We organize a scientific seminar on "Soft Sensors: Applications and Rumination". The scientific seminar will be led by Prof. Yuri Shardt - a full professor at the Department of Automation Engineering, Technical University of Ilmenau, Germany. The scientific seminar is organized in the framework of the FrontSeat project, as part of the seminar series on "Research Seminar on Smart Cybernetics".
Bio: Prof. Dr. Yuri A.W. Shardt is the chair of the Department of Automation Engineering at the Technical University of Ilmenau, Germany, working on the development of advanced system identification and fault detection and isolation methods for application to industrial problems. He teaches courses in automation engineering, statistics, and system identification. He has worked at the University of Waterloo, the University of Duisburg-Essen as an Alexander von Humboldt Fellow, in the ceramics and glass-making industry, and at the University of Alberta. In these jobs, his research focused on the development and implementation of advanced control methods for complex and uncertain systems. He has written over 50 journal papers and a book called Statistics for Chemical and Process Engineers: A Modern Approach (now in its second edition), which has also been translated into German. He has also presented numerous conference papers. He is also an associate editor for the journal ISA Transactions.
Abstract: As the world becomes increasingly interconnected and tightly controlled, the need to measure each and every variable becomes increasingly important. Unfortunately, not every variable can be measured accurately in real time, for example, concentrations of complex, multiphase mixtures or the chemical properties of films may be extremely difficult to measure accurately in real time as quickly as necessary for process monitoring and control purposes. One common solution is the use of soft sensors that take the available process information and provide a forecast or prediction of the difficult-to-measure variables. Soft sensors are essentially a mathematical model relating the easy-to-measure variables with the difficult-to-measure variables. These models are developed using methods ranging from simple regression analysis to the most complex machine learning and artificial intelligence. However, not only must the models be accurate, but the configuration of the soft-sensor system within the overall process must be considered. An improper configuration can lead to poor overall soft-sensor forecasts. Additional concerns include updating the models as the underlying process changes over time. Such methods such as adaptive learning, just-in-time modelling, or re-identification can be considered. This presentation will focus on providing an overview of soft sensors, their application, and future directions.