Assembly Line Soothsayers

by Michael Mullaney on August 10, 2009

Anyone who has ever suffered a flat tire while on the way to an important family gathering, or had their flight delayed, or had babysitter cancel at the last minute knows this all too well: time is money.

This old adage is amplified to an incredible degree in the world of service and manufacturing. Many automated production lines run for 16 or 24 hours a day, and any unplanned interruptions can results in serious overhead costs for the company. Not only do they have to pay someone to fix the problem and keep an expensive inventory of spare parts, they usually have to continue paying everyone else to wait around while the issue is resolved – and then pay those employees overtime to stay longer and help make up for lost time. Then things may lag for a while, schedules fall behind while everyone plays catch-up, and customers begin to grumble. The situation can spiral downwards quickly.

If only we had some sort of sixth sense or crystal ball to let us know ahead of time before things went awry.

That’s exactly the problem Professor Jennifer Ryan is trying to help solve. Her augury of choice, however, is nothing as arcane as a crystal ball – it’s just good decision engineering.

Ryan recently received a new grant from the NSF to further her research into exploiting data toward the goal of more efficient and less expensive management of manufacturing and logistics systems. Here’s how it works: many kinds of moving parts – whether they’re on an assembly line, or in a taxi, commercial airliner, tank, or submarine – degrade over time.  This means the operators of that machinery have to keep spare parts on hand, or on retainer, should be part in question degrade to the point of inefficiency or failure and need to be replaced.

One way to keep track of part performance is to develop a system for outfitting these degradable parts with sensors that can relay information back to a centralized computer. That means the system would know, or at least have a very good guess, how degraded any part is at any given time. This information would help to identify the most opportune times to replace parts or perform widespread maintenance.

Ryan’s challenge is to take this data, and create a mathematical model to identify the best way to manage the system. With the clear goal of minimizing cost for the operator, she is using the data to devise a set of rules to precisely identify the correct times to replace a part and order more spare parts. In this regard, the system is not unlike a crystal ball. Instead of making an educated guess, managers will be able to know exactly, without a doubt, the best time to take these actions.

To alleviate the need for purchasing a big, fast computer to crunch all of the numbers, Ryan is also working to simplify optimization rules so they are less computationally intensive.

The end result – more strategic and more cost-effective inventory and maintenance systems – saves both time and money. Sounds like a win-win situation to me.