Abstract 
Maintenance scheduling is the central theme of this presentation. Maintenance, preventive or restorative, plays a crucial role in the smooth and safe operation of many industries, such as the chemical process industry. Essentially, maintenance has to be scheduled in a cost optimal way so that an item of equipment can be planned for 'cleaning' or replacement ahead in time. The usual way such schedules are implemented is by a “receding horizon” planning, in which a long time horizon of maintenance actions is considered, and then taking the first period actions and implementing them. The process is repeated continuously so as to account in this way long term effects of costs and ensure operability of the process.

Considering maintenance scheduling in its basic form, without other operational constraints and decisions other than “what to maintain and when to maintain it”. It is shown that this basic scheduling problem can be rendered as a multiperiod optimal control formulation, in broad general terms. This in turn allows for its solution using standard optimization tools. Since the decisions made in such an optimization problem are of “on-off” nature (binary in other words, deciding to restore/replace or not some unit at a given period of time) this is a Mixed-Integer Nonlinear Programming problem. Such problems require special solvers, such as Branch & Bound, to address them and can be very costly for large scale applications as they are combinatorial in nature.

The observation for the basic maintenance scheduling problem is that the underlying optimal control problem derived is what is termed as “linear in the control”, which exhibits an optimal solution that is bang-bang, i.e. goes from bound to bound. Since the bounds are 0/1 this would solve the problem without the need for combinatorial solvers, hence allowing the efficient and general solution of complex nonlinear scheduling models of this nature. This is exploited in this work and the procedure is highlighted by the scheduling maintenance optimization of heat exchanger networks encountered in the chemical process industry. A complication of these models is that operational constraints may be present, which may affect the property of on-off optimal controls; this is addressed using practical heuristics in this work, while future work will focus on more rigorous approaches.

About the Speaker
Dr. Vassiliadis’ research interests lie in the development and application of optimisation and simulation tools in engineering and scientific domains.  His research field is Process Systems Engineering, a sub-discipline within Chemical Engineering.

He obtained his Diploma in Chemical Engineering (M.Eng.) in the School of Chemical Engineering at the National Technical University of Athens in 1989, having graduated with distinction and top of his class.  He then studied for his Ph.D. in Process Systems Engineering, in the Department of Chemical Engineering and Chemical Technology at Imperial College, London, from where he graduated in 1993.  He then spent a year working as a postdoctoral associate in New Jersey in the United States, at the Department of Chemical Engineering at Princeton University.

He joined the Department of Chemical Engineering at Cambridge as an Assistant Lecturer in 1995 and is now a Senior Lecturer, where he established and leads an independent Process Systems Engineering research group. He has acted as a consultant to AspenTech LTD for the development of an optimal control solver code, and his Ph.D. code for optimal control formed a prototype solver for gPROMS, the dynamic simulator by PSE LTD.  He is reviewer for several technical and scientific journals, in the areas of engineering science, applied computational mathematics and their overlapping domains.  Vassilis is also the CEO of a spin-out company, Cambridge Simulation Solutions.

Contact: This email address is being protected from spambots. You need JavaScript enabled to view it.

This event is part of the CyI Colloquium Series.  View all CyI events.