Control Engineering Practice 22 (2014) 163–164
Contents lists available at ScienceDirect
Control Engineering Practice journal homepage: www.elsevier.com/locate/conengprac
Editorial
Editorial overview
Despite over twenty years of commercial applications of APC technology on various industrial processes, there are still many unresolved challenges at every stage of application design and deployment (controller structure design, test signal design, systems identification, ensuring stability, handling infeasibilities, performance assessment, etc.). APC practice has become partly ‘an art’ because of many unpublished/unreported/scantily referenced ‘best practices'. Meanwhile, the number of new processes that are being automated and optimized every year is growing exponentially and so are the challenges associated with successful APC application development and deployment. In the process of large-scale industrial APC application deployment, design of step testing signals for generating representative data to build high-fidelity control relevant models forms the first and very important stage. Improperly designed step testing, which may last for extended periods of time, using deliberate process parameter perturbations during routine process operations, might result in significant losses in profitability. Darby et al. reviewed the current status of industrial practice on test signal design in their paper. They proposed a multivariable step testing methodology with correlated signals utilizing D-optimality. Multiple case studies have shown that a moderate amount of input correlation results in improved parameter accuracy compared to an uncorrelated input design. The next step in APC deployment is control-relevant model development. Despite many advances in model predictive control relevant identification (MRI), handling nonsquare and ill-conditioned systems still pose a challenging task. Liang et al. proposed a generalized predictive control (GPC) scheme under a dynamic partial least squares (PLS) framework. A methodology involving PLS framework for identification and control is presented. With decoupling MIMO system into several SISO sub-systems alongside consistent objective functions in MPC and MRI, they showed that this methodology can naturally handle non-square and ill-conditioned systems better via few simulation examples. Competitive pressure, prohibitively increasing costs of energy and tighter environmental regulations made it imperative to explore various avenues to identify potential areas of improvement in process industry. Identifying trends that lead to performance degradation in this context is also important in carrying out preventative maintenance. Because of these reasons performance monitoring has become an integral part of APC applications. Few recent developments in this area are presented in the special edition. Zhang et al. presented their work on modeling and monitoring of nonlinear multi-model processes based on subspace separation. The process is modeled using the input–output dataset as a combination two subspaces: a common subspace which is invariable across 0967-0661/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.conengprac.2013.09.001
the modes, and a specific subspace which is specific to particular mode. This subspace separation is used to establish an integrated monitoring system, which would simplify the monitoring model structure and enhance its reliability. An experimental validation is performed on extracting the inherent process-related features across multiple data sets in the electro-fused magnesia furnace (EFMF). Wang et al. presented an online monitoring approach using filtering kernel independent component analysis-principal component analysis (FKICA-PCA) which can be applied on nonlinear multivariate industrial processes. As a part of FKICA-PCA, a method to calculate the variance of independent component (VIC) is proposed to make Gaussian features and non-Gaussian features comparable and select dominant components legitimately. Genetic Algorithm is used to determine the kernel parameter. Then exponentially weighted moving average (EWMA) is used to filter the monitoring indices of KICA-PCA to improve monitoring performance and a contribution analysis scheme is developed for FKICA-PCA to diagnosis faults. The feasibility and effectiveness of the proposed method are validated on the Tennessee Eastman (TE) process. Alongside the algorithmic and methodology advancements the number of new processes that are being automated and optimized every year is growing exponentially and so are the challenges associated with successful APC application development and deployment. Such two novel APC applications are presented in this special issue, including hematite benefaction control and ethanol fermentation control. Hematite ore exhibits low grade, fine-grained and non-homogenous distribution, which causes frequent fluctuations in its key variables during regrinding process. These fluctuations lead to reduced classification efficiency of hydrocyclone and pose a complex control challenge. Zhao et al. proposed and successfully applied a hybrid intelligent control (HIC) method to a regrinding process in a largescale hematite beneficiation plant. The HIC includes a fuzzy switching controller of the sump level interval, a Multi-PI switching controller of the hydrocyclone feeding pressure and a conventional controller of the hydrocyclone feeding density. Sporadic quality variable measurements alongside sensitive microbial growth makes ethanol fermentation process control and optimization a challenging task. Hahn et al. proposed an energy balance based fermentation model, which can exploit the available continuous online temperature measurements to infer and optimize fermentation yields. It is shown on simulation studies that modifications of the input profiles for the cooling rate and the glucoamylase addition can lead to an approximately 10% increase in ethanol yield. Classical model predictive controller execution usually tracks the targets from real time optimization (RTO), which is based on steady state model of the process. In recent years, several Economic MPC
164
Editorial / Control Engineering Practice 22 (2014) 163–164
(EMPC) approaches were proposed in literature to combine economic optimization layer with feedback control, utilizing dynamic models for economic optimization. Despite the demonstrated closed-loop economic performance improvement, single layered methodology for economic optimization and feedback control makes these strategies difficult to apply with existing control architectures. While combining the benefits of closed-loop stability, feasibility imparted by Lyapunovbased techniques and easy adaptability imparted by a two-layered EMPC, Christofides et al. in their paper presented a conceptual framework for integrating dynamic economic optimization and stable linear model predictive control for optimal operation of nonlinear process systems. Via extensive sets of simulations, using a continuously stirred tank reactor (CSTR) chemical process model with a timedependent economic cost function, they validated their claims that the integrated control framework can result in a stable closed-loop timevarying state evolution and can perform economically better than traditional RTO with steady-state economic optimization. On the other hand, controller performance assessment becomes increasingly important. Liu et al. presented a performance assessment methodology for decentralized control systems based on a general quadratic performance index involving both system states and inputs. The proposed methodology involves an iterative optimization framework while accounting for the block diagonal structure of decentralized systems. The applicability and effectiveness of the proposed approach is demonstrated on two illustrative examples. As evidenced by the contributed theoretical and application papers, model-based control and monitoring is an active area of research and development. The focus of this special issue is to highlight contributions on the latest developments that overcome known limitations and advance the technology. The papers that have been collected involve innovations for optimal step testing,
model identification, economic control, controller assessment, multi-model methods, and novel applications in ethanol production and ore refining. This collection and others in the future will continue to advance the state-of-the-art in control practice.
Guest Editors Jie Yu n Department of Chemical Engineering, McMaster University, Hamilton, Canada E-mail address:
[email protected] Srinivas Karra 1 Applied Manufacturing Technologies, Houston, TX, USA E-mail address:
[email protected] Rui Huang 2 United Technologies Research Center, East Hartford, CT, USA E-mail address:
[email protected] John Hedengren 3 Department of Chemical Engineering, Brigham Young University, Provo, UT 84604, USA E-mail address:
[email protected]
Available online 27 September 2013
n
Corresponding author. Tel.: þ 1 905 525 9140x27702. Tel.: þ 1 806 282 7626. 2 Tel.: þ 1 860 610 7655. 3 Tel.: þ 1 801 477 7341. 1
