Abstract:
FMEA – Failure Mode Effect Analysis method is used as a critical tool for identifying
and mitigating risks to ensure efficiency, cost-effectiveness, and high product reliability.
FMEA is usually done with the traditional methods that are often manual, time-
consuming, subjective, biased, and prone to inconsistencies due to non-standardized
failure records. The purpose of this study is to develop and propose a data-driven FMEA
framework to reduce manual effort, increase objectivity in risk assessment, and enable
proactive quality improvement.
The study will use data science with the integration of K-Nearest neighbors (KNN) and
fuzzy logic to enhance the identification, classification, and prioritization of failure
modes. Collected real-world FMEA data from PT X, a screw manufacturing company in
Indonesia, and gained expert insights into AI integration, which provides a foundation for
developing adaptive FMEA models.
Traditional FMEA at PT X is reactive, manual, with non-standardized failure modes. The
proposed enhanced FMEA framework will enable automated failure classification,
objective Risk Priority Number, faster scoring, and manage more data without manual
intervention, and it will be faster and efficient in the meantime. It will increase
consistency and objectivity.
The integration of KNN and Fuzzy logic methods into the FMEA process can improve
risk management practices in product development in manufacturing, making it more
adaptive, data-driven, and proactive. For PT X, it can reduce the time to respond when
addressing potential failure. The approach is scalable and can be applied across industries,
aiming to modernize the FMEA process.
The proposed method will enable decision makers to use a new way for application of
machine learning techniques in FMEA as align with industrial 4.0 to leverage
digitalization.
