Identification Method For Thermal Resistance Of Northern Residential Building Exterior Walls Assisted With Infrared Thermography

Exterior wall is an important part of building envelope structure,and the on-site identification of its heat transfer coefficient is a key link in the energy-saving assessment and energy-saving reconstruction of buildings.In this paper,the research objective is limited to the exterior walls of residential buildings in severe cold area and cold area.In view of the shortcomings of the current on-site identification methods for heat transfer coefficient of building exterior walls,and based on the advantages of infrared thermography,this paper explores and studies the application of machine learning algorithm to establish a convenient and non-destructive on-site identification method for thermal resistance of exterior walls of residential buildings,thereby acquiring the heat transfer coefficient.Based on the theory of inverse problem of unsteady heat conduction in buildings,this paper proposes a method for system identification of thermal resistance of exterior walls by using machine learning algorithm.Through theoretical analysis,it is determined that the time series,the average indoor temperature,outdoor temperature and the average temperature of inside and outside wall surface are taken as input parameters to establish an identification model,and the thermal resistance value of the exterior walls can be obtained.The numerical model of exterior wall heat transfer is established and verified.Through investigation,the typical or common structural forms of the exterior walls of residential buildings in severe cold area and cold area are obtained.CFD method is used to carry out numerical experiments on the typical working conditions(constancy and variation of room temperature)of the exterior walls in winter,thereby acquiring the temperature distribution and variation law of the walls.At the same time,the numerical experimental results provide sample data for the identification and modeling of thermal resistance of exterior walls.The calibration model of infrared thermography temperature data of exterior walls is established by using the method of neural network.In this paper,the temperature data measured by infrared thermal imager and the emissivity of material are taken as input variables,while the data measured by thermocouple is taken as output variables,and Back Propagation Neural Network(BP),Radial Basis Function Neural Network(RBF)and Multiple Nonlinear Regression(MNLR)method are selected respectively to establish the modified model of infrared thermography temperature measurement data and carries out error comparison.The results show that the RBF neural network calibration model has the highest correction accuracy.In addition,a fast extraction program for the temperature data of the effective area on the wall surface is compiled.Through this program,the effective temperature information of the wall which excludes the occlusion and auxiliary components can be obtained,and the visualization of the calibrated infrared data and the display image of the effective area can be realized.A thermal resistance identification system for building exterior walls is established by machine learning algorithm.In this paper,CFD simulation data is used as training samples,and BP neural network,RBF neural network,General Regression Neural Network(GRNN)and Support Vector Machine Based on Particle Swarm Optimization(PSO-SVM)are respectively used to build the model of thermal resistance identification system for building exterior walls.In the process of establishing the model,the influence of the total length of the time series(identification period)on the identification error is considered firstly,and 12 hours is determined as the optimal time series.Then the prediction ability of the four sets of model samples is compared with the generalization ability of the test sets,the results of which show that the generalization ability of PSO-SVM model is significantly better than the other three methods,and the accuracy of PSO-SVM model is as high as 94.7%.Therefore,PSO-SVM model is selected to build the thermal resistance identification system of exterior walls,and its noise resistance is examined.Laboratory test and field test are used to verify the feasibility of the thermal resistance identification system of exterior walls.The experimental results show that the average relative error of the wall thermal resistance value and theoretical value identified by PSO-SVM is 2.6%,and the average relative error compared with the heat transfer coefficient detector is 3.5%,and the test value of PSO-SVM is lower than the theoretical value,so the test result of the thermal resistance identification system is more reasonable.In the six groups of field tests,the error of Experiment 6 is large due to the excessively large shooting angle of the infrared thermal imager.This is revised in this paper and the final average relative error of the six groups of experiments is 7.1%.Overall,the test results verify the feasibility of PSO-SVM model to test the thermal resistance of exterior walls.The research results of this paper have important theoretical significance and application value for on-site identification of heat transfer coefficient of building exterior walls.