An intelligent prediction method and system for the life cycle of building materials

Abstract

The present invention discloses an intelligent prediction method and system for the life cycle of building materials, including obtaining building material failure data and status data, fitting the failure data to obtain the basic predicted life of the building materials, calculating the comprehensive damage degree according to the building material status data, calculating the environmental impact factor according to the environmental parameters, calculating the first predicted life, life deviation coefficient, environmental damage degree, loss life and second predicted life, constructing an intelligent prediction model for the life cycle of building materials according to the calculation results, and inputting the failure data and status data of the building materials to be predicted into the intelligent prediction model for the life cycle of building materials to obtain the predicted life and risk factors. This method can not only improve the efficiency and accuracy of the intelligent prediction of the life cycle of building materials, but also has good interpretability and can be directly applied to the intelligent prediction system for the life cycle of building materials.

Description

Technical Field

[0001] The present invention relates to the field of life prediction, and particularly to an intelligent prediction method and system for the life cycle of building materials. Background Art

[0002] In the construction field, building materials are widely used in engineering structures and external decorations. After the building materials are put into use, their performance gradually deteriorates with the changes of time, environment and usage conditions. Their service life and performance status are directly related to the safety, durability and service function of the building. The building material life evaluation and prediction technology aims to evaluate the service life of building materials and predict their future states through scientific means, so as to timely detect the degradation of materials and provide a basis for personnel maintenance and decision-making, thereby extending the service life of building materials and improving the reliability and economy of the overall project.

[0003] However, traditional building material life evaluation methods usually have many deficiencies, such as the evaluation process is cumbersome and time-consuming, and may cause certain damage to building materials or structures. With the development of technology, the prediction of the building material life cycle is developing towards intelligence. By integrating intelligent sensors and Internet of Things technologies, accurate and efficient prediction of the health status and remaining service life of building materials is realized by using professional data analysis models and algorithms. Therefore, combining modern information technology, artificial intelligence technology and building material science knowledge, by collecting and analyzing information such as the performance parameters, usage environment and historical data of building materials, establishing an accurate life prediction model, and designing an intelligent prediction method and system for the life cycle of building materials to overcome the deficiencies of existing evaluation models and achieve efficient, accurate and non-destructive prediction of the building material life cycle is of great significance for improving the economy of building operation and maintenance and ensuring the safety of building structures. Summary of the Invention

[0004] The object of the present invention is to provide an intelligent prediction method and system for the life cycle of building materials.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention includes the following steps:

[0007] Obtain the failure data and status data of building materials, fit the failure data to obtain the shape parameter and scale parameter of the Weibull model for estimating the life of building materials, and obtain the basic predicted life of building materials according to the Weibull model;

[0008] The state data includes building material state data and environmental state data. The comprehensive damage degree is calculated based on the building material state data. The first predicted life is determined according to the comprehensive damage degree and the basic predicted life. The life deviation coefficient is obtained through deviation analysis of the first predicted life.

[0009] The correlation between the environmental state data and the comprehensive damage parameter is analyzed to determine the environmental index. The abnormal values of the environmental index are screened and labeled, and the environmental impact factor is constructed based on the environmental index.

[0010] The environmental damage coefficient is determined according to the comprehensive damage degree and the environmental impact factor. The loss of life is determined according to the environmental damage coefficient. The second predicted life is determined according to the first predicted life, the loss of life, and the life deviation coefficient.

[0011] An intelligent prediction model for the life cycle of building materials is constructed according to the second predicted life, the environmental damage coefficient, the life deviation coefficient, and the loss of life. The failure data and state data of the building materials to be predicted are input into the intelligent prediction model for the life cycle of building materials to obtain the predicted life and the hazard factor.

[0012] Furthermore, the method for obtaining the basic predicted life of the building material includes:

[0013] The failure data consists of test failure data and simulated failure data.

[0014] Geotechnical tests are carried out on material specimens under the same mix ratio, construction technology, and curing conditions to obtain the basic parameters and test failure data of the materials.

[0015] A finite element model of the material specimen is constructed based on the basic parameters of the material, and finite element analysis is carried out on the material specimen to obtain simulated failure data.

[0016] The scale parameter and shape parameter are obtained by fitting the failure data according to the three-parameter Weibull model. The basic predicted life of the building material is calculated through the three-parameter Weibull model. The expression of the three-parameter Weibull model is:

[0017] F(t) = 1 - exp{-[λH(t - k 0 λ -1 )] α}

[0018]

[0019] where F(t) is the three-parameter Weibull distribution function, f(t) is the three-parameter Weibull density function, t is time, t 0 =(t - k 0 λ -1 ) is the time threshold, λ is the scale parameter, k0 is the proportionality parameter of the scale parameter and the time threshold, and β is the shape parameter;

[0020] Determine the reliability function R(t) and the failure function s(t) according to the distribution function F(t), and calculate the median failure time corresponding to F(t)=0.632. The median failure time is the basic predicted life T 0 .

[0021] Further, the method for calculating the comprehensive damage degree according to the building material state data includes:

[0022] The building material state data includes the cross-sectional area and porosity of the building material;

[0023] Clean the aged and peeled parts on the surface of the building material to be predicted, measure the cross-sectional area of the building material, and calculate the peeling damage parameter. The expression is:

[0024]

[0025] where w 1 is the peeling damage parameter, m 0 is the initial mass of the building material, S 0 is the initial cross-sectional area of the building material, S S is the cross-sectional area of the building material after peeling, a 1 、a 2 are mass constants;

[0026] Use an ultrasonic detector to measure the porosity inside the building material to be predicted, and calculate the elastic modulus damage parameter. The expression is:

[0027] w 2 = [E 0 (1 - P) / (1 + P) - a 3 / a 4

[0028] where w 2 is the elastic modulus damage parameter, E 0 is the initial elastic modulus of the building material, P is the porosity, a 3 、a 4 are elastic modulus constants;

[0029] Determine the comprehensive damage degree according to the peeling damage parameter and the elastic modulus damage parameter. The expression is:

[0030] w A = lg|Aw 1 2 ·Bw 2 2 |

[0031]

[0032] where w A is the comprehensive damage degree, A is the effective damage coefficient of the peeling damage parameter w 1 and B is the effective damage coefficient of the elastic modulus damage parameter w 2 .

[0033] Furthermore, the method for obtaining the life deviation coefficient includes:

[0034] Inputting the comprehensive damage degree and the basic predicted life into the first life prediction function to obtain the first predicted life, and the expression is:

[0035]

[0036] where T 1 is the first predicted life, b 1 , b 2 , b 3 are the model parameters of the first life prediction function, which are obtained by fitting the failure data, w A ′ = dw A / dt is the relative change rate of the comprehensive damage degree, and σ p is the peak value of the fatigue stress intensity;

[0037] Establishing a life deviation field based on the basic predicted life and the first predicted life, and the expression is:

[0038] f(m, n) = T 1 (m, n) - T 0 (m, n)

[0039] where f(m, n) is the predicted life deviation field, T 1 (m, n) is the first predicted life, T 0 (m, n) is the basic predicted life, m is the mean value of the predicted life and also represents the number in the x direction, and n is the standard deviation of the predicted life and also represents the number in the y direction;

[0040] Performing a 2D-DCT positive transformation on the life deviation field to obtain the life deviation modal function, and the expression is:

[0041]

[0042] where M is the total number of sampling points in the x direction, N is the total number of sampling points in the y direction, C(μ, v) is the orthogonal transformation coefficient matrix, μ and v are frequency values, and c(μ) and c(v) are normalization coefficients;

[0043] Performing key modal identification on the life deviation modal function to obtain the orthogonal transformation matrix At the same time, for Perform inverse DCT to obtain the reconstructed life deviation field Calculate the relative reconstruction error to obtain the life deviation coefficient:

[0044]

[0045] where δ 2 is the relative reconstruction error of the reconstructed life deviation field and the life deviation coefficient takes δ.

[0046] Furthermore, the method for constructing the environmental impact factor according to the environmental parameters includes:

[0047] Perform clustering operation on the environmental state data to obtain the peeling damage environmental index and the elastic modulus damage environmental index, calculate the correlation coefficient between the peeling damage environmental index and the peeling damage parameter, calculate the correlation coefficient between the elastic modulus damage environmental index and the elastic modulus damage parameter, and select the peeling damage environmental index and the elastic modulus damage environmental index with higher correlation as the main environmental indicators;

[0048] Use the hypersphere distribution model to perform anomaly screening on the main environmental indicators, mark the main indicators located outside the hypersphere as abnormal data and perform labeling operations;

[0049] Input the main environmental indicators into the environmental impact function to obtain the environmental impact factor, and the expression is:

[0050]

[0051] where w B is the environmental impact factor, c 1 , c 2 , c 3 , c 4 , c 5 are the environmental impact weight coefficients, C SO2 (t) is the concentration of atmospheric SO 2 at time t, C CO2 (t) is the concentration of atmospheric CO 2 at time t, t s is the starting time of atmospheric monitoring, t f is the ending time of atmospheric monitoring, N max is the maximum noise intensity, V max is the maximum vibration intensity, ΔT is the temperature difference within the monitoring period, T s is the standard curing temperature, ΔH is the humidity difference within the monitoring period, H s is the standard curing humidity, R UV is the external ultraviolet index.

[0052] Further, the method for determining the environmental damage coefficient, the loss of life, and the second predicted life includes:

[0053] According to the comprehensive damage degree w A and the environmental impact factor w B Determine the environmental damage coefficient, and the expression is:

[0054]

[0055] where w C is the environmental damage coefficient, and d 1 is the weight of the environmental damage coefficient;

[0056] Input the basic predicted life T 0 , the first predicted life T 1 and the comprehensive damage degree w A into the loss of life function to obtain the loss of life, and the expression is:

[0057]

[0058] where T loss is the loss of life, d 2 is the weight of the first predicted life, is the maximum environmental damage coefficient, and β is the loss of life coefficient;

[0059] Determine the second predicted life T 2 based on the basic predicted life, the first predicted life, the loss of life, and the life deviation coefficient; the second predicted life is negatively correlated with the loss of life; the second predicted life is positively correlated with the first predicted life; the second predicted life is positively correlated with the product of the life deviation coefficient and the basic predicted life.

[0060] Further, the method for obtaining the predicted life and the hazard factor includes:

[0061] Form a comprehensive data set with the second predicted life, the loss of life, the life deviation coefficient, and the environmental damage coefficient, and divide the comprehensive data set into a training set and a test set;

[0062] Construct an intelligent prediction model for the life cycle of building materials, and the specific structure includes an input layer, a base model layer, a policy layer, and an output layer;

[0063] The input layer is connected to the machine model layer, and the preprocessed training set data is input into the machine model layer. The base model layer is connected in parallel with a random forest regression model, a support vector machine model, and a linear regression model for prediction, and the prediction results are output to the strategy layer. The strategy layer selects a hybrid method to use a new Holdout set to train the meta-learner to synthesize the final predicted life and input it into the output layer. The two fully connected layers of the input layer generate the prediction results and extract the labels respectively, and output the predicted life and risk factors;

[0064] In the support vector machine base model, the hinge loss function is adopted to ensure the accuracy of classification, and the sequential minimal optimization algorithm is used to optimize the model weights; in the linear regression base model, the least squares method is used as the loss function to calculate the difference between the predicted value and the true value, and the stochastic gradient descent algorithm is used to improve the calculation efficiency and convergence speed;

[0065] The test set is used to evaluate the model and output the intelligent prediction model for the life cycle of building materials;

[0066] The failure data and status data of the building materials to be predicted are input into the intelligent prediction model for the life cycle of building materials to obtain the predicted life and risk factors.

[0067] In a second aspect, an intelligent prediction system for the life cycle of building materials includes:

[0068] A simulation module: used to perform finite element simulation of building materials according to the geotechnical test results to obtain building material failure data;

[0069] A data acquisition module: used to acquire building material failure data and status data, and preprocess the building material failure data and the status data;

[0070] A data processing module: used to calculate the basic predicted life of building materials according to the failure data, used to calculate the comprehensive damage degree according to the building material status data, used to calculate the environmental impact factor according to the environmental parameters, used to calculate the first predicted life, the life deviation coefficient, the environmental damage degree, the loss life, and the second predicted life, and used to transmit the calculation results to the prediction model module and the intelligent supervision module;

[0071] A prediction model module: used to construct an intelligent prediction model for the life cycle of building materials according to the second predicted life, the environmental damage degree, the life deviation coefficient, and the loss life, and input the failure data and the status data of the building materials to be predicted into the intelligent prediction model for the life cycle of building materials to obtain the predicted life and risk factors;

[0072] Intelligent supervision module: used to store, view, and manage the failure data, the status data, the predicted life, and the risk factors, visually display the predicted life and the risk factors of building materials to the user, and match recommended maintenance measures according to the risk factors.

[0073] The beneficial effects of the present invention are:

[0074] The present invention is an intelligent prediction method and system for the life cycle of building materials. Compared with the prior art, the present invention has the following technical effects:

[0075] Through steps such as finite element simulation, parameter fitting, correlation analysis, deviation mode analysis, calculating the environmental damage coefficient, calculating the loss life, calculating the second predicted life, and constructing a model, the present invention can improve the data preprocessing ability and enhance the model adaptability in the intelligent prediction of the life cycle of building materials, thereby improving the efficiency and accuracy of the intelligent prediction of the life cycle of building materials. Optimizing the intelligent prediction technology of the life cycle of building materials can greatly save resources and improve work efficiency. It can realize the prediction of the life cycle of building materials, provide a scientific basis for predicting the life cycle of building materials, and is of great significance to the safety of building structures. It can meet the prediction needs of intelligent prediction systems for different building material life cycles and intelligent prediction systems for different users' building material life cycles, and has a certain universality. Description of the Drawings

[0076] Figure 1 It is a flowchart of the steps of an intelligent prediction method for the life cycle of a building material according to the present invention. Detailed Embodiments

[0077] The present invention will be further described below through specific embodiments. The illustrative embodiments and explanations of the present invention are used to explain the present invention, but do not limit the present invention.

[0078] An intelligent prediction method and system for the life cycle of a building material according to the present invention include the following steps:

[0079] As Figure 1 shown, in this embodiment, it includes the following steps:

[0080] Obtain the failure data and status data of building materials, fit the failure data to obtain the shape parameter and scale parameter of the Weibull model for estimating the life of building materials, and obtain the basic predicted life of building materials according to the Weibull model;

[0081] The state data includes building material state data and environmental state data. Calculate the comprehensive damage degree according to the building material state data, determine the first predicted life according to the comprehensive damage degree and the basic predicted life, and perform deviation analysis on the first predicted life to obtain a life deviation coefficient;

[0082] Analyze the correlation between the environmental state data and the comprehensive damage parameters to determine environmental indicators, screen out the abnormal values of the environmental indicators and label them, and construct an environmental impact factor according to the environmental indicators;

[0083] Determine the environmental damage coefficient according to the comprehensive damage degree and the environmental impact factor, determine the loss of life according to the environmental damage coefficient, and determine the second predicted life according to the first predicted life, the loss of life and the life deviation coefficient;

[0084] Construct an intelligent prediction model for the building material life cycle according to the second predicted life, the environmental damage coefficient, the life deviation coefficient and the loss of life, and input the failure data and state data of the building material to be predicted into the intelligent prediction model for the building material life cycle to obtain the predicted life and the hazard factor.

[0085] In this embodiment, the method for obtaining the basic predicted life of the building material includes:

[0086] The failure data consists of test failure data and simulated failure data;

[0087] Conduct geotechnical tests on material specimens under the same mix ratio, construction technology and curing conditions to obtain the basic parameters and test failure data of the materials;

[0088] Construct a finite element model of the material specimen according to the basic parameters of the material, and perform finite element analysis on the material specimen to obtain simulated failure data;

[0089] Fit the failure data according to the three-parameter Weibull model to obtain the scale parameter and the shape parameter, and calculate the basic predicted life of the building material through the three-parameter Weibull model. The expression of the three-parameter Weibull model is:

[0090] F(t) = 1 - exp{ - [λH(t - k 0 λ -1 )] α}

[0091]

[0092] where F(t) is the three-parameter Weibull distribution function, f(t) is the three-parameter Weibull density function, t is time, t 0 = (t - k 0 λ -1) is the time threshold, λ is the scale parameter, and k 0 is the ratio parameter of the scale parameter to the time threshold, and α is the shape parameter;

[0093] Determine the reliability function R(t) and the failure function s(t) according to the distribution function F(t), and calculate the median failure time corresponding to F(t)=0.632. The median failure time is the basic prediction life T 0 ;

[0094] In actual evaluation, taking the pylon of a cable-stayed bridge with an operation period of 5 years as an example, geotechnical tests are carried out on material specimens under the same mix ratio and curing conditions to obtain: tensile strength of 3.45 MPa, compressive strength of 40.33 MPa, elastic modulus of 30 GPa, and fracture elongation of 0.22%;

[0095] Construct a finite element model of the material specimen according to the geotechnical test results, set the environmental conditions and load conditions to carry out a finite element simulation of the failure of the material specimen to obtain simulated failure data, fit the failure data with a three-parameter Weibull model to obtain a scale parameter α = 2.5 and a shape parameter λ = 130, and predict the basic expected life at a stress level of 90 MPa according to the three-parameter Weibull model to be 39.6 years.

[0096] In this embodiment, the method for calculating the comprehensive damage degree according to the building material state data includes:

[0097] The building material state data includes the cross-sectional area and porosity of the building material;

[0098] Clean the aged and peeled parts on the surface of the building material to be predicted, measure the cross-sectional area of the building material, and calculate the peeling damage parameter. The expression is:

[0099]

[0100] where w 1 is the peeling damage parameter, m 0 is the initial mass of the building material, S 0 is the initial cross-sectional area of the building material, S S is the cross-sectional area of the building material after peeling, a 1 、a 2 are mass constants;

[0101] Use an ultrasonic detector to measure the porosity inside the building material to be predicted, and calculate the elastic modulus damage parameter. The expression is:

[0102] w 2 = [E 0 (1 - P) / (1 + P) - a 3 / a 4

[0103] where w 2 is the elastic modulus damage parameter, E 0 is the initial elastic modulus of the building material, P is the porosity, a 3 , a 4 are elastic modulus constants;

[0104] Determine the comprehensive damage degree according to the peeling damage parameter and the elastic modulus damage parameter. The expression is:

[0105] w A = lg|Aw 1 2 ·Bw 2 2 |

[0106]

[0107] where w A is the comprehensive damage degree, A is the effective damage coefficient of the peeling damage parameter w 1 , B is the effective damage coefficient of the elastic modulus damage parameter w 2 ;

[0108] In actual evaluation, non-destructive measurement of the pylon of a cable-stayed bridge with an operating period of 5 years obtains: the cross-sectional area S S = 1.43m 2 after the building material is peeled off, the initial cross-sectional area S 0 = 1.5m 2 , the porosity P = 0.7%, the initial mass m 0 = 0.0024g / cm 3 , the mass constant is taken as a 1 = 0.95, a 2 = 0.05, the elastic modulus constant is taken as a 3 = 0.6, a 4 = 0.4. The peeling damage parameter is calculated to be -18.9977, the elastic modulus damage parameter is 72.4572, and the comprehensive damage degree is 2.5574.

[0109] In this embodiment, the method for obtaining the life deviation coefficient includes:

[0110] Input the comprehensive damage degree and the basic predicted life into the first life prediction function to obtain the first predicted life. The expression is:

[0111]

[0112] where T 1 is the first predicted life, b 1 , b 2 , b 3is the parameter of the first life prediction function model, obtained by fitting failure data, w A ′ = dw A / dt is the relative change rate of the comprehensive damage degree, σ p is the peak value of the fatigue stress intensity;

[0113] A life deviation field is established based on the basic predicted life and the first predicted life, and the expression is:

[0114] f(m, n) = T 1 (m, n) - T 0 (m, n)

[0115] where f(m, n) is the predicted life deviation field, T 1 (m, n) is the first predicted life, T 0 (m, n) is the basic predicted life, m is the mean value of the predicted life and also represents the number in the x direction, n is the standard deviation of the predicted life and also represents the number in the y direction;

[0116] The 2D-DCT forward transform is performed on the life deviation field to obtain the life deviation modal function, and the expression is:

[0117]

[0118] where M is the total number of sampling points in the x direction, N is the total number of sampling points in the y direction, C(μ, v) is the orthogonal transform coefficient matrix, μ and v are frequency values, and c(μ) and c(v) are normalization coefficients;

[0119] Key mode identification is performed on the life deviation modal function to obtain the orthogonal transform matrix At the same time, perform Perform the inverse DCT to obtain the reconstructed life deviation field Calculate the relative reconstruction error to obtain the life deviation coefficient:

[0120]

[0121] where δ 2 is the relative reconstruction error of the reconstructed life deviation field , and the life deviation coefficient takes δ;

[0122] In actual evaluation, the calculated value of the life deviation coefficient of the pylon of a cable-stayed bridge with an operation period of 5 years is 0.699.

[0123] In this embodiment, the method for constructing the environmental impact factor according to the environmental parameters includes:

[0124] Perform clustering operations on environmental status data to obtain peeling damage environmental indicators and elastic modulus damage environmental indicators. Calculate the correlation coefficient between the peeling damage environmental indicator and the peeling damage parameter, and calculate the correlation coefficient between the elastic modulus damage environmental indicator and the elastic modulus damage parameter. Select the peeling damage environmental indicator and the elastic modulus damage environmental indicator with higher correlations as the main environmental indicators;

[0125] Use the hypersphere distribution model to perform anomaly screening on the main environmental indicators, and mark the main indicators located outside the hypersphere as abnormal data and perform a tagging operation;

[0126] In the actual evaluation, perform anomaly screening on the main environmental indicators of the pylon of a cable-stayed bridge with an operating period of 5 years in this group. The ultraviolet index and the atmospheric SO 2 concentration data are located outside the hypersphere, and they are marked as risk factors;

[0127] Input the main environmental indicators into the environmental impact function to obtain the environmental impact factor, and the expression is:

[0128]

[0129] where w B is the environmental impact factor, c 1 、c 2 、c 3 、c 4 、c 5 are the environmental impact weight coefficients, C SO2 (t) is the atmospheric SO 2 concentration at time t, C CO2 (t) is the atmospheric CO 2 concentration at time t, t s is the starting time of atmospheric monitoring, t f is the ending time of atmospheric monitoring, N max is the maximum noise intensity, V max is the maximum vibration intensity, ΔT is the temperature difference within the monitoring period, T s is the standard curing temperature, ΔH is the humidity difference within the monitoring period, H s is the standard curing humidity, R UV is the external ultraviolet index;

[0130] In the actual evaluation, the main environmental indicators of the pylon of a cable-stayed bridge with an operating period of 5 years: atmospheric SO 2 concentration of 15 ppb, atmospheric CO 2 concentration of 450 ppm, maximum noise intensity of 70 dB, maximum vibration intensity of 0.15 m / s 2 、temperature difference within the monitoring period of 12 °C, humidity difference within the monitoring period of 5%, external ultraviolet index of 4, and the environmental impact weight coefficient is taken as c 1= 0.2, c 2 = 0.4, c 3 = 0.1, c 4 = 0.1, c 5 = 0.2, substituting into the environmental impact function gives an environmental impact factor of 1.10695.

[0131] In this embodiment, the method for determining the environmental damage coefficient, the loss of life, and the second predicted life includes:

[0132] Based on the comprehensive damage degree w A and the environmental impact factor w B determine the environmental damage coefficient, and the expression is:

[0133]

[0134] where w C is the environmental damage coefficient, and d 1 is the weight of the environmental damage coefficient;

[0135] Substitute the basic predicted life T 0 , the first predicted life T 1 and the comprehensive damage degree w A into the loss of life function to obtain the loss of life, and the expression is:

[0136]

[0137] where T loss is the loss of life, d 2 is the weight of the first predicted life, is the maximum environmental damage coefficient, and β is the loss of life coefficient;

[0138] Determine the second predicted life T 2 based on the basic predicted life, the first predicted life, the loss of life, and the life deviation coefficient; the second predicted life is negatively correlated with the loss of life; the second predicted life is positively correlated with the first predicted life; the second predicted life is positively correlated with the product of the life deviation coefficient and the basic predicted life;

[0139] In actual evaluation, take the weight of the environmental damage coefficient d 1 = 0.5, calculate the environmental damage coefficient as 1.5414 according to the above comprehensive damage degree and environmental impact factor, take the weight of the first predicted life d 2 = 2, and the loss of life coefficient β = 0.5, and calculate the loss of life as 4.6124 years.

[0140] In this embodiment, the method for obtaining the predicted life and the hazard factor includes:

[0141] A comprehensive dataset is formed by combining the second predicted lifespan, loss lifespan, lifespan deviation coefficient, and environmental damage coefficient, and the comprehensive dataset is divided into a training set and a test set;

[0142] An intelligent prediction model for the life cycle of building materials is constructed, and its specific structure includes an input layer, a base model layer, a strategy layer, and an output layer;

[0143] The input layer is connected to the machine model layer, and the preprocessed training set data is input into the machine model layer. The base model layer is connected in parallel with a random forest regression model, a support vector machine model, and a linear regression model for prediction and outputs the prediction results to the strategy layer. The strategy layer selects a hybrid method to use a new Holdout set to train the meta-learner to synthesize the final predicted lifespan and input it into the output layer. The two fully connected layers of the input layer respectively generate the prediction results and extract the labels, and output the predicted lifespan and risk factors;

[0144] In the support vector machine base model, the hinge loss function is adopted to ensure the accuracy of classification, and the sequential minimal optimization algorithm is used to optimize the model weights; in the linear regression base model, the least squares method is used as the loss function to calculate the difference between the predicted value and the true value, and the stochastic gradient descent algorithm is used to improve the calculation efficiency and convergence speed;

[0145] The test set is used to evaluate the model and output the intelligent prediction model for the life cycle of building materials;

[0146] The failure data and status data of the building materials to be predicted are input into the intelligent prediction model for the life cycle of building materials to obtain the predicted lifespan and risk factors;

[0147] In actual evaluation, a dataset composed of multiple groups of second predicted lifespan, loss lifespan, lifespan deviation coefficient, and environmental damage coefficient is used to train the model to obtain the intelligent prediction model for the life cycle of building materials. The failure data and status data of the pylon of a cable-stayed bridge with an operating period of 5 years are input into the intelligent prediction model for the life cycle of building materials, and the predicted lifespan of the pylon of the cable-stayed bridge with an operating period of 5 years is 37.8206 years, and the risk factors are the ultraviolet index and atmospheric SO 2 concentration. Therefore, the remaining predicted lifespan of the pylon of the cable-stayed bridge with an operating period of 5 years is 32.8206 years. At the same time, maintenance measures can be taken according to the risk factor labels to enhance the durability and ultraviolet resistance of the building materials.

[0148] In the second aspect, an intelligent prediction system for the life cycle of building materials includes:

[0149] A simulation module: used to perform finite element simulation of building materials according to the geotechnical test results to obtain building material failure data;

[0150] Data acquisition module: used to obtain building material failure data and status data, and preprocess the building material failure data and the status data;

[0151] Data processing module: used to calculate the basic predicted life of building materials according to the failure data, calculate the comprehensive damage degree according to the building material status data, calculate the environmental impact factor according to the environmental parameters, calculate the first predicted life, the life deviation coefficient, the environmental damage degree, the loss life and the second predicted life, and transmit the calculation results to the prediction model module and the intelligent supervision module;

[0152] Prediction model module: used to construct an intelligent prediction model for the building material life cycle according to the second predicted life, the environmental damage degree, the life deviation coefficient and the loss life, and input the failure data and the status data of the building material to be predicted into the intelligent prediction model for the building material life cycle to obtain the predicted life and the risk factor;

[0153] Intelligent supervision module: used to store, view and manage the failure data, the status data, the predicted life and the risk factor, visually display the predicted life and the risk factor of the building material to the user, and match the recommended maintenance measures according to the risk factor.

[0154] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent prediction method for the life cycle of building materials, characterized in that: The following steps are involved: S1. Obtaining failure data and state data of building materials, fitting the failure data to obtain shape parameters and scale parameters of a Weibull model for estimating the life of building materials, and obtaining a basic predicted life of building materials according to the Weibull model; S2, the state data includes building material state data and environmental state data, calculating the comprehensive damage degree according to the building material state data, determining the first predicted life according to the comprehensive damage degree and the basic predicted life, and performing deviation analysis on the first predicted life to obtain a life deviation coefficient; S3, analyzing the correlation between the environmental status data and the comprehensive damage parameters to determine environmental indicators, screening abnormal values ​​of the environmental indicators and labeling them, and constructing environmental impact factors according to the environmental indicators; S4, determining an environmental damage coefficient according to the comprehensive damage degree and the environmental impact factor, determining a lost life according to the environmental damage coefficient, and determining a second predicted life according to the first predicted life, the lost life and the life deviation coefficient; S5. Construct an intelligent prediction model for the life cycle of building materials based on the second predicted life, the environmental damage coefficient, the life deviation coefficient and the lost life, and input the failure data and status data of the building materials to be predicted into the intelligent prediction model for the life cycle of building materials to obtain the predicted life and risk factors; the risk factors are abnormal values ​​of the environmental indicators, which are determined by label extraction operations.

2. According to claim 1, a method for intelligent prediction of the life cycle of building materials is characterized in that: The method for obtaining the basic predicted life of the building material comprises: The failure data consists of test failure data and simulation failure data; Conduct geotechnical tests on material specimens under the same mix ratio, construction technology and curing conditions to obtain basic material parameters and test failure data; Construct a finite element model of the material specimen based on the basic parameters of the material, and perform finite element analysis on the material specimen to obtain simulated failure data; The failure data is fitted according to the three-parameter Weibull model to obtain the scale parameter and shape parameter. The basic predicted life of building materials is calculated by the three-parameter Weibull model. The expression of the three-parameter Weibull model is: F(t)=1-exp{-[λH(t-k0λ -1 )] α } Where F(t) is the three-parameter Weibull distribution function, f(t) is the three-parameter Weibull density function, t is time, t0 = (t-k0λ -1 ) is the time threshold, λ is the scale parameter, k0 is the ratio parameter between the scale parameter and the time threshold, and α is the shape parameter; The reliability function R(t) and the failure function s(t) are determined according to the distribution function F(t), and the median failure time corresponding to F(t)=0.632 is calculated. The median failure time is the basic predicted life T0.

3. According to claim 1, a method for intelligent prediction of the life cycle of building materials is characterized in that: The method for calculating the comprehensive damage degree according to the building material status data comprises: The building material status data includes the cross-sectional area and porosity of the building material; Clean the aging and peeling parts of the surface of the building material to be predicted, measure the cross-sectional area of ​​the building material, and calculate the peeling damage parameter, which is expressed as: Where w1 is the peeling damage parameter, m0 is the initial mass of the building material, S0 is the initial cross-sectional area of ​​the building material, S S is the cross-sectional area of ​​the building material after stripping, a1 and a2 are mass constants; The internal porosity of the building material to be predicted is measured by an ultrasonic detector, and the elastic modulus damage parameter is calculated. The expression is: w2=[E0(1-P) / (1+P)-a3] / a4 Where w2 is the elastic modulus damage parameter, E0 is the initial elastic modulus of the building material, P is the porosity, and a3 and a4 are elastic modulus constants; The comprehensive damage degree is determined according to the peeling damage parameter and the elastic modulus damage parameter, and the expression is: w A =lg|Aw1 2 ·Bw2 2 | where w A is the comprehensive damage degree, A is the effective damage coefficient of the peeling damage parameter w1, and B is the effective damage coefficient of the elastic modulus damage parameter w2.

4. According to claim 1, a method for intelligent prediction of the life cycle of building materials is characterized in that: The method for obtaining the life deviation coefficient comprises: The comprehensive damage degree and basic predicted life are input into the first life prediction function to obtain the first predicted life, the expression is: Where T1 is the first predicted life, b1, b2, b3 are the first life prediction function model parameters obtained by fitting the failure data, and w A '=dw A / dt is the relative change rate of comprehensive damage degree, σ p is the peak value of fatigue stress intensity; The life deviation field is established based on the basic predicted life and the first predicted life, and the expression is: f(m,n)=T1(m,n)-T0(m,n) Where f(m, n) is the predicted life deviation field, T1(m, n) is the first predicted life, T0(m, n) is the basic predicted life, m is the predicted life mean and also represents the number in the x direction, and n is the predicted life standard deviation and also represents the number in the y direction; The life deviation field is transformed by 2D-DCT positive term to obtain the life deviation modal function, which is expressed as: Where M is the total number of sampling points in the x direction, N is the total number of sampling points in the y direction, C(μ, v) is the orthogonal transformation coefficient matrix, μ and v are frequency values, and c(μ) and c(v) are normalization coefficients; The orthogonal transformation matrix is ​​obtained by performing key mode identification on the life deviation mode function. At the same time, Perform inverse DCT transform to obtain the reconstructed lifetime deviation field Calculate the relative reconstruction error to get the lifetime deviation coefficient: where δ 2 To reconstruct the lifetime deviation field The relative reconstruction error of , and the lifetime deviation coefficient is δ.

5. According to claim 1, the intelligent prediction method for the life cycle of building materials is characterized in that: The method for constructing an environmental impact factor according to the environmental indicator comprises: Perform clustering operation on environmental status data to obtain peeling damage environmental index and elastic modulus damage environmental index, calculate the correlation coefficient between peeling damage environmental index and peeling damage parameter, and the correlation coefficient between elastic modulus damage environmental index and elastic modulus damage parameter, and select peeling damage environmental index and elastic modulus damage environmental index with higher correlation as main environmental index; The hypersphere distribution model is used to screen out abnormalities in the main environmental indicators, and the main indicators outside the hypersphere are marked as abnormal data and labeled; Input the main environmental indicators into the environmental impact function to obtain the environmental impact factor, the expression is: where w B is the environmental impact factor, c1, c2, c3, c4, c5 are environmental impact weight coefficients, C SO2 (t) is the atmospheric SO2 concentration at time t, C CO2 (t) is the atmospheric CO2 concentration at time t, t s is the starting time of atmospheric monitoring, t f is the end time of atmospheric monitoring, N max is the maximum noise intensity, V max is the maximum vibration intensity, ΔT is the temperature difference during the monitoring period, T s is the standard curing temperature, ΔH is the humidity difference during the monitoring period, H s is the standard maintenance humidity, R UV It is the external UV index.

6. The intelligent prediction method for the life cycle of building materials according to claim 1 is characterized in that: The method for determining the environmental damage coefficient, the lost life and the second predicted life comprises: According to the comprehensive damage degree w A and the environmental impact factor w B Determine the environmental damage coefficient, the expression is: where w C is the environmental damage coefficient, d1 is the weight of the environmental damage coefficient; The basic predicted life T0, the first predicted life T1 and the comprehensive damage degree w A Input the loss life function to obtain the loss life, the expression is: Where T loss is the life lost, d2 is the first predicted life weight, is the maximum environmental damage coefficient, β is the loss life coefficient; The second predicted life T2 is determined based on the basic predicted life, the first predicted life, the lost life and the life deviation coefficient; the second predicted life is negatively correlated with the lost life; the second predicted life is positively correlated with the first predicted life; the second predicted life is positively correlated with the product of the life deviation coefficient and the basic predicted life.

7. The intelligent prediction method for the life cycle of building materials according to claim 1 is characterized in that: The method for obtaining the predicted life span and risk factors comprises: The second predicted life, the lost life, the life deviation coefficient and the environmental damage coefficient are combined into a comprehensive data set, and the comprehensive data set is divided into a training set and a test set; Construct an intelligent prediction model for the life cycle of building materials. The specific structure includes input layer, base model layer, strategy layer and output layer; The input layer is connected to the machine model layer, and the preprocessed training set data is input to the base model layer. The base model layer uses the random forest regression model, support vector machine model and linear regression model in parallel to predict and output the prediction results to the strategy layer. The strategy layer uses a hybrid method to use the new holdout set to train the meta-learner to synthesize the final predicted lifespan and input it to the output layer. The two fully connected layers of the output layer generate prediction results and extract labels respectively, and output the predicted lifespan and risk factors. In the support vector machine base model, the hinge loss function is used to ensure the accuracy of classification, and the sequential minimum optimization algorithm is used to optimize the model weights; in the linear regression base model, the least squares method is used as the loss function to calculate the difference between the predicted value and the true value, and the stochastic gradient descent algorithm is used to improve the computational efficiency and convergence speed; Use the test set to evaluate the model and output the intelligent prediction model of the building material life cycle; The failure data and status data of the building materials to be predicted are input into the building materials life cycle intelligent prediction model to obtain the predicted life span and risk factors.

8. An intelligent prediction system for the life cycle of building materials, used to execute the method according to any one of claims 1 to 7, comprising: Simulation module: used to perform finite element simulation of building materials based on geotechnical test results to obtain building material failure data; Data acquisition module: used to obtain building material failure data and status data, and pre-process the building material failure data and status data; Data processing module: used to calculate the basic predicted life of building materials according to the failure data, used to calculate the comprehensive damage degree according to the building material status data, used to calculate the environmental impact factor according to the environmental index, used to calculate the first predicted life, the life deviation coefficient, the environmental damage degree, the lost life and the second predicted life, and used to transmit the calculation results to the prediction model module and the intelligent supervision module; Prediction model module: used for constructing an intelligent prediction model of the life cycle of building materials according to the second predicted life, the environmental damage degree, the life deviation coefficient and the lost life, and inputting the failure data and the state data of the building material to be predicted into the intelligent prediction model of the life cycle of building materials to obtain the predicted life and risk factors; Intelligent supervision module: used to store, view and manage the failure data, the status data, the predicted life and the risk factors, visualize the predicted life and risk factors of building materials to users, and recommend maintenance measures based on risk factor matching.

Images