摘要
通过沿薄板运动反方向以单元格的形式平移薄板中的热源来取代薄板的连续运动 ,使得温度场的每次计算可以采用相同的热源分布 ,无需重新计算涡流场 ,大大减少了横向磁通感应加热耦合场数值仿真的时间。利用有限元数值计算结果作为神经网络的训练样本 ,分别建立了横向磁通感应加热装置出口处平均温度及其平均相对误差的预测模型。检验样本的结果表明所训练的神经网络具有很高的准确性。利用神经网络分析了频率和电流对平均温度及其平均相对误差的影响 。
Instead of sheet movement,to move the heat source in elements in the opposite direction was used for the coupled field analysis in transverse flux induction heating(TFIH),which resulted in that successive temperature field computations could apply the same heat source distribution resulted from an eddy current field calculation in the whole procedure,and so the time for numerical simulation of TFIH was greatly reduced.With the results of three-dimensional finite element method(FEM) calculation as training examples,two neural networks were set up for predicting average temperature and its average relative error at the outlet of TFIH equipment respectively.The results of tested examples showed that the network predictions had high accuracy and could be used for the study of the effect of current and frequency on average temperature and its average relative error.Simulated annealing method was used to get the least average relative error for a given temperature.
出处
《金属热处理》
CAS
CSCD
北大核心
2004年第4期53-57,共5页
Heat Treatment of Metals
基金
霍英东教育基金会青年教师基金项目 ( 710 5 5 )
河北省自然科学基金重点项目 ( 60 2 0 78)