【摘要】 在高压下,压力传感器往往会表现出非线性的特性。本文基于反向传播神经网络对Volterra级数表示的非线性系统进行了研究。在分析了神经网络分解后的结构与Volterra级数表示的非线性系统之间的类似关系后,将所使用神经网络中的激励函数在阀值处进行泰勒级数分解,解算出了Volterra级数的各阶核,从而实现了对非线性系统(传感器)的建模。实例建模结果表明,通过使用神经网络方法求解Volterra级数核来对非线性系统进行建模的方法非常有效。更多还原

【Abstract】 At the point of high pressure,nonlinear characteristics are always shown in pressure sensors. Based onback propagation (BP) neural network, a method of modeling in Volterra nonlinear systems is proposed. Thismethod is using the Taylor series to expand each sigmoid function of three-layer feedforward perceptrons aboutits offset value. The conditions under which the two approaches yield equivalent representations of the input-out-put relation are explored. Then each-order kernels are solved. Experimental results show the efficiency of thismethod.更多还原