GE Research (GRC) is General Electric’s corporate research group, based in Israel. It supports the work of GE’s business groups by performing research aligned with GE’s business needs.
Among the research interests of the GE Research team is collaborating with GE’s Aviation business group to innovate in the domain of jet engine maintenance. With more than 33,000 engines in commercial service, GE is a world leader not just in engine manufacturing but in also in jet engine maintenance, repair, and overhaul (MRO).
Within KYKLOS 4.0, GE Research focus is on overhaul, where a jet engine is sent to an overhaul facility (GE has several such facilities around the world). Broadly speaking, the overhaul process consists of the following steps. First, the engine is disassembled into its component parts. Second, engineers inspect each part and decide whether the part is fully functional, needs some repair, or needs to be scrapped and replaced with a new part of the same type. Following that, parts that need to be repaired are sent for repair in the facility’s part-repair area or to external facilities, and parts that need to be replaced trigger an order for a replacement part. Lastly, the parts are reassembled into an engine, and the engine is tested and sent back to the customer.
This use case involves two intertwined processes: part failure prediction, and part repair scheduling.
- Part failure prediction: One potential cause for delay in MRO is the lack of replacement parts. Such parts are expensive so storing a high level of stock for each of the thousands of parts is prohibitively expensive, as well as not matching the goal of low-waste circular manufacturing. However, lead times for ordering parts through the supply chain might be longer than the time in which GE committed that the engine overhaul would be completed. Therefore, it would be very valuable if we could predict which parts would be needed by each engine, months before the engine comes into the shop. Such prediction will give the facility enough time to order the necessary parts. While we cannot expect the prediction to be completely accurate, even a reasonably probable prediction could be valuable. To perform the prediction, data is collected from each engine after each flight. Once the overhaul date for the engine is decided, we run the prediction and inform the overhaul facility what parts it may need to order.
- Part repair scheduling: Another potential cause for delay in MRO is the finite capacity of the part repair resources. These resources include highly experienced engineers, as well as complex and expensive machinery. Scheduling the repair processes is challenging, since tens of engines may be undergoing overhaul, each being disassembled into thousands of parts. Each of those parts has a certain sequence of maintenance operations it needs to undergo.