Cold chain break detection and analysis: Can machine learning help?
Résumé
Background: The impact of the cold chain breaks on food products is widely documented with multiple stakes: health, environmental and economic. The emergence of Internet of Things (IoTs) will enable more rigorous temperature monitoring in real time but raises new questions about the processing of the generated data. Scope and approach: Different definitions and challenges associated with the detection of cold chain breaks are presented and discussed. Machine learning methods applied to cold chains are described in order to highlight the issues related to these data. In addition, these studies allow us to bring out the different data sources that can be used to train the learning models. Key findings and conclusions: The field of cold chain generates experimental and numerical data that have a great potential to train machine learning models. To our knowledge, although machine learning methods have been used to predict temperature, these methods have not been used to detect breaks in the cold chain. However, several methods already exist to detect anomalies in time series data. Learning from these data would be a step forward: on one hand, to get a better knowledge of cold chain breaks, and on the other hand to alert operators at the right time.