A computer program for the detection of present and prediction of future discrete states of a complex, real-time engineering system utilizes a combination of symbolic processing and numerical model-based reasoning. One of the biggest weaknesses of a purely symbolic approach is that it enables prediction of only future discrete states while missing all unmodeled states or leading to incorrect identification of an unmodeled state as a modeled one. A purely numerical approach is based on a combination of statistical methods and mathematical models of the applicable physics and necessitates development of a complete model to the level of fidelity required for prediction. In addition, a purely numerical approach does not afford the ability to qualify its results without some form of symbolic processing.
The present software implements numerical algorithms to detect unmodeled events and symbolic algorithms to predict expected behavior, correlate the expected behavior with the unmodeled events, and interpret the results in order to predict future discrete states. The approach embodied in this software differs from that of the BEAM methodology (aspects of which have been discussed in several prior NASA Tech Briefs articles), which provides for prediction of future measurements in the continuous-data domain.
This program was written by Mark James of Caltech for NASA’s Jet Propulsion Laboratory.
This software is available for commercial licensing. Please contact Karina Edmonds of the California Institute of Technology at (626) 395-2322. Refer to NPO-45172.
This Brief includes a Technical Support Package (TSP).
Symbolic Processing Combined With Model-Based Reasoning
(reference NPO-45172) is currently available for download from the TSP library.
Don't have an account?
Overview
The document discusses an innovative approach developed by NASA's Jet Propulsion Laboratory (JPL) for predicting future discrete states of real-time systems, particularly in the context of maintenance predictions for the Deep Space Network (DSN). The method combines symbolic processing with numeric model-based reasoning, addressing the limitations of traditional prediction systems that rely solely on either numeric or symbolic approaches.
Traditional systems often struggle with the complexity of real-time data and the need for exhaustive checklists to anticipate all potential failure and non-failure modes. This can become resource-intensive and may not effectively correlate predictions with the rationale behind them. The new approach aims to overcome these challenges by integrating real-time measurements with expected behavior and past performance to enhance predictive accuracy.
The architecture utilizes multiple knowledge representations, including engineering models and causal system models, to monitor operational stresses and detect anomalies. By combining numeric algorithms that identify unmodeled events with symbolic algorithms that predict expected behavior, the system can correlate these elements to generate accurate forecasts of future states. This dual methodology allows for a more comprehensive understanding of system dynamics, enabling the identification of both modeled and unmodeled states.
One of the key advantages of this approach is its ability to refine inference strategies continuously. By eliminating incorrect hypotheses based on incomplete knowledge, the system improves its predictive capabilities over time. This is particularly important in complex systems where traditional methods may fail to account for all variables.
The document emphasizes that the integration of symbolic reasoning with numeric techniques not only enhances the accuracy of predictions but also provides a framework for qualifying results. This is crucial for ensuring that predictions are grounded in a solid understanding of the system's behavior.
In summary, the document outlines a sophisticated method for predicting future discrete states in real-time systems by merging symbolic processing with model-based reasoning. This innovative approach addresses the limitations of existing methodologies, offering a more effective solution for complex predictive analysis in aerospace applications. The work represents a significant advancement in the field, with potential implications for various technological, scientific, and commercial applications beyond aerospace.
