基于SCNN的转子热应力场的实现研究

热应力是监测机组寿命损耗的重要量 ,但现在对热应力的实时计算方法大多是采用解析方法 ,由于解析方法是对转子二维轴对称热传导方程的简化 ,这给热应力的计算结果带来了误差。基于这种背景 ,本文利用了有限元计算热应力场的中间计算结果 ,应用结构可控的前向神经网络 (SC NN)方法 ,仿真了热应力场的整体刚度矩阵代数方程组的求解 ,应用所得的节点位移 ,可求出转子的热应力场。

A Study of the Thermal Stress Field Based on SCNN (Structure-Controllable Neural Network)

Thermal stress constitutes a major parameter for monitoring the loss of steam turbine service life. To date, for the real time calculation of thermal stresses employed in most cases is an analytical method. As the analytical method involves a simplification of the two dimensional axial symmetric heat conduction equation for a rotor, it will lead to some errors in evaluating the thermal stress. In view of this, by way of the intermediate calculation results of a finite element analysis for thermal stress field the authors have utilized a SCNN (structure controllable neural network) method to simulate the solution of an algebraic equation of integral rigidity matrix for the thermal stress field. By using the displacement of nodes thus obtained, one can identify the thermal stress field of a rotor.