电控发动机怠速RBF神经网络控制

Idle Speed Control of the Electronic Control Engine Based on RBF Neural Network

提出发动机怠速控制的神经网络方法,给出了RBF神经网络模型,并将带遗忘因子的梯度下降法应用于RBF神经网络的参数调整。利用RBF神经网络良好的非线性映射能力,通过对发动机转速及转速变化率映射,得到步进电机相应驱动信号,从而实现怠速控制。实验结果表明,神经网络控制响应快、鲁棒性强,可有效提高发动机怠速品质,改善发动机的性能指标。

It is difficult to establish an accurate mathematical model for idle speed control of engine because of non-linear and changeable feature of the system. We adopted the neural network method for idle speed control of engine, and a radial basis function （RBF） neural network model was set up. The gradient descend algorithm with forgetful functor was applied to adjust the parameters of the network. The RBF neural network was trained quickly for its good abilities of non-linear mapping, learning and generalizing. Control of engine idle speed was realized by the outputs of stepping motor, which was taken as driven signals and obtained by mapping the rotating speed of the engine and its changing rate. The experimental result showed that a control with quick response, good robustness, as well as an improved idle running quality of engine could be achieved by NNC.