Human Immune System Inspires NASA Machine-Software Fault Detector
Using the human immune system as an inspiration, NASA scientists are developing software to find faults in complex machines.
The software 'tool' - called an algorithm, or mathematical recipe - looks for abnormalities in a machine's hardware and software. The mathematical recipe, which engineers may well someday put in spacecraft as well as other complex systems, is part of the Multi-level Immune Learning Detection (MILD) software 'tool,' under development at NASA Ames Research Center in California's Silicon Valley.
To the left, NASA Ames scientist, Kalmanje Krishnakumar and, to the right, co-investigator on the MILD project is Dipankar Dasgupta of the University of Memphis, Tenn., who is spending a year as a visiting faculty member at NASA Ames. Click on image for publication size.
"The human immune system doesn't try to identify what is good, only what is bad," said MILD principal investigator Kalmanje Krishnakumar, a scientist at Ames. "Similarly, MILD software only tries to identify what is bad, and that's one of the main ideas behind MILD, which is similar to biological immune systems," Krishnakumar said. Co-investigator on the MILD project is Dipankar Dasgupta of the University of Memphis, Tenn., who is spending a year as a visiting faculty member at NASA Ames.
"You can have identical MILD software recipes distributed throughout the machine that look at different potential abnormalities," Krishnakumar explained. "Typically, a problem will show up in more than one place in a machine, and comparisons of different parts of the machine help us to more accurately identify problems early," he added.
MILD uses data from sensors in machines to find patterns of system faults and damage to clarify if systems are working properly. In an aircraft, sensors may include gyroscopes and instruments to measure acceleration. Spacecraft and other machines may have temperature sensors, gas sensors and similar devices that report on the condition of the machine or its environment.
"Another advantage of the MILD tool is its ability to associate detectors to known and probable faults. This signature can then be used to identify future occurrences of similar faults. Similarly, the biological immune system quickly recognizes diseases to which it has been exposed previously or has been 'immunized' to some known diseases," Krishnakumar said.
"Another advantage of using the immune system as an inspiration is that we can program the MILD software tool to recognize known faults that occur in a machine. Similarly, a biological immune system recognizes diseases to which it has been exposed," Krishnakumar said.
The Crew-Vehicle Systems Research Facility. Click on image for publication size.
So far, scientists have tested the MILD software in a C-17 aircraft flight simulator at NASA Ames to collect normal and simulated airplane failures. "We used the aircraft simulator as a proof-of-concept experiment to test how well the MILD algorithm worked," Krishnakumar explained. The software is still in the research phase. Later, scientists hope to modify it so it will work as stand-alone software.
In the near future, when engineers use MILD software on another machine, they will need to set up the software so it will monitor data from that machine. "However, we now are enhancing the MILD software 'tool' so it can more easily be used for other machines," Krishnakumar said. "Eventually, engineers could use MILD algorithms in any kind of software and hardware in machine environments - from machines in a shop to flying airplanes and spacecraft," Krishnakumar ventured.
"We expect future machines to have their own immune systems so that they could be used for long-duration space missions, or any other use where technical support would be limited," Krishnakumar said.
Credit: John Bluck (John.G.Bluck@nasa.gov), Ames Research Center