Friction plate fatigue life prediction method considering multi-factor influence

Abstract

The invention discloses a friction plate fatigue life prediction method considering multi-factor influence, and the method comprises the steps: obtaining the performance parameters of a friction plate under the actual working condition of multi-factor influence, and introducing an average friction coefficient to calculate key parameters; constructing a multi-dimensional feature input vector, and establishing a nonlinear mapping relationship between the fatigue life of the friction plate and each influence factor based on a deep neural network model; the performance of the friction plate is monitored in real time, and a corresponding fatigue life prediction value is obtained; a fuzzy comprehensive evaluation method is adopted to analyze the change trend of the optimal characterization parameters, a fuzzy evaluation model for state evaluation is established, and a predicted value of the abrasion degree of the friction plate is obtained; a confidence coefficient analysis module is adopted for the friction plate wear degree prediction value, confidence coefficient evaluation is conducted on a prediction result of the friction plate wear degree prediction value, a fatigue life prediction interval containing an error range is generated, the prediction precision and the model generalization ability are remarkably improved, and the credibility and interpretability of the prediction result are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of friction plates, and in particular to a method for predicting the fatigue life of a friction plate taking into account the influence of multiple factors. Background Art

[0002] With the development of modern industry, the reliability and maintenance cost of mechanical equipment have become the focus of attention. Friction plates are important components in many key mechanical systems, such as brake systems and clutches. Their fatigue life directly affects the safety and service life of the equipment. In order to accurately predict the fatigue life of friction plates, various methods and technologies have been adopted. However, existing technologies still have shortcomings in dealing with fatigue life prediction under the influence of multiple factors: (1) Single factor analysis ignores complex interactions. Most current friction plate fatigue life prediction methods rely on modeling the impact of a single factor, such as temperature, pressure, or sliding speed, on performance. However, in actual operating conditions, the wear and fatigue life of friction plates are the result of the combined effects of multiple factors, and there may be complex nonlinear relationships between these factors. (2) There is a lack of effective methods for multi-factor comprehensive modeling. When dealing with high-dimensional data and complex nonlinear relationships, linear relationship modeling is used, which cannot cope with the nonlinear coupling between multiple factors and is prone to large errors. Existing multi-factor comprehensive modeling methods usually require a large amount of experimental data for calibration, and obtaining sufficient high-quality data in practical applications is often challenging. Summary of the Invention

[0003] The purpose of the present invention is to provide a friction plate fatigue life prediction method that takes into account the influence of multiple factors, so as to solve the technical problems in the prior art of ignoring complex interactions in single factor analysis and lacking an effective method for comprehensive modeling of multiple factors.

[0004] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: The present invention provides a method for predicting the fatigue life of a friction plate taking into account the influence of multiple factors, comprising the following steps: Obtaining performance parameters of the friction plate under actual working conditions that are affected by multiple factors, and introducing an average friction coefficient into the performance parameters as a key parameter for state monitoring of the friction plate; Normalizing the key parameters to construct a multidimensional feature input vector, training the deep neural network based on a deep neural network model and combining historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors; The performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current operating parameters of the friction plate to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value; Simulating the wear of the friction plate according to the fatigue life prediction value, establishing a corresponding relationship between the optimal characterization parameters of the friction plate and the state evaluation, analyzing the change trend of the optimal characterization parameters using a fuzzy comprehensive evaluation method, establishing a fuzzy evaluation model for state evaluation, and obtaining a predicted value of the wear degree of the friction plate; A confidence analysis module is used for the predicted value of the degree of wear of the friction plate, and a confidence assessment is performed on the prediction result of the predicted value of the degree of wear of the friction plate to generate a fatigue life prediction interval including an error range.

[0005] As a preferred solution of the present invention, the performance parameters of the friction plate under actual working conditions that are affected by multiple factors are obtained, and the average friction coefficient is introduced into the performance parameters as a key parameter for state monitoring of the friction plate, including: Obtaining performance parameters of the friction plate under actual working conditions influenced by multiple factors, wherein the influencing factors include at least: friction material properties, operating temperature, contact pressure, sliding speed, lubrication state, load cycle frequency, and friction force variation curve during the friction process; During the operation of the friction plate, instantaneous friction force and normal pressure data at multiple time points are collected in real time to calculate the average friction coefficient of the friction, which is expressed as: ; in, represents the friction force at the i-th sampling point, represents the positive pressure at the i-th sampling point, Indicates the total number of sampling points; The average friction coefficient is used as a characteristic parameter reflecting the wear state and friction stability of the friction plate surface, and as a key parameter for state monitoring of the friction plate.

[0006] As a preferred solution of the present invention, the key parameters are normalized to construct a multidimensional feature input vector. Based on a deep neural network model, the deep neural network is trained in combination with historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors, including: The key parameters are normalized to the minimum and maximum values, and the normalized key parameters are sorted according to the degree of wear of the friction plate. The degree of wear of the friction plate is determined by the friction coefficient on the corresponding friction plate. The expression is: ; in, represents the number of friction pairs of the friction plate, q represents the pressure ratio between the friction plates, 、 Indicates the inner and outer diameters of the friction plate, Indicates the friction torque of the friction plate; Normalize the key parameters after sorting to generate a feature input vector of uniform dimension , where n represents the number of feature dimensions; A fully connected feedforward neural network model is constructed, and the attention mechanism is integrated to train historical fatigue test data. The operating parameters under different working conditions are recorded, and the parameters of the fully connected feedforward neural network model are optimized based on the fatigue failure time and cycle number corresponding to the friction plate. The influencing factors of the allowable process of the friction plate under different industrial and mining conditions are input into the fully connected feedforward neural network model to establish a highly nonlinear, multivariable coupling relationship between the fatigue life of the friction plate and its influencing factors, and obtain the interaction effect between multiple factors.

[0007] As a preferred solution of the present invention, the performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current operating parameters of the friction plate to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value, including: The friction plate operating parameters under the current working conditions are collected in real time by sensors, and the operating parameters are input into the fully connected feedforward neural network model according to the nonlinear mapping relationship to obtain a multi-factor multi-dimensional feature input vector ,in represents the average friction coefficient, Indicates the operating temperature, Indicates the contact pressure, Indicates the sliding speed of the friction plate during actual operation. Indicates the thickness of the lubricating oil film on the friction plate surface. Indicates the friction plate load cycle frequency, represents the material density, Indicates the thermal conductivity of the material, etc. The multidimensional features are input into the vector The ReLU activation function is used to enhance the nonlinear expression capability when inputting the deep neural network model. After forward propagation calculation, the deep neural network model outputs the corresponding predicted value, which represents the estimated remaining fatigue life of the friction plate under current conditions.

[0008] As a preferred solution of the present invention, simulating the wear amount of the friction plate according to the fatigue life prediction value includes: The Archard wear model is used to simulate the wear process of the friction plate under different working conditions, and its expression is: ; in, represents the average friction coefficient, W represents the normal load of the friction plate, d represents the relevant sliding friction distance, and V represents the wear volume; According to the multi-dimensional feature input vector under the current working condition Calculate the wear amount of the friction plate in each cycle based on the fatigue life prediction value; The wear of the friction plate throughout its life cycle is dynamically simulated using the ANSYS simulation tool to generate a wear curve graph and obtain the wear variation trend over time or the number of cycles.

[0009] As a preferred solution of the present invention, a corresponding relationship between the optimal characterization parameters of the friction plate and the state evaluation is established based on the wear amount of the friction plate, including: According to the wear trend of the friction plate over time, n data points are randomly extracted. , , , establish a linear regression equation for the n data points to obtain the wear amount of the friction plate during the stable wear period and wear life The offline relationship between them is expressed as: ; in, represents the initial wear amount, Indicates the growth rate of wear; The degree of fit between the regression model and the actual data is determined by calculating the standard error of the n data points, and the warning baseline and abnormal baseline of the wear element changing over time are obtained according to the linear regression equation and the standard error; The wear amount is analyzed using random forest algorithm Conduct analysis and establish a mapping relationship between characterization parameters and wear status; Based on the characterization parameter values ​​obtained from real-time monitoring, sensitivity analysis is performed on each characterization parameter to identify the key parameters that have the greatest impact on the wear state. Based on the weighted fusion strategy, weights are assigned to different characterization parameters, and a comprehensive scoring system is constructed to quantify the overall wear state of the friction plate. Its expression is: ; in, Represent the weights of the corresponding characterization parameters, Indicates the comprehensive score.

[0010] As a preferred solution of the present invention, a fuzzy comprehensive evaluation method is used to analyze the weights affecting the optimal characterization parameters, including: Create a factor set containing each indicator for the comprehensive score , Represents the mth element that affects the evaluation object, and establishes an evaluation set for the comprehensive score , the evaluation set The middle element represents the evaluation result of its corresponding influencing factor; The positive evaluation index is used to calculate the weight of the corresponding sample under the i-th factor index, and the entropy value of the j-th index is obtained. The expression is: ; Where n represents the number of samples, Representation factor set The number of elements in It represents the weight of the jth indicator under the i-th factor indicator; The difference of each indicator is calculated according to the entropy value, the weights of multiple indicators in the comprehensive evaluation system are calculated, and the weights affecting the optimal characterization parameters are obtained. The expression is: .

[0011] As a preferred solution of the present invention, a fuzzy judgment model for state evaluation is established for the weight of the optimal characterization parameter to obtain a predicted value of the friction plate wear degree, including: For each characterization parameter, its fuzzy membership function under different wear states is defined, and a fuzzy evaluation matrix R is constructed, where each row represents a characterization parameter, each column represents a wear state level, and the matrix elements are It represents the membership value of the i-th characterization parameter under the j-th wear state level, and its expression is: ; Where m is the number of characterization parameters and n is the number of wear state levels; The fuzzy comprehensive evaluation formula is calculated according to the weight of the optimal characterization parameter to obtain a comprehensive evaluation result vector B, which is expressed as follows: ; in, , represents the comprehensive membership value of the friction plate at each wear state level; Comprehensive evaluation result vector Perform normalization processing to obtain the membership value and determine the current wear state level of the friction plate; The wear degree prediction value is obtained according to the wear degree range corresponding to the wear state level with the highest membership value.

[0012] As a preferred solution of the present invention, a confidence analysis module is used for the predicted value of the friction plate wear degree, and a confidence evaluation is performed on the predicted result of the predicted value of the friction plate wear degree, including: The wear degree prediction value is combined with each sample in the historical data set, the residual between the wear degree prediction value and the true value is calculated, and the distribution characteristics of the residual are statistically analyzed; The distribution characteristics of the residuals are calculated using the properties of the normal distribution to calculate the confidence interval and set the confidence level , calculate the upper and lower limits of the confidence interval, the expression is: ; in, represents the lower limit of the confidence interval, represents the upper limit of the confidence interval, represents the predicted value of wear degree, Indicates the corresponding confidence level under normal distribution The critical value of Represents the standard deviation of the residuals.

[0013] As a preferred solution of the present invention, combining the wear degree prediction value and its confidence interval to generate a fatigue life prediction interval containing an error range includes: Calculating the remaining useful life of the wear degree prediction value at the current wear rate using the Archard wear model; Converting the confidence interval of the wear degree prediction value into the upper and lower limits of the fatigue life prediction interval according to the wear change trend over time; Using a state assessment algorithm to perform online analysis and monitoring of the confidence level, and to track the state of the friction plate in real time; The current wear level prediction value and its confidence interval are displayed in real time. When it is detected that the friction plate enters a state of moderate wear or more serious wear, the system automatically triggers an alarm.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention introduces the average friction coefficient as a key state monitoring parameter, combines multidimensional feature normalization processing with deep neural network modeling, and establishes a nonlinear mapping relationship between the fatigue life of the friction plate and its operating conditions, significantly improving the prediction accuracy and model generalization ability.

[0015] The wear process of the friction plate is dynamically simulated based on the Archard wear model, and a state assessment model is constructed using the fuzzy comprehensive evaluation method. The entropy weight method is combined to scientifically weight the optimal characterization parameters, effectively identify the current wear stage of the friction plate, and output a quantitative wear degree prediction value. The confidence analysis module is introduced to perform statistical analysis on historical residuals and calculate the confidence interval to generate a fatigue life prediction interval containing the error range, thereby enhancing the credibility and interpretability of the prediction results. At the same time, the state assessment algorithm is combined to realize online confidence monitoring and real-time early warning functions. When the friction plate is detected to enter a moderate wear or more serious state, the system automatically triggers an alarm to assist operation and maintenance personnel in taking timely intervention measures to avoid sudden equipment failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0017] Figure 1 A flow chart of a friction plate fatigue life prediction method considering the influence of multiple factors provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, the present invention provides a method for predicting the fatigue life of a friction plate taking into account the influence of multiple factors, comprising the following steps: Obtaining performance parameters of the friction plate under actual working conditions that are affected by multiple factors, and introducing an average friction coefficient into the performance parameters as a key parameter for state monitoring of the friction plate; In this embodiment, by collecting multiple performance parameters under actual working conditions and introducing the average friction coefficient, a core indicator that can reflect friction stability and wear status, the model's ability to perceive the real-time status of the friction plate is improved, and the physical significance and engineering practicality of the data input are enhanced.

[0020] Normalizing the key parameters to construct a multidimensional feature input vector, training the deep neural network based on a deep neural network model and combining historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors; In this embodiment, a normalization processing method is used to unify input parameters of different dimensions into the same scale range, thereby avoiding model training deviations caused by large differences in feature values; the constructed multidimensional feature input vector provides a structured basis for the subsequent establishment of nonlinear mapping relationships.

[0021] The performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current operating parameters of the friction plate to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value; In this example, a deep neural network model, trained with historical fatigue test data, successfully established a highly nonlinear mapping relationship between friction plate fatigue life and various influencing factors. This model not only captures the influence of a single factor but also effectively identifies the interactions between multiple factors, significantly improving prediction accuracy and generalization capabilities.

[0022] Simulating the wear of the friction plate according to the fatigue life prediction value, establishing a corresponding relationship between the optimal characterization parameters of the friction plate and the state evaluation, analyzing the change trend of the optimal characterization parameters using a fuzzy comprehensive evaluation method, establishing a fuzzy evaluation model for state evaluation, and obtaining a predicted value of the wear degree of the friction plate; In this embodiment, the friction plate wear amount is simulated and analyzed based on the fatigue life prediction results, and the representative optimal characterization parameters are extracted. The fuzzy comprehensive evaluation method is used to model the changing trends of these parameters, and a fuzzy evaluation model for state assessment is constructed to obtain a quantitative prediction value of the friction plate wear degree, making the state assessment more scientific and operational.

[0023] A confidence analysis module is used for the predicted value of the degree of wear of the friction plate, and a confidence assessment is performed on the prediction result of the predicted value of the degree of wear of the friction plate to generate a fatigue life prediction interval including an error range.

[0024] In this embodiment, by introducing a confidence analysis module, uncertainty assessment is performed on the wear degree prediction results, and a fatigue life prediction interval including an error range is generated, which improves the credibility and engineering applicability of the prediction results and helps to realize intelligent operation and maintenance and preventive replacement strategies for friction plates.

[0025] Obtaining performance parameters of the friction plate under actual working conditions that are affected by multiple factors, and introducing the average friction coefficient into the performance parameters as a key parameter for state monitoring of the friction plate, including: Obtaining performance parameters of the friction plate under actual working conditions influenced by multiple factors, wherein the influencing factors include at least: friction material properties, operating temperature, contact pressure, sliding speed, lubrication state, load cycle frequency, and friction force variation curve during the friction process; During the operation of the friction plate, instantaneous friction force and normal pressure data at multiple time points are collected in real time to calculate the average friction coefficient of the friction, which is expressed as: ; in, represents the friction force at the i-th sampling point, represents the positive pressure at the i-th sampling point, Indicates the total number of sampling points; In this embodiment, the average friction coefficient can effectively reflect the degree of wear of the surface material of the friction plate during long-term operation. As the friction plate gradually ages or wears, its friction coefficient will change significantly. Traditional friction plate monitoring methods usually rely on a single physical parameter, such as temperature and pressure, and it is difficult to accurately capture the dynamic changes of the friction process. The average friction coefficient combines the changing trends of friction force and positive pressure, and can more comprehensively reflect the stability of the friction system.

[0026] The average friction coefficient is used as a characteristic parameter reflecting the wear state and friction stability of the friction plate surface, and as a key parameter for state monitoring of the friction plate.

[0027] In this embodiment, the average friction coefficient can be directly used to construct the input vector of the deep neural network model, which helps to improve the model training efficiency and prediction accuracy.

[0028] In this embodiment, by combining the average friction coefficient with other key parameters, the nonlinear mapping relationship between the fatigue life of the friction plate and its operating status can be more accurately modeled, thereby achieving more scientific and reasonable life prediction and maintenance decisions.

[0029] The key parameters are normalized to construct a multidimensional feature input vector. Based on a deep neural network model, the deep neural network is trained in combination with historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors, including: The key parameters are normalized to the minimum and maximum values, and the normalized key parameters are sorted according to the degree of wear of the friction plate. The degree of wear of the friction plate is determined by the friction coefficient on the corresponding friction plate. The expression is: ; in, represents the number of friction pairs of the friction plate, q represents the pressure ratio between the friction plates, 、 Indicates the inner and outer diameters of the friction plate, Indicates the friction torque of the friction plate; Normalize the key parameters after sorting to generate a feature input vector of uniform dimension , where n represents the number of feature dimensions; In this embodiment, the introduction of normalization processing and attention mechanism enables the model to capture key features more effectively, avoiding prediction deviations caused by large differences in the scales of certain parameters, thereby significantly improving prediction accuracy and stability.

[0030] In this embodiment, the feature input vector is constructed in a way that takes into account both physical meaning and data-driven characteristics, so that the model is not only applicable to a single material or working condition, but can also be extended to multiple types of friction plates and complex operating environments.

[0031] A fully connected feedforward neural network model is constructed, and the attention mechanism is integrated to train historical fatigue test data. The operating parameters under different working conditions are recorded, and the parameters of the fully connected feedforward neural network model are optimized based on the fatigue failure time and cycle number corresponding to the friction plate. In this embodiment, a large amount of friction plate fatigue test data from laboratory tests or actual service environments is collected, including: operating parameter records under different working conditions and the corresponding fatigue failure time or number of cycles. The data is divided into a training set, a validation set, and a test set. The mean square error is used as the loss function, and the model parameters are continuously optimized through the back propagation algorithm until the validation set error is stable.

[0032] The influencing factors of the allowable process of the friction plate under different industrial and mining conditions are input into the fully connected feedforward neural network model to establish a highly nonlinear, multivariable coupling relationship between the fatigue life of the friction plate and its influencing factors, and obtain the interaction effect between multiple factors.

[0033] In this embodiment, after sufficient training, the deep neural network can automatically learn and model the complex nonlinear relationship between the fatigue life of the friction plate and its influencing factors, including: the nonlinear dependence of material properties on fatigue life, the attenuation effect of temperature increase on friction stability and life, the impact of changes in lubrication state on friction coefficient fluctuations, and the dynamic correlation between the average friction coefficient and wear rate.

[0034] The performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current friction plate operating parameters to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value, including: The friction plate operating parameters under the current working conditions are collected in real time by sensors, and the operating parameters are input into the fully connected feedforward neural network model according to the nonlinear mapping relationship to obtain a multi-factor multi-dimensional feature input vector ,in represents the average friction coefficient, Indicates the operating temperature, Indicates the contact pressure, Indicates the sliding speed of the friction plate during actual operation. Indicates the thickness of the lubricating oil film on the friction plate surface. Indicates the friction plate load cycle frequency, represents the material density, Indicates the thermal conductivity of the material, etc. In this embodiment, the multi-dimensional feature input vector integrates multiple key parameters such as the average friction coefficient, operating temperature, and contact pressure, so that the model can identify the coupling relationship between different factors and improve the robustness and adaptability of the prediction.

[0035] The multidimensional features are input into the vector The ReLU activation function is used to enhance the nonlinear expression capability when inputting the deep neural network model. After forward propagation calculation, the deep neural network model outputs the corresponding predicted value, which represents the estimated remaining fatigue life of the friction plate under current conditions.

[0036] In this embodiment, the trained deep neural network model can accurately capture the complex nonlinear relationship between the fatigue life of the friction plate and its influencing factors, thereby improving the prediction accuracy. The system can complete a complete fatigue life prediction in a short time.

[0037] The wear amount of the friction plate is simulated according to the fatigue life prediction value, including: The Archard wear model is used to simulate the wear process of the friction plate under different working conditions, and its expression is: ; in, represents the average friction coefficient, represents the normal load of the friction plate, d represents the related sliding friction distance, represents the wear volume; According to the multi-dimensional feature input vector under the current working condition Calculate the wear amount of the friction plate in each cycle based on the fatigue life prediction value; The wear of the friction plate throughout its life cycle is dynamically simulated using the ANSYS simulation tool to generate a wear curve graph and obtain the wear variation trend over time or the number of cycles.

[0038] In this embodiment, the Archard wear model is introduced and combined with the ANSYS simulation tool to model and simulate the friction plate wear process. This can accurately simulate the wear evolution process of the friction plate under different working conditions, and display the wear trend through a graphical interface, which is convenient for engineers to understand and apply. On the basis of data-driven prediction, the physical wear model is integrated, so that fatigue life prediction not only relies on statistical learning, but also has a clear mechanical basis, which enhances the credibility and interpretability of the model.

[0039] In this embodiment, the wear curve can be used to estimate when the friction plate will reach the maximum allowable wear, so that a replacement plan can be formulated in advance to avoid equipment downtime or safety accidents caused by sudden failure.

[0040] According to the wear amount of the friction plate, a corresponding relationship between the optimal characterization parameter of the friction plate and the state evaluation is established, including: According to the wear trend of the friction plate over time, n data points are randomly extracted. , , , establish a linear regression equation for the n data points to obtain the wear amount of the friction plate during the stable wear period and wear life The offline relationship between them is expressed as: ; in, represents the initial wear amount, Indicates the growth rate of wear; The degree of fit between the regression model and the actual data is determined by calculating the standard error of the n data points, and the warning baseline and abnormal baseline of the wear element changing over time are obtained according to the linear regression equation and the standard error; In this embodiment, the standard error of the regression model is calculated to measure the degree of fit of the linear regression model to the actual wear data. According to the size of the standard error, it can be determined whether the current wear process conforms to the expected stable wear law.

[0041] The wear amount is analyzed using random forest algorithm Conduct analysis and establish a mapping relationship between characterization parameters and wear status; Based on the characterization parameter values ​​obtained from real-time monitoring, sensitivity analysis is performed on each characterization parameter to identify the key parameters that have the greatest impact on the wear state. Based on the weighted fusion strategy, weights are assigned to different characterization parameters, and a comprehensive scoring system is constructed to quantify the overall wear state of the friction plate. Its expression is: ; in, Represent the weights of the corresponding characterization parameters, Indicates the comprehensive rating.

[0042] In this embodiment, through linear regression modeling and standard error analysis, the wear stage and development trend of the friction plate can be accurately identified. The warning baseline and abnormal baseline generated by the regression model can be used to provide timely warnings before the wear amount reaches the failure threshold, effectively avoiding the occurrence of sudden failures.

[0043] In this embodiment, the random forest algorithm can capture the complex dependency between the friction plate wear state and multiple operating parameters, improve the accuracy and robustness of state recognition, and integrate multiple characterization parameters into a highly interpretable numerical indicator, making it easier for engineers to quickly determine the current state level of the friction plate.

[0044] The fuzzy comprehensive evaluation method is used to analyze the weights affecting the optimal characterization parameters, including: Create a factor set containing each indicator for the comprehensive score , Represents the mth element that affects the evaluation object, and establishes an evaluation set for the comprehensive score , the evaluation set The middle element represents the evaluation result of its corresponding influencing factor; The positive evaluation index is used to calculate the weight of the corresponding sample under the i-th factor index, and the entropy value of the j-th index is obtained. The expression is: ; Where n represents the number of samples, Representation factor set The number of elements in It represents the weight of the jth indicator under the i-th factor indicator; In this embodiment, the information content and difference of each indicator are automatically calculated based on the actual data distribution to obtain the entropy value, which avoids the deviation caused by subjective experience weighting and makes the weight distribution more objective and reasonable.

[0045] The difference of each indicator is calculated according to the entropy value, the weights of multiple indicators in the comprehensive evaluation system are calculated, and the weights affecting the optimal characterization parameters are obtained. The expression is: .

[0046] In this embodiment, by performing system modeling and weight analysis on multiple influencing factors, the key parameters that have the most significant impact on the wear state of the friction plate can be effectively identified, which helps to improve the accuracy and stability of the evaluation results.

[0047] A fuzzy judgment model for state evaluation is established for the weights of the optimal characterization parameters to obtain a predicted value of the degree of friction plate wear, including: For each characterization parameter, its fuzzy membership function under different wear states is defined, and a fuzzy evaluation matrix R is constructed, where each row represents a characterization parameter, each column represents a wear state level, and the matrix elements are It represents the membership value of the i-th characterization parameter under the j-th wear state level, and its expression is: ; Where m is the number of characterization parameters and n is the number of wear state levels; In this embodiment, by defining the fuzzy membership function of each characterization parameter under different wear states and constructing a fuzzy evaluation matrix, the originally fuzzy and uncertain wear state judgment process is transformed into a computable and quantifiable mathematical model, thereby improving the accuracy and consistency of the evaluation results.

[0048] The fuzzy comprehensive evaluation formula is calculated according to the weight of the optimal characterization parameter to obtain a comprehensive evaluation result vector B, which is expressed as follows: ; in, , represents the comprehensive membership value of the friction plate at each wear state level; Comprehensive evaluation result vector Perform normalization processing to obtain the membership value and determine the current wear state level of the friction plate; In this embodiment, the fuzzy comprehensive evaluation method can effectively handle the uncertainty and nonlinear problems existing in the operation of the friction plate, is applicable to various working conditions and material types, and has good generalization ability and engineering practicality.

[0049] The wear degree prediction value is obtained according to the wear degree range corresponding to the wear state level with the highest membership value.

[0050] In this embodiment, by collecting a large amount of actual wear data of the friction plate under different working conditions, the membership value of each characterization parameter under different wear states is calculated, and a fuzzy evaluation matrix R is constructed. The optimal characterization parameter weights obtained based on methods such as the entropy weight method are introduced into the fuzzy comprehensive evaluation process, so that the model can pay more attention to the key factors that have a greater impact on the wear state during the evaluation process. Through the normalization processing of the fuzzy evaluation result vector, the current wear state level of the friction plate can be clearly identified.

[0051] In this embodiment, a predicted value of the wear degree of the friction plate is outputted based on the wear degree range corresponding to the wear state level with the highest membership, which can be used to estimate the remaining service life of the friction plate.

[0052] A confidence analysis module is used for the predicted value of the friction plate wear degree to perform a confidence assessment on the predicted result of the predicted value of the friction plate wear degree, including: The wear degree prediction value is combined with each sample in the historical data set, the residual between the wear degree prediction value and the true value is calculated, and the distribution characteristics of the residual are statistically analyzed; In this embodiment, by performing statistical analysis on the residuals between the predicted value and the true value of each sample in the historical data set, the distribution characteristics of the residuals, such as the mean and standard deviation, are obtained, so that the prediction error is no longer a "black box" but an uncertainty range that can be quantified, thereby enhancing the credibility and interpretability of the model output.

[0053] The distribution characteristics of the residuals are calculated using the properties of the normal distribution to calculate the confidence interval and set the confidence level , calculate the upper and lower limits of the confidence interval, the expression is: ; in, represents the lower limit of the confidence interval, represents the upper limit of the confidence interval, represents the predicted value of wear degree, Indicates the corresponding confidence level under normal distribution The critical value of Represents the standard deviation of the residuals.

[0054] In this embodiment, the upper and lower limits of the confidence interval calculated based on the normal distribution assumption can provide equipment maintenance personnel with a clear error range, assisting them in making more reasonable maintenance or replacement decisions under different risk preferences, and avoiding excessive maintenance or sudden failures due to misjudgment.

[0055] In this embodiment, the friction plate may encounter various non-steady-state conditions during actual operation, and the prediction error will fluctuate with environmental changes. By introducing confidence analysis, the system can dynamically reflect the prediction stability, improving the model's adaptability and robustness under changing conditions.

[0056] In this embodiment, the introduction of confidence intervals can not only be used to numerically express prediction accuracy, but can also be used as part of a visual display, presented in the form of error bands in the monitoring interface, to help operation and maintenance personnel intuitively judge the state of the friction plate and set different warning levels accordingly.

[0057] Combine the wear degree prediction value and its confidence interval to generate a fatigue life prediction interval with an error range, including: Calculating the remaining useful life of the wear degree prediction value at the current wear rate using the Archard wear model; In this embodiment, the wear degree prediction value is converted into the remaining service life based on the Archard wear model, and a physical correlation between the wear rate and fatigue life is established, which improves the mechanism support of the life prediction model and enhances the interpretability of the prediction results.

[0058] Converting the confidence interval of the wear degree prediction value into the upper and lower limits of the fatigue life prediction interval according to the wear change trend over time; In this embodiment, by converting the wear degree confidence interval into the fatigue life prediction interval, the system can identify the wear stage of the current friction plate, such as initial wear, stable wear or accelerated failure, and divide it into different risk levels accordingly, providing a basis for operation and maintenance decisions.

[0059] Using a state assessment algorithm to perform online analysis and monitoring of the confidence level, and to track the state of the friction plate in real time; The current wear level prediction value and its confidence interval are displayed in real time. When it is detected that the friction plate enters a state of moderate wear or more serious wear, the system automatically triggers an alarm.

[0060] In this embodiment, the current wear degree prediction value and its confidence interval are displayed in real time, allowing operation and maintenance personnel to intuitively grasp the changing trend of the friction plate health status, assisting in quickly determining whether replacement or adjustment of operating parameters is needed, and improving the operability and intelligence level of the system.

[0061] In this embodiment, a confidence interval is introduced on the basis of traditional single-point life prediction, which not only provides the predicted value itself but also gives its possible error range, making the prediction result more valuable for reference and engineering operability, and helping to formulate a scientific and reasonable maintenance plan.

[0062] The present invention introduces the average friction coefficient as a key state monitoring parameter, combines multidimensional feature normalization processing with deep neural network modeling, and establishes a nonlinear mapping relationship between the fatigue life of the friction plate and its operating conditions, significantly improving the prediction accuracy and model generalization ability.

[0063] The wear process of the friction plate is dynamically simulated based on the Archard wear model, and a state assessment model is constructed using the fuzzy comprehensive evaluation method. The entropy weight method is combined to scientifically weight the optimal characterization parameters, effectively identify the current wear stage of the friction plate, and output a quantitative wear degree prediction value. The confidence analysis module is introduced to perform statistical analysis on historical residuals and calculate the confidence interval to generate a fatigue life prediction interval containing the error range, thereby enhancing the credibility and interpretability of the prediction results. At the same time, the state assessment algorithm is combined to realize online confidence monitoring and real-time early warning functions. When the friction plate is detected to enter a moderate wear or more serious state, the system automatically triggers an alarm to assist operation and maintenance personnel in taking timely intervention measures to avoid sudden equipment failure.

[0064] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. A method for predicting the fatigue life of a friction plate considering the influence of multiple factors, characterized in that: The following steps are involved: Obtaining performance parameters of the friction plate under actual working conditions that are affected by multiple factors, and introducing an average friction coefficient into the performance parameters as a key parameter for state monitoring of the friction plate; Normalizing the key parameters to construct a multidimensional feature input vector, training the deep neural network based on a deep neural network model and combining historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors; The performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current operating parameters of the friction plate to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value; Simulating the wear of the friction plate according to the fatigue life prediction value, establishing a corresponding relationship between the optimal characterization parameters of the friction plate and the state evaluation, analyzing the change trend of the optimal characterization parameters using a fuzzy comprehensive evaluation method, establishing a fuzzy evaluation model for state evaluation, and obtaining a predicted value of the wear degree of the friction plate; A confidence analysis module is used for the predicted value of the degree of wear of the friction plate, and a confidence assessment is performed on the prediction result of the predicted value of the degree of wear of the friction plate to generate a fatigue life prediction interval including an error range.

2. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 1, characterized in that: Obtaining performance parameters of the friction plate under actual working conditions that are affected by multiple factors, and introducing the average friction coefficient into the performance parameters as a key parameter for state monitoring of the friction plate, including: Obtaining performance parameters of the friction plate under actual working conditions influenced by multiple factors, wherein the influencing factors include at least: friction material properties, operating temperature, contact pressure, sliding speed, lubrication state, load cycle frequency, and friction force variation curve during the friction process; During the operation of the friction plate, instantaneous friction force and normal pressure data at multiple time points are collected in real time to calculate the average friction coefficient of the friction, which is expressed as: ； in, represents the friction force at the i-th sampling point, represents the positive pressure at the i-th sampling point, Indicates the total number of sampling points; The average friction coefficient is used as a characteristic parameter reflecting the wear state and friction stability of the friction plate surface, and as a key parameter for state monitoring of the friction plate.

3. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 2, characterized in that: The key parameters are normalized to construct a multidimensional feature input vector. Based on a deep neural network model, the deep neural network is trained in combination with historical fatigue test data to establish a nonlinear mapping relationship between the fatigue life of the friction plate and various influencing factors, including: The key parameters are normalized to the minimum and maximum values, and the normalized key parameters are sorted according to the degree of wear of the friction plate. The degree of wear of the friction plate is determined by the friction coefficient on the corresponding friction plate. The expression is: ； in, represents the number of friction pairs of the friction plate, q represents the pressure ratio between the friction plates, 、 Indicates the inner and outer diameters of the friction plate, Indicates the friction torque of the friction plate; Normalize the key parameters after sorting to generate a feature input vector of uniform dimension , where n represents the number of feature dimensions; A fully connected feedforward neural network model is constructed and trained on historical fatigue test data by integrating an attention mechanism. The operating parameters under different working conditions are recorded and the parameters of the fully connected feedforward neural network model are optimized based on the fatigue failure time and cycle number corresponding to the friction plate. The influencing factors of the allowable process of the friction plate under different industrial and mining conditions are input into the fully connected feedforward neural network model to establish a highly nonlinear, multivariable coupling relationship between the fatigue life of the friction plate and its influencing factors, and obtain the interaction effect between multiple factors.

4. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 3, characterized in that: The performance of the friction plate is monitored in real time according to the nonlinear mapping relationship, and the current friction plate operating parameters to be predicted are input into the trained deep neural network model to obtain the corresponding fatigue life prediction value, including: The friction plate operating parameters under the current working conditions are collected in real time by sensors, and the operating parameters are input into the fully connected feedforward neural network model according to the nonlinear mapping relationship to obtain a multi-factor multi-dimensional feature input vector ,in represents the average friction coefficient, Indicates the operating temperature, Indicates the contact pressure, Indicates the sliding speed of the friction plate during actual operation. Indicates the thickness of the lubricating oil film on the friction plate surface. Indicates the friction plate load cycle frequency, represents the material density, Indicates the thermal conductivity of the material, etc. The multidimensional features are input into the vector The ReLU activation function is used to enhance the nonlinear expression capability when inputting the deep neural network model. After forward propagation calculation, the deep neural network model outputs the corresponding predicted value, which represents the estimated remaining fatigue life of the friction plate under current conditions.

5. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 3, characterized in that: The wear amount of the friction plate is simulated according to the fatigue life prediction value, including: The Archard wear model is used to simulate the wear process of the friction plate under different working conditions, and its expression is: ； in, represents the average friction coefficient, W represents the normal load of the friction plate, d represents the relevant sliding friction distance, and V represents the wear volume; According to the multi-dimensional feature input vector under the current working condition Calculate the wear amount of the friction plate in each cycle based on the fatigue life prediction value; The wear of the friction plate throughout its life cycle is dynamically simulated using the ANSYS simulation tool to generate a wear curve graph and obtain the wear variation trend over time or the number of cycles.

6. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 5, characterized in that: According to the wear amount of the friction plate, a corresponding relationship between the optimal characterization parameter of the friction plate and the state evaluation is established, including: According to the wear trend of the friction plate over time, n data points are randomly extracted. 、 , , establish a linear regression equation for the n data points to obtain the wear amount of the friction plate during the stable wear period and wear life The offline relationship between them is expressed as: ； in, represents the initial wear amount, Indicates the growth rate of wear; The degree of fit between the regression model and the actual data is determined by calculating the standard error of the n data points, and the warning baseline and abnormal baseline of the wear element changing over time are obtained according to the linear regression equation and the standard error; The wear amount is analyzed using random forest algorithm Conduct analysis and establish a mapping relationship between characterization parameters and wear status; Based on the characterization parameter values ​​obtained from real-time monitoring, sensitivity analysis is performed on each characterization parameter to identify the key parameters that have the greatest impact on the wear state. Based on the weighted fusion strategy, weights are assigned to different characterization parameters, and a comprehensive scoring system is constructed to quantify the overall wear state of the friction plate. Its expression is: ； in, Represent the weights of the corresponding characterization parameters, Indicates the comprehensive rating.

7. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 6, characterized in that: The fuzzy comprehensive evaluation method is used to analyze the weights affecting the optimal characterization parameters, including: Create a factor set containing each indicator for the comprehensive score , Represents the mth element that affects the evaluation object, and establishes an evaluation set for the comprehensive score , the evaluation set The middle element represents the evaluation result of its corresponding influencing factor; The positive evaluation index is used to calculate the weight of the corresponding sample under the i-th factor index, and the entropy value of the j-th index is obtained. The expression is: ； Where n represents the number of samples, Representation factor set The number of elements in It represents the weight of the jth indicator under the i-th factor indicator; The difference of each indicator is calculated according to the entropy value, the weights of multiple indicators in the comprehensive evaluation system are calculated, and the weights affecting the optimal characterization parameters are obtained. The expression is: 。 8. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 7, characterized in that: A fuzzy judgment model for state evaluation is established for the weights of the optimal characterization parameters to obtain a predicted value of the degree of friction plate wear, including: For each characterization parameter, its fuzzy membership function under different wear states is defined, and a fuzzy evaluation matrix R is constructed, where each row represents a characterization parameter, each column represents a wear state level, and the matrix elements are It represents the membership value of the i-th characterization parameter under the j-th wear state level, and its expression is: ； Where m is the number of characterization parameters and n is the number of wear state levels; The fuzzy comprehensive evaluation formula is calculated according to the weight of the optimal characterization parameter to obtain a comprehensive evaluation result vector B, which is expressed as follows: ； in, , represents the comprehensive membership value of the friction plate at each wear state level; Comprehensive evaluation result vector Perform normalization processing to obtain the membership value and determine the current wear state level of the friction plate; The wear degree prediction value is obtained according to the wear degree range corresponding to the wear state level with the highest membership value.

9. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 8, characterized in that: A confidence analysis module is used for the predicted value of the friction plate wear degree to perform a confidence assessment on the predicted result of the predicted value of the friction plate wear degree, including: The wear degree prediction value is combined with each sample in the historical data set, the residual between the wear degree prediction value and the true value is calculated, and the distribution characteristics of the residual are statistically analyzed; The distribution characteristics of the residuals are calculated using the properties of the normal distribution to calculate the confidence interval and set the confidence level , calculate the upper and lower limits of the confidence interval, the expression is: ； in, represents the lower limit of the confidence interval, represents the upper limit of the confidence interval, represents the predicted value of wear degree, Indicates the corresponding confidence level under normal distribution The critical value of Represents the standard deviation of the residuals.

10. The method for predicting fatigue life of a friction plate considering multiple factors according to claim 9, characterized in that: Combining the wear degree prediction value and its confidence interval, a fatigue life prediction interval with an error range is generated, including: Calculating the remaining useful life of the wear degree prediction value at the current wear rate using the Archard wear model; Converting the confidence interval of the wear degree prediction value into the upper and lower limits of the fatigue life prediction interval according to the wear change trend over time; Using a state assessment algorithm to perform online analysis and monitoring of the confidence level, and to track the state of the friction plate in real time; The current wear level prediction value and its confidence interval are displayed in real time. When it is detected that the friction plate enters a state of moderate wear or more serious wear, the system automatically triggers an alarm.

Images