Simulation Supporting a Tender Application by Wayne O'Rourke

I was recently asked by a client to help them prepare for a tender application to provide coaching transport for a major airport. Previously, the client had won the tender for another major airport but when they came to actually running the service they found out very quickly that they had miscalculated their needs to meet the service levels they had committed themselves to.  This time they wanted to be prepared!

Like any project we started by defining the requirements of the simulation by identifying the results the simulation would need to produce to allow us to compare different options.

Fundamentally this problem came down to a basic optimization:

How many coaches (and what size) do I need to ensure that passengers wait no more than 10 minutes to be collected from a stop?

By answering this question and taking into account the cost of different size coaches we would be able to calculate the costs necessary for the tender application.

However like all good problems it’s never as easy as it seems!  In this case we had four clearly different demand profiles that we needed to satisfy:

1. Staff arriving at dedicated car-parks needing to be shuttled to the terminals.
2. Passengers arriving at a number of different car-parks (spread all over the airport) to be shuttled to the terminals.
3. The movement of staff from Terminal to departing Aircraft & arriving Aircraft to Terminal.
4. The movement of passengers from Terminal to departing Aircraft & arriving Aircraft to Terminal.

Each of these profiles varied by time of day, day of week and week of year!

I’m sure you’re familiar with similar types of issues in your own industry.  When you start collecting the data and analyzing it the task can seem over-whelming.

In this case I had an easy approach to take.  It was essential for the tender application that we could maintain service levels throughout the year whether the airport was busy or quiet.

For each of our four demand profiles we produced a “Busy”, “Quiet” and “Normal” profile.  These could then be fed into the simulation to understand the utilization of the resources (and subsequent costs) over different fluctuations in demand.

The building of the simulation was relatively straight-forward:

• I used four work entry points to simulate the arrival of passengers by demand profile (sourced from EXCEL);
• Each work item that entered the simulation was routed to a collection point (about 200 in total);
• I simulated the coaches using Work Items with labels to record occupancy, current location and next destination.
• Storage bins with “Minimum Wait Time” were used to simulate the movement of the coaches from one location to the next (the time dependent on the collect point and the next-drop-off point);
• We wrote an algorithm in Visual Logic to mimic the decision criteria for assigning coaches to jobs

After verifying the simulation and proving to the client the simulation was following the rules they specified, we ran the simulation with the different demand profiles.

We started with the busiest demand profiles for all four demands to give us the “worst case” in terms of the number of coaches required that would be needed.  We ran this scenario with a number of different coach sizes.

We then repeated the exercise with the lowest demand profiles and this provided upper and lower numbers of coaches required.

As we continued with our experimentation we discovered that it wasn’t a “one-size-fits-all” circumstance and we began to fine tune the numbers and sizes of coaches by each demand profile.

After much experimentation (and use of OptQuest) we produced an ideal fleet of coaches and a plan for how this would vary through the course of the year.

We tested the robustness of the fleet by running hundreds of trials where we varied the demand and arrival profiles as well as journey time and maintenance regimes until we were finally satisfied that we had a robust solution.

This solution was used as the back-bone of the tender.

The Airport operator hasn’t yet made the decision of who to award the contract to but at the very least my client has now has confidence in their proposition and they have a tool that they can use again in the future for other contracts, just by changing the data!

If you have any questions generated from, or comments about this newsletter then email Newsletter@SIMUL8.com