Industrial Process Monitoring Technology and Its Application Based on Partial Least Square
|School||East China University of Science and Technology|
|Course||Control Science and Engineering|
|Keywords||process monitoring Kernel partial least squares adaptive monitoring concurrent projection to latent structures|
In the past few years, the scale of modern industrial production systems are becoming increasingly large and the complexity of systems are also doubled increasing with the development of science and technology. However, once these systems in the event of failure, it will cause huge property losses and casualties. In order to ensure the safe operation of production process, some methods need to be used to monitor industrial process abnormal condition or fault. Meanwhile, with the filed bus technology and the distributed control systems widely used, a large number of the historical operating data can be measured and stored automatically. In this context, data-based multivariate statistical methods become an active research area in process monitoring.Unfortunately, most traditional data-based fault detection methods often contain some assumptions such as linear, time-invariant process. These assumptions cannot able to describe the actual industrial process accurately, thus affecting the monitoring effect. To deal with the practical monitoring challenges, some novel ideas and strategies are proposed, which are summarized as follows:(1) To handle the nonlinear problem for process monitoring, a new technique based on kernel partial least squares (KPLS) is developed. KPLS is to first map the input space into a high-dimensional feature space via a nonlinear kernel function and then to use the standard PLS in that feature space. For process monitoring, two statistics of KPLS method are constructed. Compared to linear PLS, KPLS can effectively capture the nonlinear relationship between the input variables and output variables.(2) To solve the time-varying problem of industrial process, a novel monitoring strategy was proposed. When a new data sample came, first judged whether it was normal sample or not, if it was normal, then the new data was accepted and the model was updated using new data and old model. Once the model achieved a certain scale, we excluded extremely old data whilst accepted new data to update model, so the scale of the model remained unchanged. The simulation results showed the effectiveness of the proposed approaches compared to the conventional PLS and RPLS methods.(3) In view of the incompletion of conventional PLS method, a new concurrent projection to latent structures for the monitoring of output-relevant faults that affect the quality and input-relevant process faults is proposed. The input and output data spaces are concurrently projected to five subspaces, a joint input-output subspace that captures covariations between input and output, an output-principal subspace, an output-residual subspace, an input-principal subspace, and an input-residual subspace. The proposed CPLS monitoring method offers complete monitoring of faults that happen in the predictable output subspace and the unpredictable output-residual subspace, as well as faults that affect the input spaces and could be incipient for the output.Finally, several monitoring methods are compared with the proposed novel methods, and the TE process and CSTR process are applied to illustrate the efficient of the proposed methods.