WO2022052333A1

WO2022052333A1 - Polymer material service life prediction method based on environmental big data and machine learning

Info

Publication number: WO2022052333A1
Application number: PCT/CN2020/133112
Authority: WO
Inventors: 覃家祥; 李淮; 陶友季; 时宇; 张晓东
Original assignee: 中国电器科学研究院股份有限公司
Priority date: 2020-09-08
Filing date: 2020-12-01
Publication date: 2022-03-17
Also published as: CN112331281A; KR102613725B1; KR20230092889A; CN112331281B

Abstract

A polymer material service life prediction method based on environmental big data and machine learning. According to the method, natural environment feature data is extracted and used as feature parameters; the feature parameters undergo dimensionality reduction and noise reduction using principal component analysis algorithms; the processed result is used as an input parameter, the aging degree of the corresponding material is used as an input parameter, and the test data in some regions is used as a training dataset; machine learning on the relationship between the environment and the performance is performed using Python software; and a service life prediction model is constructed to predict the service life of the polymer material in different regions in the next step. The method is convenient, fast and highly accurate, can effectively reduce test workload, and can be used for improving the weather resistance of the material and designing the weather resistance performance of the product.

Description

Service life prediction method of polymer materials based on environmental big data and machine learning

technical field

The invention belongs to the technical field of service life prediction of polymer materials, in particular to a service life prediction method of polymer materials based on environmental big data and machine learning.

Background technique

Due to a series of excellent properties and high cost performance, polymer materials have been widely used in equipment products in the fields of automobiles, electronics, electrical appliances, construction, packaging, chemical engineering, and biological engineering. In the process of processing, storage and service, the polymer materials of equipment products will gradually lose their luster and color under the combined action of various internal and external factors and various internal and external factors. , yellowing, cracking, peeling, embrittlement, and the decline of various physical and chemical properties, and finally lead to the loss of performance. Compared with metal materials, polymer materials are more prone to aging, and their service life determines the service life of equipment. Therefore, the service life prediction of polymer materials has always been a concern of the industry.

At present, the existing service life prediction methods of polymer materials are as follows:

(1) Linear relationship method. Under different temperature conditions, when the material property P reaches a critical value, the logarithm of the aging time t has a linear relationship with the inverse of the aging temperature T. The linear relationship method is derived from the fact that the change of material properties P with the aging time t obeys the first-order reaction law. The aging reaction rate constant K and aging temperature T obey the Arrhenius equation. Therefore, the calculation formula of the linear relationship method is as follows:

f(P)=Kt=Ae- ^E/RT (Formula 1)

In the formula, K is the aging reaction rate constant, t is the time, A is the pre-exponential factor, E is the reaction activation energy, T is the temperature, and R is the gas constant. However, this method has a serious disadvantage that there may be different aging mechanisms for aging at different temperatures, resulting in different aging rate constants; at the same time, this method also ignores the influence of environmental humidity, which greatly reduces the accuracy of the test results. accuracy.

(2) Variable conversion method. The variable reduction method is essentially a method of drawing. According to the equivalence principle of time and temperature, the test data under high temperature conditions are converted into data under lower temperature conditions. There is a corresponding relationship between temperature and time as follows:

P(T ₁ ,t)=P(T ₂ ,t/α _T ) (Equation 2)

In the formula, P is the performance, T is the temperature, t is the time, and the α _T transformation factor. The limitation of this method is that the aging mechanism and reaction rate at different temperatures are set to remain unchanged, and other factors do not affect the aging reaction or have little effect, while the polymer materials in the actual service state are often gradually under the synergistic effect of the comprehensive environment. Aging, which can lead to a large difference between the predicted results and the actual service life.

(3) Mathematical model method. By fitting and parameterizing the effects of temperature, humidity and irradiation on the aging of most polymer materials, a mathematical relationship model based on various environmental factors and material service life is obtained.

In the formula: f _A is the acceleration factor, T _f is the acceleration factor of the material for every 10℃ rise in temperature, determined according to the type of material, x is the effective solar radiation factor, determined according to the type of material, y is the effective relative humidity factor, Determined according to the type of material, I _r1 is the total annual solar ultraviolet radiation in the sun-tracking concentrating accelerated aging test, in megajoules per square meter (MJ/m ² ), and RH ₁ is the sun-tracking concentrating accelerated aging test The annual average relative humidity in , T ₁ is the annual average temperature in the sun-tracking concentrating accelerated aging test, in degrees Celsius (°C), I _rA is the total annual solar ultraviolet radiation in area A, in megajoules per square m (MJ/m ² ), RH _A is the annual average relative humidity in area A, and T _A is the annual average temperature in area A, in degrees Celsius (°C). The limitation of this method is that the parameterized fitting of the influence factors of humidity and irradiation on materials is based on the average value of many kinds of materials, and there are great differences in the influence of light irradiation and humidity on different materials. This also leads to lower accuracy of model predictions.

Therefore, the above methods have various defects, and it is necessary to develop a new method for predicting the service life of polymer materials.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention is based on the aging test in the actual environment, forms environmental big data through real-time monitoring of environmental factors, and mines and builds the relationship between environmental big data and material performance changes based on machine learning algorithms, so as to mine the impact of various environmental factors on polymers. Based on the actual influence factors of materials, a method for predicting the service life of polymer materials with higher accuracy and more universality has been developed.

The above technical problems to be solved by the present invention can be achieved by the following technical solutions: a method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms, comprising the following steps:

(1) Select polymer materials, carry out aging tests in different regions, obtain the changes in performance parameters of the polymer materials during the test period during the aging process, and use the performance parameter changes as life evaluation indicators;

(2) collecting environmental data in the corresponding experimental period in step (1), including temperature, humidity and irradiation;

(3) Extract the characteristic data in the environmental data obtained in step (2) as characteristic parameters, and use the principal component analysis algorithm to reduce the dimension and noise of the characteristic parameter data, wherein the characteristic data is the cumulative sum of time in different environmental data stages, which is defined as cumulative damage time;

(4) Grouping the cumulative damage time-material property changes in different regions, using some regions as the training set for the construction of the life prediction model, and using the remaining regions as the test set for the verification of the life prediction model;

(5) Taking the cumulative damage time of the training set as the input parameter, and the material property change of the training set as the output parameter, using Python software to build a machine learning algorithm to train the environmental big data life prediction model to form a life prediction model;

(6) Using the cumulative damage time of the test set as an input parameter, predict the material performance changes under different cumulative damage times, and obtain its service life. At the same time, calculate the square value R ² of the correlation coefficient with the experimental value, when R ² ≥95%, the prediction results are credible.

Therefore, the method of the present invention obtains the performance data under different aging degrees through the natural aging of the polymer material in the test station, and simultaneously records the environmental data such as temperature, humidity, and irradiation of the test station during the natural aging process. By extracting the characteristic data of the natural environment (such as the accumulation time of 35-36℃ in half a year, the accumulation time of high temperature and humidity, the accumulation time of irradiation intensity greater ^than 1000W/m2, etc.) as characteristic parameters, the principal component analysis (PCA algorithm) is used. The algorithm performs dimension reduction and noise reduction processing on the characteristic parameter data, takes the processed results as input parameters, and the aging degree of the corresponding material (such as material yellowing index, tensile strength, impact strength, etc.) as output parameters. The test data of the station is used as a training set, and Python software is used to perform machine learning on the relationship between environment and performance changes, and build a life prediction model for the next step in the prediction of the service life of polymer materials in different regions.

Among the above methods for predicting the service life of polymer materials based on environmental big data and machine learning algorithms:

Preferably, the polymer material described in step (1) is a composite material of one or more of polystyrene, polycarbonate, polyethylene and polypropylene.

Preferably, the different regions described in step (1) include Qionghai, Sanya, Guangzhou in China and Jeddah in Saudi Arabia, Sanya in France and Chennai in India abroad.

Preferably, the aging test in step (1) is a natural aging test, a natural accelerated aging test or an artificial accelerated aging test.

Preferably, the performance parameters of the polymer material in step (1) include optical properties, mechanical properties and thermal properties.

Preferably, the performance parameters of the polymer material in step (1) are one or more of yellow index, transparency, tensile strength, melting temperature, glass transition temperature and initial decomposition temperature.

In steps (1) to (2), the experimental period, sampling interval and environmental data recording time can be adjusted according to the actual experimental process, and the sampling interval and environmental data recording time can be reasonably selected according to the length of the experimental period. The sampling interval and the environmental data recording time can be extended accordingly. When the experimental period is short, the sampling interval and the environmental data recording time can be shortened, mainly because enough representative data can be obtained in the corresponding experimental period.

Preferably, the experimental period in steps (1) to (2) is 1 to 5 years, the sampling interval for changes in performance parameters of the polymer material is 1 to 3 months, and the environmental data is recorded every 1 to 10 hours.

Preferably, the acquisition of the environmental data in step (2) can be obtained through actual measurement at the test site, or through the information published on the website to obtain local climate and environmental data such as temperature, humidity, and irradiation data.

The characteristic data in step (3) includes the cumulative damage time of a single factor and the cumulative damage time of a multi-factor synergy, wherein the single factor includes temperature, humidity or irradiation, and the multi-factor includes two or three of temperature, humidity and irradiation. .

Preferably, the acquisition process of the accumulated damage time in step (3) includes: performing statistical analysis on the temperature, humidity and irradiation data through Python software, and acquiring the performance change data of the polymer material and the corresponding temperature, humidity and irradiation data during the experimental period The cumulative sum of time under the conditions of high temperature, high humidity and high temperature and high irradiation.

The performance change data of polymer materials during the experiment period and the corresponding cumulative sum of time under temperature, humidity and irradiation are listed as follows: temperature cumulative damage time (such as the cumulative time within a month of greater than 30 °C), humidity cumulative damage time (if greater than Accumulated time at 80% humidity), cumulative damage time of irradiation (such as cumulative time of irradiation intensity greater ^than 1000W/m2), cumulative damage time under high temperature and high humidity (such as temperature ≥ 30 ℃, humidity ≥ 80%), high temperature and high irradiation Accumulated damage time (such as temperature≥30℃, irradiation intensity≥850W/m ² ), etc.

For example, in step (3), the characteristic data of the natural environment (such as the accumulated time of high temperature and high humidity, the accumulated time of irradiation intensity greater ^than 1000W/m2, etc.) are extracted as characteristic parameters, and the principal component analysis (PCA algorithm) algorithm is used to analyze the characteristics. The parameter data is subjected to dimensionality reduction and noise reduction.

Further, as a preferred embodiment of the present invention:

When the cumulative damage time of temperature is selected as the characteristic parameter, the temperature can be divided into the interval of ≤0℃, 0～10℃, 10～20℃, 20～30℃ and higher than 30℃, and the accumulated temperature in the above temperature interval can be calculated respectively damage time.

When selecting the cumulative damage time of humidity as a characteristic parameter, the humidity can be divided into ≤40%, humidity>40% and ≤80%, and humidity>80% for interval division, and the cumulative damage time in the above humidity interval is calculated respectively.

When the cumulative damage time of irradiation is selected as the characteristic parameter, the irradiation can be divided into: irradiance range≤30W/m ^{2 , irradiance range 30-100W/m 2} ^, irradiance range 100-300W/m ² , The irradiance range is 300-500W/m ² , the irradiance range is 500-700W/m ² , the irradiance range is 700-1000W/m ² and the irradiance range is >1000W/m ² . According to the cumulative damage time in the interval.

When selecting the cumulative damage time of high temperature and high humidity as the characteristic parameter, select the cumulative damage time when the temperature is greater than 30°C and the humidity is greater than 80%.

When the cumulative damage time of high temperature and high irradiation is selected as the characteristic parameter, the cumulative damage time is selected when the temperature is greater than 30℃ and the irradiance range is greater than 850W/m ² .

Take one or more of temperature, humidity and irradiation as characteristic data, first perform dimension reduction and denoising processing, and then use them as input parameters to correspond to the aging degree of the material (such as the yellowing index of the material, tensile strength, impact strength, melting temperature, glass transition temperature, initial decomposition temperature, etc.) as the output parameters, using the test data of some regions (such as domestic test stations) as the training set, using Python software to perform machine learning on the relationship between environment and performance changes, and build life expectancy The prediction model is used to predict the service life of polymer materials in different regions (such as foreign regions) in the next step.

As one of the preferred implementations, in step (4), domestic Qionghai, Sanya, and Guangzhou are used as the training set, and foreign countries such as Jeddah, Saudi Arabia, Chennai, India, and Sanya, France are used as the test set.

Preferably, the machine learning algorithm in step (5) is one or more of neural network, support vector machine, random forest, regression analysis, deep learning and XGboost.

Further, as a preferred embodiment of the present invention, the method for predicting the service life of polymer materials based on machine learning provided by the present invention includes the following steps:

(1) Select any polymer material as the test object to carry out the natural aging test, the material sample has no defects and high consistency; at the same time, set the relevant properties (such as yellow index, transparency, tensile strength, melting temperature, glass transition temperature, initial decomposition temperature, etc.) as the life evaluation index, and test the initial performance of the sample to be investigated before the test is carried out;

(2) Carry out natural aging tests of polymer materials in natural environment test stations such as Qionghai, Sanya and Guangzhou in China, and Jeddah in Saudi Arabia, Sanya in France or Chennai in India. Tensile strength, melting temperature, glass transition temperature or initial decomposition temperature, etc. At the same time, obtain environmental data recorded every 1h, including temperature, humidity, irradiation, etc.;

(3) Statistical data such as temperature, humidity, and irradiation are carried out through Python software to obtain material performance changes in each time period and the corresponding cumulative damage time under temperature and humidity; at the same time, due to the aging mechanism of high temperature, high humidity and high irradiation It has the effect of accelerating the aging of polymer materials. The cumulative damage time under high temperature, high humidity and high temperature and high irradiation conditions is extracted respectively. For the obtained data set, the PCA (Principal Component Analysis) algorithm is used to reduce the dimension and denoise the data set. , to complete the characteristic processing of environmental big data;

(4) Grouping based on the characterized environmental data and the material property change data after the test, the environmental data and material property change data of the domestic Qionghai, Sanya, Guangzhou and other test stations are used as the training set for the construction of machine learning models; The test data and material property change data of foreign test stations such as Jeddah, Saudi Arabia, Chennai, India, and Sanarea, France are used as test sets for subsequent model verification;

(5) Based on Python software, the machine learning algorithm is used to train the life prediction model of the training set to construct the prediction model. The input parameters of the training set are mainly the cumulative damage time of temperature, the cumulative damage time of humidity, high temperature and high humidity, high temperature and high irradiation, etc. Acting cumulative damage time;

(6) Using the characteristically processed foreign environmental cumulative damage time data as the input layer, predict its service performance changes under different cumulative damage times, and obtain its service life, and at the same time, calculate the square value of its correlation coefficient with the experimental value (R ² ), when R ² ≥ 95%, the prediction result is credible.

In this preferred embodiment, the polymer material samples are used as the test objects, and natural aging tests are carried out in natural environment test stations such as Qionghai, Sanya, Guangzhou in China, and Jeddah in Saudi Arabia, Chennai in India, and Sanare in France. The climate and environmental data, including temperature, humidity, irradiation, etc., during the entire test period of each site at home and abroad are processed based on Python to form and characterize the accumulated damage time data. The feature data set is subjected to dimensionality reduction and denoising processing; then, the cumulative damage time of the domestic environment - the performance change of polymer materials is used as the training set, and the model is trained by machine learning algorithm, and a service life model is formed, which is used to predict the damage in different regions abroad. Performance change law and service life. This method can predict the performance change law and service life of polymer materials of equipment products used in foreign countries through the test data of domestic test sites and foreign environmental data, which can effectively reduce the test cost, shorten the product development cycle, and improve the weather resistance quality of products. , has high application value.

Compared with the prior art, the present invention has the following advantages:

(1) The method of the present invention is based on the aging test in the actual environment, forms environmental big data through real-time monitoring of environmental factors, and mines and constructs its relationship with material performance changes based on machine learning algorithms, so as to mine the actual impact of various factors on polymer materials. factor, and developed a more accurate service life prediction method for polymer materials;

(2) The method of the present invention is based on the characteristic processing of environmental big data based on machine learning, fully obtains the influence of environmental factors on the aging of polymer materials, and realizes the prediction of service life of polymer materials in different regions; at the same time, the method has stronger generalization The adaptability can be extended to the life prediction of different polymer materials and different performance indicators, so as to guide the weather resistance design of equipment products, provide equipment product quality, and serve China's equipment products "going out" and quality power strategy;

(3) The present invention uses environmental big data for the service life prediction of polymer materials for the first time, and fully excavates the impact information of the environment on the material aging, so that the accuracy of the model prediction results is higher;

(4) The method of the present invention can predict the performance change law and service life of the polymer materials of equipment products used abroad through the test data of domestic test sites and foreign environmental data, which can effectively reduce the test cost, shorten the product development cycle, and improve the weather resistance of the product. Sexual quality and other advantages, with high application value;

(5) The method of the present invention has the advantages of convenience, speed and high accuracy, can effectively reduce the test workload, and can be used to guide the improvement of the weather resistance of materials and the design of the weather resistance of products.

Description of drawings

1 is a flowchart of a method for predicting the service life of a polymer material based on neural network big data provided in Embodiments 1-3 of the present invention;

Fig. 2 predicts the change of material properties in Jeddah, Saudi Arabia by domestic test data provided in Example 1 of the present invention;

Fig. 3 is the domestic test data provided in the embodiment of the present invention 2 predicts the material property change situation of France Sanares area;

Fig. 4 is the domestic test data provided in Example 3 of the present invention to predict the change of material properties in Chennai, India.

detailed description

Example 1

As shown in FIG. 1 , the method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms provided in this embodiment includes the following steps:

The natural aging test of polystyrene swatch samples was carried out in Qionghai, Sanya, Guangzhou and foreign countries, such as Qionghai, Sanya, Guangzhou, and Jeddah, Saudi Arabia. The size of the swatch is 50×80×4mm. The sample surface is defect-free and has high transparency. The yellow index was set as the life evaluation index, and the difference between the yellow index and the initial value was 50 as the end of life, and the initial yellow index of the sample was tested before the test was carried out.

During the natural aging process of polystyrene swatch samples, sampling tests were carried out every other month to analyze their natural environment in January, February, 3, 4, 5, 6, 7, 8, 9, 10, 11, and 12. The corresponding yellow index below (shown in Figure 2).

At the same time, the temperature, humidity and irradiation data recorded in the test area every 1 h were obtained through real-time monitoring.

The obtained temperature, humidity, irradiation and other data are counted through Python software, and the material performance changes in each time period in the test area and the corresponding accumulated temperature damage time (for example, the accumulated time within one month of greater than 30 °C, see the following table) 1), the cumulative damage time of humidity (such as the cumulative time of more than 80% humidity, see Table 1 below), the cumulative damage time of irradiation (such as the cumulative time of irradiation intensity greater ^than 1000W/m2), the cumulative damage time of high temperature and high humidity (Temperature≥30℃, Humidity≥80%), cumulative damage time under high temperature and high irradiation (temperature≥30℃, irradiation intensity≥850W/m ² ), build a mapping relationship with the polystyrene yellow index after the corresponding aging time, Complete the characterization of environmental big data.

Table 1 Statistical analysis of cumulative damage time of environmental data in Jeddah in one year

Based on the characteristic processing of the accumulated environmental damage time and the material property change data set after the corresponding aging time, the data set is grouped, and the test data and material property change data of the domestic Qionghai, Sanya, Guangzhou and other test stations are used as the training set for the construction of machine learning models. ; At the same time, the test data and material property change data of foreign test stations such as Jeddah, Saudi Arabia are used as the test set for subsequent model verification.

Machine learning is performed on the training set through Python software, and the algorithm used is the BP neural network to build the prediction model; the parameters are set as follows: the initial weight is set to 0; the training algorithm uses "trainlm"; the number of neurons is set to 8; The ratio of training set and validation set is 0.85:0.15, and the learning rate is set to 0.1; the maximum number of failure iteration steps of validation set data is 26; other parameters use the system default parameters.

(7) Using the characterized accumulated damage time data of foreign environments as the input layer, the service performance changes under different accumulated damage times are predicted, so as to obtain its service life. as shown in picture 2. Entering the yellow index at the end of the PS life as 50 in the prediction model, the time corresponding to the yellow index 50 is 9.8 months, that is, the service life of PS in Jeddah, Saudi Arabia is about 10 months, and the correlation coefficient between the PS and the experimental value is calculated. The squared value R ² . R ² =96%, more than 95%, the prediction result is accurate.

Example 2

As shown in Figure 1, the method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms provided by this embodiment includes the following steps:

The natural aging test of polycarbonate swatch samples was carried out in Qionghai, Sanya, Guangzhou and Sanya, France. The sample surface is defect-free and has high transparency. Set the color difference as the life evaluation index, the difference between the color difference and the initial value of 35 as the end point of life, and test the initial color difference of the sample before carrying out the test.

During the natural aging process of polycarbonate swatch samples, sampling tests were carried out every other month to analyze their natural environment in January, February, 3, 4, 5, 6, 7, 8, 9, 10, 11 and 12 respectively. The corresponding chromatic aberration below (Figure 3).

The temperature, humidity and irradiation data recorded every 1 h in the test area were obtained through real-time monitoring, as shown in Table 2 below.

Table 2 Statistical analysis of accumulated damage time of environmental data in Sanares, France in one year

The obtained temperature, humidity, irradiation and other data are counted through Python software, and the material performance changes in each time period in the test area and the corresponding accumulated temperature damage time (such as the accumulated time within a month of greater than 30°C), and the accumulated humidity are obtained. Damage time (such as cumulative time greater than 80% humidity), cumulative damage time of irradiation (such as cumulative time of irradiation intensity greater ^than 1000W/m2), cumulative damage time under high temperature and high humidity (temperature ≥ 30 ℃, humidity ≥ 80% ), the accumulated damage time under high temperature and high irradiation (temperature≥30℃, irradiation intensity≥850W/m ² ), and build a mapping relationship with the polycarbonate color difference after the corresponding aging time to complete the characterization of environmental big data.

Based on the characteristic processing of the accumulated environmental damage time and the material property change data set after the corresponding aging time, the data set is grouped, and the test data and material property change data of the domestic Qionghai, Sanya, Guangzhou and other test stations are used as the training set for the construction of machine learning models. ; At the same time, the test data and material property change data of foreign test stations such as Saarre in France are used as the test set for subsequent model verification.

(7) Using the characterized accumulated damage time data of foreign environments as the input layer, predict the service performance changes under different accumulated damage times, and obtain its service life. As shown in Figure 3, inputting the color difference at the end of the PC life as 35 in the prediction model, the time corresponding to the color difference 35 is 8.6 months, that is, the service life of polycarbonate in Saare, France is 8.5 months. The squared value R ² of the correlation coefficient of the experimental value. R ² =97%, more than 95%, the prediction result is accurate.

Example 3

The natural aging test of high-density polyethylene dumbbell-shaped samples was carried out in Qionghai, Sanya, Guangzhou and Chennai, India. The sample is intact and the surface is free of defects and scratches. The tensile strength is set as the life evaluation index, the tensile strength is 30% of the initial value as the end of life, and the initial tensile strength of the sample is tested before the test is carried out.

During the natural aging process of the high-density polyethylene dumbbell-shaped samples, sampling tests were carried out every other month. corresponding tensile strength. At the same time, the temperature, humidity and irradiation data recorded in the test area every 1 h were obtained through the open-source meteorological website. The results are shown in Table 3.

The obtained temperature, humidity, irradiation and other data are counted through Python software, and the material performance changes in each time period in the test area and the corresponding accumulated temperature damage time (such as the accumulated time within a month of greater than 30°C), and the accumulated humidity are obtained. Damage time (such as cumulative time greater than 80% humidity), cumulative damage time of irradiation (such as cumulative time of irradiation intensity greater ^than 1000W/m2), cumulative damage time under high temperature and high humidity (temperature ≥ 30 ℃, humidity ≥ 80% ), the cumulative damage time under high temperature and high irradiation (temperature ≥ 30°C, irradiation intensity ≥ 850W/m ² ), build a mapping relationship with the tensile strength of HDPE after the corresponding aging time, and complete the characterization of environmental big data. .

Based on the characteristic processing of the accumulated environmental damage time and the material property change data set after the corresponding aging time, the data set is grouped, and the test data and material property change data of the domestic Qionghai, Sanya, Guangzhou and other test stations are used as the training set for the construction of machine learning models. ; At the same time, the test data and material properties of test stations such as Chennai, India

The variable data is used as a test set for subsequent model validation.

(7) Using the characterized accumulated damage time data of foreign environments as the input layer, the service performance changes under different accumulated damage times are predicted, so as to obtain its service life. As shown in Figure 4. The tensile strength is used as the life evaluation reference performance, and the tensile strength is 30% as the end of life (the initial tensile strength is 20.74MPa, the end-of-life tensile strength is 6.22MPa, and the ordinate difference corresponding to the prediction model is -14.52). Entering the PE end of life -14.52 into the model, the time corresponding to 30% tensile strength at the end of life is 20.7 months, that is, the service life of HDPE in Chennai, India is 20.7 months. At the same time, the square value R ² of the correlation coefficient thereof with the experimental value was calculated. R ² =98%, more than 95%, the prediction result is accurate.

Table 3 Statistical analysis of cumulative damage time of environmental data in Chennai, India in three years

Several common polymer materials and three foreign regions are listed above to verify the accuracy of the method of this application. Other polymer materials and other regions can also input the characteristic parameters and aging performance parameters into the model established in this application, and carry out life expectancy.

The above-mentioned embodiments of the present invention are not intended to limit the scope of protection of the present invention, and the embodiments of the present invention are not limited thereto. According to the above-mentioned contents of the present invention, according to common technical knowledge and conventional means in the field, without departing from the present invention Under the premise of the above-mentioned basic technical idea, other various modifications, substitutions or changes made to the method of the present invention shall fall within the protection scope of the present invention.

Claims

A method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms, which is characterized by comprising the following steps:

(1) Select polymer materials, carry out aging tests in different regions, obtain the changes in performance parameters of the polymer materials during the test period during the aging process, and use the performance parameter changes as life evaluation indicators;

(2) collecting environmental data in the corresponding experimental period in step (1), including temperature, humidity and irradiation;

(3) Extract the characteristic data in the environmental data obtained in step (2) as characteristic parameters, and use the principal component analysis algorithm to reduce the dimension and noise of the characteristic parameter data, wherein the characteristic data is the cumulative sum of time in different environmental data stages, which is defined as cumulative damage time;

(4) Grouping the cumulative damage time-material property changes in different regions, using some regions as the training set for the construction of the life prediction model, and using the remaining regions as the test set for the verification of the life prediction model;

(5) Taking the cumulative damage time of the training set as the input parameter, and the material property change of the training set as the output parameter, using Python software to build a machine learning algorithm to train the environmental big data life prediction model to form a life prediction model;

(6) Using the cumulative damage time of the test set as an input parameter, predict the material performance changes under different cumulative damage times, and obtain its service life. At the same time, calculate the square value R 2 of the correlation coefficient with the experimental value, when R 2 ≥95%, the prediction results are credible.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the polymer materials described in step (1) are polystyrene, polycarbonate, polyethylene And one or more composite materials of polypropylene.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the aging test in step (1) is a natural aging test, a natural accelerated aging test or an artificial accelerated aging test. experiment.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the performance parameters of the polymer materials in step (1) include optical properties, mechanical properties and thermal properties .
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 4, wherein the performance parameters of the polymer materials in step (1) are yellowness index, transparency, tensile strength , one or more of melting temperature, glass transition temperature and initial decomposition temperature.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the experimental period in steps (1) to (2) is 1 to 5 years, and the polymer The sampling interval for material performance parameter changes is 1 to 3 months, and the environmental data is recorded every 1 to 10 hours.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the acquisition process of the accumulated damage time in step (3) includes: using Python software to measure temperature, humidity and radiation According to the statistics of the irradiation data, the performance change data of the polymer material during the experiment period and the corresponding cumulative sum of time under temperature, humidity and irradiation, as well as the cumulative sum of time under high temperature, high humidity and high temperature and high irradiation conditions are obtained.
The method for predicting the service life of polymer materials based on environmental big data and machine learning algorithms according to claim 1, wherein the machine learning algorithms in step (5) are neural networks, support vector machines, random forests, regression One or more of analytics, deep learning and XGboost.