Ravindra D Gudi

Personal Information
Full Name: Ravindra D Gudi
Room No: 243, CAD Center
+91 (22) 2576 7231 (O)
+91 (22) 2572 6895 (Fax)
Email Contact Form

Detailed Information / Research Group Web-Page


  • B.Tech., I.I.T. Bombay, 1985
  • Ph.D., University of Alberta, 1995

Awards & Fellowships

  • Chair, IFAC Technical Committee 8.4 (Since July 2014)
  • Associate Editor, IFAC Journal of Process Control (since January 2010)
  • Herdillia Award for Excellence in Basic Research in chemical Engineering (2009)
  • Manudhane Applied Research Award, Department of Chemical Engg, IIT Bombay. (2006).
  • Visiting Professor, Department of Chemical and Biological Engineering, University of Wisconsin-Madison, USA (2003-2004)
  • Lovraj Kumar Memorial Award for promotion of Industry Academia Interaction, (July 1998- January 1999).
  • Canadian Commonwealth Fellowship by the Government of Canada (1991-1995).


A list of publications is available in this link.

R&D Areas/Projects

  • Control relevant identificationSome of the work done in our group focuses on extending the control relevant identification methodologies proposed in literature to enhance their applicability to practical problems. In the first extension, we have looked at a two degree of freedom structure and analyzed issues related to control relevant identification for this structure. Further, steady state gain match (of the identified model) is an important requirement from stability and practicality viewpoints. Towards this end, we have proposed to extend control relevant identification to achieve steady state gain matching as well. For multivariable systems, we have proposed a relatively simpler extension of the SISO based control relevant strategy, to MIMO systems. Further, we have also obtained some interesting results in control relevant model reduction for nonlinear plants.
  • Nonlinear IdentificationFor nonlinear model identification, we have sought to extend the traditional canonical variate analysis techniques to iteratively capture the nonlinearities through bilinear approximations. We have obtained some encouraging results in this area. In a second approach, we have proposed decomposition of the operating space using fuzzy segregation techniques to obtain locally linear models. The predictions from these models are aggregated suitably and used for closed loop control. In a third approach, we have also looked at extending traditional black box model structures and enhancing their performance through appropriate apriori knowledge and input design. In a third approach, we have also looked at Just-in-Time modeling (rapid identification on demand) as a method of adaptive modeling and identification under closed loop for nonlinear chemical processes. Also, the data based CART (Classification and Regression Trees) approach has been extended through simple fuzzification to adapt it to nonlinear and multivariable systems.
  • Scheduling and Decision SupportFrom a productivity enhancement viewpoint, scheduling of various production operations has attracted lot of attention amongst researchers across various disciplines (e.g. Mechanical Engineering, Chemical Engineering) and has seen a variety of applications from job-shop type to flow-shop types of scenarios. In the chemical process industry, considerable work has been done in the area of scheduling of batch processes. Of late, there has been interest in scheduling of continous time chemical processes wherein the nature of the problems posed are a little different. Specifically, scheduling in serial or parallel lines have been considered. However, in a good number of chemical process operations, the processing could take place in mixed lines, i.e. both in serial and parallel lines. We have looked at scheduling strategies for such processes involving mixed lines. The considerations of such mixed lines has thrown up a number of interesting research problems which have been tackled through use of heuristics and multilevel decomposition. Of particular interest is the issue of handling NP-hardness that typically is posed up by these formulations in the solution of the resulting Mixed-Integer-Nonlinear-Programming (MINLP) problems.
  • Disturbance/ Fault AccomodationFaults in operating process plants bring in their own signatures in terms of the patterns in the operating data. To exploit this information and use it in fault diagnosis and detection, we have proposed to use clustering/ classifying techniques to characterize the abnormal/ aberrant operation separately from the normal operating region. Fuzzy segregation techniques have been employed for this purpose. Since the data from operating plants are inevitably correlated, these classification methods have been applied to the PCA-transformed space. We have obtained some interesting results for single and double fault detection of the Tennesse Eastman problem and other representative problems.
  • Optimal Control of Fermentation ProcessesEnhancing Lactic Acid Productivity using Simultaneous Saccharification and Fermentation. Enhancing oxygen transfer through optimal peroxide additionsFermentation processes are usually characterized by reactions involving feedback repression/inhibition mechanisms by the intermediates and products. Thus, to enhance the productivity of such fermentations, suitable manipulation of the nutrient or inducer additions are used. Strategies that have been currently proposed for determining optimal additions of the nutrients, are restrictive in their formulations and are difficult to employ when two or more nutrient additions are involved. We have looked at alternative strategies to determine the optimal policy for nutrient additions. Specifically, we have looked at two processes viz. the Simultaneous Saccharification and Fermentation of Starch using L. delbreuckii wherein, starch and glucose additions are considered as manipulative variables. In another application, we have looked at enhancing oxygen transfer through optimal additions of hydrogen peroxide. This latter application also posed an interesting problem in that there is a constraint on the permissible levels of the peroxide concentrations, i.e. an optimal control problem with a path constraint is posed. We have also obtained some interesting results by using evolutionary optimization techniques such as differential evolution methods to solve such problems that are typically difficult to solve using gradient based optimization techniques.