一种非线性过程监控方法

A monitoring method for nonlinear process

实际工业过程都具有非线性等特征。传统的监控方法有将降维后的非线性数据映射到高维线性空间再进行数据处理,实现过程的监控。本文是在一种否定选择算法的基础上,首先利用最大方差展开(MVU)方法对正常高维数据进行降维,再利用否定选择算法直接对降维后的多维非线性数据建立"超球体群"模型,实现对过程的监控,保证工业过程的平稳运行。仿真实验是基于TE模型进行的,仿真结果表明该方法较传统方法及其他改进方法具有更好的监控能力,说明了该方法的有效性。

All of the industrial processes in our society have such characteristics as non-linearity. Some traditional monitoring methods mapped the nonlinear data to the high-dimensional linear space for data dealing, and then monitored the process. This paper is based on a negative selection algorithm, and firstly we deal the normal high-dimensional data with dimension reduction method of Maximum Variance Unfolding （MVU）. Then we use the negative selection algorithm to make the ＇hyper-sphere group＇ model with the multidimensional nonlinear data directly after dimension reduction. Thus the monitoring of the process is realized and the smooth operation of the industrial process is ensured. The simulation is based on the TE model, and the result shows the method gets a better monitoring capability comparing with the traditional method and other advanced methods. Eventually, this paper illustrates the effectiveness of the method.