EMB clamping force prediction method considering multi-source uncertainty

Abstract

The invention discloses an EMB clamping force prediction method considering multi-source uncertainty, and belongs to the technical field of automobile industry, and the method comprises the steps: obtaining a historical mapping relation between clamping force and control input, building prior distribution, and employing normal distribution as the prior distribution to describe an expected range of the clamping force; the method comprises the following steps: acquiring observation data of clamping force historically collected by a sensor, introducing the observation data into a likelihood function, and setting a relationship between the observation data and real clamping force to be influenced by multi-source noise to generate an error model so as to describe probability distribution of the observation data under specific clamping force by maximizing the likelihood function; on the basis of a Bayesian inference framework, updating the trust degree of the clamping force predicted by the EMB through prior distribution and a likelihood function, and calculating posterior distribution by using known prior information and real-time observation data; and extracting 95% confidence interval and uncertainty index in the posterior distribution, and when the confidence interval or the system entropy value is greater than the uncertainty index, rejecting or optimizing the prediction result of the clamping force predicted by the EMB.

Description

Technical Field

[0001] The invention belongs to the technical field of automobile industry, and in particular relates to an EMB clamping force prediction method considering multi-source uncertainty. Background Art

[0002] The electronic mechanical brake (EMB) is a complex mechanical system. The clamping force prediction accuracy is limited in complex dynamic environments (temperature, humidity, nonlinear changes in friction coefficient, model errors, and observation noise). Under complex multi-source uncertainty conditions, the technical challenges that this method aims to address are high-precision EMB clamping force prediction, sensor noise quantification, accurate modeling of dynamic nonlinear systems, and giving the EMB adaptive prediction and risk perception capabilities under dynamic working conditions.

[0003] To address these technical challenges, existing technologies propose the following approaches: 1. Direct measurement using high-precision sensors, enabling real-time estimation of clamping force data based on pressure or displacement sensors; 2. Utilizing multi-sensor fusion algorithms, including redundant sensors and Kalman filtering, to improve measurement accuracy; and 3. Employing data-driven prediction methods, based on artificial intelligence (AI) algorithm models, to establish and predict nonlinear mapping relationships within the system. However, these approaches have limitations. Sensor-based methods lack iterative self-learning and adaptive capabilities, while AI-based methods rely heavily on large amounts of high-quality training data. Combining these approaches leads to exponentially increased complexity and cost. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide an EMB clamping force prediction method considering multi-source uncertainty, a clamping force prediction and uncertainty quantification technology based on the Bayesian method, which integrates multi-source sensor data and Bayesian theory, accurately predicts the clamping force, quantifies the credibility of the prediction, and provides adaptive optimization capabilities under dynamic working conditions.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for predicting EMB clamping force considering multiple sources of uncertainty includes:

[0007] Step 1: Obtain the historical mapping relationship between the clamping force and the control input to establish a prior distribution, and use the normal distribution as the prior distribution to describe the expected range of the clamping force;

[0008] Step 2: Obtain the historical clamping force observation data collected by the sensor and introduce it into the likelihood function. Set the relationship between the observation data and the actual clamping force to be affected by multiple sources of noise to generate an error model to describe the probability distribution of the observation data under a specific clamping force by maximizing the likelihood function.

[0009] Step 3: Based on the Bayesian inference framework, the confidence level of the EMB predicted clamping force is updated through the prior distribution and likelihood function, and the posterior distribution is calculated using the known prior information and real-time observation data;

[0010] Step 4: Extract the 95% confidence interval and uncertainty index from the posterior distribution. When the confidence interval or the system entropy value is greater than the uncertainty index, reject or optimize the prediction result of the EMB clamping force.

[0011] Furthermore, step 1 specifically includes:

[0012] Obtain historical sensor data of the mapping relationship between the clamping force measurement value and the control input in the EMB, and perform data denoising and normalization to obtain system historical data;

[0013] The mean and standard deviation of the clamping force are calculated through the system's historical data, the parameters of the normal distribution are estimated, and the control input is used as the conditional variable of the prior distribution to improve the composition of the prior distribution of the clamping force.

[0014] The rationality of the prior distribution is verified by obtaining real system data. When there is a deviation, a deviation model is constructed through incremental learning to obtain the prior distribution model of the clamping force:

[0015]

[0016] Where P(F) represents the prior probability distribution of the clamping force, μ0 represents the mean of the clamping force data, and σ0 2 Indicates the standard deviation of the clamping force data.

[0017] Furthermore, step 2 specifically includes:

[0018] Obtaining the observation data of the clamping force collected historically by the sensor and introducing it into the likelihood function; wherein the observation data includes the braking force, pressure and displacement;

[0019] The relationship between the assumed observed data and the true clamping force is affected by noise from multiple sources;

[0020] When the relationship between the observed data and the true clamping force is affected by multi-source noise, an error term ∈ is generated; where the error term ∈ has a mean of 0 and a variance of σ obs 2 The normal distribution of is used to reflect the multi-source uncertainty influence of measurement error and system noise;

[0021] An error model is generated based on the error term ∈ to describe the probability distribution of the observed data under a specific clamping force by maximizing the likelihood function:

[0022]

[0023] Where D is the observed data of the clamping force, F is the true clamping force, and ∈ is the error term that takes into account the influence of multi-source uncertainty.

[0024] Furthermore, the posterior distribution calculated in step 3 includes:

[0025]

[0026] Where P(F│D) represents the posterior distribution estimate of the clamping force, P(D) represents the marginal likelihood or evidence term, which serves as a normalization constant to ensure that the total probability of the posterior distribution is 1; P(F) represents the prior probability distribution of the clamping force, which describes the probability distribution of the clamping force when there are no observations; P(D│F) represents the likelihood function, which represents the probability of observing data D given the clamping force F.

[0027] Furthermore, the method for calculating the system entropy value in step 4 includes:

[0028] H=-∫P(F│D)log P(F│D)dF

[0029] Among them, H represents the current system entropy value of the system, P(F│D) represents the current posterior distribution estimate of the system, and logP(F│D) represents the logarithmic probability of the system measuring local uncertainty.

[0030] The beneficial effects of the present invention are:

[0031] The present invention proposes an EMB clamping force prediction method that considers multi-source uncertainty. Compared with the existing technology, this method comprehensively quantifies the uncertainty information of multiple sources based on the Bayesian method, converts the predicted value into a probability distribution, and intuitively reflects the prediction credibility. Under high multi-source uncertainty, the EMB prediction results can be rejected or optimized, thereby enhancing the adaptability and robustness of the system under dynamic working conditions; the sensor information identified / acquired by the EMB can be constructed as new prior information, so that the prediction proposed by this method has sustainable iteration and self-optimization capabilities.

[0032] Other advantages, objectives, and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or may be taught by those skilled in the art from the practice of the present invention. The purposes and other advantages of the present invention may be realized and obtained through the structures particularly pointed out in the written description and the accompanying drawings.

[0033] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0035] Figure 1 4 is a flow chart of a method for predicting EMB clamping force considering multi-source uncertainties in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0037] like Figure 1 As shown, the present invention proposes an EMB clamping force prediction method considering multi-source uncertainty, including:

[0038] Step 1: Obtain the historical mapping relationship between the clamping force and the control input to establish a prior distribution, and use the normal distribution as the prior distribution to describe the expected range of the clamping force;

[0039] Step 2: Obtain the historical clamping force observation data collected by the sensor and introduce it into the likelihood function. Set the relationship between the observation data and the actual clamping force to be affected by multiple sources of noise to generate an error model to describe the probability distribution of the observation data under a specific clamping force by maximizing the likelihood function.

[0040] Step 3: Based on the Bayesian inference framework, the confidence level of the EMB predicted clamping force is updated through the prior distribution and likelihood function, and the posterior distribution is calculated using the known prior information and real-time observation data;

[0041] Step 4: Extract the 95% confidence interval and uncertainty index from the posterior distribution. When the confidence interval or the system entropy value is greater than the uncertainty index, reject or optimize the prediction result of the EMB clamping force.

[0042] The working principle and beneficial effects of the above technical solution are as follows: To overcome the limitations of existing technologies, this method proposes a clamping force prediction and uncertainty quantification technology based on the Bayesian method. It integrates multi-source sensor data and Bayesian theory to accurately predict the clamping force, quantify the reliability of the prediction, and provide adaptive optimization capabilities under dynamic working conditions;

[0043] First, the system historical data (the mapping relationship between clamping force and control input) is used to establish a prior distribution, and the normal distribution is used to describe the expected range of clamping force as follows:

[0044]

[0045] Where P(F) represents the prior probability distribution of the clamping force, μ0 represents the mean of the clamping force data, and σ0 2 Indicates the standard deviation of the clamping force data;

[0046] By constructing a prior distribution model of clamping force, prior information is provided for subsequent Bayesian inference;

[0047] Then, the clamping force observation data (braking force, pressure, displacement) collected from the sensor is introduced as a likelihood function to describe the probability distribution of the observation data under a specific clamping force, as shown in the following formula:

[0048]

[0049] The above steps further quantify uncertainty, providing observation data and probability model basis for subsequent Bayesian inference;

[0050] Based on the Bayesian inference framework, the confidence level in the clamping force is updated using the prior distribution and likelihood function obtained from the above calculations, and the posterior distribution is inferred using the known prior information and observed data, as shown in the following formula:

[0051]

[0052] Where P(F│D) represents the posterior distribution estimate of the clamping force, P(D) represents the marginal likelihood or evidence term, which serves as a normalization constant to ensure that the total probability of the posterior distribution is 1; P(F) represents the prior probability distribution of the clamping force, which describes the probability distribution of the clamping force when there are no observations; P(D│F) represents the likelihood function, which represents the probability of observing data D given the clamping force F;

[0053] Finally, the uncertainty of the clamping force estimate is quantified by extracting confidence intervals from the posterior distribution. First, the 95% confidence interval of the posterior distribution is calculated. This interval represents a 95% probability that the estimated value falls within this range. The width of the confidence interval reflects the uncertainty of the prediction. A wider confidence interval means that there is greater uncertainty in the system's prediction results.

[0054] Secondly, the uncertainty of the EMB clamping force prediction system can be further quantified by calculating the entropy of the lag distribution. The larger the entropy, the more ambiguous the system's understanding of the prediction results and the greater the uncertainty of the prediction. Combining the above two results, it can be determined that if the confidence interval is too wide or the entropy value is too large, it indicates that the model has a high uncertainty in estimating the clamping force under the current working conditions, and the system may not be able to make reliable decisions. The uncertainty index set in step 4 is used to limit the specific threshold for a confidence interval that is too wide or an entropy value that is too large.

[0055] In situations where the system has multiple sources and high uncertainty, the clamping force prediction system can choose to reject the current prediction results, or adjust the control strategy based on the current uncertainty, take more conservative measures for optimization, or require more current sensor observation data to further reduce uncertainty. This method enables the system to dynamically adapt to changing working conditions and adopt reasonable response strategies in situations with large uncertainty to ensure the reliability and accuracy of the EMB clamping force prediction results.

[0056] Through the above technical solution, the posterior distribution of clamping force prediction is constructed by combining prior distribution and observation data. At the same time, the uncertainty of the shape and range of the posterior distribution is quantified, providing a confidence interval and credibility assessment of the EMB prediction value. When the system uncertainty is too high, an alarm can be issued and the EMB prediction result can be rejected or optimized. Compared with the existing technology, this method comprehensively quantifies the uncertainty information of multiple sources based on the Bayesian method, converts the prediction value into a probability distribution, and intuitively reflects the prediction credibility. Under high uncertainty from multiple sources, the EMB prediction result can be rejected or optimized, thereby enhancing the adaptability and robustness of the system under dynamic working conditions. The sensor information identified / acquired by the EMB can be constructed as new prior information, so that the prediction proposed by this method has the ability of sustainable iteration and self-optimization.

[0057] In one embodiment, step 1 specifically includes:

[0058] Obtain historical sensor data of the mapping relationship between the clamping force measurement value and the control input in the EMB, and perform data denoising and normalization to obtain system historical data;

[0059] The mean and standard deviation of the clamping force are calculated through the system's historical data, the parameters of the normal distribution are estimated, and the control input is used as the conditional variable of the prior distribution to improve the composition of the prior distribution of the clamping force.

[0060] The rationality of the prior distribution is verified by obtaining real system data. When there is a deviation, a deviation model is constructed through incremental learning to obtain the prior distribution model of the clamping force:

[0061]

[0062] Where P(F) represents the prior probability distribution of the clamping force, μ0 represents the mean of the clamping force data, and σ0 2 Indicates the standard deviation of the clamping force data;

[0063] The working principle and beneficial effects of the above technical solution are as follows: first, historical sensor data on the correspondence between the measured value of the clamping force and the control input in the EMB is collected, and data denoising and normalization are performed; second, based on the statistical characteristics of the historical observation data, a normal distribution is assumed as the prior distribution of the clamping force, that is, the value of the clamping force fluctuates around a certain mean, and the degree of fluctuation is measured by the standard deviation. The mean and standard deviation of the clamping force are calculated through historical data, and the parameters of the normal distribution are estimated; at the same time, the control input is used as the conditional variable of the prior distribution to further improve the composition form of the prior distribution of the clamping force; finally, the rationality of the prior distribution is verified by comparison with the real system data. If there is a deviation, a deviation model can be constructed through methods such as incremental learning. The prior distribution model of the clamping force is constructed by the above method, which can provide prior information for subsequent Bayesian inference.

[0064] In one embodiment, step 2 specifically includes:

[0065] Obtaining the observation data of the clamping force collected historically by the sensor and introducing it into the likelihood function; wherein the observation data includes the braking force, pressure and displacement;

[0066] The relationship between the assumed observed data and the true clamping force is affected by noise from multiple sources;

[0067] When the relationship between the observed data and the true clamping force is affected by multi-source noise, an error term ∈ is generated; where the error term ∈ has a mean of 0 and a variance of σ obs 2 The normal distribution of is used to reflect the multi-source uncertainty influence of measurement error and system noise;

[0068] An error model is generated based on the error term ∈ to describe the probability distribution of the observed data under a specific clamping force by maximizing the likelihood function:

[0069]

[0070] Where D is the observed data of the clamping force, F is the true clamping force, and ∈ is the error term considering the influence of multi-source uncertainty;

[0071] The working principle and beneficial effects of the above technical solution are as follows: based on the clamping force observation data obtained by the sensor, including but not limited to braking force, pressure and displacement data, it is introduced as a likelihood function. Assuming that the relationship between the observation data and the clamping force is affected by multi-source noise, the observed value of the clamping force can be represented by adding an error term, which is assumed to obey a mean of 0 and a variance of σ obs 2The normal distribution of the error term reflects the influence of multiple sources of uncertainty such as measurement error and system noise. The observed value D of the clamping force can be expressed by the sum of the actual clamping force F and the error term. Based on this error model, the parameters in the system can be estimated by maximizing the likelihood function, and the uncertainty can be further quantified, providing the observation data and probability model basis for subsequent Bayesian inference.

[0072] In one embodiment, calculating the posterior distribution in step 3 includes:

[0073]

[0074] Where P(F│D) represents the posterior distribution estimate of the clamping force, P(D) represents the marginal likelihood or evidence term, which serves as a normalization constant to ensure that the total probability of the posterior distribution is 1; P(F) represents the prior probability distribution of the clamping force, which describes the probability distribution of the clamping force when there are no observations; P(D│F) represents the likelihood function, which represents the probability of observing data D given the clamping force F;

[0075] The working principle and beneficial effects of the above technical solution are as follows: the posterior distribution in step 3 is the product of the assumed prior distribution and the constructed likelihood function, which is then normalized by a standardization constant. By maximizing the posterior distribution and then by the mean, a point estimate of the clamping force can be obtained, and finally the uncertainty of the estimate is obtained by the variance. More specifically, the method reflects the estimated range of the clamping force. The larger the variance, the more uncertain the estimate of the clamping force, and the smaller the variance, the more accurate the estimate. At the same time, the traditional predicted clamping force value estimate is converted into a predicted clamping force distribution estimate, which increases the reliability of the EMB system. The mean estimate and the posterior distribution of uncertainty quantification described above provide a credibility estimate for further decision-making and optimization, and give the EMB the ability of dynamic adjustment and adaptive optimization in practical applications.

[0076] In one embodiment, the method for calculating the system entropy value in step 4 includes:

[0077] H=-∫P(F│D)log P(F│D)dF

[0078] Among them, H represents the current system entropy value of the system, that is, the current conditional differential entropy of the system, P(F│D) represents the current posterior distribution estimate of the system, that is, the current posterior probability density function of the system, and logP(F│D) represents the logarithmic probability of the system to measure local uncertainty;

[0079] The working principle and beneficial effects of the above technical solution are as follows: the system entropy value in step 4 is used to further quantify the uncertainty of the EMB clamping force prediction system. By adding an estimate of the system entropy value, the system can be more dynamically adapted to changing working conditions.

[0080] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A method for predicting EMB clamping force considering multiple sources of uncertainty, characterized in that: include: Step 1: Obtain the historical mapping relationship between the clamping force and the control input to establish a prior distribution, and use the normal distribution as the prior distribution to describe the expected range of the clamping force; Step 2: Obtain the historical clamping force observation data collected by the sensor and introduce it into the likelihood function. Set the relationship between the observation data and the actual clamping force to be affected by multiple sources of noise to generate an error model to describe the probability distribution of the observation data under a specific clamping force by maximizing the likelihood function. Step 3: Based on the Bayesian inference framework, the confidence level of the EMB predicted clamping force is updated through the prior distribution and likelihood function, and the posterior distribution is calculated using the known prior information and real-time observation data; Step 4: Extract the 95% confidence interval and uncertainty index from the posterior distribution. When the confidence interval or the system entropy value is greater than the uncertainty index, reject or optimize the prediction result of the EMB clamping force.

2. The EMB clamping force prediction method considering multiple sources of uncertainty according to claim 1 is characterized in that: Step 1 specifically includes: Obtain historical sensor data of the mapping relationship between the clamping force measurement value and the control input in the EMB, and perform data denoising and normalization to obtain system historical data; The mean and standard deviation of the clamping force are calculated through the system's historical data, the parameters of the normal distribution are estimated, and the control input is used as the conditional variable of the prior distribution to improve the composition of the prior distribution of the clamping force. The rationality of the prior distribution is verified by obtaining real system data. When there is a deviation, a deviation model is constructed through incremental learning to obtain the prior distribution model of the clamping force: Where P(F) represents the prior probability distribution of the clamping force, μ0 represents the mean of the clamping force data, and σ0 2 Indicates the standard deviation of the clamping force data.

3. The EMB clamping force prediction method considering multiple sources of uncertainty according to claim 1 is characterized in that: Step 2 specifically includes: Obtaining the observation data of the clamping force collected historically by the sensor and introducing it into the likelihood function; wherein the observation data includes the braking force, pressure and displacement; The relationship between the assumed observed data and the true clamping force is affected by noise from multiple sources; When the relationship between the observed data and the true clamping force is affected by multi-source noise, an error term ∈ is generated; where the error term ∈ has a mean of 0 and a variance of σ obs 2 The normal distribution of is used to reflect the multi-source uncertainty influence of measurement error and system noise; An error model is generated based on the error term ∈ to describe the probability distribution of the observed data under a specific clamping force by maximizing the likelihood function: Where D is the observed data of the clamping force, F is the true clamping force, and ∈ is the error term that takes into account the influence of multi-source uncertainty.

4. The EMB clamping force prediction method considering multiple sources of uncertainty according to claim 1 is characterized in that: The posterior distribution calculated in step 3 includes: Where P(F│D) represents the posterior distribution estimate of the clamping force, P(D) represents the marginal likelihood or evidence term, which serves as a normalization constant to ensure that the total probability of the posterior distribution is 1; P(F) represents the prior probability distribution of the clamping force, which describes the probability distribution of the clamping force when there are no observations; P(D│F) represents the likelihood function, which represents the probability of observing data D given the clamping force F.

5. The EMB clamping force prediction method considering multiple sources of uncertainty according to claim 1 is characterized in that: The calculation method of the system entropy value in step 4 includes: H=-∫P(F│D)logP(F│D)dF Among them, H represents the current system entropy value of the system, P(F│D) represents the current posterior distribution estimate of the system, and logP(F│D) represents the logarithmic probability of the system measuring local uncertainty.

Images