利用有限元神经网络计算漏磁场方法研究

摘要

To reduce computational cost of FEM, a neural network was adopted to simulate the process of finite element analysis, a finite element neural network ( FENN ) model to calculate the leakage field was established, and a conjugate gradient (CG) method was introduced as a learning algorithm. The magnetic leakage field of a rectangle defect was calculated by using FENN. The magnetic field intensity, magnetic flux density and the x and y components of leakage magnetic flux were obtained. The results indicate that the method has the advantages of rapidness and stability and can be applied to leakage field' s parallel resolving.%针对有限元法计算量大的不足,用神经网络模拟有限元的分析过程,建立了求解漏磁场计算的有限元神经网络模型,并采用共轭梯度学习算法,对矩形缺陷的漏磁场进行了计算.通过计算得到了磁场强度、磁感应强度矢量图以及漏磁通密度x、y分量图.结果表明,有限元神经网络能够实现漏磁场的并行求解,具有速度快、稳定性好等优点,是一种漏磁场的快速计算方法.