Supply Chain Scheduling - White Paper
Transform the monthly LP Plan into a Feasible Schedule
Keeping Your Optimal Operations Plan On Track
It is a familiar scenario in the refining and petrochemical world. Companies have done a significant effort in adopting advanced planning and scheduling systems. Optimization models based on mathematical programming have been in use for many years and they have become part of the standard procedure to generate the desired operation plan. One could argue that almost every company in the business has the ability to create an economically efficient operations plan to balance supply and demand.
It could be observed as well that very few companies have the ability to execute their operations while keeping a close adherence to their optimal plan. Immediately after the operation plan is released for execution, the critical activity of managing the environment variability, detecting events that disrupt the coordination and adjusting the actual operation to keep adherence to the planned targets, takes center stage.
Managing Scheduled Operations
There is wide recognition on how critical the task of achieving the expected benefits of optimally planned operations is. After all, what's the value of an optimal plan that fails to be executed? Despite this strong argument, a systematic procedure to perform this task is seldom found in practice. Neither the systems used to generate the optimal operations plan, nor the manufacturing execution systems provide adequate support.
Even those companies that have reached a level of sophistication with Advanced Planning and Scheduling Systems and Advanced Plant Control recognize they have a broken link among these two functions. When doing event management and exception handling, Operation Managers are poorly supported and prompted to make decisions with a rather high degree of improvisation.
On exploring this apparent gap in the business process, the first thing that should be addressed is, why isn't the system that is used to create the optimal operations plan fitted to support the managing of the execution.
There is a tempting idea along this line: "we already have a model for creating an optimal operations plan, so if something goes different as planned, let's run it again." If the change in the environment has significantly modified the basic assumptions driving the economics of the previous plan, there is no doubt that a new optimal operation plan must be generated. However, a much more frequent scenario is that the desired targets remain the same but the operation execution is challenged by unpredictable variations that hinder their completion as planned.
Most of the time, these variations are associated to logistics operations both in feedstock replenishment and finished products expedition. As compared with process unit operations, logistics operations are far less reliable and have inferior control capabilities. They are the source of most of the disruptive events and where the need for adjusting the schedule becomes a daily task. When facing this sort of operational variability, experienced managers will try to take surgical actions to fix the on-going operations plan and put it back on track. Making a complete recreation of the optimal plan is not the right answer.
The model used to decide an optimal operations plan is not suited for making short-term adjustments to on-going operations. It is a model designed to capture the economical trade-offs by deciding on variables allowed to move within feasibility boundaries. Any optimal solution will have many of those decision variables actually reaching those bounds.
Typically, operation managers will never plan to operate too tight against true operational constraints, especially those involving some sort of operational risk. Therefore the boundaries in the model for operations planning will be set to define a "comfort zone" for smooth operation rather than expressing the actual operational constraints. However, in order to cope with unexpected events that disrupt the plan under execution, they will be willing to expand the boundaries, even outside of their comfort zone, for short periods of time or to enable limited actions.
Adjusting operations by expanding operational boundaries in a limited and controlled way is a critical feature to support event management during execution. The models used to create an optimal operations plan are too rigid to accommodate such flexibility. Even if the model already includes some parametric constraints that allow adaptation by the user, it is hard to anticipate where and when to expand a boundary in order to circumvent a problem unless you can simulate the scheduled operations under the new conditions to detect the infeasibilities and assess their magnitudes. Thus, the model that was suited for deciding an optimal operations plan is unfitted for adjusting the short-term operation to support exception handling. But at the same time, proper analytical support is crucial to making the necessary corrections. These actions are meant to reduce the disruptions aiming to keep "as close as possible" adherence to the targets, distributing the impact of the disruption among the many buffers that any rational plan does provide.