Program addresses need for enhanced diagnostic capabilities.
The U.S. Air Force is developing a software-based system that will allow aircrews to diagnose and predict equipment failure with greater speed and accuracy, keeping more aircraft in the air, not the hangar. In a renewed effort to maintain operational readiness through enhanced systems integration, the service is emphasizing the need for greater precision and efficiency across the spectrum of aviation maintenance.
A part of the human engineering branch of the Air Force Research Laboratory (AFRL), Wright-Patterson Air Force Base, Ohio, the Predictive Failures and Advanced Diagnostics (PFAD) program, introduced in the summer of 2000, addresses the high cost of replacement parts for F-15, F-16 and C-130 aircraft. The program is currently in the first year of a 5-year, $9.4 million contract between Northrop Grumman, Los Angeles, and the AFRL. Teamed with Northrop Grumman, DATAMAT Systems Research, McLean, Virginia, is studying how information taken from multiple aircraft sensors can help determine why and when system failures occur.
According to Northrop Grumman’s program manager for PFAD, Joe Castrigno, the problem that military aviation maintenance faces today is that on the ground it is difficult to duplicate the airborne conditions that cause equipment failures aboard many legacy-part-dependent aircraft. As a result, machinery suspected of malfunction is removed from the airframe for costly troubleshooting repairs. In many cases, the initial misdiagnoses of elusive system bugs lead to the ordering of parts for which the original components are later determined to be operable, he adds. To remedy this situation, the Air Force is concentrating on integrating maintenance indicators and increasing system autonomy within aircraft to raise repair reliability.
“The new technology will ‘front end’ existing diagnostic tools and take the guesswork out of fault analysis,” Castrigno explains. “Reducing aircraft life-cycle costs while simultaneously increasing flight readiness is the main goal of the PFAD program.” Two divisions of Northrop Grumman, the Integrated Systems Sector (ISS), Dallas, Texas, and the Electronic Sensors and Systems Sector (ES3), Dayton, Ohio, have coordinated their efforts on the joint PFAD program. The company’s ISS Airborne Early Warning and Electronic Warfare Systems sector is leading the technology’s development, using six testbed BAC One-Eleven aircraft to experiment with potential techniques for the future integration of the program’s software into existing systems.
Dr. Barbara Gilmartin, Northrop Grumman’s principal investigator for the PFAD program, states, “The PFAD idea is to upgrade rather than replace existing equipment so that the military can get more out of aircraft already in service. We want to know more intelligently when parts will likely be failing. This is predicated on the supply cycle being based on actual equipment conditions as opposed to reliability predictions.” The program seeks to compare previously gathered computer-stored data with performance monitoring data from legacy systems aboard designated aircraft in an effort to pinpoint the cause of a failure. In theory, this analysis will prevent expensive, time-consuming troubleshooting and increase each aircraft’s turnaround time from the hangar to the flight line.
Operating under the guidelines of the Air Force maintenance standard, known as the Reliability and Maintainability Information System, PFAD is using data mining techniques to develop pattern-based predictive capabilities. By comparing results from old and new diagnostic tests, technicians will be able to paint a picture of the operational trends in a piece of equipment. In doing so, a functional history can be mapped, telling specialists where and when the next malfunction is likely to occur.
Looking at parts histories on the system level, rather than the characteristics level, is a key program initiative. “To diagnose a problem accurately, you need to know how each part responds under certain conditions,” Gilmartin indicates. “The more you know about past performance, the better you can pinpoint a system irregularity.” Standard aircraft maintenance involves swapping faulty line replacement units (LRUs) with properly functioning ones. However, ground verification of problematic LRUs is often difficult because stresses occurring in flight cannot be precisely duplicated in a hangar. PFAD plans to simulate this airborne environment through the use of a combination of machine intelligence techniques such as model-based reasoning and neural networks. Using existing tools to compare like systems, aircraft technicians can determine where subtle differences in temperature, fluid flow or electrical current are located. “Prognosticating the gradual degradation of specific components by applying node-based network monitoring to the maintenance equation can help you understand what an initial failure means,” Gilmartin explains.
Fault indication is another area of focus. When a component begins to show signs of failure, indicators will alert the operator to potential corruption within the system. “The majority of legacy equipment will not recognize a problem until it has matured to a certain degree within a system,” Castrigno points out. “PFAD technology will give operators information at the onset of trouble, enabling them to make failure rate estimates using the appropriate indicators.” Diagnosing a fault’s origin becomes more difficult the longer it progresses, so early prediction saves both the money to replace the whole unit and the time needed to identify the problem at a later date, he adds.
The primary demonstration platform for the PFAD program has been the F-16 AN/APG-68 radar system. George Rovnack, PFAD program director for Northrop Grumman’s ES3, states, “One of the chief problems that we are experiencing with regard to the AN/APG-68 radar and other onboard tracking systems is the intermittent failure of certain components under dynamic stress conditions such as temperature changes, vibrations or shock. These conditions are difficult to replicate on the ground because of the complex nature of the gravitational forces involved.” Maintenance approaches currently being used in many radar systems diagnose potential problems to a higher degree of accuracy using lighter versions of existing LRUs. This helps operators locate the source of a failure by excluding the clutter presented by extraneous components.
A downside to this type of fault isolation exists, however. As system failures become more widespread throughout an LRU, techniques using lower-level units are not as effective in locating more complex problems. “Ambiguities that are present in current fault isolation methods will be significantly reduced by PFAD,” Rovnack indicates. “In situations where the swapping out of LRUs might be the standard procedure, PFAD will enable real-time testing of components on the aircraft to keep turnaround times short.” DATAMAT is developing algorithmic formulas using data mining techniques. These methods will enable aircraft maintenance personnel to more readily locate the cause of a system failure by breaking down the data input of an LRU. “If you know what went in, then you can more easily characterize a component’s response,” Rovnack adds.
Studies by AFRL researchers have examined variations on model- and case-based reasoning to determine where existing fault indication and isolation techniques need improvement. Paul Faas, PFAD program manager, AFRL, explains, “Our main concern is in finding ways to reduce maintenance-related costs so that money can be reallocated into ongoing research and development projects. PFAD will eliminate much of the unnecessary expense that is associated with the inaccuracy of predictive reliability maintenance.”
PFAD aims to shift military aviation from standard phase maintenance to a form of condition-based, predicted maintenance. “The goal is to perform checks on aircraft using data that indicates they need attention in certain areas,” Faas explains. “If you know the specific details of each system’s life cycle, then you can maintain each LRU based on its actual state of operation rather than by a suggested maintenance calendar. This keeps hangars full of only the aircraft that need to be there, allowing for the more efficient use of maintenance time and money.” In recognition of the fact that every aircraft undergoes different stresses in flight, AFRL officials are seeking ways to apply PFAD technology to aircraft mission sensors so that each aircraft is checked according to its individual condition.
A more immediate program goal is to enhance maintenance tools by combining existing built-in and manual test capabilities. “Data collected onboard the aircraft is largely confined within independent systems,” Faas notes. “Finding ways of integrating related system information to increase output detail is one of the challenges PFAD is addressing.”
According to Capt. Ken Eizenga, USAF, deputy director of PFAD, AFRL, monitoring gravitational forces on an aircraft in flight could yield more data on the specific effects that stresses have on multiple independent systems. By integrating information from a number of affected systems, operators can collect more data on how gravity influences particular subsystems within the aircraft, he adds.
Beside the F-16 AN/APG-68 radar system, PFAD technology also is slated for potential application with legacy power generation and fuel subsystems across a variety of aircraft platforms. “With regard to electrical operations onboard aircraft, effects on one system often lead to altered conditions in others,” Dr. Gilmartin remarks. “It is not unusual for sympathetic failures to occur in a combination of radar, avionics and display-mounted subsystems in response to a loss of circuit power. One of our key approaches is to prevent the propagation of simultaneously cascading effects across multiple systems by looking beyond the boundaries of individual subsystems to see better how each affects the other.”
One issue that the program is currently investigating is the streamlining of multiple-level maintenance to enhance diagnostics by bridging the gap from one level of analysis to the next. According to Faas, model- and case-based reasoning are being examined as viable alternatives to standard equipment swapping methods. PFAD is focusing on introducing an integrated software package to do the work done by external auxiliary devices.
In cases when the standard F-16-mounted integrated warning system provides cautions, advisories and alerts to pilots about system failures, the lack of an onboard integrated analysis capability prevents the operator from knowing any specifics about the problem. “We hope that PFAD will provide the intelligence needed to diagnose a situation fully so that the pilot knows the extent of a condition even before the aircraft lands,” Faas declares. The idea is to give pilots as much information as possible so that appropriate actions can be taken that increase not only pilot safety but also aircraft reliability, he explains. Research by the ISS Airborne Early Warning and Electronic Systems sector continues into how PFAD might be integrated into existing onboard sensory equipment.
Program testing is ongoing at the AFRL’s Dayton facility and at the Northrop Grumman Flight Development Laboratory in Baltimore, Maryland. The flight test phase of the program is slated to begin in 2004. Program officials estimate that a mature software product will appear in the first quarter of 2005. As researchers perform further analysis on existing sensor systems, a better picture of which ones are compatible for potential integration with the new software will become evident, he adds.