NASA Logo, National Aeronautics and Space Administration

HyDE Commercialization by Impact Technologies

Impact Technologies, LLC (www.impact-tek.com) has recently introduced a commercial tool suite with HyDE at its core called ReasonProX™, which includes a one click installer, complete documentation and integration with MATLAB/Simulink, and a remote deployment tool and runtime environment.

The ReasonProX™ software suite integrates diagnostics early in the development cycle by integrating the powerful Hybrid Diagnostic Engine (HyDE) from NASA with the widely used Matlab/Simulink environment. It also eases the transition to deployment by providing tools that create runtime instances ready for use in the ultimate target system.

Application developers can leverage existing physical/dynamic Matlab models or create new ones. As part of the diagnostic lifecycle, they also create new HyDE diagnostic models and integrate them with the Matlab physical models using ReasonProX.

ReasonProX seamlessly integrates with Matlab/Simulink allowing developers to stimulate, validate and verify their diagnostic models. Once the diagnostic model is verified ReasonProX provides tools to ease the deployment of the resulting diagnostic engine to the target physical system.

HyDE Applications

HyDE Commercialization by Impact Technologies

Impact Technologies is currently involved in an effort to create a commercial version of HyDE which will include a one click installer, complete documentation and integration with MATLAB. As part of this effort HyDE will be used to build diagnostic systems for the green building at NASA Ames Research Center and for printers at Palo Alto Research Center.

HyDE Oracle in DX Competition 2010

HyDE was used as an oracle in the 2010 Diagnostic Competition (DXC-10). HyDE could be queried about possible recovery actions for given fault conditions and it would respond with the set of recovery actions with the least cost. For this the HyDE model of the ADAPT system (used in the competition) was modified to switch the meanings of commanded transitions and fault transitions. The fault transitions became commands and the commanded transitions become "fault" transtions in the new model.

Advanced Diagnostics and Prognostics test-bed (ADAPT)

ADAPT

ADAPT was developed at NASA Ames Research Center with the goal to test, measure, evaluate, and mature diagnostic and prognostic health-management technologies. The test-bed hardware generates, stores, distributes, and monitors electrical power. The initial test-bed configuration is functionally representative of an exploration vehicle’s electrical power system. HyDE was used to build an 86-component enumeration constraint model of the test-bed components. The ADAPT is in stark contrast to the DAME model (described later) in that it contains a large number of components but very few variables and constraints on the components. Diagnosis was purely consistency-based. In addition, modeling used a special feature of HyDE that allows observations to be made input variables, enabling the modelers to build consistency models instead of predictive models. HyDE passed all the acceptance tests, which included a set of pre-defined fault scenarios and acceptance bounds on the time to diagnosis. Additional tests were run on other fault scenarios (not included in acceptance tests) and HyDE was able to diagnose all but one of these scenarios (resulting from inadequate information in the model). More details of HyDE on ADAPT are presented in [Poll, et al., 2007].

Drilling Automation for Mars Exploration (DAME)

DAME

The DAME project, led by NASA Ames Research Center, is aimed at developing a lightweight, low-power drill prototype that can be mounted on a Mars lander and drill several meters below the Mars surface for conducting geology and astrobiology research. Three kinds of diagnosis technologies were used on this project: HyDE for model-based diagnosis, a rule-based diagnosis system, and a neural-network diagnosis system. The drill and the diagnostic application using HyDE are illustrated in figure above. HyDE was to model only the two major components of the drill, the bit and the auger . Nevertheless the model was complex, requiring 73 variables and 162 differential and algebraic constraints. The model had nominal (nominal, idle), faulty (jamming, inclusion unknownFault), and unobserved (HardMaterial) locations.

There were four rounds of testing over a period of two years. In 2005, laboratory experiments were run at Honeybee Robotics, the company that constructed the drill. Later in 2005, field tests were performed at Houghton Crater on Devon Island, in the Canadian arctic, chosen to approximate the Martian terrain and climate. Based on the results, laboratory tests were performed at NASA Ames Research Center in 2006 to improve the models for better diagnosis. Finally, there was a second field test at the Houghton Crater. [Balaban, et al., 2007] summarize HyDE’s final performance:

All of the modeled fault modes were encountered in the field; some, such as choking, binding, and hard material, numerous times. The model-based diagnostic system was able to successfully identify the faults in roughly 85% of the cases. The rate of false positive diagnoses was approximately 5%.

Smart Spacecraft Working Group (SSWG)

ALDER

The SSWG project demonstrated autonomy capabilities relevant to a small spacecraft mission on traditional flight hardware integrated with traditional flight software. Diagnosis systems, including HyDE, were ported to VxWorks and executed on the Aitech S950 processor, interoperating with autonomy technologies for planning, diagnosis, computer vision and adaptive control. A HyDE component model of a simple propulsion system (tanks, valves, regulators, attitude control system and main engine) detected common faults such as stuck valves and regulator failures.

Other Projects

Several other projects are considering or have considered applying HyDE for offline diagnosis using simulated data. The International Space Station Electrical Power System recovery procedure automation project at NASA Ames Research Center uses HyDE to supply current hybrid state to an execution system that will attempt to partially automate recovery procedures based on this information. The Spacecraft Engine Diagnosis project at NASA Marshall Space Flight Center uses HyDE to model components of the J2X engine, which is expected to power the upper stage of the NASA Crew Launch Vehicle. HyDE was also integrated with the CLARAty (Coupled Layer Architecture for Robotic Autonomy) architecture at NASA Jet Propulsion Laboratories.

First Gov logo
NASA Logo - nasa.gov