Cable accessory silicone rubber mechanical property prediction method, system, equipment and medium

Abstract

The invention provides a cable accessory silicone rubber mechanical property prediction method, system and device and a medium, and belongs to the field of cable accessories, and the method comprises the following steps: collecting near infrared spectrums of a plurality of cable accessory silicone rubber samples; calculating the importance weight of the wavelength of each near infrared spectrum, determining the most relevant features according to the importance weight, selecting random features from other features except the most relevant features in each near infrared spectrum as feature perturbation items, and constructing a plurality of feature subspaces by using the most relevant features and the feature perturbation items; taking the plurality of feature subspaces as input data, taking elongation at break and tensile strength as output data, constructing a plurality of base models based on Gaussian process regression, and endowing the plurality of base models with different weights according to output results of the plurality of base models to obtain an integrated model; and predicting the elongation at break and tensile strength of the silicone rubber sample by using the integrated model. The prediction precision of the mechanical properties of the silicone rubber can be improved.

Description

Technical Field

[0001] The invention belongs to the field of cable accessories, and in particular relates to a method, system, equipment and medium for predicting the mechanical properties of silicone rubber of cable accessories. Background Art

[0002] With the continuous growth of electricity demand, the application scope of cables has expanded rapidly, and the use of wires and cables has become an important trend in the development of power systems. At present, the scale of the cable network has grown rapidly, and the number of cable and accessory suppliers has become increasingly large, resulting in uneven quality. Some suppliers have problems such as inferior products and poor production processes. If the quality of cables and accessories is not properly controlled in the early stage of cable access, allowing defective or unqualified products to mix into the power grid system may become a potential threat to power grid security. Therefore, it is particularly important to strictly control the quality of cables and accessories.

[0003] Silicone rubber has been widely used in the field of cable accessories due to its excellent insulation and mechanical properties. It has become an indispensable key material for ensuring the safe and stable operation of cable systems. Therefore, its performance directly affects the quality of cable accessories. Among the many parameters that characterize the performance of silicone rubber materials for cable accessories, mechanical properties are the most commonly used key indicators. The mechanical properties of insulating materials are not only closely related to the chemical properties and microstructure of the materials, which can sensitively reflect the microscopic changes of the materials, but also directly affect the function and reliability of cable accessories. The relevant standards promulgated by the International Electrotechnical Commission (IEC) also clearly recommend that mechanical properties be used as a key parameter for evaluating the performance of polymer insulating materials. Therefore, it is of great significance to achieve accurate detection of the mechanical properties of silicone rubber for the evaluation of its quality.

[0004] At present, the mechanical property testing method of silicone rubber materials for cable accessories is single and mainly relies on tensile testing. This method usually requires the material to be made into dumbbell-shaped specimens, and then tested on a tensile testing machine to measure its elongation at break and tensile strength. Although this method is currently an international standard test method, it has the following problems: (1) The sample preparation is complicated and the test steps are cumbersome; (2) The results are highly dispersed, and the national standard GB / T2951.11-2008 requires at least 5 specimens to be tested for mechanical properties; (3) Sampling is required, and the test is destructive; (4) The equipment is large and is often carried out in the laboratory, which is not suitable for on-site testing. The above problems have led to the fact that the quality inspection of silicone rubber materials for cable accessories is mostly small-scale sampling, which seriously restricts the realization of large-scale rapid screening and intelligent testing needs.

[0005] Near-infrared spectroscopy analysis technology is a rapid and non-destructive in-situ detection method, which is widely used in many industries such as agriculture, food, petrochemical, and pharmaceutical. Near-infrared spectra are generated by the absorption of light during the overtone and combination frequency energy level transitions of molecular substances, reflecting the microscopic characteristics of substances at the molecular level. It is a commonly used means for material identification, component analysis, and process monitoring. In the detection of polymers, near-infrared spectroscopy analysis technology shows certain advantages. Near-infrared spectra can sensitively capture the chemical and physical information inside materials. By combining chemometric algorithms, the explicit and implicit information in near-infrared spectra can be fully utilized, thereby indirectly reflecting the physical and chemical properties of materials. Therefore, near-infrared spectroscopy technology becomes an ideal choice for the detection of the mechanical properties of silicone rubber. However, in near-infrared spectroscopy analysis, spectral data often presents a high-dimensional feature trend, resulting in a large amount of redundant information and potential interference factors in the dataset. To improve the performance of the model, it is often necessary to select an appropriate feature subset from the high-dimensional space, remove irrelevant or redundant features, thereby reducing the negative impact of interference information on the model, and then improving the accuracy and generalization ability of the model. However, there is a risk of information loss during the feature extraction process, which may lead to the extracted features failing to fully express the original data information, thus making the model prone to falling into local optima and affecting the final prediction accuracy. Summary of the Invention

[0006] In order to overcome the deficiencies of the above-mentioned existing technologies, the present invention provides a method for predicting the mechanical properties of silicone rubber for cable accessories, including the following steps:

[0007] Collect the near-infrared spectra of multiple silicone rubber samples of cable accessories;

[0008] Calculate the importance weights of the wavelengths of multiple near-infrared spectra, determine the features most relevant to the mechanical properties of silicone rubber for cable accessories according to the importance weights, then select random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and use the most relevant features and feature perturbation terms to construct multiple feature subspaces;

[0009] Take multiple feature subspaces as input data, take the elongation at break and tensile strength as output data, construct multiple base models based on Gaussian process regression, and assign different weights to the multiple base models according to the output results of the multiple base models to obtain an ensemble model;

[0010] Collect the near-infrared spectrum of the silicone rubber sample of the cable accessory to be predicted, input it into the ensemble model, and obtain the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

[0011] Preferably, the calculation of the importance weights of the wavelengths of each near-infrared spectrum includes the following steps:

[0012] Calculate the average wavelength of multiple near-infrared spectra, and use the average wavelength as the wavelength of the standard spectrum;

[0013] Subtract the wavelength of each near-infrared spectrum from the wavelength of the standard spectrum to obtain the wavelength of the difference spectrum;

[0014] Calculate the importance weight of the wavelength of each near-infrared spectrum according to the wavelength of the difference spectrum.

[0015] Preferably, randomly select random features from other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, specifically: according to the importance weight, use the random weighted sampling method without replacement to select random features from other features except the most relevant features in each near-infrared spectrum, and use the random features as feature perturbation terms.

[0016] Preferably, before calculating the importance weight of each near-infrared spectrum, it also includes preprocessing each near-infrared spectrum, specifically: performing normalization processing, smoothing processing and baseline correction on each near-infrared spectrum.

[0017] The present invention also provides a mechanical property prediction system for silicone rubber materials of cable accessories, including:

[0018] A data acquisition module for acquiring the near-infrared spectra of multiple silicone rubber samples of cable accessories;

[0019] A feature subspace construction module for calculating the importance weight of the wavelength of multiple near-infrared spectra, determining the features most relevant to the mechanical properties of silicone rubber of cable accessories in the near-infrared spectra according to the importance weight, then randomly selecting random features from other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and using the most relevant features and feature perturbation terms to construct multiple feature subspaces;

[0020] An integrated model construction module for using multiple feature subspaces as input data, using the elongation at break and tensile strength as output data, constructing multiple base models based on Gaussian process regression, and assigning different weights to the multiple base models according to the output results of the multiple base models to obtain an integrated model;

[0021] A mechanical property prediction module for acquiring the near-infrared spectrum of a silicone rubber sample of a cable accessory to be predicted, inputting it into the integrated model, and obtaining the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

[0022] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the mechanical property prediction method of silicone rubber of cable accessories.

[0023] The present invention also provides a computer-readable storage medium storing a computer program, which is adapted to be loaded by a processor to execute the method for predicting the mechanical properties of silicone rubber for cable accessories.

[0024] The method for predicting the mechanical properties of silicone rubber for cable accessories provided by the present invention has the following beneficial effects:

[0025] By calculating the importance weights of the wavelengths of each near-infrared spectrum, the present invention can determine the features most relevant to the mechanical properties of silicone rubber for cable accessories in each near-infrared spectrum, and by introducing a feature perturbation mechanism, a diverse feature subspace with both difference and effectiveness can be constructed, so as to fully explore the potential information of spectral data; by using multiple feature subspaces as input data and elongation at break and tensile strength as output data, multiple base models based on Gaussian process regression are constructed, and different weights are assigned to the multiple base models according to the output results of the multiple base models, an integrated model with excellent performance can be constructed; this integrated model can accurately predict the elongation at break and tensile strength of silicone rubber samples for cable accessories while reducing the risk of information loss, significantly improving the prediction accuracy. Description of the Drawings

[0026] In order to more clearly illustrate the embodiments of the present invention and their design schemes, the accompanying drawings required for the present embodiments will be briefly introduced below. The accompanying drawings in the following description are only partial embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0027] Figure 1 It is a flowchart of the method for predicting the mechanical properties of silicone rubber for cable accessories according to the embodiment of the present invention;

[0028] Figure 2 It is the original near-infrared spectrum of the silicone rubber sample;

[0029] Figure 3 It is the spectrum of the silicone rubber sample after pretreatment;

[0030] Figure 4 It is the result of feature wavelength extraction; among them, Figure 4 (a) of it is the key band for predicting the tensile strength of silicone rubber; Figure 4 (b) of it is the key band for predicting the elongation at break of silicone rubber;

[0031] Figure 5 It is a schematic diagram of a diverse feature subspace;

[0032] Figure 6 It is the prediction result of the mechanical properties of silicone rubber based on the GPR model; among them, Figure 6 (a) of it is the prediction result of the elongation at break;Figure 6 (b) is the prediction result of the tensile strength.

[0033] Figure 7 are the prediction results of the mechanical properties of silicone rubber based on the integrated model; among them, Figure 7 (a) is the prediction result of the elongation at break; Figure 7 (b) is the prediction result of the tensile strength. Detailed implementation manners

[0034] In order to enable those skilled in the art to better understand the technical solutions of the present invention and be able to implement them, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0035] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", etc. are based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the technical solutions of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention.

[0036] In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that unless otherwise clearly specified or limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In the description of the present invention, unless otherwise stated, the meaning of "a plurality" is two or more, and details are not described herein again.

[0037] Embodiment

[0038] The present invention provides a method for predicting the mechanical properties of silicone rubber for cable accessories, specifically as Figure 1 shown, including the following steps:

[0039] Step 1: Collect the near-infrared spectra of multiple silicone rubber samples of cable accessories.

[0040] The present invention prepares samples of silicone rubber materials with different mechanical properties by controlling the vulcanization temperature and time. The first vulcanization temperature is set at 110°C - 135°C, the temperature gradient is 5°C, and the vulcanization time is 20 minutes. The second vulcanization conditions are respectively set at 200°C for 2 hours and 170°C for 1 hour. A total of 12 silicone rubber samples with different properties are prepared, and the sample thickness is 1 mm.

[0041] The present invention uses a NIR Quest type near-infrared spectrometer produced by Ocean Optics to collect the near-infrared spectra of the prepared silicone rubber samples. The wavelength range collected by the near-infrared spectrometer is 900 - 2200 nm, the number of pixels is 512, and the WS-1 diffuse reflection standard plate is used as the standard white board reference. Each spectrum is the average result obtained by scanning 12 times. During collection, each sample is sampled according to the five-point sampling method, and the spectra are collected five times on both the front and back sides to obtain the near-infrared spectra of different silicone rubber materials, as Figure 2 shown.

[0042] To verify the prediction effect of the mechanical properties of the method of the present invention, the present invention prepares silicone rubber samples with different properties under the same controlled processing conditions, and uses the tensile test required by the current national standard as the standard method to detect the elongation at break and tensile strength of the silicone rubber samples, and takes this value as the standard value. By comparing the elongation at break and tensile strength values obtained by detecting the mechanical properties of the same sample using two different methods, the effectiveness of the method of the present invention is judged.

[0043] The specific test method is as follows: The prepared silicone rubber samples with different properties are made into dumbbell-shaped specimens, and a universal testing machine is used for tensile testing. Each sample is ensured to have at least 5 valid tensile test data, and the median value is taken as the final test result. The numbers of different samples and their corresponding elongation at break and tensile strength are shown in Table 1.

[0044] Table 1 Elongation at break and tensile strength of silicone rubber samples

[0045] Sample Number Elongation at Break / % Tensile Strength / Mpa Sample Number Elongation at Break / % Tensile Strength / Mpa 1 1032.12 9.14 7 1008.16 9.35 2 870.18 8.04 8 944.96 8.87 3 943.95 8.61 9 915.92 9.37 4 1024.38 9.36 10 1014.3 9.51 5 927.5 8.61 11 988.81 9.7 6 981.75 9.48 12 972.71 9.92

[0046] Step 2: Calculate the importance weights of the wavelengths of multiple near-infrared spectra, determine the features most relevant to the mechanical properties of the silicone rubber for cable accessories according to the importance weights, and then select random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and use the most relevant features and the feature perturbation terms to construct multiple feature subspaces. The specific process is as follows:

[0047] (1) Near-infrared spectrum preprocessing.

[0048] The near-infrared spectrum data of the silicone rubber samples is subjected to normalization processing, smoothing processing, and baseline correction to reduce the interference of noise signals on the spectrum data. The method of normalization processing is as follows:

[0049]

[0050] Wherein, x i is the original spectral data, μ is the average value of all the original spectral data, σ is the standard deviation of all the original spectral data, and x i * is the spectral data after normalization processing.

[0051] The function of smoothing processing is to filter out high-frequency noise. The principle of smoothing processing is to slide a window on the original data, fit the data within the window to a polynomial by the least squares method, and use the fitted value as the value of the data in the window. The specific calculation method is as follows:

[0052] For each spectral data, fit the data points within the window with a polynomial of degree k - 1 to obtain 2n + 1 equations of degree k - 1. The fitting equation at the j-th characteristic wavelength x j is:

[0053]

[0054] If we denote:

[0055] Y = (y j-n … y j … y j+n ) T ;

[0056]

[0057] A = (a0 … a k-1 ) T ;

[0058] Then we have:

[0059] Y = XA;

[0060] Determine the estimated value of the fitting coefficient matrix through the least squares method Specifically as follows:

[0061]

[0062] Obtain the smoothed data set Specifically as follows:

[0063]

[0064] B = X · (X T · X) -1 · X T ;

[0065] Wherein, is the smoothed dataset, Y is the original dataset, X is the wavelength dataset, and X T is the transpose of X; x j is the j-th characteristic wavelength, and y j is the spectral value corresponding to the j-th characteristic wavelength; A is the polynomial fitting coefficient matrix, is the estimated value of the fitting coefficient matrix; k is the order of the polynomial; n is the window width, and a i is the polynomial fitting coefficient to be solved.

[0066] The function of baseline correction is to remove the influence of instrument background or drift on the signal. The baseline background can be removed by differentiation to improve the spectral resolution. The basic formula is as follows:

[0067]

[0068] In the formula, is the first derivative of the spectral value corresponding to the j-th characteristic wavelength, and x j+d is the spectral value corresponding to the (j + d)-th characteristic wavelength, and x j is the spectral value corresponding to the j-th characteristic wavelength, and d is the window width.

[0069] The preprocessed near-infrared spectrum is as Figure 3 shown.

[0070] (2) Extraction of diverse feature subspaces of near-infrared spectra.

[0071] In near-infrared spectrum analysis, spectral data often exhibit a high-dimensional feature trend, resulting in a large amount of redundant information and potential interference factors in the dataset. To improve the model performance, it is often necessary to select an appropriate feature subset from the high-dimensional space, remove irrelevant or redundant features, thereby reducing the negative impact of interference information on the model, and further improving the accuracy and generalization ability of the model. However, there is a risk of information loss during the feature extraction process, which may lead to the extracted features failing to fully express the original data information, thus making the model prone to falling into local optima and affecting the final prediction effect.

[0072] To overcome this problem, an ensemble method can be used to improve the performance of the model. By combining the prediction results of multiple base models trained with different feature subspaces, the ensemble method can effectively alleviate the overfitting or underfitting problems that may be caused by a single model, thereby improving the overall prediction accuracy and generalization ability. In particular, by training multiple models with different feature subspaces, the ensemble method can introduce diversity and effectively avoid the over-reliance of the model on certain specific features.

[0073] The performance of the integrated model depends to a large extent on the quality of each base model, and the quality of the base model is closely related to the diversity and effectiveness of the feature subspaces it uses. If the feature subspaces lack differences, the base models may tend to be the same, resulting in limited integration effects; while if the feature subspaces are not representative and effective, it may affect the learning ability of the base models, thereby reducing the overall performance of the integrated model. Therefore, ensuring that the feature subspaces have both differences and effectiveness is the key to improving the performance of the integrated model. To address the above problems, a method for integrating differential feature subspaces is proposed. This method constructs diverse feature subspaces by extracting core features (the most relevant features) and introducing a feature perturbation mechanism, ensuring that they have both differences and effectiveness. On this basis, multiple base models are trained and integrated, reducing the risk of information loss while further improving the generalization ability of the model.

[0074] The goal of core feature extraction is to extract key information from the original data as the "core" of each feature subspace to ensure the effectiveness of each subspace. The core feature extraction process is as follows:

[0075] First, calculate the average spectrum of all spectra (the average of the reflectivities corresponding to each wavelength point of all spectra) as the standard spectrum

[0076]

[0077] In the formula, x i (ν) is the reflectivity at wavelength ν in the i-th spectrum, and n is the total number of spectra.

[0078] Subtract each spectrum from the standard spectrum to obtain the difference spectrum The difference spectrum can characterize the spectral change trends of silicone rubber materials with different mechanical properties:

[0079]

[0080] Calculate the importance weights of different spectral features, which can characterize the fluctuation intensity of the spectra with different mechanical properties in different bands of multiple silicone rubber materials. The calculation formula is as follows:

[0081]

[0082] In the formula: w(v) is the importance weight of the feature at wavelength v; n is the total number of spectra; y i is the output value of the i-th spectrum; is the average value of the target outputs.

[0083] In the present invention, the importance index sizes of each spectral feature are calculated for the spectral data of silicone rubber materials with different mechanical properties asFigure 4 as shown, where Figure 4 (a) of which is the key band for predicting the tensile strength of silicone rubber; Figure 4 (b) of which is the key band for predicting the elongation at break of silicone rubber. By peak searching, the wavelengths corresponding to 11 intensity peaks at 905.3nm, 910.2nm, 1176.1nm, 1185.7nm, 1405.9nm, 1510.5nm, 1788.7nm, 1819.2nm, 1849.6nm, 1970.9nm, and 2077.6nm are used as the core features for predicting the mechanical properties of silicone rubber.

[0084] After determining the core features, a random weighted sampling method without replacement is used to select random features from the remaining wavelength features as feature perturbation terms, and combine them with the core features to construct diverse feature subspaces (as shown in Figure 5 ), so as to enhance the diversity and difference between different feature subspaces.

[0085] Step 3: Use multiple feature subspaces as input data, and use the elongation at break and tensile strength as output data to construct multiple base models based on Gaussian process regression. Different weights are assigned to the multiple base models according to the output results of the multiple base models to obtain an ensemble model.

[0086] In the process of integrating and modeling diverse feature subspaces (multiple feature subspaces), the selection of the base learner (base model) is crucial. Considering the non-linear and complex characteristics of spectral data, the Gaussian process regression model (GPR) is selected as the base learner. Gaussian process regression is a non-parametric Bayesian regression method. Its core lies in deriving the distribution of the predicted values of the corresponding samples based on the distribution of the given sample spectral data, and the mathematical expectation of the distribution function is the prediction result. This method does not depend on a specific functional form and can flexibly handle complex non-linear relationships according to the data distribution. The steps of Gaussian process regression modeling are as follows.

[0087] Assume that the predicted target value y consists of a Gaussian distribution function f(x) and noise ε, and satisfies the following formula:

[0088] y = f(x) + ε;

[0089] f(x) ∼ GP(m(x), k(x, x'));

[0090]

[0091] In the formula, the Gaussian distribution function f(x) is determined by its expectation m(x) and variance k(x, x′). The specific form of k(x, x′) is the kernel function. In the present invention, a quadratic rational kernel is selected; σ is the kernel amplitude, α is the smoothing parameter, and l is the length scale; σ n2 is the variance of the noise distribution function, and the distribution of y also conforms to the Gaussian process. This distribution function can be obtained by solving f(x) and ε:

[0092]

[0093] where δ(x, x′) is the Kronecker function. When x = x′, δ(x, x′) = 1; otherwise, δ(x, x′) = 0. When the input test data is x * , the joint distribution of the training set and the predicted target value of the sample to be measured can be calculated by the following formula:

[0094]

[0095] where K is the variance matrix, and y * is the distribution function of the predicted target value. By solving the mathematical expectation of the target value, the prediction result of the sample to be measured can be obtained.

[0096] Based on the Gaussian process regression model, this invention constructs an ensemble model. Multiple feature subspaces are used as input data, and the elongation at break and tensile strength are used as output data to construct a base model based on Gaussian process regression. Finally, the prediction results are integrated to obtain the final predicted values of the elongation at break and tensile strength. The prediction results are integrated in a weighted average manner, and the calculation method is as follows.

[0097] First, different weights are assigned to the prediction results of each model. The weights are dynamically adjusted according to the model performance, and the calculation formula is as follows:

[0098]

[0099]

[0100] where: w j is the weight of the jth model, L j is the output result of the jth model, and L is the output result of the final ensemble model.

[0101] Step 4: Collect the near-infrared spectrum of the silicone rubber sample of the cable accessory to be predicted, and input it into the ensemble model to obtain the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

[0102] The preprocessed near-infrared spectrum data of the silicone rubber material is randomly divided into a training set and a test set at a ratio of 7:3. The training set is used to train the ensemble model, and the test set is used to evaluate the prediction effect of the ensemble model. The near-infrared spectrum is used to predict the elongation at break and tensile strength of the silicone rubber respectively. The root mean square error RMSE, mean absolute error MAE, and determination coefficient R 2As an index for evaluating the prediction performance of the model, the calculation method is as follows.

[0103]

[0104] The prediction results of the Gaussian process regression model are as shown in Figure 6 and Table 2, where Figure 6 (a) of Figure 6 is the prediction result of the elongation at break; Figure 7 (b) of Figure 7 is the prediction result of the tensile strength; the prediction results of the integrated model proposed by the present invention are as shown in Figure 7 and Table 3, where Figure 7 (a) of Figure 7 is the prediction result of the elongation at break; Figure 6 (b) of Figure 7 and Figure 6 The true values of the ordinates in Figure 7 are the results of the tensile test, and the predicted values of the abscissas are the prediction results of the model. The closer the scatter points in the figure are to the diagonal line, the better the prediction effect of the model; the prediction indexes in Table 2 and Table 3 quantify the difference degree between the tensile test values and the model prediction values. It can be seen from Figure 6 , Figure 7 , Table 2 and Table 3 that the prediction model based on Gaussian process regression has shown good prediction results, and the determination coefficients R 2 are all above 0.87, but the performance of the integrated model is better than that of a single Gaussian process regression model. The root mean square error RMSE, mean absolute error MAE, and determination coefficient R 2 of the elongation at break prediction of the integrated model are 13.7841, 11.0982, and 0.9102 respectively, and the root mean square error RMSE, mean absolute error MAE, and determination coefficient R 2 of the tensile strength prediction are 0.1234, 0.0903, and 0.9528 respectively. The prediction errors of the elongation at break and the tensile strength are both less than 10%. This result not only shows that the integrated model proposed by the present invention has excellent prediction performance, but also proves that the mechanical properties of the silicone rubber material for cable accessories can be accurately evaluated by using near-infrared spectroscopy.

[0105] Table 2 Prediction Results of Mechanical Properties of Silicone Rubber (GPR Model)

[0106] Prediction Index Root Mean Square Error RMSE Mean Absolute Error MAE <![CDATA[Coefficient of determination R 2 > Elongation at Break / % 18.7458 10.4693 0.8746 Tensile Strength / MPa 0.1437 0.0938 0.8763

[0107] Table 3 Prediction Results of Mechanical Properties of Silicone Rubber (Integrated Model)

[0108] Prediction Index Root Mean Square Error RMSE Mean Absolute Error MAE Coefficient of Determination R2 Elongation at Break / % 13.7841 11.0982 0.9102 Tensile Strength / MPa 0.1234 0.0903 0.9528

[0109] The present invention also provides a system for predicting the mechanical properties of a silicone rubber material for cable accessories, including:

[0110] A data acquisition module for acquiring the near-infrared spectra of a plurality of silicone rubber samples of cable accessories;

[0111] A feature subspace construction module, which is used to calculate the importance weights of the wavelengths of each near-infrared spectrum, determine the features most relevant to the mechanical properties of the silicone rubber of the cable accessory according to the importance weights, then select random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and use the core features and feature perturbation terms of each near-infrared spectrum to construct multiple feature subspaces;

[0112] An integrated model construction module, which is used to take multiple feature subspaces as input data, take the elongation at break and tensile strength as output data, construct multiple base models based on Gaussian process regression, and assign different weights to the multiple base models according to the output results of the multiple base models to obtain an integrated model;

[0113] A mechanical property prediction module, which is used to collect the near-infrared spectrum of the silicone rubber sample of the cable accessory to be predicted, input it into the integrated model, and obtain the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

[0114] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for predicting the mechanical properties of the silicone rubber of the cable accessory.

[0115] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute the method for predicting the mechanical properties of the silicone rubber of the cable accessory.

[0116] The above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of technical solutions that can be obviously obtained by those skilled in the art within the technical scope disclosed by the present invention all belong to the protection scope of the present invention.

Claims

1. A method for predicting the mechanical properties of silicone rubber for cable accessories, characterized in that, It includes the following steps: Collect the near-infrared spectra of multiple silicone rubber samples of cable accessories; Calculate the importance weights of the wavelengths of multiple near-infrared spectra, determine the features most relevant to the mechanical properties of the silicone rubber of cable accessories according to the importance weights, then select random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and use the most relevant features and the feature perturbation terms to construct multiple feature subspaces; Take multiple feature subspaces as input data, take the elongation at break and tensile strength as output data, construct multiple base models based on Gaussian process regression, and assign different weights to the multiple base models according to the output results of the multiple base models to obtain an ensemble model; Collect the near-infrared spectrum of the silicone rubber sample of the cable accessory to be predicted, input it into the ensemble model, and obtain the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

2. The method for predicting the mechanical properties of silicone rubber for cable accessories according to claim 1, wherein The calculation of the importance weights of the wavelengths of each near-infrared spectrum includes the following steps: Calculate the average wavelength of multiple near-infrared spectra, and take the average wavelength as the wavelength of the standard spectrum; Subtract the wavelength of each near-infrared spectrum from the wavelength of the standard spectrum to obtain the wavelength of the difference spectrum; Calculate the importance weights of the wavelengths of each near-infrared spectrum according to the wavelength of the difference spectrum.

3. The method for predicting the mechanical properties of silicone rubber for cable accessories according to claim 1, characterized in that The step of then selecting random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms is specifically: according to the importance weights, use the random weighted sampling method without replacement to select random features from the other features except the most relevant features in each near-infrared spectrum, and take the random features as feature perturbation terms.

4. The method for predicting the mechanical properties of silicone rubber for cable accessories according to claim 1, wherein Before calculating the importance weights of each near-infrared spectrum, it also includes preprocessing each near-infrared spectrum, specifically: performing normalization processing, smoothing processing and baseline correction on each near-infrared spectrum.

5. A silicone rubber mechanical property prediction system for cable accessories, characterized in that, It includes: A data acquisition module for collecting the near-infrared spectra of multiple silicone rubber samples of cable accessories; A feature subspace construction module for calculating the importance weights of the wavelengths of multiple near-infrared spectra, determining the features most relevant to the mechanical properties of the silicone rubber of cable accessories according to the importance weights, then selecting random features from the other features except the most relevant features in each near-infrared spectrum as feature perturbation terms, and using the most relevant features and the feature perturbation terms to construct multiple feature subspaces; An ensemble model construction module for taking multiple feature subspaces as input data, taking the elongation at break and tensile strength as output data, constructing multiple base models based on Gaussian process regression, and assigning different weights to the multiple base models according to the output results of the multiple base models to obtain an ensemble model; A mechanical property prediction module for collecting the near-infrared spectrum of the silicone rubber sample of the cable accessory to be predicted, inputting it into the ensemble model, and obtaining the elongation at break and tensile strength of the silicone rubber sample of the cable accessory to be predicted.

6. A computer device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for predicting the mechanical properties of the silicone rubber of cable accessories according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the method for predicting the mechanical properties of silicone rubber for cable accessories according to any one of claims 1-4.

Images