Mani Bhushan

Biography

B.Tech., I.I.T. Bombay, 1997 Ph.D., I.I.T. Bombay, 2001 Postdoc, Purdue University, 2001-2002 Assistant Prof., University of Alberta, Canada, 2003-2005

Research Interest

Sensor Network DesignChoosing appropriate variables for measurement in the process is essential to a variety of tasks, such as process control, diagnosis, inventory management, process safety, and quality control, amongst others. The sensor network design problem is concerned with issues related to deciding which variables to be measured, the spatial location of the measurements (in case of distributed parameter systems), hardware redundancy (number of sensors used to measure any given variable), and sampling frequency of the chosen sensor. Currently, we are focusing on the tradeoff between issues such as accuracy, reliability and cost while designing sensor network problems. The problem has been posed as (implicit) multiobjective constrained optimization problem, and a systematically-tuned robust genetic algorithm is being used to solve the same. Another research effort is in the direction of understanding the tradeoff between the various sensor features for multirate systems in a state-space formulation (lumped parameter systems), and also for distributed parameter systems Optimal Alarm ManagementFor reasons of process safety and optimality, in a chemical process, alarm thresholds are set for certain variables. When such a variable crosses its threshold, an alarm is generated informing the process operator of some abnormality in the process. This research involves the development of an alarm generation (which variables should generate an alarm and what should be their threshold) and operation strategy (detection and diagnosis of the process abnormality based on the observed alarms) in a stochastic framework which minimizes the expected loss or maximizes the expected profit obtained due to a particular alarm configuration. Further, most approaches in literature are based on univariate data. Formulation of this alarm management problem in a multivariate framework is also of interest. Process Planning and SchedulingThe research interest in this area is to pose the process planning and scheduling problems in a multiobjective optimization framework, and generate the tradeoff between various issues (such as profitability versus inventory requirements). Another interest is in using smart enumeration techniques, such as constrained programming for generating all optimal solutions for some specific optimization problems. Process Safety AnalysisThe interest in this area is to automate the procedure of process safety analysis. In particular, use of �enhanced� digraph models (enhanced by considering time delay and time constants for various arcs) for this purpose is being investigated.