Data Driven Shelf Life Prediction

Data Driven Shelf Life Prediction

Ceelen, Marcel (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Transport Engineering and Logistics)

Polinder, Henk (mentor)
Garofano, Vittorio (mentor)
Negenborn, Rudy (graduation committee)
Jarquin Laguna, Antonio (graduation committee)

Delft University of Technology

Mechanical Engineering | Transport Engineering and Logistics

2019-07-18

Artificial intelligence is used in this research to predict the shelf life of strawberries. The prediction of shelf life is based on temperature measurements from the moment a package of strawberries is harvested till the moment this same package is bought by a customer in a local PLUS supermarket. The strawberries are harvested in the south of Spain, near Huelva and distributed to local PLUS Supermarkets near Rotterdam. After the packages with strawberries, including temperature loggers, arrive at the supermarket shelf, the packages are moved to a shelf life room for visual inspection. During this daily inspection, the actual shelf life of the strawberries is determined by a classified inspector. The combination of the actual shelf life and the temperature profile through the supply chain is used to train, validate and test different machine learning algorithms. The most reliable shelf life prediction algorithm is the Exponential Gaussian Process Regression Algorithm, with the smallest confidence interval and an average deviation of 14.1 \%. To conclude, the possible improvements in the supply chain based on shelf life prediction, like traceability, food date labeling and quality grading are evaluated.

Artificial intelligence
Shelf Life Prediciton
Food

http://resolver.tudelft.nl/uuid:e27cdddf-b97d-4550-b3bc-e59c2d22b48d

Student theses

master thesis

© 2019 Marcel Ceelen