A large-size data reduction/fusion algorithm for spacecraft vehicle health management systems

Being a standard part of next generation spacecraft, the onboard integrated vehicle health management system processes real-time and historical measurement data to make failure diagnoses and corrective decisions. However, this task is complicated by the extremely large amount of the available data, the existence of uncertainties, and vehicle interactive operational conditions. Therefore, a data reduction/fusion algorithm with an object-oriented approach is designed based on fuzzy inference and statistical analysis. Representative data from each object, i.e., from each measurement source, are checked at a certain sampling frequency. The importance levels of individual objects with respect to system failures are determined with fuzzy inference rules and a relational database. Data reduction is then achieved by reassigning a lower sampling frequency to insignificant data groups. In this way, we only select critical data groups for regular detection and diagnoses. Using a fuzzy network as a classifier, the data fusion function extracts useful information for the vehicle health management system