Explanation constraint programming for model-based diagnosis of engineered systems

Abstract

Abstract-We can expect to see an increase in the deployment of unmanned air and land vehicles for autonomous exploration of space. In order to maintain autonomous control of such systems, it is essential to track the current state of the system. When the system includes safety-critical components, failures or faults in the system must be diagnosed as quickly as possible, and their effects compensated for so that control and safety are maintained under a variety of fault conditions. The Livingstone fault diagnosis and recovery kernel and its temporal extension L2 are examples of model-based reasoning engines for health management. Livingstone has been shown to be effective, it is in demand, and it is being further developed. It was part of the successful Remote Agent demonstration on Deep Space One in 1999. It has been and is being utilized by several projects involving groups from various NASA centers, including the In Situ Propellant Production (ISPP) simulation at Kennedy Space Center, the X-34 and X-37 experimental reusable launch vehicle missions, Techsat-21, and advanced life support projects. Model-based and consistency-based diagnostic systems like Livingstone work only with discrete and finite domain models. When quantitative and continuous behaviors are involved, these are abstracted to discrete form using some mapping. This mapping from the quantitative domain to the qualitative domain is sometimes very involved and requires the design of highly sophisticated and complex monitors. We propose a diagnostic methodology that deals directly with quantitative models and behaviors, thereby mitigating the need for these sophisticated mappings. Our work brings together ideas from model-based diagnosis systems like Livingstone and concurrent constraint programming concepts. The system uses explanations derived fiom the propagation of quantitative constraints to generate conflicts. Fast conflict generation algorithms are used to generate and maintain multiple candidates whose consistency can be tracked across multiple time steps.

The Diagnostic Competitions

“…The hybrid diagnosis engine (HyDE) (Narasimhan and Brownston 2007) is a model‐based diagnosis engine that uses consistency between model predictions and observations to generate conflicts that in turn drive the search for new fault candidates. HyDE uses discrete models of the system and a discretization of the sensor observations for diagnosis.…”

Model-Based Fault Detection and Diagnosis System for NASA Mars Subsurface Drill Prototype

“…HyDE [4][5] [6], the model-based diagnostic engine used for HyDE is the successor to the Livingstone and Livingstone 2 diagnostic engines [3], also developed at NASA Ames and its name stands for "Hybrid Diagnostic Engine". The Livingstone family of diagnostic engines relied on discrete modeling methods best suitable to such domains as digital electrical systems or computer networks.…”

A Survey of Data-Driven Prognostics

“…Livingstone is a model-based diagnosis engine developed at NASA Ames Research Center that reasons about system-wide interactions to detect and isolate failures. 25,26,27 Livingstone uses a hierarchical model of components and modules. Each component is modeled using a finite state machine.…”