Abstract: In many industrial control and diagnostic applications, a reliable mathematical model of the process is of great advantage in reducing component cost in achieving the design objectives. While reliable and dependent virtual sensors require more complex models and computing architectures, recent developments in the information theory as well as advances in multi-core computing have opened new gates to the dream of developing a real-time capability for adaptive and learning techniques implemented in many-core processors efficiently. We will present examples from the statistical theory, recurrent neural networks, and wave networks to compare and conclude the promise and pitfalls of each method. We specifically point out the advantages neural-net based techniques inherently offer in an efficient computational implementation given their naturally parallelizable structure. We will focus on specific automotive applications of the presented software techniques and hardware technologies.
Biography: Hossein Javaherian is a Technical Fellow in the Propulsion Systems Research Lab and a member of the Science Labs Technical Council. He has expertise in engine control systems and vehicle emissions control. His research focus is on advanced control algorithm development for next generation propulsion systems to fulfill GM's vision of mobility. He also has collaborative research projects globally with academic institutes and industrial organizations to expedite the development process. He has broad interest and experience in adaptive and learning systems for powertrain control, nonlinear systems analysis and control, acoustics and vibrations, embedded systems, neural networks for control and diagnostics, robotics and industrial automation. Hossein received his bachelor's degree in Mechanical Engineering from Tehran University of Technology, Iran, in 1970 and M.Sc. and Ph.D. in Control Systems from Imperial College, London, England in 1974 and 1978, respectively. He was a Research Fellow at Imperial College in 1978-79, a faculty member at Tehran University of Technology for 4 years and the Head of Power Plant Division at the Electric Power Research Center, Tehran, for 2 years before joining GM R&D staff in 1985. He holds twenty patents and has authored 45 technical publications. He is also the recipient of the prestigious GM "Boss" Kettering Award in 2000 and two McCuen Awards in 1996 and 1999.