Modified FMEA using machine learning for food supply chain

Abstract

Food waste has been a major issue in the food supply chain in recent years. Many FMEA methods have been developed to identify supply chain failures, but the problem remains the same. This paper describes a modified FMEA that can identify the most serious problem in seconds. One of the major issues with FMEA is its subjectivity. The value of the FMEA for the same failure in the supply chain changes as the person changes. A machine-learning model is created in order to eliminate subjectivity. To train the data, the deep neural networks approach, which is part of the machine-learning model, is used. A secondary data source is used for this. According to the findings, RPN values for the selected food supply chain stages of harvesting, handling & storage, processing & packaging, and distribution & retail are roughly the same in both conventional and modified FMEA methods. To predict RPN values at periodic time intervals, the conventional method requires human intervention. However, in the modified FMEA method, the programme automatically updates and predicts the RPN values. The continuous training of the model with periodic time allows the algorithm to accurately predict the RPN values for any new failure mode that occurs in the supply chain.