NASA has developed a new method for analyzing complex system behavior that also may be viewed as a type of data visualization and decision support tool. Large complex control systems may have thousands or even millions of sensors, each providing some type of signal that ultimately integrates into a larger organization. For each signal, behavior is represented by a sequence of pairs, with each pair containing a change value (monotonic) and a time interval length over which each of these changes occurs. Signal amplitudes and first derivatives serve as markers for these time intervals. This approach permits a finer scale characterization of the signal(s). The novelty of this approach is in using human visual interpretation in combination with computer signal analysis to monitor the behavior of complex systems in an enhanced manner.
This technology is a software method for performing visual and/or numerical comparison of behavior of a system using decomposition of a function of bounded variation as sums and differences of monotonic functions. It takes the output of the monotonic signal analysis and feeds it into another part of the algorithm that translates the data into a graphic symbol language that finally becomes a set of images meaningful to human interpretation. This technology is used to determine if a system is behaving normally or is varying into abnormalities through comparisons with reference values. Using this decomposition, the signal may then be transformed into a fixed-size collection of graphic symbols that may in turn be combined into a larger visual representation. The fixed-size alphabet of symbols also allows easy neural net pattern recognition if desired.
The conversion into visual display information is a key aspect of the invention, since it takes advantage of the human skill of visual pattern recognition that exceeds computer capabilities. Using human visual recognition in combination with numerical computer analysis provides a new level of system monitoring capability that can recognize faults at the sub-threshold level, and provide early warnings as well as intelligence towards reaction. Another unique attribute of the technology is its ability to reverse the visual translation and provide access to the underlying data. This may allow a user to dig deeper into a problem if an anomaly is detected.