Magnetorheological grease performance prediction method and system in complex environment

Abstract

The invention provides a magneto-rheological grease performance prediction method and system in a complex environment, and aims to overcome the defects that the rheological property of a magneto-rheological material is difficult to predict and a constitutive parameterized model is poor in applicability. The method comprises the steps that influence factors of the magneto-rheological grease performance in the complex environment are analyzed and determined, and data information of the influence factors is collected; importing original data and carrying out normalization processing; mapping the sample by using a kernel function; the hyper-parameters are optimized through cross validation; and finally, using experimental data which does not participate in training as a test set, and verifying the reliability of the prediction method. According to the method, based on a support vector regression (SVR) algorithm, information such as material ratio, temperature, environment information and magnetic field intensity is taken as model input, material viscosity and shear stress are taken as model output, and reliability performance prediction of the magnetorheological material under a complex environment and a high shear condition is realized.

Description

Technical Field

[0001] The invention belongs to the field of material performance prediction, and in particular relates to a method and system for predicting magnetorheological grease performance under complex environments. Background Art

[0002] As a controllable intelligent material, the rheological properties of magnetorheological materials are affected by factors such as temperature and magnetic field strength. It is necessary to establish a concise and reliable method for predicting the performance of magnetorheological materials. Most of the existing methods for predicting the performance of magnetorheological materials use various types of parameter models to fit the experimental data and then realize the prediction. Since magnetorheological grease exhibits non-Newtonian fluid properties under steady-state shear, its commonly used constitutive models include Bingham, Herba, Carlo, etc., but they consider fewer parameters. In order to enable the constitutive model to consider more factors in the actual use of the material, many researchers have proposed various modified constitutive models, adding parameters such as temperature and magnetic field strength. Although most of these constitutive models are concise enough to clearly describe the relationship between shear stress and various parameters, they still have defects in performance prediction. Obviously, it is difficult for the constitutive model to predict the performance of materials with different ratios. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for predicting the performance of magnetorheological grease in a complex environment, so as to solve the defects of parameter models in performance prediction.

[0004] The technical solution for achieving the purpose of the present invention is as follows:

[0005] A method for predicting magnetorheological grease performance under complex environment comprises the following steps:

[0006] Step S1. Analyze the factors affecting the performance of magnetorheological grease in a complex environment and collect data information on the factors affecting the performance;

[0007] Step S2. Divide and process the collected data information into data sets; the data sets include test samples and training samples;

[0008] Step S3. Obtaining the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm;

[0009] Step S4. Establish a model with the optimal parameter combination, import the test sample in step S2 into the model for training, and obtain a prediction model for the performance of magnetorheological grease under complex environment;

[0010] Step S5. Predict the test sample through the prediction model, obtain the magnetorheological grease performance results under different input parameter conditions, and verify the prediction results.

[0011] Furthermore, in step S1, factors affecting the performance of magnetorheological grease in a complex environment include material ratio, temperature, environmental information and magnetic field strength.

[0012] Furthermore, data information of influencing factors is collected, including:

[0013] Collection of material ratio data: record during the material preparation process, transmit the data to the PC host computer and store it;

[0014] Shear rate data collection: A commercial rheometer (PhysicaMCR 302, Anton Paar, Austria) was used to measure the rheological properties of magnetorheological materials, record the experimental data of the shear rate of the material under different working conditions, and transmit the data to the PC host computer for storage;

[0015] Temperature data collection: When measuring the rheological properties of materials, the temperature data of the temperature control unit (JULABO F25) is recorded and the data is transmitted to the PC host computer for storage;

[0016] Magnetic field strength data collection: When measuring the rheological properties of materials, the magnetic field strength data of the magnetron accessory (MRD 180) is recorded and the data is transmitted to the PC host computer for storage;

[0017] Furthermore, the specific steps of step S2 include:

[0018] Step S21. Count the data collected in step S1 and save them in an Excel file, select material ratio, temperature, environmental information and magnetic field strength as independent variables, i.e., input parameters; select shear stress as dependent variable, i.e., output parameter;

[0019] Step S22. Arrange the data collected in step S1 into a data set file and import it into Matlab software, use the rowrank function to rearrange it, select 7 / 8 of the data from the data set as training samples, and select 1 / 8 of the data as test samples;

[0020] Step S23. Perform global normalization on the sample data, use the mapminmax function to scale the data to between [-1, 1], and then use the apply function to apply the previously saved normalization parameters to normalize the new data.

[0021] Furthermore, the specific steps of step S3 include:

[0022] Step S31. Based on the prediction target of shear stress, establish the following regression model: y = f(x i )+e,

[0023] In the formula, are sample characteristics, including graphite mass fraction, temperature, magnetic field intensity, shear rate, etc.; is the output data, i.e. the estimated value of shear stress; e is a random variable.

[0024] The 800 data obtained from the above characterization experiment and processing is written as G = {(x i ,y),i=1,2,…,800}. In the process of algorithm implementation, in order to make the data separable, it is necessary to map the original data to a high-dimensional space through a function, and the function to achieve this goal is the kernel function. The present invention implements data mapping based on the Gaussian kernel function, such as:

[0025]

[0026] In the formula, x i 、x j is the eigenvalue, σ 2 Represents the bandwidth of the Gaussian kernel function.

[0027] Introducing Lagrange multipliers, f(x i ) can be written as:

[0028]

[0029] The training process of the model is the process of obtaining the appropriate w and b. During the training process, the SVR algorithm based on the Gaussian kernel function has two key influencing parameters: C and σ. Among them, C represents the penalty coefficient, that is, the tolerance to error. The larger C is, the higher the error requirement is, but it is easy to overfit. σ represents the maximum allowable error and determines the number of supported vectors. The larger σ is, the fewer supported vectors are, which will lead to a decrease in variance but an increase in deviation.

[0030] Step S32: Use the meshgrid function to generate two grid matrices to represent the logarithmic values ​​of the influencing parameters C and σ respectively.

[0031] Step S33. Traverse each pair of C and σ value combinations, generate the corresponding SVR model and calculate the cross-validation error. Use the cmd string to construct the SVR training command.

[0032] Step S34: Use the svrtrain function to train the SVR model and store the cross-validation error to obtain the optimal parameter combination.

[0033] Furthermore, the specific steps of step S4 include:

[0034] Step S41. Construct a parameter string cmd for storing command line parameters, wherein the required parameters include the kernel function type, the SVR type, the cross-validation fold, and the optimal parameter combination obtained in step S3;

[0035] Step S42. Construct a model variable for storing a support vector regression model, call the svrtrain function for training the support vector regression prediction model, the parameter list of the svrtrain function includes the training samples obtained in step S2 and the cmd string, and obtain the trained prediction model.

[0036] Furthermore, the specific steps in step S5 include:

[0037] Step S51. Call the model variable, use the svrpredict function, use the trained model to predict the training set, test set and other data sets, and obtain the predicted value and error. And use denormalization to convert the predicted results to the scale of the original data. Combine the results of the training set and test set with the predicted value and take the absolute value. Write the results of the test set to an Excel file.

[0038] Step S52. Make a comparison chart of the actual value and predicted value of the shear stress of the material, and verify the authenticity and accuracy of the prediction result through the evaluation index of the regression model; the performance index parameters of the evaluation regression model include mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE) and determination coefficient (R 2 ).

[0039] A magnetorheological grease performance prediction system under complex environment, comprising:

[0040] A collection unit is used to analyze and determine the factors affecting the performance of magnetorheological grease in complex environments and collect data information on the factors;

[0041] A data processing unit divides the collected data information into data sets and processes the data;

[0042] An optimal parameter combination determination unit, which obtains the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm;

[0043] A prediction model training unit is used to establish a prediction model with an optimal parameter combination, and the data set in step S2 is imported into the prediction model for training to obtain a prediction model for the performance of magnetorheological grease under a complex environment;

[0044] The output unit predicts the test samples through the prediction model to obtain the magnetorheological grease performance results under different input parameter conditions.

[0045] Compared with the prior art, the present invention has the following advantages: the present invention uses information such as material ratio, temperature, environmental information, magnetic field strength, etc. as model input, and uses material viscosity and shear stress as model output, and based on the support vector regression (SVR) algorithm, realizes reliable performance prediction of magnetorheological materials in complex environments and high shear conditions; the prediction error of the support vector regression function of the present invention is within a given tolerance range (ε-insensitivity interval), and the boundary of the regression function is determined by the support vector, so the prediction accuracy is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Figure 1 It is the technical roadmap of the performance prediction of magnetorheological grease in complex environment based on support vector regression of the present invention.

[0048] Figure 2 The present invention is a flow chart of an algorithm for predicting the performance of magnetorheological grease in complex environments based on support vector regression.

[0049] Figure 3 It is a training flow chart of the support vector regression model of the present invention.

[0050] Figure 4 It is a comparison chart of the shear stress experimental data of the present invention and the performance prediction results of the SVR-based GMRG under different working conditions.

[0051] Figure 5 It is a comparison diagram between the actual value and the predicted value of shear stress of the support vector regression model of the present invention. DETAILED DESCRIPTION

[0052] See attached Figure 1-3 The present invention provides a method for predicting magnetorheological grease performance under complex environment, comprising the following steps:

[0053] Step S1. Analyze the factors affecting the performance of magnetorheological grease in complex environments and collect data information required for the performance prediction model. The specific steps include:

[0054] Step S11: Investigate the performance of magnetorheological grease under complex environment, master the multiple factors affecting its performance, and determine the data required to establish a magnetorheological grease performance prediction model under complex environment, such as material ratio, temperature, magnetic field strength, shear rate and other parameters.

[0055] Step S12. Measure and store the required data. When preparing the magnetorheological grease material, record its material ratio and calculate its mass fraction. Use a rheometer (Physica MCR 302, Anton Paar, Austria) to measure the rheological properties of the magnetorheological material, and record the experimental data of the shear rate and shear stress of the material under different working conditions. While measuring the rheological properties of the material, record the temperature data of the temperature control unit (JULABO F25) and the magnetic field strength data of the magnetron accessory (MRD 180).

[0056] Step S2: divide the collected data into data sets and process the data; the data sets include test samples and training samples. The specific steps include:

[0057] Step S21. Count the data collected in step S1 and save them in an Excel file, select material ratio, temperature, shear rate and magnetic field strength as independent variables, i.e., input parameters; select shear stress as dependent variable, i.e., output parameter;

[0058] Step S22. Arrange the data collected in step S1 into a data set file and import it into Matlab software, use the rowrank function to rearrange it, select 7 / 8 of the data from the data set as training samples, and select 1 / 8 of the data as test samples;

[0059] Step S23. In order to better feedback the mapping relationship between input parameters and output parameters, before creating the model, the sample data is globally normalized. The mapminmax function is used to scale the data to between [-1, 1], and then the apply function is used to apply the previously saved normalization parameters to normalize the new data.

[0060] Step S3. Obtain the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm. The specific steps include:

[0061] Step S31. Based on the prediction target of shear stress, establish the following regression model: y = f(x i )+e,

[0062] In the formula, are sample characteristics, including graphite mass fraction, temperature, magnetic field intensity, shear rate, etc.; is the output data, i.e. the estimated value of shear stress; e is a random variable.

[0063] The 800 data obtained from the above characterization experiment and processing is written as G = {(x i,y),i=1,2,…,800}. In the process of algorithm implementation, in order to make the data separable, it is necessary to map the original data to a high-dimensional space through a function, and the function to achieve this goal is the kernel function. The present invention implements data mapping based on the Gaussian kernel function, such as:

[0064]

[0065] In the formula, x i 、x j is the eigenvalue, σ 2 Represents the bandwidth of the Gaussian kernel function.

[0066] Introducing Lagrange multipliers, f(x i ) can be written as:

[0067]

[0068] The training process of the model is the process of obtaining the appropriate w and b. During the training process, the SVR algorithm based on the Gaussian kernel function has two key influencing parameters: C and σ. Among them, C represents the penalty coefficient, that is, the tolerance to error. The larger C is, the higher the error requirement is, but it is easy to overfit. σ represents the maximum allowable error and determines the number of supported vectors. The larger σ is, the fewer supported vectors are, which will lead to a decrease in variance but an increase in deviation.

[0069] Step S32. Use meshgrid function to generate two grid matrices representing the logarithmic values ​​of the influencing parameters C and σ, respectively, with a range of [-10, 10]. The cross-validation fold is set to 5, the initial optimal parameters are all 0, and the comparison error threshold is set to 10. -4 .

[0070] Step S33. Traverse each pair of C and σ value combinations, generate the corresponding SVR model and calculate the cross-validation error. Use the cmd string to construct the SVR training command, where the parameters include; C and σ, the number of cross-validation folds, the kernel function type, the SVR model and the ε-insensitive interval.

[0071] Step S34. Use the svrtrain function to train the SVR model and store the cross-validation error, update the best parameter combination, and consider the priority of the error and the parameters to finally obtain the best parameter combination.

[0072] Step S4. Establish a model with the optimal parameter combination, import the test sample in step S2 into the model for training, and obtain a prediction model for the performance of magnetorheological grease under complex environment. The specific steps include:

[0073] Step S41. Construct a parameter string cmd for storing command line parameters so as to be subsequently passed to the svrtrain function, where the required parameters include the kernel function type, the SVR type, the number of cross-validation folds, and the optimal parameter combination obtained in step S3;

[0074] Step S42. Construct a model variable for storing a support vector regression model, call the svrtrain function for training the support vector regression prediction model, the parameter list of the svrtrain function includes the training samples obtained in step S2 and the cmd string, and obtain the trained prediction model.

[0075] Step S5. Use the prediction model to predict the test sample, obtain the magnetorheological grease performance results under different input parameter conditions, and verify the prediction results. The specific steps include:

[0076] Step S51. Call the model variable, use the svrpredict function, use the trained model to predict the training set, test set and other data sets, and obtain the predicted value and error. And use denormalization to convert the predicted results to the scale of the original data. Combine the results of the training set and test set with the predicted value and take the absolute value. Write the results of the test set to an Excel file.

[0077] Step S52. Make a comparison chart of the actual value and predicted value of the material shear stress, as shown in the attached figure. Figure 3 , 4 , Figure 3 The corresponding parameters are: (a) GMRG-2, H = 194 kA / m -1 ; (b) GMRG-2, H = 391 kA / m -1 ; (c) GMRG-5, H = 194 kA / m -1 ; (d) GMRG-5, H = 391 kA / m -1 The authenticity and accuracy of the prediction results are verified by the evaluation indicators of the regression model; the performance indicator parameters of the regression model include the root mean square error (RMSE) and the coefficient of determination (R 2 ). It can be summarized into the following formula;

[0078]

[0079] Where: τ EXP represents the test value, τ EST represents the estimated value, n represents the sample size, SSE is the mean value of each group of shear stress test values, SST is the sum of squares of the total sum.

[0080] The performance index parameters for evaluating the regression model are shown in Table 1:

[0081] Table 1

[0082]

[0083] A magnetorheological grease performance prediction system under complex environment, comprising:

[0084] A collection unit is used to analyze and determine the factors affecting the performance of magnetorheological grease in complex environments and collect data information on the factors;

[0085] A data processing unit divides the collected data information into data sets and processes the data;

[0086] An optimal parameter combination determination unit, which obtains the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm;

[0087] A prediction model training unit is used to establish a prediction model with an optimal parameter combination, and the data set in step S2 is imported into the prediction model for training to obtain a prediction model for the performance of magnetorheological grease under a complex environment;

[0088] The output unit predicts the test samples through the prediction model to obtain the magnetorheological grease performance results under different input parameter conditions.

[0089] Aiming at the disadvantages of difficulty in predicting the rheological properties of magnetorheological materials and poor applicability of constitutive parameterized models, the present invention proposes a magnetorheological material property prediction model based on the support vector regression (SVR) algorithm. The material ratio, temperature, environmental information, magnetic field strength and other information are used as model inputs, and the material viscosity and shear stress are used as model outputs to achieve reliable performance prediction of magnetorheological materials under complex environments and high shear conditions.

[0090] The basic principles, main features and advantages of the present invention are shown and described above. It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and therefore, various changes and improvements of the present invention fall within the scope of the present invention as claimed without departing from the spirit and scope of the present invention. The scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting magnetorheological grease performance under complex environments, characterized in that: The following steps are involved: Step S1. Analyze and determine the factors affecting the performance of magnetorheological grease in a complex environment, and collect data information on the factors affecting the performance; Step S2. dividing the collected data information into data sets and performing data processing; Step S3. Obtaining the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm; Step S4. Establish a prediction model with the optimal parameter combination, import the data set in step S2 into the prediction model for training, and obtain a prediction model for the performance of magnetorheological grease under complex environment; Step S5: predict the test sample through the prediction model to obtain the magnetorheological grease performance results under different input parameter conditions.

2. The method for predicting magnetorheological grease performance under complex environment according to claim 1, characterized in that: The influencing factors include material ratio, temperature, shear rate and magnetic field strength.

3. The method for predicting magnetorheological grease performance under complex environment according to claim 1, characterized in that: The data information of the collected influencing factors specifically includes: Record material proportion data during material preparation; Use a commercial rheometer to measure the rheological properties of magnetorheological materials and record the experimental data of the shear rate of the material under different working conditions; When measuring the rheological properties of materials, record the measured temperature data; When measuring the rheological properties of materials, record the magnetic field strength data of the magnetron attachment.

4. The method for predicting magnetorheological grease performance under complex environment according to claim 1, characterized in that: The specific steps of step S2 include: Step S21. Count and save the data collected in step S1, taking material ratio, temperature, shear rate and magnetic field strength as independent variables, i.e., input parameters; taking shear stress as dependent variable, i.e., output parameter; Step S22. Arrange the data collected in step S1 into a data set, rearrange the data using a ranking function, select 7 / 8 of the data from the data set as training samples, and select 1 / 8 of the data as test samples; Step S23: Scale the sample data to between [-1, 1], and then normalize the new data.

5. The method for predicting magnetorheological grease performance under complex environment according to claim 1, characterized in that: The optimal parameter combination of support vector regression hyperparameters is obtained through cross-validation algorithm, including: Step S31. Establishing a support vector regression model based on the prediction target of shear stress; Step S32. Use the meshgrid function to generate two grid matrices, which respectively represent the logarithmic values ​​of the influencing parameters C and σ; Step S33. Traverse each pair of C and σ value combinations, generate the corresponding SVR model and calculate the cross-validation error, and use the cmd string to construct the SVR model training command; Step S34: Use the svrtrain function to train the SVR model and store the cross-validation error to obtain the optimal parameter combination.

6. The method for predicting magnetorheological grease performance under complex environment according to claim 5, characterized in that: The support vector regression model is: y=f(xi)+e, In the formula, are sample characteristics, including material ratio, temperature, shear rate, and magnetic field strength; is the output data, i.e. the estimated value of shear stress; e is a random variable, σ 2 Represents the bandwidth of the Gaussian kernel function, and w, b and αi are coefficients.

7. The method for predicting magnetorheological grease performance under complex environment according to claim 1, characterized in that: The specific steps of step S4 include: Step S41. Construct a parameter string cmd for storing command line parameters, wherein the required parameters include the kernel function type, SVR type, the number of cross-validation folds, and the optimal parameter combination obtained in step S3; Step S42. Construct a model variable for storing a support vector regression model, call the svrtrain function to train and store the support vector regression model, the parameter list of the svrtrain function includes the training samples obtained in step S2 and the character string cmd, and obtain a trained prediction model.

8. A system for predicting magnetorheological grease performance in a complex environment for implementing any of the methods described in claims 1-7, characterized in that: include: A collection unit is used to analyze and determine the factors affecting the performance of magnetorheological grease in complex environments and collect data information on the factors; A data processing unit divides the collected data information into data sets and processes the data; An optimal parameter combination determination unit, which obtains the optimal parameter combination of support vector regression hyperparameters through a cross-validation algorithm; A prediction model training unit is used to establish a prediction model with an optimal parameter combination, and the data set in step S2 is imported into the prediction model for training to obtain a prediction model for the performance of magnetorheological grease under a complex environment; The output unit predicts the test samples through the prediction model to obtain the magnetorheological grease performance results under different input parameter conditions.

Images