 |
| Elements of an artificial neural network. |
Logistics support of Army After Next (AAN) forces will differ radically from today's logistics system. AAN battle forces will be self-sufficient for long periods of time, move two to three times faster than current units, and operate at great distances from higher level sources of support. This will require weapon systems and other equipment to be significantly more reliable than those in use today. System reliability of 90 percent is acceptable for current equipment, but 99 percent or higher will be required for AAN equipment. It will have to be "ultrareliable." Thus, ultrareliability is one of the 6 pillars of the AAN support concept.
In this context, the term "ultrareliability" means to provide failure-free operations. New equipment and weapon systems need to be designed to be inherently reliable, which will require advances in design engineering and greater emphasis on reliability standards by materiel developers. However, this approach may prove to be too costly to attain the improvements in reliability needed for AAN forces. Another approach, which may be much less expensive, is to embed automated diagnostics and prognostics into weapon systems. The goal would be to predict failures in all weapon subsystems and components and to make that information available to commanders and support personnel in real time. Achieving this goal will require the use of emerging and future technologies and a revolution in the way the Army maintains its equipment.
Horizontal Technology Integration
The Army's commitment to its plan to remain the dominant military force in the period 2010 to 2025 was emphasized by Major General Robert H. Scales, Jr., Commandant of the Army War College—
We believe the Army has seized upon a highly compelling vision of its future role in land warfare. It has also carefully thought through a comprehensive process that will determine the key science and technology investments enabling it to achieve this vision.
In 1997, the Program Manager for Test, Measurement, and Diagnostic Equipment (PM TMDE) was tasked to create a plan for embedding diagnostics in all Army equipment. The result called for a horizontal technology integration approach that would allow new technologies to be applied to multiple weapon systems. In January 1998, the Automated Diagnostics Improvement Program (ADIP) was approved as a horizontal technology program, and embedded diagnostics and prognostics were identified as tenants of Army XXI and AAN. Now all weapon system PM's must coordinate with the PM TMDE to ensure that diagnostics and prognostics are incorporated early in the new system acquisition process.
Facing reduced budgets, the Army must find ways to reduce the operating and support costs of current and future equipment while continuing to purchase technologically superior weapon systems. The ADIP planners recognize that the Army is lagging behind the commercial sector in the use of diagnostics and prognostics. Thus, the ADIP master plan stresses the need to use commercial off-the-shelf and nondevelopmental items.
|
| Elements of an artificial neural network. |
ADIP Master Plan
The ADIP master plan describes phases, or thrusts, that will take place concurrently, with early results being used to generate even greater advances in the later stages.
The first thrust involves inserting available commercial diagnostic technology into current weapon systems and replacing current support equipment. An example of diagnostics that already have been incorporated into a weapon system is the turbine engine diagnostic expert system that diagnoses faults in the M1 Abrams tank's turbine engine. Using artificial neural network technology, sensors in the engine analyze fuel flow and provide data to the crew and support personnel. The system has made diagnosis easier and reduced the number of "no evidence of failure" faults in turbine engines returned for repair.
The goal of thrust two is to develop an anticipatory maintenance system. This will be accomplished by the Failure Analysis and Maintenance Planning System, which will be a submodule of the Global Combat Support System-Army (GCSS-Army)—the Army's system that will digitize the support structure. Data will be collected using electronic sensors and exported using improved data buses, the tactical Internet, and satellites. Analysis of the data will give users the ability to predict maintenance needs, as information from across the Army is placed in a data base and made available to personnel at all levels. In addition, ordering parts and processing work orders will be automated fully.
Thrust three will result in the fielding of an embedded diagnostic system in all new Army equipment. Advanced research and development projects begun in the first two phases will yield a true prognostic capability and complete the revolution in weapon systems maintenance. Programmable sensors will monitor and collect data on all parts of a system continuously. These data will be analyzed by on-board processors and will be made available, through interactive interfaces, to the crew, commanders, and logisticians. The data also will be transmitted in real time to central data bases at theater level and to the continental United States. It may even be possible to provide a level of self-healing to a weapon system.
Required Technologies
Technological advances in several key areas are needed if the goals outlined in the ADIP master plan are to be realized. These include—
Incorporating advances made in these areas by the commercial sector will enable Army materiel developers to field weapon systems that have a true prognostic capability.
Sensor Technology
Sensors can gather data on weapon system performance and status. Factors such as temperature, noise, pressure, and vibration can be measured and used as diagnostic signals. Some notable innovations in this area include—
Data Processing and Storage Technology
One technology that has great promise is the artificial neural network. An artificial neural network mimics the way the human brain functions. A node receives input from a number of sources, the input is evaluated using weighted factors, and a total output is found (see chart on page 33). This output is compared to an expected value. The neural network then "learns" by repeating the calculations either forward or backward through the system. In effect, it becomes "smarter," or more accurate, over time. Researchers are finding ways to make these artificial neural networks more complex by adding additional processors or layers of nodes. The ability of the neural network to "learn" is what makes it such a powerful tool.
The turbine engine diagnostic system mentioned earlier makes use of this type of artificial intelligence to monitor the M1 tank's turbine engine. With M1 tank engines, each turbine has slightly different operating characteristics due to varying tolerances in machining and other factors. The turbine engine diagnostic system on each tank comes to recognize the peculiar operating characteristics of its engine and bases its analysis of that turbine's operation on those specific factors. This use of artificial intelligence has hundreds of commercial applications and has great potential in the military.
After a sensor collects data, the data must be processed to be useful. Advances in digital storage capacity are necessary to make efficient use of the large amounts of data that will be collected from hundreds or thousands of sensors. Computer processing power is expanding at a phenomenal rate. In the 1950's, it took a billion billion atoms to store one bit of information; by 2025, it could take 1,000 atoms. In fact, in the laboratory, researchers have been able to store a bit of information on a single atom. Similarly, in 1956 it cost $10,000 to store a megabyte of information, while today it can be done for 10 cents. By 2025, it may be possible to store a terabyte (1,000 gigabytes) for only $1. Advances like these will make it possible to have on-board data storage devices that collect the vast amounts of digital information required to realize a truly prognostic capability.
Communications Technology
Communications technology must be improved for diagnostic systems to work properly. Huge amounts of data will have to be transmitted securely and rapidly around the battlefield. While developing the hardware or software to achieve this capability is the responsibility of the signal community, not ADIP, some emerging technologies are worth noting—
Digital cellular networks are beginning to appear in North America and Europe. In addition, much larger satellite communications networks will come on line in the next decade.
Short-range radio frequency links may become an effective way for logisticians to transfer data over short distances.
Data compression technology, encryption, and methods of using the full spectrum of bandwidth will further improve our ability to move information from point to point.
Internet transfer of information is being researched. The Navy, in conjunction with the Electric Power Research Institute and Pennsylvania State University's Applied Research Laboratory, is testing the data transfer capabilities of the web. Data obtained remotely are sent over the Internet to an expert system and evaluated.
To support the rapidly moving, self-sufficient AAN forces, equipment will have to be nearly failure free. Embedded diagnostic and prognostic systems will be needed to monitor the equipment and identify problems before they occur. The ADIP lays out a plan for achieving this goal using horizontal technology integration. It provides the means to use commercially developed products within Army systems. Using current and emerging technologies, the Army will provide the ultrareliable equipment that AAN forces will need to get the job done.
Major Steve March, USAR, is an Active Guard/Reserve officer assigned to the 310th Theater Support Command at Camp Zama, Japan. He is a graduate of the Ordnance Officer Advanced Course, the Combined Arms and Services Staff School, and the Army Logistics Management College's Logistics Executive Development Course, for which he prepared this article.