【摘要】 该文针对梯度下降BP算法在网络训练时易陷入局部最小值及共轭梯度BP算法的适用范围窄的问题,将两者有机结合起来,应用于火电厂高压加热器系统故障诊断中。经仿真试验证明,这种结合能有效地避免局部极小值的问题,极大提高网络的训练速度。同时通过试验,搜寻到了判断高加系统故障诊断网络训练进入极值点的最优标准值。更多还原

【Abstract】 Aiming at the question that the Grads Descend BP Arithmetic is easy to get into the local least value and the Conjugate Grads BP Arithmetic can be use d in a small scope, this paper combines the two kinds of BP Arithmetic and appli es them to the fault diagnosis of high pressure heater. The imitation test tells us that the combination can effectively avoid the local least value and greatly en hance the net training velocity. Meanwhile through the test ,we can get the best standard value of judge the network training of high pressure heater system fau lt diagnosis to enter into the local least value.更多还原