Real-Time Assessment and Diagnosis of Process Operating Performance

Distributed Operating Performance Assessment of the Plant-Wide Process Based on Data-Driven Hybrid Characteristics Decomposition

“…The above phenomenon is termed nonoptimal, and it is different from the process fault, which is an unpermitted deviation of the system from the acceptable operation conditions. In the past decade, more and more scholars have devoted themselves to the study of data-driven assessment of industrial process operation performance. − Ye et al proposed a Gaussian mixture model-based operating performance assessment method for multimodal industrial processes, in which the performance assessment was equivalent to the classification issue; nevertheless, it may not necessarily obtain sufficient historical data of non-OOP in practical production processes (PPPs). Aiming at this situation, the multivariate statistical process monitoring technique whose essence is data description or one-class classification provides a good solution.…”

“…In terms of extracting dynamic characteristics of processes, canonical variate analysis (CVA), − dynamic principal component analysis (DPCA), , and dynamic partial least square demonstrated superior performances. However, the processes were often assumed to be either static or dynamic in advance in the aforementioned methods, − which might be one-sided for many cases. Once a static assessment model was established for the dynamic process, the assessment accuracy would be greatly reduced and vice versa.…”

Operating Performance Assessment for Transition State of Industrial Processes

“…Partial least squares were used to predict the CEI of a steady state, and the weighted average of the CEIs based on similar historic transitions was treated as the CEI for a transition mode. Sedghi and Huang [11] presented a systematic assessment and diagnosis technique for the operating performance of industrial processes, in which mixture probabilistic principal component regression was employed to tackle the behavior of the steady-state modes, while the dynamic principal component regression was applied for investigating the transitions among different modes. Then, both steady state and transition state were evaluated based on an optimality index.…”