带料纠偏的模糊预测控制

Fuzzy predictive control of tape rectification

带料纠偏是高度非线性过程,传统的模型预测控制(MPC)无法有效地处理这种过程。模糊神经网络(FNN)方法可以实现非线性过程模型。通过测量得到的数据作为样本来训练神经网络。预测准确度由前馈网络的插值能力保证。多维搜索技术用来解决非线性最优化问题,最优结果被嵌入BP神经网络预测控制器中。BP神经网络的快速计算能满足实时控制需要。带料纠偏试验结果已经证明了FNN预测控制的有效性。

Tape Rectification is a highly nonlinear process.Conventional Model Predictive Control （MPC） may not be able to cope with such process. The Fuzzy Neural Network （FNN） me proposed to realize the model of the nonlinear process. The measured data serve as samples to train the neural network. The prediction accuracy is guaranteed by the interpolation ability of the feedforward network. Multi-dimensional searching technique is used to solve the nonlinear optimization problem. The optimization results are embedded into a BP neural network predictive controller. On-line computation of BP neural network can satisfy the real-time control demand. The experimental results of tape rectification have shown the effectiveness of the FNN predictive control.