Data Based Operating Performance Assessment Method And Its Application For Complex Industrial Processes

In order to grasp the operation status of industrial production process timely and accurately,ensure the quality of products,and maximize the economic benefits of enterprises,the research on operating performance assessment of complex industrial processes has been paid more and more attention by academia and industry in recent years.With the improvement of automation technology,a large number of industrial process data can be saved.Therefore,the research on data-driven operating performance assessment method of complex industrial processes will have important guiding significance and application value for the optimization and adjustment of modern industrial production process.In this paper,the problem of noise interference and outliers in data of the industrial field is studied in depth.Firstly,a kernel partial robust M-regression based on the kernel density estimation method(KDE-KPRM)is proposed.Then,based on this,a weighted kernel total latent structure projection method(WKT-PLS)is proposed to evaluate the operation status of complex industrial processes,and it is applied to the heavy medium coal preparation process.This study includes the following contents:(1)On the basis of understanding the characteristics of data in complex industrial process,in order to handle the problem that data are seriously disturbed by noises and the relationship between data is nonlinear,a kernel partial robust M-regression based on the kernel density estimation method(KDE-KPRM)is proposed.In this method,the weights of input data are calculated by the kernel density function and the principal component analysis,and the weights of output data are calculated by output residual value and the kernel density function.Under the application precision of industrial process,the method can assign appropriate weights to samples without iteration,which can reduce the amount of calculation and effectively improve the efficiency of modeling.Finally,the effectiveness of the proposed method is verified by simulation experiments.(2)Based on the robust modeling method,the operating performance assessment method based on weighted kernel total latent structure projection(WKT-PLS)for complex industrial processes is proposed.This method combines the KDE-KPRM method with the KT-PLS method,uses the KDE-KPRM method to pre-process the process data to remove noise and reduce redundancy of information,so as to reduce the impact of data disturbed by noise on the evaluation model.Then,the KT-PLS method is used for feature extraction of process data to obtain the process information that can directly reflect changes,and an off-line evaluation model is established.Finally,the proposed method is applied to dense medium coal preparation process.Based on the analysis of the influencing factors of the operation state of the dense medium coal preparation process and the actual operation data of the coal preparation plant,the off-line evaluation model is established.In order to ensure the reliability and real-time of the evaluation results,a robust online evaluation strategy based on forgetting weights is proposed,which combines sliding window technology and forgetting weights.The effectiveness of the proposed method is verified by the simulation experiment of dense medium coal preparation process.