An EMB system brake clearance identification and adaptive clamping force estimation method

Abstract

The present application belongs to the technical field of EMB braking, and relates to a kind of EMB system brake gap identification and adaptive clamping force estimation method, including according to EMB contact time, brake gap is divided into normal brake gap, large brake gap and small brake gap;According to motor torque balance equation, the electromagnetic torque, angle and angular velocity of motor are used to establish brake clamping force dynamics model;According to brake clamping force dynamics model, extended state observer of fusion neural network and angle compensation extended state observer A and B are respectively constructed, and brake clamping force is estimated according to brake gap by using extended state observer of fusion neural network, angle compensation extended state observer A and B;The present application estimates the size of brake gap by using pressure sensorless estimation algorithm, and realizes adaptive brake clamping force estimation under different brake gaps.

Description

Technical Field

[0001] This invention belongs to the technical field of EMB braking systems, specifically relating to a method for identifying braking gaps and estimating adaptive clamping force in EMB systems. Background Technology

[0002] With the major trend of automotive electrification and intelligentization, EMB, driven by an electric motor, boasts an extremely fast response time (80-100 milliseconds), a significant advantage over EHB's 150 milliseconds, resulting in higher safety. It requires no hydraulic oil, is environmentally friendly, easily integrates new functions, is convenient to assemble, simplifies system structure, and enhances driving comfort.

[0003] Currently, the placement and application of pressure sensors in EMB integrated structures face numerous challenges: First, miniature pressure sensors meeting accuracy standards are expensive, and their stringent assembly precision requirements further increase production costs. Second, the integrated layout and calibration of pressure sensors are difficult; placing them at the front end of the lead screw can cause interference with measurement accuracy due to the high-temperature environment near the brake pads; embedding the sensor inside the actuator may lead to frictional hysteresis, further affecting vehicle braking performance and compromising driving safety. Furthermore, the braking clearance changes during vehicle operation, and traditional estimation algorithms cannot guarantee accurate estimation of clamping force under different braking clearances.

[0004] Therefore, in order to avoid the problem of reduced braking force accuracy and impact on braking safety caused by changes in the arrangement of pressure sensors and braking clearance, it is necessary to invent a clamping force estimation algorithm that can identify contact time without pressure sensors and adapt to the braking clearance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for identifying the brake gap and estimating the adaptive clamping force in an EMB system.

[0006] To achieve the objectives of this invention, the following technical solutions are adopted.

[0007] A method for identifying brake clearance and estimating adaptive clamping force in an EMB system includes the following steps:

[0008] S1. Based on the EMB contact time, the EMB braking gap is divided into normal braking gap state, large braking gap state, and small braking gap state; where: the EMB contact time is determined according to the change of motor angular velocity;

[0009] S2. Based on the motor torque balance equation of EMB during braking, a dynamic model of braking clamping force of EMB during braking is established using the electromagnetic torque, motor angle and motor angular velocity of the motor.

[0010] S3. Based on the dynamic model of the braking clamping force during the braking process of EMB, construct an extended state observer with fused neural network and an extended state observer with angle compensation respectively.

[0011] When the EMB brake clearance is in the normal brake clearance state, the extended state observer with fused neural network is used to estimate the brake clamping force.

[0012] When the EMB brake clearance is in a large brake clearance state, the extended state observer A with angle compensation is used to estimate the brake clamping force.

[0013] When the EMB brake clearance is in a small brake clearance state, the extended state observer B with angle compensation is used to estimate the brake clamping force.

[0014] As a preferred embodiment of the present invention, the specific implementation process of step S1 includes the following steps:

[0015] S101, Angular velocity sensor is used to collect the motor angular velocity The rate of change of the motor's angular velocity is obtained through calculation;

[0016] S102, Pre-calibrate the threshold for motor angular velocity change during EMB contact. When the rate of change of motor angular velocity obtained in step S101 is less than or equal to At that time, that is At that time, the EMB contact time was collected. ;

[0017] S103. Based on the reasonable braking gap range set by EMB, pre-calibrate the contact time. reasonable range ;like This is judged to be a normal braking clearance; if The problem is determined to be due to excessive braking clearance; if The problem was determined to be due to insufficient braking clearance.

[0018] As a preferred embodiment of the present invention, the specific implementation process of step S2 includes the following steps:

[0019] S201. Set the motor electromagnetic torque as follows: ,

[0020] In the formula, This represents the number of pole pairs of the motor. , , and The magnetic flux on the stator side and Coordinate system current;

[0021] S202. The EMB braking process is divided into two processes: clamping and releasing.

[0022] During the clamping process, based on the torque balance equation of the motor, we can derive:

[0023] ;

[0024] In the formula, The angular velocity of the motor. For time, The moment of inertia of the motor. For the electromagnetic torque of the motor, This refers to the dynamic offset of the inherent parameters of the motor. Let Coulomb friction coefficient be the coefficient of friction of the motor. This is the magnification factor;

[0025] During the release process, we can conclude that:

[0026] ;

[0027] In the formula, and It was measured using the EMB clamping force test with the motor torque as the abscissa.

[0028] S203. Based on the motor torque balance equations for the clamping and releasing processes, a dynamic model of the braking clamping force during the clamping and releasing processes is established, wherein:

[0029] The dynamic model of the braking clamping force during the clamping process is as follows:

[0030] ;

[0031] In the formula, , This represents the sum of clamping torque and frictional torque. The derivative of the sum of clamping torque and frictional torque. express The derivative, For the motor angle, express The derivative, This is the braking clamping force;

[0032] The dynamic model of the braking clamping force during the release process is as follows:

[0033] ;

[0034] In the formula, .

[0035] As a preferred embodiment of the present invention, the extended state observer of the fusion neural network includes extended state observers of the fusion neural network for the clamping process and the release process, wherein:

[0036] The expression for the extended state observer of the fusion neural network in the clamping process is:

[0037] ;

[0038] The expression for the extended state observer of the fusion neural network during the release process is:

[0039] ;

[0040] in, for The estimated value, For the error term, i.e. , for The estimated value, for The estimated value, For time step, This refers to the gain parameter output by the neural network during the clamping process. The gain parameter is the output of the neural network during the release process. express function, The function is:

[0041] .

[0042] As a preferred embodiment of the present invention, the process of obtaining the gain parameter is as follows: the number of nodes in the neural network input layer is... The input to the neural network is ;

[0043] The inputs and outputs of the hidden layer are as follows:

[0044] ;

[0045] ;

[0046] In the formula, Let be the weights representing the influence of the i-th input node of the neural network on the j-th hidden layer. For the neural network One input, , For the hidden layer The input and output of each node; , These are the input layer and hidden layer nodes, respectively; , These are the translation factor and the scaling factor, respectively.

[0047] The output of the output layer is:

[0048]

[0049] ;

[0050] In the formula, , The output layer is respectively The actual output and expected output of each node; For error; These are the weights of the neural network; This represents the number of nodes in the output layer.

[0051] At the end of each step, if the EMB is in the clamping process, the gain parameters of the neural network output during the clamping process are obtained. If the EMB is in the release process, then the gain parameter of the neural network output during the clamping process is obtained. .

[0052] As a preferred embodiment of the present invention, the angle-compensated extended state observer A includes an extended state observer A for clamping process angle compensation and an extended state observer A for release process angle compensation, wherein:

[0053] The expression for the extended state observer A for clamping process angle compensation is: ;

[0054] The expression for the extended state observer A for angle compensation during the release process is:

[0055] ;

[0056] In the formula, Based on contact time Nonlinear angle compensation curve of large braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the large braking clearance estimation model. The gain parameter is determined by the reference angle in the large braking clearance estimation model.

[0057] As a preferred embodiment of the present invention, the process of establishing the angle-compensated extended state observer A includes the following steps:

[0058] S701, respectively establish as For the estimation model of wear interval, with corresponding As the reference angle in the large braking clearance estimation model;

[0059] S702. If the braking clearance continues to increase, the motor angle should be adjusted. In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle.

[0060] S703, Fitting the contact time The x-axis represents the motor angle. Nonlinear compensation curve of the large braking clearance estimation model with ordinate as the vertical axis The influence parameters caused by different wear amounts are unified into the angle compensation to obtain the extended state observer A of angle compensation.

[0061] As a preferred embodiment of the present invention, the angle-compensated extended state observer B includes an extended state observer B for clamping process angle compensation and an extended state observer B for release process angle compensation, wherein:

[0062] The expression for the extended state observer B for clamping process angle compensation is:

[0063] ;

[0064] The expression for the extended state observer B for angle compensation during the release process is:

[0065] ;

[0066] In the formula, Based on contact time Nonlinear angle compensation curve of small braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the model with a small braking clearance estimation. The gain parameter is determined by the reference angle in the model with a small braking gap estimation.

[0067] As a preferred embodiment of the present invention, the process of establishing the angle-compensated extended state observer B includes the following steps:

[0068] S901, respectively establish as For the estimation model of the interval, corresponding As the reference angle in the small braking gap estimation model;

[0069] S902. If the braking gap continues to decrease, the motor angle... In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle.

[0070] S903, Fitting the contact time The x-axis represents the motor angle. The nonlinear compensation curve of the small braking clearance estimation model is shown on the ordinate. The influence parameters caused by excessively small braking clearance are unified into the angle compensation to obtain the extended state observer B of angle compensation.

[0071] As a preferred embodiment of the present invention, the training set data update process of the neural network includes the following steps:

[0072] S1001. During each braking action, the neural network updates the weights within the network using the gradient descent method to obtain the gain parameters; only when the vehicle is parked is the neural network updated as a whole by updating the training set data.

[0073] S1002. The training set data acquisition method for the neural network is as follows: The compliance of the clamping force corresponding to each step length during the braking process is detected, and the compliance detection standard is:

[0074] On level roads, record the vehicle speed when the brake pedal is depressed. When the brake pedal is released, braking is considered complete, and the vehicle speed at that moment is recorded. During braking, at each interval Record vehicle speed at all times Until braking ends; the actual clamping force is calculated using the formula: , , The data collection interval is where For longitudinal resistance, For vehicle quality, The number of brake wheels, To generate braking torque for a single wheel, The effective radius of the tire. The radius of the brake disc, The coefficient of friction between the brake pads and the brake disc. For real braking force, , The rolling resistance and wind resistance experienced by the vehicle at the current speed are measured experimentally; if In the formula, To specify a standard, the input parameters and output gain parameters corresponding to the step size braking are retained as the dataset for neural network updates.

[0075] Beneficial effects

[0076] This invention can detect the braking gap of the EMB during each braking process and use different estimation models according to the current braking gap, thus solving the problems of high cost, high failure rate and inability to guarantee the accuracy of different braking gaps, which affect the braking of the whole vehicle.

[0077] This invention first obtains the required braking clamping force using a brake pedal position sensor. Next, by identifying the motor angular velocity The sudden change signal is used to determine the contact time and braking gap; secondly, the torque balance equation of the motor is considered, and the electromagnetic torque of the motor is utilized. And the motor (angle) A basic observer model is established, and the clamping force is estimated using estimation models for normal clearance, large braking clearance, and small braking clearance, respectively, based on the state of the braking clearance. The conventional brake gap estimation model uses an observer fused with a neural network, while the large brake gap estimation model and the small brake gap estimation model employ nonlinear angle compensation.

[0078] This invention can accurately identify the contact time and braking gap, and based on the braking gap, establish estimation models for normal gap, large braking gap and small braking gap respectively based on the extended state observer, so as to realize adaptive clamping force estimation under the current braking gap of EMB, and solve the problems that may occur when using pressure sensors and using a single estimation model. Attached Figure Description

[0079] Figure 1 This is a detailed flowchart of the method for brake gap identification and adaptive clamping force estimation of an EMB system based on a pressure sensor-free system according to the present invention.

[0080] Figure 2 This is a contact time identification diagram in an embodiment of the present invention;

[0081] Figure 3 This is the topology graph optimized by the wavelet neural network in an embodiment of the present invention;

[0082] Figure 4 This is a flowchart of the ESO process that integrates wavelet neural networks in the conventional braking gap estimation model in this embodiment of the invention. Detailed Implementation

[0083] The present invention will be further described in conjunction with the embodiments and accompanying drawings.

[0084] As an embodiment of the present invention, such as Figure 1 As shown, a method for brake gap identification and adaptive clamping force estimation in an EMB system based on a pressure sensorless system includes the following steps:

[0085] S1: The EMB contact time is identified by measuring the motor angular velocity, and the specific process for classifying the brake clearance into three categories—normal brake clearance, large brake clearance, and small brake clearance—based on the contact time is as follows:

[0086] S1: The EMB contact time is identified by measuring the motor angular velocity, and the specific process for classifying the brake clearance into three categories—normal brake clearance, large brake clearance, and small brake clearance—based on the contact time is as follows:

[0087] (1) Acquire the angular velocity of the motor through an angular velocity sensor ;

[0088] (2) The threshold for angular velocity change during EMB contact is pre-calibrated as follows: When the rate of change of the motor's angular velocity is less than or equal to At that time, that is At that time, the EMB contact time was collected. ;

[0089] (3) Based on the reasonable braking gap range set by EMB, pre-calibrate the contact time. reasonable range ;like This is judged to be a normal braking clearance; if The problem is determined to be due to excessive braking clearance; if The problem was determined to be due to insufficient braking clearance.

[0090] S2: The specific process for establishing the EMB dynamics model is as follows:

[0091] (1) Set the electromagnetic torque of the motor to be ,

[0092] in This represents the number of pole pairs of the motor. , , and The magnetic flux on the stator side and Coordinate system current;

[0093] (2) Divide the EMB braking into two processes: clamping and releasing.

[0094] For the clamping process, according to the torque balance equation of the motor, we can derive:

[0095]

[0096] In the formula, For the angular velocity, For time, Let be the moment of inertia of the motor. For the electromagnetic torque of the motor, This refers to the dynamic offset of the inherent parameters of the motor. Let Coulomb friction coefficient be the coefficient of friction of the motor. This is the magnification factor;

[0097] Regarding the release process, we can conclude that:

[0098]

[0099] In the above formula and The clamping force can be measured using the motor torque as the abscissa through an EMB clamping force test.

[0100] (3) Based on the torque balance equation, an extended state observer is built to estimate the clamping force:

[0101] During the clamping process, the dynamic model of the braking clamping force is as follows:

[0102]

[0103] In the formula, , This represents the sum of the clamping torque and the frictional torque;

[0104] During the release process, the dynamic model of the braking clamping force is as follows:

[0105]

[0106] In the formula, ;

[0107] Therefore, regarding the clamping process Release process ;

[0108] S3: The specific process for estimating the braking clamping force using an extended state observer with a fusion neural network for conventional braking gaps and an extended state observer with angle compensation for large and small braking gaps is as follows:

[0109] (1) An extended state observer of the braking clamping force is built based on the EMB dynamic model. The observer is based on the motor (angle). Electromagnetic torque As input, estimate Then obtain :

[0110] Observer of the expansion state of the clamping force during the clamping process:

[0111]

[0112] After discretization:

[0113]

[0114] In the formula: for The estimated value, For the error term, i.e. , for The estimated value, for The estimated value, For time step, This refers to the gain parameter output by the neural network during the clamping process. The gain parameter is the output of the neural network during the release process. express function, The function is:

[0115] ;

[0116] Similarly, an extended state observer of the braking clamping force during the release process:

[0117]

[0118] After discretization:

[0119]

[0120] In the formula: ;

[0121] For the clamping process Release process ;

[0122] (2) If actual contact Within the specified range, i.e., when the EMB is in the normal braking gap, based on the above observer expression, the extended state observer method that integrates an updatable neural network is used to adapt to small-range deviations. In this embodiment, a wavelet neural network will be specifically used to achieve accurate estimation of clamping force.

[0123] Among them, the conventional braking gap estimation model uses a neural network to obtain the gain parameters of the extended state observer. , , A wavelet neural network is employed to optimize the tracking effect on clamping force through training. The hidden layers of the wavelet neural network use wavelet functions as activation functions, and the learning method is gradient descent. The topology of the wavelet function in this embodiment is as follows: Figure 3 As shown:

[0124] The input layer has 4 nodes, namely: In the formula , ,in: To meet the driver's braking clamping force requirements;

[0125] The hidden layer is 1, and the number of nodes is ;

[0126] Output layer node The output contains the parameters required in ESO. , , ;

[0127] Wavelet function selection: Morlet mother wavelet basis function

[0128]

[0129] Let the objective function of the wavelet neural network be:

[0130]

[0131] In the formula, ;

[0132] Gradient descent is used to correct the weights of the wavelet neural network. and and wavelet basis function coefficients , ;

[0133]

[0134] In the formula: , , , For the wavelet neural network, the first The weights and coefficients are updated at each step. The learning rate is used; when adjusting the weights and coefficients, a gradient descent learning method is needed to derive the values ​​sequentially. , , , ;

[0135]

[0136] In the formula: , The first The implicit layer of time The input and output of each node; For the output layer The output of each node; due to Unknown can be replaced by approximations. Finally, we can conclude that:

[0137]

[0138] At the end of each step, if the EMB is in the clamping process, the gain parameters of the neural network output during the clamping process are obtained. If the EMB is in the release process, then the gain parameter of the neural network output during the clamping process is obtained. ;

[0139] The ESO flowchart fused with wavelet neural networks in this embodiment is as follows: Figure 4 As shown, the specific process is as follows: After the wavelet neural network is trained, the network is initialized, at time t. Based on the network output, an initial learning rate is set relatively small. Subsequently, the learning rate is appropriately increased when the error curve becomes relatively flat and the error decreases, and decreased otherwise. Updated neural network weights are obtained; gain parameters are then obtained. , , and system output ;when At any given moment, the learning rate in that direction is increased in a tiered manner until the error no longer decreases, and then the threshold and weights at that moment are recorded; conversely, the learning rate is decreased until the error decreases, and the threshold and weights are recorded; then the process proceeds to the next moment.

[0140] The expression for the expanded state observer after incorporating the neural network, and the clamping process are as follows:

[0141]

[0142] The release process is as follows:

[0143]

[0144] in, This refers to the gain parameter output by the neural network during the clamping process. This refers to the gain parameter of the neural network output during the release process;

[0145] For the clamping process Release process ;

[0146] (3) If the actual contact time If the EMB is outside the specified range, i.e., under abnormal braking clearance, the estimation model is corrected using a nonlinear angle compensation method based on the extended state observer expression mentioned above. Specifically, the braking clamping force estimation model established under the large braking clearance state is the large braking clearance estimation model. The method for obtaining the nonlinear angle compensation curve in the large braking clearance estimation model is as follows:

[0147] ① Establish separately with For the estimation model of wear interval, with correspond As the reference angle in the large braking clearance estimation model;

[0148] ② If the braking gap continues to increase, the motor (angle) should be adjusted. In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle.

[0149] ③ Fit the contact time The x-axis represents the motor angle. The nonlinear compensation curve of the large braking clearance estimation model with the vertical axis as the ordinate. The influence parameters caused by different wear amounts are unified into the angle compensation.

[0150] By using angle compensation to adapt to changes in the braking clearance, accurate estimation of the clamping force is achieved; the observer expression corresponding to the large braking clearance estimation model is as follows:

[0151]

[0152] The release process is as follows:

[0153] ;

[0154] In the formula, Based on contact time Nonlinear angle compensation curve of large braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the large braking clearance estimation model. The gain parameter is determined by the reference angle in the large braking clearance estimation model;

[0155] For the clamping process Release process ;

[0156] Among them, the brake clamping force estimation model established under the small brake gap state is the small brake gap estimation model. The method for obtaining the nonlinear angle compensation curve in the small brake gap estimation model is as follows:

[0157] ① Establish separately with For the estimation model of the interval, correspond As the reference angle in the small braking gap estimation model;

[0158] ② If the braking clearance continues to decrease, the motor (angle) should be adjusted. In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle.

[0159] ③ Fit the contact time The x-axis represents the motor angle. The nonlinear compensation curve of the small braking clearance estimation model is shown on the ordinate. The influence parameters caused by excessively small braking clearance are unified into the angle compensation.

[0160] By using angle compensation to adapt to changes in the braking clearance, accurate estimation of clamping force can be achieved.

[0161] The clamping process is as follows:

[0162] ;

[0163] The release process is as follows:

[0164] ;

[0165] In the formula, Based on contact time Nonlinear angle compensation curve of small braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the model with a small braking clearance estimation. The gain parameter is determined by the reference angle in the model with a small braking gap estimation.

[0166] For the clamping process For the release process ;

[0167] Therefore, when the braking gap is within the specified range, an extended state observer with a fused neural network is used, while when the braking gap is too large or too small, an extended state observer with angle compensation is used to achieve accurate estimation of the clamping force under different braking gap conditions.

[0168] S4: Update the neural network training set data in the conventional gap estimation model according to the vehicle status.

[0169] (1) When the vehicle brakes, the wavelet neural network updates the weights in the network to obtain the gain parameters by using the gradient descent method; only when the vehicle is parked, the neural network is updated as a whole by updating the training set data to ensure that the network adapts to the current vehicle state.

[0170] (2) The training set data acquisition method of the wavelet neural network is as follows: the clamping force qualification is detected for each step length during the braking process. The qualification detection standard is: under level road conditions, when the brake pedal is pressed, the vehicle speed at that moment is recorded. When the brake pedal is released, braking is considered complete, and the vehicle speed at that moment is recorded. ; each interval Record vehicle speed at all times The actual clamping force is calculated using the following formula: until the braking ends.

[0171] ,

[0172] In the formula, , The data collection interval is denoted as: For longitudinal resistance, For vehicle quality, The number of brake wheels, To generate braking torque for a single wheel, The effective radius of the tire. The radius of the brake disc, The coefficient of friction between the brake pads and the brake disc. For real braking force, , The rolling resistance and wind resistance experienced by the vehicle at the current speed can be measured experimentally.

[0173] like In the formula: To specify a standard, the input parameters and output gain parameters corresponding to the step size braking are retained as the dataset for wavelet neural network updates;

[0174] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for identifying brake clearance and estimating adaptive clamping force in an EMB system, characterized in that: Includes the following steps: S1. Based on the EMB contact time, the EMB braking gap is divided into normal braking gap state, large braking gap state, and small braking gap state; where: the EMB contact time is determined according to the change of motor angular velocity; S2. Based on the motor torque balance equation of EMB during braking, a dynamic model of braking clamping force of EMB during braking is established using the electromagnetic torque, motor angle and motor angular velocity of the motor. S3. Based on the dynamic model of the braking clamping force during the braking process of EMB, construct an extended state observer with fused neural network and an extended state observer with angle compensation respectively. When the EMB brake clearance is in the normal brake clearance state, the extended state observer with fused neural network is used to estimate the brake clamping force. When the EMB brake clearance is in a large brake clearance state, the extended state observer A with angle compensation is used to estimate the brake clamping force. When the EMB brake clearance is in a small brake clearance state, the extended state observer B with angle compensation is used to estimate the brake clamping force. Wherein: the angle-compensated extended state observer A includes an extended state observer A for clamping process angle compensation and an extended state observer A for release process angle compensation, wherein: The expression for the extended state observer A for clamping process angle compensation is: ; The expression for the extended state observer A for angle compensation during the release process is: ； In the formula, Based on contact time Nonlinear angle compensation curve of large braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the large braking clearance estimation model. The gain parameter is determined by the reference angle in the large braking clearance estimation model; The process of establishing the angle-compensated extended state observer A includes the following steps: Step 1: Establish separate systems based on... For the estimation model of wear interval, with corresponding As the reference angle in the large braking clearance estimation model; Step 2: If the braking gap continues to increase, the motor angle should be adjusted. In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle. Step 3: Fit the contact time The x-axis represents the motor angle. Nonlinear compensation curve of the large braking clearance estimation model with ordinate as the vertical axis The influence parameters caused by different wear amounts are unified into the angle compensation to obtain the extended state observer A of the angle compensation. The angle-compensated extended state observer B includes an extended state observer B for clamping process angle compensation and an extended state observer B for release process angle compensation, wherein: The expression for the extended state observer B for clamping process angle compensation is: ; The expression for the extended state observer B for angle compensation during the release process is: ; In the formula, Based on contact time Nonlinear angle compensation curve of small braking gap estimation model Corrected motor angle The Fal function is determined by the reference angle in the model with a small braking clearance estimation. The gain parameter is determined by the reference angle in the model with a small braking gap estimation.

2. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 1, characterized in that: The specific implementation process of step S1 includes the following steps: S201, Angular velocity sensor is used to collect motor angular velocity The rate of change of the motor's angular velocity is obtained through calculation; S202, Pre-calibrate the threshold for motor angular velocity change during EMB contact. When the rate of change of motor angular velocity obtained in step S101 is less than or equal to At that time, that is At that time, the EMB contact time was collected. ; S203. Based on the reasonable braking clearance range set by EMB, pre-calibrate the contact time. reasonable range ;like This is judged to be a normal braking clearance; if The problem is determined to be excessive braking clearance; if The problem was determined to be due to insufficient braking clearance.

3. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 1, characterized in that: The specific implementation process of step S2 includes the following steps: S301, Set the motor electromagnetic torque as follows: , In the formula, This represents the number of pole pairs of the motor. , , and The magnetic flux on the stator side and Coordinate system current; S302. The EMB braking process is divided into two processes: clamping and releasing. During the clamping process, based on the torque balance equation of the motor, we can derive: ； In the formula, The angular velocity of the motor. For time, The moment of inertia of the motor. For the electromagnetic torque of the motor, This refers to the dynamic offset of the inherent parameters of the motor. Let Coulomb friction coefficient be the coefficient of friction of the motor. This is the magnification factor; During the release process, we can conclude that: ； In the formula, and It was measured using the EMB clamping force test with the motor torque as the abscissa. S303. Based on the motor torque balance equations for the clamping and releasing processes, a dynamic model of the braking clamping force for the clamping and releasing processes is established, wherein: The dynamic model of the braking clamping force during the clamping process is as follows: ； In the formula, , This represents the sum of clamping torque and frictional torque. The derivative of the sum of clamping torque and frictional torque. express The derivative of For the motor angle, express The derivative of This is the braking clamping force; The dynamic model of the braking clamping force during the release process is as follows: ； In the formula, .

4. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 1, characterized in that: The extended state observer of the fusion neural network includes extended state observers for the clamping and release processes of the fusion neural network, wherein: The expression for the extended state observer of the fusion neural network in the clamping process is: ； The expression for the extended state observer of the fusion neural network during the release process is: ； in, for The estimated value, For the error term, i.e. , for The estimated value, for The estimated value, For time step, This refers to the gain parameter output by the neural network during the clamping process. This refers to the gain parameter of the neural network output during the release process. express function, The function is: 。 5. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 4, characterized in that: The process of obtaining the gain parameter: The number of nodes in the neural network input layer is The inputs to the neural network are respectively ; The inputs and outputs of the hidden layer are as follows: ； ； In the formula, Let be the weights representing the influence of the i-th input node of the neural network on the j-th hidden layer. For the neural network One input, , For the hidden layer The input and output of each node; , These are the input layer and hidden layer nodes, respectively; , These are the translation factor and the scaling factor, respectively. The output of the output layer is: ； ； In the formula, , The output layer is respectively The actual output and expected output of each node; For error; These are the weights of the neural network; This represents the number of nodes in the output layer. At the end of each step, if the EMB is in the clamping process, the gain parameters of the neural network output during the clamping process are obtained. If the EMB is in the release process, then the gain parameter of the neural network output during the clamping process is obtained. .

6. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 1, characterized in that: The process of establishing the angle-compensated extended state observer B includes the following steps: S601, respectively establish as For the estimation model of the interval, corresponding As the reference angle in the small braking gap estimation model; S602. If the braking gap continues to decrease, the motor angle... In addition, the Fal function in the ESO algorithm is modified; through experiments, the same estimation accuracy is achieved by fixing the Fal function and only modifying the motor angle. S603, Fit the contact time The x-axis represents the motor angle. The nonlinear compensation curve of the small braking clearance estimation model is shown on the ordinate. The influence parameters caused by excessively small braking clearance are unified into the angle compensation to obtain the extended state observer B of angle compensation.

7. The method for identifying brake clearance and estimating adaptive clamping force in an EMB system according to claim 1, characterized in that: The training set data update process of the neural network includes the following steps: S701. During each braking action, the neural network updates the weights within the network using the gradient descent method to obtain the gain parameters; only when the vehicle is parked is the neural network updated as a whole by updating the training set data. S702. The training set data acquisition method for the neural network is as follows: The compliance of the clamping force corresponding to each step length during the braking process is detected, and the compliance detection standard is: On level roads, record the vehicle speed when the brake pedal is depressed. ; When the brake pedal is released, braking is considered complete, and the vehicle speed at this moment is recorded. ; During braking, at each interval Record vehicle speed at all times Until braking ends; the actual clamping force is calculated using the formula: , , The data collection interval is where For longitudinal resistance, For vehicle quality, The number of brake wheels, To generate braking torque for a single wheel, The effective radius of the tire. The radius of the brake disc, The coefficient of friction between the brake pads and the brake disc. For real braking force, , The rolling resistance and wind resistance experienced by the vehicle at the current speed are measured experimentally; if In the formula, To specify a standard, the input parameters and output gain parameters corresponding to the step size braking are retained as the dataset for neural network updates.

Images