Concluding Remarks

Plant turnarounds and maintenance activities are essential to continued reliability and productivity. At the same time, they also represent production disruptions for significant periods of time, and this effect is exacerbated in large chemical site networks. Though each plant or envelope in a site runs its own preferred routine and preventive maintenance over short time horizons, planning and coordinating large-scale preventive maintenance tasks across the entire integration network has significant potential benefits^[1,2].

We propose a methodology of how to make adjustments to an existing turnaround plan in order to respond to unexpected changes in the system over a short and medium-term horizon. In practice, plants are constantly monitored as the processes and markets change with time. The states of the systems may diverge more than that anticipated at the time of the original turnaround planning. For example, a plant may maintain high catalyst activity longer than usual and the rescheduling of an upcoming turnaround may have the highest economic benefit for that individual plant, but other plants in the network may actually be impacted by a delayed turnaround. In another scenario, the demand for products and market prices for raw materials could change sufficiently to prompt ramp-up in production. The converse market conditions would prompt a ramp-down in production and also change turnaround economics, which are governed by both direct resource costs of turnarounds, as well as loss in profits due to down-time. Further, the time value of money of future delayed turnarounds, sunk costs towards the originally scheduled turnaround, and reliability consequences of altered schedules all make the implementation of an alternative turnaround schedule a significant financial decision. 

This presentation will describe an extension of earlier work^[3]. We include comparisons of different notions of financial risk and weigh their relative benefits. The incorporation of risk measures is particularly delicate in the context of multi-stage stochastic programs, as we may need to consider a convolution of risk functions over all the stages, as well as the issue of time-consistency of the measures. We employ recent methods from the literature that tackle these issues.

Further, while our previous study focused on the adjustment of a single turnaround using a multi-stage stochastic linear program, we also extend the problem to look at multiple possible adjustments, giving rise to a multi-stage stochastic mixed-integer program. We discuss modeling and computational challenges, and discuss results from an industrial case study.

References:

[1] S. Amaran, N. V. Sahinidis, B. Sharda, M. Morrison, S. J. Bury, S. Miller, and J. M. Wassick. Long-term turnaround planning for integrated chemical sites. Computers & Chemical Engineering  72, 145-158, 2015.

[2] S. Amaran, T. Zhang, N. V. Sahinidis, B. Sharda, S. J. Bury. Medium-term maintenance turnaround planning under uncertainty for integrated chemical sites. Computers & Chemical Engineering, doi:10.1016/j.compchemeng.2015.09.007.

[3] S. Rajagopalan, N. V. Sahinidis, B. Sharda, S. Amaran, and S. J. Bury. Flexible turnaround planning in integrated chemical site networks. 2015 AIChE Annual Meeting, Salt Lake City.