ORBITAL TRANSPORTS INC — National Aeronautics and Space Administration SBIR Phase I: S17

Description

To address NASArsquo;s need for advanced model-based systems engineering (MBSE) methods and tools that integrate digital engineering and science activities across the entirety of the mission and program life cycle, Orbital Transports proposes to develop AI for Systems Engineering (AISE), an innovative digital design assistant to provide semantic search and classification of systems design work products enabling reuse of systems models in new missions and contexts. Historically, knowledge about a system or its context would be siloed within a single mission or even a single project. AISE will make system models and institutional systems engineering knowledge broadly accessible across the organization for use with other missions and programs, greatly accelerating risk-informed and evidence-based decision making.The proposed digital design assistant utilizes OpenAIrsquo;s Generative Pre-Trained Transformer 3 (GPT-3) to identify and classify systems engineering models expressed in SysML v2 for retrieval from a repository of SysML v2 models using symmetric and asymmetric queries. Symmetric queries use an engineerrsquo;s current work context to recommend similar models from the repository developed within the organization in different projects and contexts, reducing redundancy and errors, thus streamlining development by encouraging reuse of previously validated work products. Asymmetric queries enable NASA engineers to use naturalistic, free-form prompts to pursue effective lines of inquiry through the repository, easily finding relevant engineering designs and work products even when the exact target of search is not clearly known or well-formed.The Phase I effort will validate methods for extracting and saving embeddings of SysML v2 work products, and for processing symmetric and asymmetric searches of a repository of systems models. The proposed effort will deliver a proof-of-concept demonstrating these methods to support efficient retrieval and usage of organization work products.nbsp;nbsp;