent substitutions of the technical solutions of this invention do not depart from the spirit and scope of the technical solutions of this invention and should be covered within the scope of the claims of this invention.

Claims

1. A data-driven online estimation method for braking clearance in an EMB system, characterized in that, Includes the following steps: S1. Data anomaly preprocessing: Acquiring motor current during braking loading phase. Motor rotation angle during loading phase And perform outlier preprocessing on it; S2. Construction of the Legendre Orthogonal Basis Regression Model: Based on the preprocessed current signal, a current-rotation angle regression model is constructed using Legendre orthogonal polynomials as basis functions. ; S3, Enhanced RLS Parameter Recognition: Employing a forgetting factor Enhanced recursive least squares method for parameter vectors in regression models Perform online identification and updates; S4. Zero-current extrapolation estimation of braking gap: This involves estimating the parameter vector online. Substituting the values ​​into the zero-current extrapolation calculation yields the estimated value of the braking gap. .

2. The data-driven online estimation method for braking clearance in an EMB system according to claim 1, characterized in that, In step S1, the motor current during the loading stage is... Motor rotation angle during loading phase Outlier preprocessing is performed based on physical boundaries and single-point recursive statistics.

3. The data-driven online estimation method for braking clearance in an EMB system according to claim 2, characterized in that, The construction of the regression model in step S2 specifically includes: According to the braking motion law of the electromechanical braking system, during the entire loading process from the initial position, there is a definite mathematical relationship between its current-angle global mapping relationship and the current braking gap, as follows: ； in, This indicates the motor current during the loading phase. Indicates the motor rotation angle during the loading phase. express and The nonlinear function, where A, B, C, and D are constants; Motor current during loading phase Perform linear normalization to interval : ； in, This indicates the preset maximum current value under actual measurement conditions. The auxiliary variable representing linear normalization, This indicates the motor current during the loading phase; Legendre orthogonal base in the interval It has orthogonality, and the weight function The orthogonal polynomial form is as follows: ； in, and Represents the Legendre base. Auxiliary variables representing linear normalization; For the current-rotation angle mapping relationship, the first four Legendre bases are used, namely: ； in, Represents the first Legendre base. Indicates the second Legendre base. Indicates the third Legendre base, This represents the fourth Legendre base. Auxiliary variables representing linear normalization; At the same time, the Legendre basis satisfies the following orthogonality: ， ； in, and Represents the Legendre base. Auxiliary variables representing linear normalization; Constructing a regression vector Define parameter vector The current-rotation angle regression model based on the Legendre orthogonal basis is obtained as follows: ； in, The current-rotation regression model is based on a Legendre orthogonal basis. , , , Based on Legendre , , , The coefficients are those corresponding to the Legendre basis; Current-rotation regression model based on Legendre orthogonal basis Convert to matrix form: ； in, The current-rotation regression model is based on a Legendre orthogonal basis. For the regression vector, For parameter vectors.

4. The data-driven online estimation method for braking clearance in an EMB system according to claim 3, characterized in that, Step S3 employs a forgetting factor. Furthermore, an enhanced RLS algorithm is established by combining the Legendre orthogonal basis for online identification and updating, and a current-rotation angle regression model based on the Legendre orthogonal basis is constructed. The predictor and cost function are defined as follows: ； ； in, This indicates the predicted motor rotation angle. For the regression vector, Represents the parameter vector estimated online. This represents the weighted sum of squares of the prediction error. Forgetting factor, Current-rotation regression model based on Legendre orthogonal basis; Constructing an enhanced RLS with a forgetting factor: ； in, This represents the gain matrix for parameter estimation. Represents the covariance matrix. Forgetting factor, For the regression vector, Represents the parameter vector estimated online. The current-rotation regression model is based on the Legendre orthogonal basis.

5. The data-driven online estimation method for braking clearance in an EMB system according to claim 4, characterized in that, In step S4, the current parameter estimate is obtained. Then, normalize the auxiliary variables. Extrapolating to the theoretical zero current point, i.e. At that time, the estimated value of the braking clearance is obtained. : ； in, This is the estimated value for the braking clearance. , , , These are the estimated Legendre basis coefficients.

6. The data-driven online estimation method for braking clearance in an EMB system according to claim 5, characterized in that, Step S4 is followed by step S5: Update and store historical data: After each braking cycle, based on the estimated braking clearance... Trigger a recursive iteration to update the initial value and store historical results to improve the cold start speed of the electromechanical braking system.

7. A data-driven online braking clearance estimation system for an EMB system, characterized in that, include: A data preprocessing module, configured to perform step S1 in the method according to claim 2, for performing outlier preprocessing on the acquired current and angle signals; A model building module, configured to perform step S2 in the method according to claim 3, for building a Legendre orthogonal basis regression model; A parameter identification module, configured to execute step S3 in the method according to claim 4, for realizing online adaptive identification of parameters; A clearance calculation module, configured to perform step S4 in the method according to claim 5, for calculating an estimated braking clearance value; An update storage module is configured to perform step S5 in the method according to claim 6, for updating the initial state and managing historical data.

8. The data-driven online braking clearance estimation system for an EMB system according to claim 7, characterized in that, The system is integrated into the vehicle's electromechanical braking system to provide real-time estimates for adaptive adjustment of the braking clearance.

Images