Sealing material performance degradation prediction method based on GA-SVR model

Abstract

The invention discloses a sealing material performance degradation prediction method based on a GA-SVR model. The method comprises the following steps: obtaining performance data of a sealing material in different corrosion environments; the performance data are input into a degradation prediction model, penalty factors and insensitive errors are introduced, an optimal nonlinear regression model is determined, and the degradation prediction model is constructed on the basis of a support vector regression model; on the basis of the optimal nonlinear regression model, performing global search optimization by using a genetic algorithm to obtain an optimal GA-SVR (Genetic Algorithm-Support Vector Regression) model; and inputting the performance data into the optimal GA-SVR model to obtain a performance prediction result. The method can improve the prediction precision and reliability of the performance degradation of the sealing material in the acid-alkali corrosion environment, and provides technical support for the risk prevention and control of the leakage of dangerous media of the sealing structure of chlor-alkali chemical production equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of material performance prediction, and particularly relates to a sealing material performance degradation prediction method based on a GA-SVR model. BACKGROUND

[0002] In chlor-alkali chemical production, sealing materials are exposed to strong acid and alkali and other corrosive media for a long time, and the performance degradation of the sealing materials easily leads to sealing failure and leakage risk. Traditional sealing material aging evaluation methods mostly rely on empirical models (such as exponential models), and the prediction accuracy is insufficient under small sample conditions. The standard support vector regression (SVR) model is highly dependent on the kernel function and hyperparameter setting, and is easily trapped in local optimization. Therefore, a sealing material performance degradation prediction method based on a GA-SVR model is urgently needed. SUMMARY

[0003] To solve the above technical problems, the application provides a sealing material performance degradation prediction method based on a GA-SVR model, which realizes more accurate performance degradation trend prediction through intelligent algorithm optimization.

[0004] To achieve the above purpose, the application provides a sealing material performance degradation prediction method based on a GA-SVR model, which comprises the following steps:

[0005] Obtaining performance data of sealing materials under different corrosion environments;

[0006] Inputting the performance data into a degradation prediction model, introducing a penalty factor and an insensitive error, and determining an optimal nonlinear regression model, wherein the degradation prediction model is constructed based on a support vector regression model;

[0007] Based on the optimal nonlinear regression model, a genetic algorithm is used for global search optimization to obtain an optimal GA-SVR model;

[0008] Inputting the performance data into the optimal GA-SVR model to obtain a performance prediction result.

[0009] Optionally, the performance data includes quality, hardness, tensile strength, and elongation at break.

[0010] Optionally, inputting the performance data into a degradation prediction model, introducing a penalty factor and an insensitive error, and determining an optimal nonlinear regression model comprises the following steps:

[0011] Using a polynomial kernel function to initialize a nonlinear mapping relationship between the performance data and the degradation time;

[0012] Obtaining an insensitive loss function, introducing an insensitive error and a penalty factor, constructing an objective function, and obtaining an optimization problem of support vector regression;

[0013] Solving the optimization problem by using dual theory to obtain support vector coefficients and bias terms, and determining the optimal nonlinear regression model.

[0014] Optionally, the insensitive loss function is:

[0015]

[0016] Wherein, L ε is an insensitive loss function, y is the actual value of the sample, f(x) is the predicted value of SVR, and epsilon is an insensitive error.

[0017] Optionally, the optimization problem is:

[0018]

[0019] Wherein, xi i , xi * i is a slack variable, W is a weight vector, T is a transpose, C is a penalty factor, i is a sample index, and n is the number of training samples.

[0020] Optionally, the optimal GA-SVR model is obtained by using a genetic algorithm for global search optimization, including:

[0021] Calculating a fitness function of the degradation prediction model;

[0022] According to the calculation result, a termination condition is determined to obtain the optimal GA-SVR model.

[0023] Optionally, the fitness function is:

[0024]

[0025] Wherein, n is the number of training samples, yi i is the actual value of the i th sample index, is the predicted value of the i th sample.

[0026] Compared with the prior art, the present application has the following advantages and technical effects:

[0027] The application significantly improves the prediction accuracy and reliability of the performance degradation of sealing materials in acid-base corrosion environment by combining SVR and GA technology. Compared with the traditional exponential model and the standard SVR model, the GA-SVR model proposed in the application significantly improves the prediction accuracy, and the average absolute error MAE of the GA-SVR model is reduced by more than 50% compared with the traditional model; the model is effectively applicable to the modeling of corrosion environment data with scarce samples; the model can better process nonlinear and high-dimensional data, effectively overcome the prediction bottleneck problem under the condition of small samples, and provide data-driven decision support for the maintenance of the sealing structure of the chlor-alkali device. BRIEF DESCRIPTION OF DRAWINGS

[0028] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of this application and their description together with the drawings serve to explain the application. In the drawings:

[0029] Figure 1 is a GA-SVR model-based sealing material performance degradation prediction method flowchart of an embodiment of the application;

[0030] Figure 2 is a GA-SVR model-based sealing material performance degradation prediction method framework of an embodiment of the application;

[0031] Figure 3 is a GA-SVR model-based fluororubber hardness degradation evaluation in an embodiment of the application;

[0032] Figure 4 is a 90-day prediction result of fluororubber in a hydrochloric acid environment in an embodiment of the application. DETAILED DESCRIPTION

[0033] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0034] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0035] The present embodiment proposes a GA-SVR model-based sealing material performance degradation prediction method, as shown in Figure 1 The specific steps include the following steps:

[0036] Obtain the performance data of the sealing material under different corrosion environments;

[0037] The performance data is input into the degradation prediction model, and the penalty factor and insensitive error are introduced to determine the optimal nonlinear regression model. The degradation prediction model is constructed based on the support vector regression model.

[0038] Based on the optimal nonlinear regression model, a genetic algorithm is used to perform global search and optimization to obtain the optimal GA-SVR model;

[0039] The performance data is input into the optimal GA-SVR model to obtain the performance prediction results.

[0040] Specifically, such as Figure 2 As shown in the figure, a degradation prediction model is constructed for the degradation process of sealing materials in acid-base corrosion environment to achieve high-precision prediction of key performance parameters of materials, so as to improve the prediction accuracy and reliability of sealing material performance degradation in acid-base corrosion environment, and provide technical support for the prevention and control of hazardous media leakage risks in the sealing structure of chlor-alkali chemical production equipment.

[0041] Furthermore, the performance data include: mass, hardness, tensile strength and elongation at break.

[0042] Specifically, the performance data of sealing materials under different corrosive environments, including mass, hardness, tensile strength, elongation at break, etc., are collected and standardized to eliminate the influence of scale differences between different characteristic dimensions.

[0043] Furthermore, the performance data is input into the degradation prediction model, and the penalty factor and insensitive error are introduced to determine the optimal nonlinear regression model including:

[0044] A polynomial kernel function is used to initialize the nonlinear mapping relationship between performance data and degradation time;

[0045] Obtain an insensitive loss function, introduce insensitive error and penalty factors, construct the objective function, and obtain the optimization problem of support vector regression;

[0046] The duality theory is used to solve the optimization problem, obtain the support vector coefficients and bias terms, and determine the optimal nonlinear regression model.

[0047] Specifically, a degradation prediction model was constructed based on the support vector regression (SVR) model, using a polynomial kernel function to initialize the nonlinear mapping between sealing material performance and degradation time. The SVR model involves two hyperparameters that need to be adjusted: the penalty factor C and the insensitivity error ε.

[0048] Furthermore, the genetic algorithm is used to perform global search and optimization to obtain the optimal GA-SVR model, including:

[0049] Calculate the fitness function of the degradation prediction model;

[0050] According to the calculation result, it is determined that the termination condition is met, and the optimal GA-SVR model is obtained.

[0051] Specifically, the biological evolution mechanism is introduced into the parameter coding population composed of the penalty factor C and the insensitive loss ε, and GA is used for global search optimization. In the parameter optimization process, the mean absolute error (MAE) is used as the fitness function, and the parameter population is updated constantly until the termination condition is met, and the optimal parameter combination is output.

[0052] More specifically, the optimized GA-SVR model is used to train and predict the performance degradation of the sealing material. The final GA-SVR prediction curve and parameter setting are output, which are suitable for performance prediction of various sealing materials in hydrochloric acid and caustic soda environments.

[0053] The embodiment will be described in detail below in conjunction with the drawings:

[0054] Step 1: Data collection and preprocessing:

[0055] The performance data of sealing materials in different corrosion environments (such as hydrochloric acid and caustic soda) are collected. Table 1 below shows various performance parameters of fluororubber materials after degradation in a hydrochloric acid environment.

[0056] Table 1

[0057]

[0058] Step 2: Support Vector Regression (SVR) model construction:

[0059] Taking the hardness degradation performance of fluororubber in a hydrochloric acid environment as an example, support vector regression is selected as the basic model, and a polynomial kernel function is used to initialize the nonlinear mapping relationship between the performance of sealing materials and the degradation time. The hyperparameters of the SVR model are determined: the penalty factor C and the insensitive error ε.

[0060] From a geometric point of view, support vector regression is: the training set S = {(x i ,y i )|i = 1, 2..., r}, x i ∈R n , y i ∈R, x i is the input value of the i-th sample index, and y i is the actual value of the i-th sample index. Find a hyperplane function:

[0061] f(x i ) = <W·x i >+b

[0062] In the formula, <W·x i > is W and x iinner product. This function satisfies all the training set y i with a deviation no more than ε, and makes ||W|| as small as possible. 2 Therefore, the weight coefficient W and the bias term b are obtained by solving the minimum value optimization problem of the following equation:

[0063]

[0064] In the equation, C is a regularization parameter in SVR, which determines the tolerance of error in the training process, and is usually called a penalty factor.

[0065] Since the fitted curve inevitably has some error, SVR sets an insensitive interval ε to make the prediction error of most data points fall within the interval, thereby improving the fault tolerance and generalization ability of the model. In the following equation, L ε is the insensitive loss function, and ε is the insensitive error. The expression of the insensitive loss function L ε is as follows:

[0066]

[0067] To handle regression samples that exceed the error tolerance range, slack variables ξ i and ξ i * are introduced to measure their deviation from the correct value. Substituting the insensitive loss function L ε into the equation, the optimization problem is obtained as follows:

[0068]

[0069] The constraint condition is satisfied:

[0070]

[0071] When solving the above equation, the dual theory is generally used to convert it into a quadratic programming problem. For nonlinear data, a nonlinear mapping function φ is introduced, and the Lagrange equation is established. Through simplification, the dual model of the support vector regression model can be obtained:

[0072]

[0073] In the equation, α, α * are Lagrange multipliers. Let K(x i , x j ) = φ(x i ) T φ(x j ), K(x i , x j) is the inner product of the feature space, called the kernel function. The introduction of the kernel function is essentially to map the samples in the low-dimensional input space to a high-dimensional or even infinite-dimensional feature space, where complex nonlinear relationships can be transformed into linear problems for processing, thus achieving effective regression modeling of nonlinear data. The SVR model is used for degradation evaluation, and the penalty factor C=2 and insensitive error ε=1.5 are selected as parameters.

[0074] Step three: genetic algorithm (GA) optimization:

[0075] The biological evolution mechanism is introduced into the parameter coding population composed of penalty factor C and insensitive error ε, and GA is used for global search optimization. In the parameter optimization process, the mean absolute error (MAE) is used as the fitness function, and the parameter population is updated until the termination condition is met, and the optimal parameter combination is output.

[0076] In order to evaluate the performance of the model on the data set, it is necessary to measure the degree of matching between the predictions made by the model and the actual data. The most commonly used method to measure this is to use the mean absolute error (Mean Absolute Error, MAE), which is used to evaluate the performance of the trained model, and the specific expression is:

[0077]

[0078] In the formula: n is the number of samples in the data set; y i is the actual value of the i-th sample index; is the predicted value of the i-th sample index.

[0079] Where the smaller the mean absolute error, the higher the goodness of fit, indicating that the prediction effect is better and the performance of the model is superior.

[0080] The following process is mainly used to calculate the MAE of a given model:

[0081] 1. Divide the data set into training set and test set;

[0082] 2. Create a model using only the data of the training set;

[0083] 3. Use the model to predict the test set data and calculate the MAE.

[0084] Step four: model training and prediction:

[0085] The optimized GA-SVR model is used to train and predict the performance degradation of sealing materials, and compared with the traditional exponential model and the standard SVR model. The final GA-SVR prediction curve and parameter settings are output, which are suitable for performance prediction of various sealing materials in hydrochloric acid and caustic soda environments. The unknown data is predicted, and the prediction results are analyzed and evaluated.

[0086] As Figure 3 shown, the hardness performance of fluoroelastomer in hydrochloric acid environment, the two parameters of the SVR model optimized by genetic algorithm are C=1.2914 and ε=0.7431 respectively.

[0087] Deterioration prediction:

[0088] The 90-day prediction result of fluoroelastomer in hydrochloric acid environment is shown in Figure 4 , the relative error of GA-SVR model prediction is 0.0031%.

[0089] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for predicting sealing material performance degradation based on GA-SVR model, characterized in that: include: Obtain performance data of sealing materials in different corrosive environments; Inputting the performance data into a degradation prediction model, introducing a penalty factor and an insensitive error, and determining an optimal nonlinear regression model, wherein the degradation prediction model is constructed based on a support vector regression model; Based on the optimal nonlinear regression model, a genetic algorithm is used to perform global search and optimization to obtain an optimal GA-SVR model; The performance data is input into the optimal GA-SVR model to obtain a performance prediction result.

2. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 1, characterized in that: The performance data include: mass, hardness, tensile strength and elongation at break.

3. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 1, characterized in that: Inputting the performance data into the degradation prediction model, introducing a penalty factor and an insensitive error, and determining the optimal nonlinear regression model include: A polynomial kernel function is used to initialize the nonlinear mapping relationship between performance data and degradation time; Obtain an insensitive loss function, introduce insensitive error and penalty factors, construct the objective function, and obtain the optimization problem of support vector regression; The duality theory is used to solve the optimization problem, obtain the support vector coefficients and bias terms, and determine the optimal nonlinear regression model.

4. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 3, characterized in that: The insensitive loss function is: Among them, L ε is the insensitive loss function, y is the true value of the sample, f(x) is the predicted value of SVR, and ε is the insensitive error.

5. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 3, characterized in that: The optimization problem is: Among them, ξ i ,ξ * i is the slack variable, W is the weight vector, T is the transpose, C is the penalty factor, i is the sample index, and n is the number of training samples.

6. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 1, characterized in that: Using genetic algorithms to perform global search and optimization to obtain the optimal GA-SVR model includes: Calculating the fitness function of the degradation prediction model; According to the calculation results, it is determined that a preset maximum number of iterations is met or a change in the objective function is less than a preset threshold, and the optimal GA-SVR model is obtained.

7. The method for predicting sealing material performance degradation based on the GA-SVR model according to claim 6, characterized in that: The fitness function is: Where n is the number of training samples, y i is the actual value of the i-th sample index, is the predicted value of the i-th sample.

Images