In March 2011, aircraft from Britain’s Royal Air Force joined an international task force patrolling the skies above Libya, enforcing a United Nations-mandated no-fly zone as the regime of Libyan leader Muammar Gaddafi crumbled.
Among the aircraft making up the task force: 10 Eurofighter Typhoon multi-role fighters rapidly dispatched to Italy from their home airbases of RAF Coningsby in Lincolnshire, and RAF Leuchars in Fife, Scotland. Having flown to Gioia del Colle Air Base, near the country’s heel, on March 20th, along with support technicians and provisions, the Typhoon aircraft flew their first-ever combat missions on March 21.
Helping to keep the aircraft airborne and fighting were advanced predictive modeling techniques, deployed by a team of personnel from defense contractor and aircraft manufacturer BAE Systems, targeted on identifying likely maintenance needs.
Equipment maintenance is a growing area for predictive analytics, as companies such as IBM and GE use the ability to collect and analyze sensor data from machines (and parts of machines) to detect wear and predict when a key mechanical piece might fail. Apache Corp., an oil company, uses such algorithms to prevent the failure of essential underground and underwater pumping equipment to avoid spilling valuable crude.
Fighter jets are another example. But as well as relying on on-board sensors and instrumentation to flag maintenance needs, BAE Systems is also using Witness analytics and simulation software from Lanner. The team’s role is to model each Typhoon’s operational life, incorporating its maintenance history and the type and location of the missions it flies, and use this information to predict the associated consumption and re-supply requirements of the relevant spare parts—predicting maintenance requirements before on-board instrumentation signals a need for action.
Part of an initial five-year £450 million (about $700 million in today’s dollars) contract for maintenance outsourcing signed with Britain’s Ministry of Defence in 2009, the use of predictive modeling underpins an initiative to save the taxpayer £2 billion over the 25-year anticipated lifespan of the Typhoon contract.
Specifically, explains Geoff Pickering, head of modeling and simulation within the Typhoon Availability Service at BAE Systems, the predictive modeling team has two main objectives. First, it aims to reduce the cost and backup inventory requirements of aircraft maintenance—and second, it helps to ensure that when aircraft are deployed in combat roles, sometimes far from their home British airbases, aircraft uptime is maximized.
And as simulation and analytics challenges go, the Typhoon contract cuts right to the heart of how predictive modeling builds on predictive analytics and simulation in general, Pickering says.
“You can’t use predictive analytics in a conventional manner, because there probably won’t be any applicable past data relating to the operational conditions that you’re interested in,” he says. “Instead, what you have to is take the attributes of the previous scenario, and extrapolate them into the present through modeling and simulation, rather than conventional analytics.”
Consequently, Pickering explains, the team—headquartered at RAF Coningsby—models the service and operational life of each aircraft. The starting point: the hours that each aircraft flies, and where and how they are flown.
A Typhoon aircraft, for instance, might fly for 20 hours a month on training sorties from Scotland’s windswept RAF Leuchars, a mile or so from the North Sea. But a mission to Italy, repeatedly flying over the Mediterranean Sea to the desert conditions of Libya, subjects an aircraft’s engines, airframe and systems to a very different operational environment.
“The failure rate of aircraft components changes as you operate in different conditions—and the locations in which the aircraft are based will impose very different re-supply conditions, in terms of lead times and required inventory,” says Pickering. “You’ll need to have more spare parts available locally, in order to minimize the risk of not being able to fly the aircraft, but also be able to make the right decisions about how to source those spare parts.”
Simulating Wear on 2,000 Components
Using Witness, the BAE Systems team runs between 50 and 500 simulations each quarter, depending upon operational activity, modeling high volumes of complex variables to provide up-to-date information to drive better management and maintenance decisions. In all, 2,000 components are simulated on each aircraft, with each iteration of the model generating a 40-megabyte output file which becomes the input for the next run of the model.
“We do between 50 and 500 simulations, and so use between 2 gigabytes and 20 gigabytes of data storage for each option that we process,” says Pickering. “Without the use of simulation on this scale, the kinds of projections we’ve been able to achieve would simply be impossible. The team can view projections of the Typhoon service as it stands at any given time, and how changes will impact cost and operational achievement.”
For Lanner, a former subsidiary of AT&T, and a business with a 35-year history of developing computer-based simulations, BAE Systems’ use of its technology for predictive modeling is part of a growing trend.
“Traditional simulation technology takes historical data and builds a model that works in the present,” says Ian Dabney, product director at Lanner. “Predictive simulation allows you to run that model into the future, but still using historical data—but predictive modeling, on the other hand, is about making changes in the model to predict the effect of trends and changes in the environment.”
And when dealing with multi-million dollar assets at the end of long and complex supply chains, those predictions can have far-reaching consequences—not all of which have a dollar value, or which can be disclosed publicly, says BAE’s Pickering.
Certainly, he acknowledges, predictive modeling with Lanner’s Witness technology has identified capacity constraints within the provision for Typhoon maintenance—and has done so in time to put new plans in place to smooth the maintenance plan, and thereby maintain the RAF’s Typhoon capability at optimum levels.
“Through being able to make quantifiable decisions, our ability to boost service and improve operational readiness has increased significantly,” he says. “A redesign of our support solution at RAF Leuchars, for instance, reduced costs while maintaining the same level of service.”
And in situations such as the Libyan deployment, modeling has helped to answer questions such as whether it’s better to carry out component repairs on-site in Italy, for example, or ship them back to the UK—and in the process deliver cost savings.
Aircraft uptime has also improved, adds Cliff Wren, of the Defence Equipment and Support division at the Ministry of Defence.
“[There has been a] measurable performance improvement,” Wren says in an official statement. “When we started the contract, things like the ‘Aircraft on Ground Awaiting Spares’ metric was typically in double digits, whereas now it’s around 4 percent. That is a dramatic improvement, and it has been delivered with significant cost reductions.”
Envisaged to run for the entire 25-year life of the Typhoon, the first five-year contract has already passed a major milestone, when BAE Systems completed the first-ever major maintenance on a Typhoon aircraft, after it had logged 1,600 flying hours.
The aircraft in question? ZJ921, from 3 Squadron, based at RAF Coningsby—and one of the 10 Typhoons sent to Gioia Dell Colle in 2011, to patrol the Libyan no-fly zone.
Malcolm Wheatley, a freelance writer and contributing editor at Data Informed, is old enough to remember analyzing punched card datasets in batch mode, using SPSS on mainframes. He lives in Devon, England, and can be reached at email@example.com.