This software addresses the problem of how to efficiently execute a knowledge base in the presence of missing data. Computationally, this is an exponentially expensive operation that without heuristics generates a search space of 1 + 2n possible scenarios, where n is the number of rules in the knowledge base. Even for a knowledge base of the most modest size, say 16 rules, it would produce 65,537 possible scenarios. The purpose of this software is to reduce the complexity of this operation to a more manageable size. The problem that this system solves is to develop an automated approach that can reason in the presence of missing data. This is a meta-reasoning capability that repeatedly calls a diagnostic engine/model to provide prognoses and prognosis tracking. In the big picture, the scenario generator takes as its input the current state of a system, including probabilistic information from Data Forecasting. Using model-based reasoning techniques, it returns an ordered list of fault scenarios that could be generated from the current state, i.e., the plausible future failure modes of the system as it presently stands. The scenario generator models a Potential Fault Scenario (PFS) as a black box, the input of which is a set of states tagged with priorities and the output of which is one or more potential fault scenarios tagged by a confidence factor. The results from the system are used by a model-based diagnostician to predict the future health of the monitored system.
This program was written by Mark James of Caltech for NASA's Jet Propulsion Laboratory. For more information, download the Technical Support Package (free white paper) at www.techbriefs.com/tsp under the Software category.
This software is available for commercial licensing. Please contact Karina Edmonds of the California Institute of Technology at (626) 395-2322. Refer to NPO-42818.
This Brief includes a Technical Support Package (TSP).
Reducing a Knowledge-Base Search Space When Data Are Missing
(reference NPO-42818) is currently available for download from the TSP library.
Don't have an account?
Overview
The document titled "Reducing a Knowledge-Base Search Space When Data Are Missing" (NPO-42818) from NASA's Jet Propulsion Laboratory outlines a system designed to efficiently manage knowledge base operations in scenarios where data is incomplete. This is particularly relevant for applications in aerospace, where reliable decision-making is critical, especially in manned flight and autonomous exploration.
The core challenge addressed by the system is the computational complexity involved in executing knowledge bases when data is missing. Traditional methods can be exponentially expensive, making them impractical on non-parallel hardware. The proposed solution reduces this complexity through a combination of serialization techniques and heuristic methods.
Key features of the system include:
-
Guaranteed Truths: Rules that are fully true are grouped into a single meta result, allowing them to be treated as composite objects. This means that once their truth is established, they do not need to be re-evaluated, significantly reducing computational overhead.
-
Permutable Truths: The remaining rules, which are not classified as Guaranteed Truths, are termed Permutable Truths. Any hypothetical scenarios generated from these truths must always include the Guaranteed Truths and remain logically consistent with them.
-
Heuristic Scoring: A heuristic scoring method is employed to characterize the execution of rules, which helps in reducing the search space that needs to be explored. This scoring system allows for more efficient generation of potential scenarios from the knowledge base.
The document emphasizes the importance of this system for Integrated Vehicle Health Systems, which aim to enhance prognostic forecasting, reasoning on incipient faults, and operational efficiency in degraded systems. These capabilities are essential for advancing human spaceflight and enabling deeper space exploration.
The software developed as part of this system is versatile, running on various platforms including SUN, HP, Intel, and Apple MACs. It can be distributed in both source and binary code forms and requires a LISP compiler for operation. The implementation is designed to integrate seamlessly into different environments, making it adaptable for various applications.
Overall, this document highlights a significant advancement in knowledge management systems, providing a robust framework for handling incomplete data in complex aerospace applications, thereby enhancing decision-making processes in critical scenarios.
