| Nuclear fusion energy is a new clean energy, with high energy density and safety, which is not affected by climate, is considered to be one important way to solve the energy crisis. The structure material is one of the bottleneck problems of fusion reactor blanket and future commercial fusion power plant, it has relationship with realization and successful application of fusion.At present, the main problems of fusion reactor structural materials include: composition optimization study, properties study, manufacturing process and connection study, chemical compatibility corrosion and irradiation. In this dissertation, the composition optimization and performance prediction of fusion structure materials are studied based on the extreme learning machine method. The main work is as follows:(1) In this dissertation, we proposed an extreme learning machine (ELM) method MVFSA-ELM based on improved fast simulated annealing algorithm. This method maintained the advantage of extreme learning machine algorithm, optimized the parameters of extreme learning machine using the improved fast simulated annealing algorithm, accelerated the extreme learning machine number of hidden nodes in order that it can reach the global optimal solution of the extreme learning machine prediction. In comparison with I-ELM, CI-ELM, EI-ELM and EM-ELM algorithms, the MVFSA-ELM algorithm shows the advantages of the prediction accuracy. In order to further verify the feasibility of this algorithm in the performance prediction of the fusion reactor, this dissertation first uses this algorithm to predict the tensile strength of India IN-RAFM steel, and the predicted values are in good agreement with the experimental values.(2) A NMVFSA-ELM method is proposed for predicting the property of fusion reactor structural material with big data sets. This method can not only guarantee the global optimal solution, but also can converge to the optimal solution more quickly. A comparative analysis is performed on multiple data sets, the algorithm shows a faster prediction speed than the MVFSA-ELM algorithm, which helps to predict the properties of nuclear big data in the future. At the same time, the radiation hardening properties of European EUROFER97 were further verified, and the better prediction results were obtained.(3) We designed and implemented the nuclear reactor materials online performance prediction system based on MVFSA-ELM algorithm. This system has high accuracy and speed of prediction performance, and high data access speed. The system, according to the requirements of different users and different overall demands, designed and realized the function modules:the query module, the processing module of results, the information management module and the help module. The query module was designed and implemented with three different query methods of query by material, query by property and query by ingredient. The processing module implemented the download or print of performance data, and drawing or predicting performance online. The information management module realized the management of user information or data information for the administrator.We did the Ta optimization study of CLAM steel, one of the world’s three major industrialized RAFM steel, based on the online predicted system. We predicted the yield strength, tensile strength and elongation, section shrinkage rate performance with different Ta contents and different experimental temperatures. Not only verified the predicted data with the experimental data very well, also predicted the yield strength, tensile strength and elongation, section shrinkage rate under new conditions (the conditions that has never carried out experiments) with Ta quality percentage of 0.10wt%,0.12wt% and 0.14 wt%,0.16 wt%,0.17 wt%,0.20 wt%. Through comparative analysis, it concluded that, when the temperature between 350℃ and 550℃, and Ta quality percentage range in 0.18 wt%-0.20 wt%, the CLAM steel has better yield strength and tensile strength. Under the same temperature and Ta content, cross-section shrinkage rate and elongation are not significantly decreased. The analysis results will be helpful for the further study of CLAM steel Ta optimization. |
PDF Full Text Request