| Energy is the basis and guarantee for supporting national economic development and maintaining social stability.With the continuous growth of China's economy,the demand of energy for China continues to increase.Compared with traditional fossil energy,nuclear energy has the advantages of rich resources,high energy conversion efficiency,low cost and low pollution.In the face of traditional energy problems and crises,the importance of nuclear energy to China becomes more obvious.With the continuous development of nuclear energy,its safety has also become more prominent.After the Fukushima nuclear accident in Japan,accident-tolerant fuel(ATF)with higher safety and economy than the traditional fuel system of UO2 pellet-Zr cladding came into being.Improving the TC of fuel pellets is an important way to develop ATF,which can be achieved by adding high TC materials(BeO,SiC,Mo,diamond,etc.)to UO2 matrix to prepare composite fuel.At present,the optimal system,composition and structure of UO2 composite fuel are not clear at home and abroad,which need to be further studied and determined.The research and development(R&D)of UO2 composite fuel has the characteristics of long cycle,high cost,toxic and radioactive materials,so the traditional trial-and-error method is inefficient to develop the composite fuel.In recent years,machine learning is playing an important role in the rapid R&D of materials,which can greatly shorten and reduce the R&D cycle and cost of UO2 composite fuel.Finite element method(FEM)can be applied to thermal analysis of material,and calculate material thermal conductivity quickly and accurately,which can solve the problem of data scarcity and provide data basis for the application of machine learning in the R&D of UO2 composite fuel.In this work,the TC enhanced UO2 composite fuel is developed by the method combined by FEM,machine learning and experiment.The main process of the method is as follows.The variation ranges of FEM input parameters are determined according to the existing experimental and theoretical data.A certain number of parameter combinations are obtained by sampling methods(Latin hypercube,random sampling,etc.)within the parameter setting range,which are as the input of FEM modeling and analysis.Then the set of FEM samples including target performance is generated.The features appropriate to predict target performance are selected by feature selection methods(Pearson coefficient,gradient boosting decision tree),the mapping from predictive features to target performance is constructed using proper machine learning models(neural network,support vector regression,Gaussian process regression,etc.),and the accuracy of predictive models are verified by FEM and experimental samples.Combined with the appropriate optimization method(numerical optimization,genetic algorithm,multi-objective optimization),the reverse design from the target performance to the predictive feature is realized.According to the guidance of the reverse design results,the composite pellets corresponding to the composition,microstructure and design results are prepared by spark plasma sintering technology.The microscopic images of the pellets are obtained by laser confocal microscope and scanning electron microscope to verify whether the tissue structure met the design requirements;the TC of the pellets are obtained by laser TC testing instrument and the high-throughput platform of TC characterization to verify whether the performance of the pellets meet the design requirements.Through the above method,the main research results of this work are as follows:1)The irradiation stability and spent fuel reprocessing for different types of UO2 composite fuels were analyzed.According to the comprehensive analysis,two composite fuel systems,i.e.UO2-SiC with SiC dispersion and UO2-Mo with Mo continuous,have a certain application prospect,which narrows the research scope of UO2 composite fuels.2)The predictive model of TC for dispersed composite fuel was established and verified.The results of feature selection show that particle fraction,average grain size and average particle size are the three most related structural features to the TC of dispersed composite fuel.The relative error of the proposed model for the TC of experimental UO2-SiC samples was less than 5%.Combined with the numerical optimization method and the analytical formula of the predictive model,the reverse design of UO2-SiC composite from target TC to structural characteristics was realized.According to the design results,the relative error between the thermal conductivity of the UO2-SiC experimental sample and the target thermal conductivity was less than 5%.3)The predictive model of TC for continuous composite fuel was established and verified.The results of feature selection show that the second phase fraction,porosity,continuous channel area and microcrack ratio are the four most related structural features to the TC of continuous composite fuel.Through the analysis of the predictive models,it is found that under the condition of the constant second phase content,the TC of the composite fuel decreases at first and then stabilizes with the increase of the continuous channel area;and under the condition of the same theoretical density,the TC of the composite fuel decreases at first and then stabilizes with the increase of microcrack ratio.According to the results of theoretical analysis,the microstructure of experimental UO2-Mo samples with low Mo content was optimized,and the microcrack-free UO2-2 vol%Mo composite fuel was successfully prepared.The TC of the optimal UO2-2 vol%Mo is about 20%higher than that of pure UO2 at room temperature,which is close to the simulation theoretical value.4)The predictive model of TC and thermal stress of UO2-Mo-Nb multi-component composite was established and verified.The optimized composition of UO2-3 vol%Mo-1 vol%Nb was obtained by multi-objective optimization,and the balance between TC and thermal stress of UO-Mo-Nb composite fuel was realized.5)The heat conduction model based on low power laser measurement was established and verified.The model can obtain an accurate temperature-time curve of the sample surface through the parameters such as convective heat transfer coefficient,laser power,heating time,sample thickness,sample properties,etc.The TC of the material can be obtained by reverse calculation from the temperature-time curve using the optimization method.Based on the heat conduction model,the high-throughput platform of TC characterization was built,which has the ability of auto continuous feeding of 20 samples in one time.The relative error of the platform is less than 10%at room temperature. |
PDF Full Text Request