Habitat ISHM Using Non-Invasive Load Management Analytics

We believe a robust approach to integrated system health management (ISHM) design is the application of redundancy. Redundancy is often thought of in terms of hardware; however, functional, analytic, and information redundancy strategies should also be considered. Modeling sensor information is invaluable for diagnostics and critical path analysis. A total system approach is an efficient means of prognostics as well as identifying the time of failure. However, fidelity and resolution must be considered in both approaches. There are compounding errors as the subsystems are aggregated in a component model. Sensors themselves introduce a point of error and require due consideration of size, weight, and power (SWaP).Signal processing, machine learning, and data mining techniques are common approaches in ISHM to improve the accuracy of alerts for known issues and an ability to identify latent and unknown failure conditions. Such techniques are not limited to ISHM. They are also used in fraud detection, image processing, medical diagnostics, and other domains.Our innovation draws from the domain of electrical power systems with the application of non-invasive load management (NILM) models for load disaggregation. NILM is a means of extracting and analyzing discrete end-use system components from an aggregate energy signal. NILM evolution has run parallel with the developments in signal processing, machine learning, and data mining for feature extraction, classification, and action. A NILM approach for managing habitat subsystems allows for optimization of the number of sensors, mitigating points of information failure and the constraints of size, weight, and power while providing analytical redundancy to hardware systems. We submit that the application of disaggregation analytics is an innovative ISHM technology that supports NASA missions.