System key performance degradation factor prediction method for equipment system argumentation

Abstract

This invention discloses a method for predicting critical performance degradation factors in equipment systems, comprising: 1) collecting current degradation factors and estimating their end-of-life values; 2) performing a linear transformation on the degradation factors to construct a degradation factor predictor based on a linear-exponential model; 3) transforming the model into a linear recursive form of the parameter vector and identifying the model parameters online; 4) calculating the posterior probability of the degradation state transition and identifying the transition point from linear degradation to exponential degradation online. This invention, while obtaining the predicted degradation factor values, can determine whether the corresponding performance will be in the uniform degradation stage indicated by the linear degradation model or the accelerated degradation stage indicated by the exponential degradation model, thereby facilitating the assessment of the system's health status and enabling predictive maintenance and health management.

Description

Technical Field

[0001] This invention proposes a method for predicting key performance degradation factors of a system for equipment system demonstration, which belongs to the field of equipment fault diagnosis and health monitoring. Background Technology

[0002] During the long-term demonstration and design process of equipment systems, various complex equipment systems are usually composed of a large number of functional units and subsystems. There are close structural connections and functional couplings between the components, and they are often in complex, variable, and even harsh environmental conditions during actual operation. As the service life of the system extends, various key components may gradually experience performance degradation due to factors such as dirt accumulation, material corrosion and aging, mechanical wear, frequent changes in operating conditions, and improper maintenance conditions. This can lead to a decline in the overall performance of the system, and in severe cases, may even cause system failures, significantly impacting the safety, reliability, and mission capabilities of the equipment system.

[0003] System critical performance degradation factors are a set of important indicators that continuously characterize the performance change trend of equipment systems. By modeling and analyzing these indicators, the impact of these degradation factors on the system's health status and performance level can be quantitatively described. In equipment system demonstration and system modeling, degradation factors typically consist of multiple influencing factors and can be represented as a multidimensional parameter vector. During simulation model operation, the values ​​of degradation factors and their combinations affect the system's internal state variables and further influence the model output, thus reflecting the impact of system performance degradation on overall operational capability. Simultaneously, with the development of intelligent equipment systems and model-driven design concepts, model-based analysis and control methods are widely used. These methods typically require high parameter accuracy in the system model, including critical performance degradation factors. Therefore, high-precision prediction of system critical performance degradation factors is of great significance for equipment system demonstration, system modeling, and operational performance evaluation.

[0004] Classic methods for predicting degradation factors can be categorized as follows: (1) Prediction methods based on physical models. This involves using the performance degradation mechanism of the system to predict degradation factors. For example, establishing a mechanism model of material aging, strength loss, crack propagation, or structural failure in the system to predict the changing trend of degradation factors. This type of method requires a high-precision physical model, and the model establishment and prediction process is relatively complex.

[0005] (2) Prediction methods based on expert systems. These methods predict the changing trends of degradation factors through expert knowledge. By establishing a rule base, the predicted values ​​of degradation factors are inferred based on the rules. This type of method relies on the accuracy of expert knowledge, and the process of manually establishing the rule base is also quite cumbersome.

[0006] (3) Filter-based prediction methods. Such as using Kalman filtering or particle filtering to predict the state variables of a system, such as degradation factors. These methods require linearization or other transformations of the model, which leads to a loss of model accuracy and a decrease in prediction accuracy.

[0007] (4) Prediction methods based on data-driven models. These methods use historical data to train data-driven models, such as neural networks and support vector machines, and then use the trained models to predict degradation factors. This type of method requires the collection of a large amount of data to train the model, and there is a problem of mismatch between the distribution of training data and the target domain for different models and operating conditions, making it difficult to achieve accurate modeling of specific systems.

[0008] In view of the problems of the above technologies, a degradation mathematical model that is decoupled from the physical process of the system and has online prediction and parameter identification should be established, and a real-time degradation factor prediction algorithm should be designed accordingly to achieve high-efficiency, high-precision and high-applicability performance degradation factor prediction. Summary of the Invention

[0009] To address the problem of predicting critical performance degradation factors of systems as described in the background art, this invention proposes a method for predicting critical performance degradation factors of systems oriented towards equipment system demonstration.

[0010] The technical solution of this invention is as follows: A method for predicting critical performance degradation factors in equipment systems demonstration is proposed. This method comprises four parts: collecting current degradation factors, constructing a predictor, performing online parameter identification, and identifying transition points. The specific steps are as follows: S1 collects the degradation factors of the current performance; including the following steps: S1.1 Establish a component-level model for the system, specifying the decreasing degradation factor for each corresponding component. ; S1.2 Solve and optimize the component-level model to obtain the current operating cycle. Optimal degradation factor estimate ; S1.3 Estimate the value of the degradation factor at the end of its lifespan based on historical information from systems of the same model, denoted as... ,in It is the number of operating cycles at the end of its lifespan; S2 constructs a degradation factor predictor based on a linear-exponential model; including the following steps: S2.1 Degradation Factor Perform a linear transformation: in, It is the new degradation factor after linear transformation. It is a degenerative factor The value at the initial stage of degradation, thus ; and thus: in, This is an estimate of the new degradation factor at the end of its lifespan; S2.2 to Establish a linear-exponential model to obtain its performance over any operating period. Predicted value : in These are the parameters of the linear-exponential model. It is the turning point in the cycle from linear degradation to exponential degradation. It is the first During the running cycle The predicted value; S3 performs online identification of parameters for linear-exponential models, including the following steps: S3.1 Transform the linear-exponential model into a parameter vector The linear recursive form: In the formula, This is the coefficient matrix of the linear-exponential model. This is the parameter vector of the linear-exponential model before recursive updates. These are the parameters of the linear-exponential model before recursive updates; It is the first New degradation factors in the operating cycle; Problem transformed into online identification That is, starting from a certain initial value, and then obtaining the current running cycle. New degradation factors At that time, Updated to The problem; S3.2 by Given the conditions, and initial values ​​of the parameter vector of the linearly degenerate state: In the formula, Parameters of the linear-exponential model for the linear degradation stage initial value, Parameters of the linear-exponential model for the linear degradation stage The initial value; S3.3 Transition point from the operating cycle The condition for first-order continuity of a function is given by an exponentially degenerate state in... The parameter value at time, that is, when There is always In the formula, These are the parameters of the recursively updated linear-exponential model; Setting preparatory parameters for exponential degradation state Used to determine the turning point of the operating cycle. Then it switches to the exponentially degenerate state. To ensure the continuity of the function, we have... S3.4 Utilizing a forgetting factor Online parameter updates using recursive least squares; including: S3.4.1 Introducing a symmetric positive definite covariance matrix To describe the uncertainty of the parameters and perform initialization: in for unit array; S3.4.2 For each running cycle The recursive least squares method is performed iteratively to obtain the updated parameter vector. positive definite covariance matrix : Its expansion is in For the running cycle Gain factor at time; S3.5 When At that time, after each execution of the recursive least squares method in step S3.4, the following is taken out: Update according to the method described in S3.3 ; S4 Online Transition Point Identification The steps include: S4.1 Calculate the prior probability; let the set of linearly degenerate operating cycles be... and the set of exponential degradation operating cycles And set a priori transition point. and prior lifetime termination point , and The prior probabilities are respectively S4.2 Establish the error distribution; including: S4.2.1 Set the operating cycle Prediction error The prediction error of the linear model was obtained by expanding the model using both linear and exponential models. And exponential model prediction error : S4.2.2 Let and Follows a Gaussian distribution: in These are the means of the corresponding distributions. These are the standard deviations of the corresponding distributions; S4.2.3 Setting initial values ​​for the error distribution parameters of the linear model: initial value of the sample mean Initial value of sample standard deviation These are the preset values. and : In the formula, This is the initial value of the sample mean of the linear model error. This represents the initial standard deviation of the linear model error samples. S4.2.4 When Right now At that time, the initial values ​​of the error distribution parameters of the exponential model are set to the same preset values. and : S4.3 Calculate the posterior probability; including: S4.3.1 Predicting Errors Using a Linear Model The distribution is taken from the corresponding 97.5 quantile. and 2.5 quantiles ; S4.3.2 If Then, the result is calculated according to the following expression after observation. hour, Posterior probability: in yes Gaussian distribution at time The probability density function at that location, yes Gaussian distribution at time The probability density function at a given location is a Gaussian distribution. The expression for the probability density function is: In the formula, The prediction error corresponding to the linear-exponential model in S4.2.1 is converted into the linear decay phase and the exponential decay phase, respectively. and ; S4.3.3 If S4.3.2 condition If this is not true, then calculate according to the following expression based on the observed... hour, Posterior probability: S4.4 For each newly observed cycle If there is Then determine and correct Then stop the judgment: Otherwise, determine The data for the next operating cycle will be further evaluated. S4.5 Given a threshold number of running cycles ,when and Sometimes, or and At this point, the recursive iteration begins, updating the corresponding Gaussian distribution parameters: In the formula, The mean of the error samples during the c-th cycle. Let be the sample variance of the error during the c-th cycle. Let Variance be the error distribution variance during the c-th cycle. Let be the mean of the error distribution during the c-th cycle.

[0011] The beneficial effects of this invention are as follows: (1) While obtaining the predicted value of the degradation factor, it is possible to determine whether the corresponding performance will be in the uniform degradation stage indicated by the linear degradation model or the accelerated degradation stage indicated by the exponential degradation model, which is conducive to assessing the health status of the system and carrying out predictive maintenance and health management.

[0012] (2) The linear-exponential model proposed in this invention is applicable to the prediction of various degradation factors in the system and can perform various performance degradation modeling analyses.

[0013] (3) The predictor proposed in this invention does not require a large amount of data for learning, has strong versatility, and can directly carry out online parameter identification and prediction for any type and working condition of system, thereby establishing the corresponding feature model.

[0014] (4) The combination of degradation factor prediction and simulation technology can lay the foundation for estimating performance indicators under degradation state and conducting system control research. Attached Figure Description

[0015] Figure 1 A schematic diagram of the linear-exponential model and posterior probability calculation established for this invention; Figure 2 This is a flowchart of the algorithm of the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. It is understood that the embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present invention are shown in the accompanying drawings.

[0017] S1 collects the degradation factors of the current performance; including the following steps: S1.1 Establish a component-level model for the system, specifying the decreasing degradation factor for each corresponding component. ; S1.2 Solve and optimize the component-level model to obtain the current operating cycle. Optimal degradation factor estimate ; S1.3 Estimate the value of the degradation factor at the end of its lifespan based on historical information from systems of the same model, denoted as... ,in It is the number of operating cycles at the end of its lifespan; S2 constructs a degradation factor predictor based on a linear-exponential model; including the following steps: S2.1 Degradation Factor Perform a linear transformation: in, It is the new degradation factor after linear transformation. It is a degenerative factor The value at the initial stage of degradation, thus ; and thus: in, This is an estimate of the new degradation factor at the end of its lifespan; S2.2 to Establish a linear-exponential model to obtain its performance over any operating period. Predicted value : in These are the parameters of the linear-exponential model. It is the turning point in the cycle from linear degradation to exponential degradation. It is the first During the running cycle The predicted value; S3 performs online identification of parameters for linear-exponential models, including the following steps: S3.1 Transform the linear-exponential model into a parameter vector The linear recursive form: In the formula, This is the coefficient matrix of the linear-exponential model. This is the parameter vector of the linear-exponential model before recursive updates. These are the parameters of the linear-exponential model before recursive updates; It is the first New degradation factors in the operating cycle; Problem transformed into online identification That is, starting from a certain initial value, and then obtaining the current running cycle. New degradation factors At that time, Updated to The problem; S3.2 by Given the conditions, and initial values ​​of the parameter vector of the linearly degenerate state: In the formula, Parameters of the linear-exponential model for the linear degradation stage initial value, Parameters of the linear-exponential model for the linear degradation stage The initial value; S3.3 Transition point from the operating cycle The condition for first-order continuity of a function is given by an exponentially degenerate state in... The parameter value at time, that is, when There is always In the formula, These are the parameters of the recursively updated linear-exponential model; Setting preparatory parameters for exponential degradation state Used to determine the turning point of the operating cycle. Then it switches to the exponentially degenerate state. To ensure the continuity of the function, we have... S3.4 Utilizing a forgetting factor Online parameter updates using recursive least squares; including: S3.4.1 Introducing a symmetric positive definite covariance matrix To describe the uncertainty of the parameters and perform initialization: in for unit array; S3.4.2 For each running cycle The recursive least squares method is performed iteratively to obtain the updated parameter vector. positive definite covariance matrix : Its expansion is in For the running cycle Gain factor at time; S3.5 When At that time, after each execution of the recursive least squares method in step S3.4, the following is taken out: Update according to the method described in S3.3 ; S4 Online Transition Point Identification The steps include: S4.1 Calculate the prior probability; let the set of linearly degenerate operating cycles be... and the set of exponential degradation operating cycles And set a priori transition point. and prior lifetime termination point , and The prior probabilities are respectively S4.2 Establish the error distribution; including: S4.2.1 Set the operating cycle Prediction error The prediction error of the linear model was obtained by expanding the model using both linear and exponential models. And exponential model prediction error : S4.2.2 Let and Follows a Gaussian distribution: in These are the means of the corresponding distributions. These are the standard deviations of the corresponding distributions; S4.2.3 Setting initial values ​​for the error distribution parameters of the linear model: initial value of the sample mean Initial value of sample standard deviation These are the preset values. and : In the formula, This is the initial value of the sample mean of the linear model error. This represents the initial standard deviation of the linear model error samples. S4.2.4 When Right now At that time, the initial values ​​of the error distribution parameters of the exponential model are set to the same preset values. and : S4.3 Calculate the posterior probability; including: S4.3.1 Predicting Errors Using a Linear Model The distribution is taken from the corresponding 97.5 quantile. and 2.5 quantiles ; S4.3.2 If Then, the result is calculated according to the following expression after observation. hour, Posterior probability: in yes Gaussian distribution at time The probability density function at that location, yes Gaussian distribution at time The probability density function at a given location is a Gaussian distribution. The expression for the probability density function is: In the formula, The prediction error corresponding to the linear-exponential model in S4.2.1 is converted into the linear decay phase and the exponential decay phase, respectively. and ; S4.3.3 If S4.3.2 condition If this is not true, then calculate according to the following expression based on the observed... hour, Posterior probability: S4.4 For each newly observed cycle If there is Then determine and correct Then stop the judgment: Otherwise, determine The data for the next operating cycle will be further evaluated. S4.5 Given a threshold number of running cycles ,when and Sometimes, or and At this point, the recursive iteration begins, updating the corresponding Gaussian distribution parameters: In the formula, The mean of the error samples during the c-th cycle. Let be the sample variance of the error during the c-th cycle. Let Variance be the error distribution variance during the c-th cycle. Let be the mean of the error distribution during the c-th cycle.

[0018] Ultimately, a prediction algorithm for key performance degradation factors of the system is formed, such as... Figure 2 As shown.

[0019] The above examples are merely implementation results of the present invention. It should be noted that those skilled in the art can make various modifications or improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for predicting key performance degradation factors of a system for equipment system demonstration, characterized in that, Includes the following steps: S1 collects the degradation factors of the current performance; including the following steps: S1.1 Establish a component-level model for the system, specifying the decreasing degradation factor for each corresponding component. ; S1.2 Solve and optimize the component-level model to obtain the current operating cycle. Optimal degradation factor estimate ; S1.3 Estimate the value of the degradation factor at the end of its lifespan based on historical information from systems of the same model, denoted as... ,in It is the number of operating cycles at the end of its lifespan; S2 constructs a degradation factor predictor based on a linear-exponential model; including the following steps: S2.1 Degradation Factor Perform a linear transformation: in, It is the new degradation factor after linear transformation. It is a degenerative factor The value at the initial stage of degradation, thus ; and thus: in, This is an estimate of the new degradation factor at the end of its lifespan; S2.2 to Establish a linear-exponential model to obtain its performance over any operating period. Predicted value : in These are the parameters of the linear-exponential model. It is the turning point in the cycle from linear degradation to exponential degradation. It is the first During the running cycle The predicted value; S3 performs online identification of parameters for linear-exponential models, including the following steps: S3.1 Transform the linear-exponential model into a parameter vector The linear recursive form: In the formula, This is the coefficient matrix of the linear-exponential model. This is the parameter vector of the linear-exponential model before recursive updates. These are the parameters of the linear-exponential model before recursive updates; It is the first New degradation factors in the operating cycle; Problem transformed into online identification That is, starting from a certain initial value, and then obtaining the current running cycle. New degradation factors At that time, Updated to The problem; S3.2 by Given the conditions, and initial values ​​of the parameter vector of the linearly degenerate state: In the formula, Parameters of the linear-exponential model for the linear degradation stage initial value, Parameters of the linear-exponential model for the linear degradation stage The initial value; S3.3 Transition point from the operating cycle The condition for first-order continuity of a function is given by an exponentially degenerate state in... The parameter value at time, that is, when There is always In the formula, These are the parameters of the recursively updated linear-exponential model; Setting preparatory parameters for exponential degradation state Used to determine the turning point of the operating cycle. Then it switches to an exponentially degenerate state. To ensure the continuity of the function, we have... S3.4 Utilizing a forgetting factor Online parameter update using recursive least squares; including: S3.4.1 Introducing a symmetric positive definite covariance matrix To describe the uncertainty of the parameters and perform initialization: in for unit array; S3.4.2 For each running cycle The recursive least squares method is performed iteratively to obtain the updated parameter vector. and positive definite covariance matrix : Its expansion is in For the running cycle Gain factor at time; S3.5 When At that time, after each execution of the recursive least squares method in step S3.4, the following is taken out: Update according to the method described in S3.3 ; S4 Online Transition Point Identification The steps include: S4.1 Calculate the prior probability; let the set of linearly degenerate operating cycles be... and the set of exponential degradation operating cycles And set a priori transition point. and prior lifetime termination point , and The prior probabilities are respectively S4.2 Establish the error distribution; including: S4.2.1 Set the operating cycle Prediction error The prediction error of the linear model was obtained by expanding the model using both linear and exponential models. And exponential model prediction error : S4.2.2 Let and Follows a Gaussian distribution: in These are the means of the corresponding distributions. These are the standard deviations of the corresponding distributions; S4.2.3 Setting initial values ​​for the error distribution parameters of the linear model: initial value of the sample mean Initial value of sample standard deviation These are the preset values. and : In the formula, This is the initial value of the sample mean of the linear model error. This represents the initial standard deviation of the linear model error samples. S4.2.4 When Right now At that time, the initial values ​​of the error distribution parameters of the exponential model are set to the same preset values. and : S4.3 Calculate the posterior probability; including: S4.3.1 Predicting Errors Using a Linear Model The distribution is taken from the corresponding 97.5 quantile. and 2.5 quantiles ; S4.3.2 If Then, the result is calculated according to the following expression when observed. hour, Posterior probability: in yes Gaussian distribution at time The probability density function at that point, yes Gaussian distribution at time The probability density function at a given location is a Gaussian distribution. The expression for the probability density function is: In the formula, The prediction error corresponding to the linear-exponential model in S4.2.1 is converted into the linear decay phase and the exponential decay phase, respectively. and ; S4.3.3 If S4.3.2 condition If this is not true, then calculate according to the following expression based on the observed... hour, Posterior probability: S4.4 For each newly observed cycle If there is Then determine and correct Then stop the judgment: Otherwise, determine The data for the next operating cycle will be further evaluated. S4.5 Given a threshold number of running cycles ,when and Sometimes, or and At this point, the recursive iteration begins, updating the corresponding Gaussian distribution parameters: In the formula, The mean of the error samples during the c-th cycle. Let be the sample variance of the error during the c-th cycle. Let Variance be the error distribution variance during the c-th cycle. Let be the mean of the error distribution during the c-th cycle.

Images