Ensuring capacity for engine shop visits has not been a challenge during most of the pandemic, but engine work is picking up with the recovery, and labor and part supplies are always potential restraints. So engine shops must make smart plans before committing to overhauls.
Most planning systems now are determinate: managers make best estimates of the various overhaul parameters and simulate, or optimize, a schedule for meeting commitments to overhaul customers.
Now a new company, Singapore-based keepflying, is offering a different approach to engine shop planning. Its FinTwin, for Financial Digital Twin, uses historic shop data plus data on incoming engines’ utilization, environment, severity, LLP states, thrust ratings and minimum average thrust de-rates to project shop requirements. Then artificial intelligence (AI) drives dynamic capacity planning to optimize shop commitments and performance, while machine learning constantly adjusts and improves the accuracy of forecasts.
Chief Growth Office Chandresehkar Jayaramakrishnan explained how his system worked, and how it differs from traditional shop planning in a recent webinar. First, new tools are needed because engine maintenance has changed. “There is a different demand recently,” Jayaramakrishnan said. “Cash conservation is top of mind, so we are seeing a rise in shop visits but not in core performance restorations.”
Keepflying’s FinTwin depends on historic engine data, but not necessarily back-to-birth data. Accurate data on previous performance restorations allows machine learning to build algorithms to forecast both individual engine requirements and total shop requirements. Jayaramakrishnan recommends running the shop FinTwin twice a day to ensure plans are set based on the latest data on engine progress in overhaul lines.
The shop FinTwin can be run to seek three different objectives. First, it can seek an overhaul plan that maximizes shop profits. Or it can yield a plan that minimizes risks of exceeding contractual turnaround times (TATs). Or FinTwin can find a hybrid solution, of maximizing profits subject to avoiding exceeding TATs for certain customers. This hybrid approach might be desirable for a shop that seeks to increase profits while not disappointing certain long-time, critical customers.
The FinTwin model can also spot and find solutions to capacity problems. For example, it can simulate the financial and TAT consequences of bringing in workers at overtime rates. Or, it can simulate cost and timing effects of exchanges and swapping engine components, Or FinTwin can simulates the gains of outsourcing module work to third parties, thus increasing costs but speeding up work to meet TAT commitments. Using historic data, the FInTwin can even rate and recommend the vendor to be chosen.
The AI-driven approach depends on highly accurate historic data. “Quality is more important than quantity,” emphasized Chief Technology Officer Sudarsan Lakshmikumar. So the first step is what Lakshmikumar calls “data wrangling.”
Data is brought in from all sources and silos, from Excel spreadsheets to databases in the MRO’s enterprise resource planning or maintenance & engineering system. Keepflyng staff are very familiar with and can work with standard systems like AMOS, Trax and many others. Then this data is cleaned and errors removed, stored and reformatted to ATA standards, validated and tidied up in a multi-step, largely automated process. For instance, Another AI tool, natural language processing, is used to ensure data is consistent with correct definitions.
The entire process of getting an engine shop ready to use keepflying’s FinTwin capabilities takes 10-12 weeks. Four to six weeks are required for data wrangling. Setting up and configuring the model for each shops’ complicated operational and contractual arrangements takes two weeks. Testing the system and training users takes another two weeks, then testing security and other features requires another two weeks.
The engine shop FinTwin is available now on a software-as-service basis. Keepflying plans to introduce in 2022 similar FinTwins for engine lessors and asset owners and FinTwins for airframe shops and non-engine component shops. And the engine-shop FinTwin may exploit sensor data for incoming engines as well.
