GRAHAM WARWICK / WASHINGTON DC
The USA's Defense Advanced Research Projects Agency aims to develop a system that can ask an airframe or engine how it really feels
Today's maintenance industry - airframe and engines, civil and military - is built around the principle of regular inspections at intervals determined during design and testing to provide the best chance of detecting and repairing damage before it becomes critical. That could change, if a programme launched by the US Defense Advanced Research Projects Agency (DARPA) succeeds in demonstrating technology to interrogate an airframe or engine and determine its fitness.
DARPA launched its Prognosis programme in December with the award of technology development contracts to General Electric, Northrop Grumman and Pratt & Whitney. Northrop Grumman received $14.1 million to create a tool for predicting structural integrity without the need for inspections, while GE and P&W received $7.1 million each to develop a system for predicting engine capability and providing the pilot with early warning of pending failure. In addition, a $1.7 million contract was awarded to Southwest Research Institute to demonstrate a wireless sensor system for detecting and monitoring cracks in turbine engine components.
"Current ways of lifing systems all rely on a predetermined inspection methodology designed to detect a crack of a particular length in a particular location after a particular number of cycles," says Prognosis programme manager Dr Leo Christodoulou. "If you don't look in the right location, or look properly, tough luck." He cites the 1989 crash of a United Airlines McDonnell Douglas DC-10 at Sioux City, Iowa. Investigators concluded the accident was caused by inspection limitations that resulted in the failure to detect a fatigue crack originating from a previously undetected metallurgical defect in an engine fan disk.
"They were surprised by an event not because it was a surprising failure - but because they were not looking in the right place," Christodoulou says. "What if we were able to interrogate a structure, as opposed to inspecting it, and allow the material to tell us how it feels?"
The Prognosis programme aims to change the paradigm by which aircraft lives are managed. Today, fear of failure controls the design and use of aircraft, DARPA says, resulting in large safety margins that reduce performance, limit availability and drive up costs. Fleets of aircraft are managed on the statistical likelihood of failures occurring. "We inspect 1,000 parts to find the one that is likely to be damaged," says Christodoulou.
Individual capability
DARPA aims to develop technology to manage the fleet based on individual and actual aircraft capability, and not on a statistical basis. At the heart of the concept is the fact that damage begins with microstructural changes before a detectable crack has formed, and evolves with time and use in a predictable way. "The programme will attempt to identify the physics of failure and the evolution of damage, and predict what comes next," says Christodoulou.
At any point in an airframe or engine's life the system will make a prediction of its near-term capability - "in 10, 100 or 1,000 cycles, not tens of thousands", Christodoulou says - based on failure physics and damage evolution models. "The models are imperfect, and will not capture all events, so we will modulate the prediction with interrogation technology that will provide us with state awareness." In other words, sensors will indicate whether the models are overestimating or underestimating the life remaining on an airframe or engine component.
Instead of a time- or cycle-based maintenance regime determining whether an aircraft can fly or not fly, operators will be able to operate aircraft to the best of their current capabilities. A military commander, for example, could elect to continue flying an aircraft within a restricted speed and altitude profile until repairs can be performed. "The system would enable a local commander to decide which aircraft are most capable of a mission and which are restricted, when on paper they are all the same," says Christodoulou.
A prognosis system integrates several elements: physics-based damage evolution models; global and local state-awareness sensors; and more effective use of existing flight and maintenance history databases. The basic approach of combining the models, sensors and databases to predict capability is applicable to both civil and military airframes and turbomachinery, including helicopter rotors and gearboxes, says Christodoulou. The system can also provide real-time warnings of pending failures.
The Prognosis programme has two phases. In the 24-month Phase 1, contractors will demonstrate critical elements "without which the system cannot be implemented", says Christodoulou. These include the ability to extract features from complex signals. "You may have multiple cracks. Can you tell if one is masking another? We are confident we can do it in individual components, but in a 12-stage turbine or a wing box it is more difficult to find cracks or delaminations," he says. A better understanding of the interaction between failure modes, such as corrosion and fatigue, is also required, and part of the Phase 1 task is to develop linkages between physics-based failure models.
If the programme proceeds into the 24-month Phase 2, the teams will build system prototypes, probably a complete powerplant and complex wing/fuselage structure, and undergo a blind test in which their prognosis system will have to find deliberately induced flaws. On the government side, the US Federal Aviation Administration and National Transport-ation Safety Board are part of a working group to looking operational aspects of the system, including eventual certification.
In engine system prognosis, GE and P&W are being asked to relate damage evolution to throttle setting. "Throttle setting impacts the temperature, stress, time and environment on any component. We have computational fluid dynamics and finite element models that give the temperature and stress at component level. At component level we have material models that use temperature, stress, time and environment as inputs," says Christodoulou. "The big challenge is to link the models and run them together, to link throttle setting to damage evolution."
Tracking changes
Sensors will be used to bound the uncertainty in the models. These include "global" acoustic, thermal and vibration sensors and "local" laser ultrasonic, thermoacoustic and thermoelectric sensors at different locations within the engine. "It is impossible to take a totally unknown system and make a prognosis, so we define the initial state and track any changes," says Christodoulou. The acoustic signature of a bearing, for example, changes as it wears. The ultrasonic signature of a directionally solidified turbine blade changes as the material crystallises. The ultrasonic attenuation of composites changes dramatically with delamination.
Sensors provide the signature of a component's state before a flight. The planned mission profile is run through the failure physics model to determine the new state at the end of the flight, and decide if damage evolution will still be within safe limits. The predicted and actual sensor signatures after the mission are then compared. "The system may say it can't make up its mind, go do an inspection, but we consider that positive, not negative. Inspecting three instead of 1,000 is still a lot better," says Christodoulou.
Source: Flight International
