Showcasing our research contributions in Reliability Engineering and AI.
Publisher: IEEE
Authors: Kalyan Bhagavan Tadaka, Rocco Cassandro, Wei Zhang, Zhaojun Steven Li
Design for Reliability (DfR) is an engineering process aimed at enhancing product reliability by proactively identifying and analyzing potential failure modes, their causes, their effects and consequences during the early stages of engineering design. Hence, it helps mitigate risks in design and accelerates product development timelines. However, implementing DfR in complex engineering systems poses significant challenges due to factors such as limited testing data, diverse failure information, and uncertainties.
This research explores the use of Large Language Models (LLMs) to develop a customized Copilot, ChatReliability, to facilitate the DfR process. The developed Copilot is trained on an array of reliability data sources, encompassing both qualitative and quantitative such as Design Failure Mode and Effects Analysis (DFMEA), reliability standards, failure databases, and engineering instructions. Its primary function is to generate DfR outcomes including reliability estimation, risk prioritization/visualization, and test plans. By enabling designers to interact with the Copilot, they can collaboratively explore solutions for design enhancement and reliability optimization.
The robust capabilities of LLMs empower the Copilot to efficiently interpret and synthesize vast volumes of design and failure-related data. The prototype of the Copilot is presented in this research to showcase the potential of utilizing LLMs in facilitating DfR practice.
