ABSTRACT: Due to the increasing focus on energy efficiency, environmental regulations, and market competitiveness, developing technologies for innovative process operation is an active research subject in academia and industry. Process Systems Engineering (PSE) technologies play an important role in addressing these challenges. Since most of advanced PSE methodologies use model-based algorithms, developing a representative and reliable process model is a crucial step in utilizing advanced technologies in process industries. Major challenges in developing process models are complexity, dimensionality, nonlinearity, and time-varying properties of processes. An efficient approach to overcome these problems is to use reduced-dimension data-based latent variable modeling (LVM) techniques. In this seminar, latent variable models and their contributions to different PSE areas will be reviewed. First, the application of LVMs for multivariable process data analysis, process monitoring, and fault diagnosis will be presented. Then, a novel Model Predictive Control (MPC) algorithm based on LVMs will be introduced (named LV-MPC). The LV-MPC is a system of advanced MPC algorithms including different modeling and control alternatives. Users can choose the appropriate algorithm for any specific application. LV-MPC addresses special characteristics of batch processes while being applicable to continuous processes. Furthermore, the subject of identifiability of LVMs for batch processes will be investigated. Finally, theoretical and industrial open problems and future research directions will be discussed.
Bio-sketch:
Dr. Masoud Golshan is a real time optimization and data reconciliation engineer at Royal Dutch Shell. He has been a technology leader at Shell since 2013 and has been globally recognized for his contributions to technology development and implementation at Shell. Dr. Golshan received his Ph.D. from McMaster University under the supervision of Prof. John F. MacGregor, a world leader in multivariate statistical data analysis. Masoud has conducted fundamental research studies in the areas of multivariate statistical data analysis, process modeling, system identification, and process control. He has authored several refereed journal publications and has several notable research and teaching accomplishments. He is an expert in implementing applied multivariate statistical data analysis in process industries. Prior to joining Shell, he worked as an NSERC industrial post-doctoral research scientist in pattern discovery technologies (PDT) Inc., a company rooted in the department of systems design engineering of the University of Waterloo. He developed advanced algorithms for process monitoring and fault diagnosis based on different data mining and reduced dimension data driven modeling algorithms for PDT. He has also collaborated with Carnegie Mellon University as a part-time post-doctoral fellow. Dr. Golshan has completed his M.Sc. in real time optimization of plant-wide processes at Sharif University of technology. His undergraduate degree is in chemical engineering.