Industrial Process Monitoring Research Based On KOLS&RW-LSSVM

With the development of modern industrial processes into large-scale, integration, complication, efficient and stable process monitoring and fault diagnosis technologies are the key to ensure process operation safety, enhance product quality, reduce production cost increase economic benefits, and promote the enterprise competitiveness. Since it is difficult to achieve the accurate mathematical model in complex industrial process, also a large amount of process data have been sampled and collected, how to analysis these data and extract the useful features to monitor and diagnosis fault becomes a challenging issue.This paper focus on fault detection and diagnosis of process monitoring technology. Firstly, since the advantage of kernel partial least squares (KPLS) for nonlinear monitoring, a new online monitoring method based on KPLS is adopted to judge whether the process running in the stable; then also proposed a improved least squares support vector machine(LS-SVM) based on robust weighted method, to structure a fault classification machine for fault diagnosis and identification.In order to verify the validity of these methods, this method is applied to the standard process simulation model for TE process fault detection and diagnosis, and it achieved the desired effect.