GPT-4o assisted generation of PFMEA process step & process work element functions

Existing Failure Mode and Effects Analysis (FMEA) documents have information on process step name, failure modes and rootcauses. However, process step (PS) function and process work element (PWE) functions are not normally defined as per recent VDA-AIAG guidelines. This research paper explores the role of generative AI models such as GPT-4o, by taking inputs such as multiple failure modes and root causes for each process step name to generate Process Step (PS) functions and Process Work Element (PWE) functions. Utilising this information, GPT-4o updates failure modes if necessary and predicts detection controls for PS functions, as well as detection and prevention controls for PWE functions, forming a comprehensive pipeline of step-by-step interactions. The paper demonstrates the implementation of this pipeline on two FMEA sheets used as case studies: a polymerisation process and a machining process involving a 5-axis precision drilling machine, showcasing the efficient conversion from the old format to the new guidelines using self-designed prompts. Users have access to these prompts, allowing them to input their own data and generate their own results. This approach is particularly beneficial for process and quality engineers who have limited time, as it significantly reduces the effort required for FMEA updates as it provides them with a practical and time-saving solution for maintaining high-quality standards.