Predictive maintenance usually summons up visions of massive data volumes and complicated, hyper-sophisticated predictive algorithms. But before any of this happens--or at least happens with maximum effect--thorny questions of data sharing must be resolved. Rolls-Royce has now taken two major steps to answering these questions, by partnering with SITA and IFS-Maintenix, according to Manager of Digital Services Nick Ward.
Rolls has been providing TotalCare engine support for two decades but wants to make this support even better, saving both itself and its airline customers money. The old TotalCare was based chiefly on snapshots of engine performance from aircraft communications addressing and reporting system (ACARS). “Now we want continuous performance data, and engineering data, including maintenance data we don’t have, cancellations, delays, service bulletins, technical logs and reasons for MRO events,” Ward explains.
But to get that data, Rolls had to meet several airline objections, including questions about the benefits of sharing data for airlines, how airline data would be protected, pilot-union wariness about sharing quick access recorder (QAR) data and the burden imposed on airline IT departments of enabling data sharing.
Rolls had to reassure airlines on several of these concerns. But Ward says the last problem, IT burdens, has been addressed by a partnership with SITA, a neutral and well-trusted company, that will provide continuous engine operating data, and now another partnership with IFS-Maintenix, which will build plug-in application programming interfaces, or APIs, to retrieve engineering data for the aircraft it supports.
Since IFS systems now support about 20% of the commercial and military aircraft in the world, this partnership should substantially boost Rolls’s access to in-house-engineering data. About two-thirds of Rolls-powered aircraft are supported by just four MRO IT systems: TRAX, AMOS, Maintenix and SAP.
The other virtue of the Maintenix API is that Rolls-Royce can automatically and at no charge send useful data back to the customer airline, for example fuel use, deratings and estimated remaining lives of engine components.
The improvements in predictive maintenance from getting both continuous operating data and airline data could be dramatic. Ward estimates that Rolls could be up to 30% more effective in identifying engine failure modes. Presently, “we know 20% of the drivers of reliability,” Ward says. “If we can get data from all airlines we could know 100%.”
But the Maintenix agreement alone has enabled Rolls to get started by implementing the new sharing procedure with 25 airlines, and the first one will go live in November 2019. The next question is whether IFS or another IT vendor can provide the plug-in service for the other major MRO systems.
In any case, it has been a long journey toward total predictive maintenance, or "total" TotalCare, for Rolls and its customers. Two decades ago, Rolls started out by manually analyzing only 3 kilobits of data in those early ACARs snapshots. In 2013, it built diagnostic networks to automate analysis of many more bytes and bits. In 2016, Rolls began using artificial intelligence to spot new problems that had not appeared in historical data, a kind of “safety net” in Ward’s words. Now the OEM is collecting data from engines on about 30,000 aircraft and issues about 400 alerts annually. Because the OEM wants to be as cautious as possible in catching problems, it tolerates a false positive rate of 2-4%.
But there is a lot more to predictive maintenance than alerts. Ward divides the predictive effort into three parts. First, there is traditional engine health monitoring, mostly based on detecting engine deterioration. Next, Rolls can use batched data to recalculate the remaining life of engine parts, based on a more precise estimate of how much of an standard engine cycle has actually been flown in each actual cycle. ”Was it a real cycle or just 0.6 of a cycle?” Ward asks. This true cycle data can now be passed to airlines by Maintenix’s API to help airline managers understand upcoming overhaul costs or to keep engines on-wing longer.
Finally, Rolls uses engine data to simulate engine performance and project out 30 years when and how many engines will need off-wing maintenance in order to plan its own shop capacity and spare-part production.
Data sharing was hardly the only hurdle encountered during this long predictive journey. The massive amount of data to be processed required a cloud solution. Ward says it took Rolls six years to solve the various problems associated with moving all its engine data to the cloud, including data security, export controls, cost minimization and figuring out the best architecture to meet all these concerns.
Now that’s been done, and much more data is starting to flow, Rolls is ready to further power up its analytical capabilities.