Search

Data-based process monitoring has become a key technology in process industries for safety, quality, and operation efficiency enhancement. This paper provides a timely update review on this topic. First, the natures of different industrial processes are revealed with their data characteristics analyzed. Second, detailed terminologies of the data-based process monitoring method are illustrated. Third, based on each of the main data characteristics that exhibits in the process, a corresponding problem is defined and illustrated, with review conducted with detailed discussions on connection and comparison of different monitoring methods. Finally, the relevant research perspectives and several promising issues are highlighted for future work.

Compared to the steady phase, quality prediction in the transition period of batch process has been explored as a difficult task in recent years. Unlike the data behavior in the steady phase, there may be significant correlations among the data samples in each transition period. Without considering the relationships among different time slices (time pieces in the batch dataset), the performance of the quality prediction model may be degraded. In the present work, an information transfer PLS model is particularly proposed for quality prediction in transition periods of the batch process. By transferring the main data information of one time slice to the next one, different time slices in the transition period are connected. As a result, most available data information in the previous time slices can be efficiently used for quality modeling of a specific time slice in the transition period. For performance evaluation of the developed method, a case study of an industrial injection molding process is provided.

Because of expensive cost or large time delay, quality data are difficult to obtain in many batch processes, while the ordinary process variables are measured online and recorded frequently. This paper intends to build a statistical quality prediction model for batch processes under limited quality data. Particularly, the self-training strategy is introduced and combined with the partial least-squares regression model. For multiphase batch processes, a phase-based self-training PLS model is developed for quality prediction in each phase of the process. The feasibility and effectiveness of the developed method is evaluated by an industrial injection molding process.

Typically, the operation condition of the batch process is changed frequently, following different recipes or manufacturing various production grades. For quality prediction purpose, the prediction model should also be updated or rebuilt, which leads to a significant model maintenance effort, especially for those processes which have various phases. To reduce such effort, a maintenance-free method is proposed in this article, which incorporates the setting information of the batch process for modeling. The whole process variations are separated into two parts: setting information related and other quality related variations. By constructing a relationship between setting variables and other process variables, the data variations explained by the setting information can be efficiently removed. Then, a robust regression model connecting process variables to the quality variable is developed in different phases of the batch process. The feasibility and effectiveness of the proposed method is evaluated through an industrial injection molding process. (C) 2012 American Institute of Chemical Engineers AIChE J, 59: 772-779, 2013

Typically, a multiphase batch process comprises several steady phases and transition periods. In steady phases, the data characteristics remain similar during the phase and have a significant repeatability from batch to batch; thus most data nonlinearities can be removed through the batch normalization step. In contrast, in each transition period, process observations vary with time and from batch to batch, so nonlinearities in the data may not be eliminated through batch normalization. To improve quality prediction performance, an efficient nonlinear modeling methodrelevance vector machine (RVM) was introduced. RVMs were formulated for each transition period of the batch process, and for combining the results of different process phases. For process analysis, a phase contribution index and a variable contribution index are defined. Furthermore, detailed performance analyses on the prediction uncertainty and variation were also provided. The effectiveness of the proposed method is confirmed by an industrial example. (C) 2011 American Institute of Chemical Engineers AIChE J, 58: 17781787, 2012

This paper develops a modeling approach to address the end-of-batch product quality prediction problem for batch processes with limited batch-cycle data. Generally, those batch processes that have multiple phases are the focus of the present paper. Different from the traditional multiway/phase-based partial least-squares (PLS) method, which unfolds the three-way data set through the batch direction, the proposed method unfolds the data set through the variable direction, in order to generate more training data samples. Reproducing the product quality data with the noise injection method allows a statistical model to be developed in each phase of the batch process. This, however, does not remove the nonlinearity of the batch process data, as practically addressed by the typical batch normalization. Therefore, a nonlinear regression model is subsequently introduced to handle this problem for product quality prediction modeling. To compare the performance of linear and nonlinear statistical models, phase-based PLS and relevance vector machine models have both been developed for prediction of product quality in an industrial injection molding process.

A unified least-square kernel (ULK)framework is formulated for both process modeling and fault diagnosis issues. As a specific algorithmic implementation of the ULK method, an adaptive kernel learning (AKL) network classifier is developed for process diagnosis, which is a two-stage online learning algorithm with a fixed-memory strategy. A new concept of space angle index is proposed to structure the growth of node and actively control the complexity of the network. The AKL network performs a backward decreasing for forgetting an old pattern and a forward increasing for incorporating a new online pattern. The recursive algorithms for both stages are derived for quick online updating. Applications of the AKL network to two numerical cases and the Tennessee Eastman process show good performance in comparison to other established methods, and new insights on the pattern recognition for fault diagnosis arising from this flexible classifier are highlighted. (c) 2006 American Institute of Chemical Engineers.