System and method for explanation of condition predictions in complex systems

A machine system monitoring and alert apparatus, comprising:

• a computer system comprising a plurality of inputs to receive time series signals from a plurality of sensors;
• the computer system comprising a processor and a memory adapted with instructions forming:
• logic to receive a labeled point and a set of signals from which the labeled point was derived, each of the signals having a corresponding signal projection subspace;
• logic to project a model for the machine system into each signal projection subspace of the set of signals, the model comprising a plurality of model condition points;
• logic to project the labeled point into each signal projection subspace;
• logic to constrain an analysis neighborhood for the labeled point to a set number of model condition points closest to the labeled point in the signal projection subspace;
• logic to calculate a first percent of neighbors in the analysis neighborhood having a same label as the labeled point;
• logic to calculate a second percent of points in the model having a same label as the labeled point;
• logic to calculate a contribution of each signal to a machine system condition corresponding to the labeled point from the first percent and the second percent, to form a sorted list of signals and contributions;
• and
• logic to display signals that are higher in the sorted list of signals as the most likely contributors for the machine system condition corresponding to the labeled point.

A non-transitory machine-readable storage medium for monitoring and identifying causes of complex physical system behavior, the machine-readable storage medium including instructions that when executed by a processor, cause the machine comprising the processor to:

• receive a labeled point and a set of signals from which the labeled point was derived;
• generate a signal projection subspace for each signal from the set of signals;
• project condition points from a model of the physical system into each signal projection subspace of the set of signals;
• project the labeled point into each signal projection subspace;
• constrain an analysis neighborhood for the labeled point to a set number of model condition points closest to the labeled point in the signal projection subspace;
• calculate a first ratio of (a) neighbors in the analysis neighborhood having a same label as the labeled point and (b) the set number;
• calculate a second ratio of (c) points in the model having a same label as the labeled point and (d) the set number;
• assign a contribution of each signal to a machine system condition corresponding to the labeled point, based on a ratio of the first ratio and the second ratio, to form a sorted list of signals and contributions;
• and
• adapt the physical system's behavior based on the list.