Machine Learning Model for a Digital Twin for Predictive Maintenance and Optimization in A Fruit Supply Chain

The statistics released by the Logistics Bureau of US state that the shipment of fruits in a fruit supply chain bears about a 33% loss during transit due to inefficiencies in the supply chain and mismanagement of stock and resources. This is a critical issue for fruit dealers and demands a mechanism to reduce these losses. This research paper focuses on the development of a machine learning model for the implementation of a digital twin to help fruit storage warehouses prolong the shelf-life and the fruit quality of their stock by utilizing predictive maintenance and optimizing storage conditions. The integration of digital twin with predictive analysis using Decision Tree Regressor, Random Forest Classifier, Naïve Bayes and Supported Vector Machine produces overall accuracy of 82%, thus showing tremendous potential in the model thus, better decision making. Digital twin creates a virtual replica of the supply chain, which, on inputting real-time data for temperature, humidity, and other environmental factors, uses predictive analysis to determine the approximate shelf-life, overall spoilage, and degradation of the fruits. This model is quite helpful for warehouse managers to take inform decisions about strategies to rotate stock and manage the produce inventory. By applying advanced analytics, the model can also predict potential rot conditions and maintenance needs. This study provides valuable insights for future applications in the agri-food sector.