A Model-based Approach to Engineering Behavior of Complex Aerospace Systems

Abstract

One of the most challenging yet poorly defined aspects of engineering a complex aerospace system is behavior engineering, including definition, specification, design, implementation, and verification and validation of the system's behaviors. This is especially true for behaviors of highly autonomous and intelligent systems. Behavior engineering is more of an art than a science. As a process it is generally ad-hoc, poorly specified, and inconsistently applied from one project to the next. It uses largely informal representations, and results in system behavior being documented in a wide variety of disparate documents. To address this problem, JPL has undertaken a pilot project to apply its institutional capabilities in Model-Based Systems Engineering to the challenge of specifying complex spacecraft system behavior. This paper describes the results of the work in progress on this project. In particular, we discuss our approach to modeling spacecraft behavior including 1) requirements and design flowdown from system-level to subsystem-level, 2) patterns for behavior decomposition, 3) allocation of behaviors to physical elements in the system, and 4) patterns for capturing V&V activities associated with behavioral requirements. We provide examples of interesting behavior specification patterns, and discuss findings from the pilot project.

A Tool for Model-Based Generation of Scenario-Driven Electric Power Load Profiles

“…To start filling this existing need, initial pilot efforts were performed to prototype ontologies and modeling patterns that would support the storage of power behavior information, along with the scripting necessary for transforming the resulting models into solvable constraint sets 5,6 . This early work was founded on even earlier developments in behavioral modeling approaches 7,8 , as well as the Timeline representation of time-varying information 9 . SPLAT was then the next step -the combining of these earlier capabilities into a single software package (implemented as a Java plugin in the MagicDraw ‡ tool), with the addition of features allowing users to more easily configure and execute the analysis.…”