Compact supervisory system for cold chain logistics

Abstract

In this paper, a compact supervisory system for cold chain logistics organisers is designed to track refrigerated transportation and detect improper practices only with temperature data. First, to simulate the trailer of a refrigerated vehicle, a box-like testbed is built. Second, a completed transportation task of a refrigerated vehicle is categorised into four modes according to the guidance for industry. And a compact supervisory system with four types of corresponding supervisory modules is designed to identify the current mode and detect improper practices with temperature data, the situations of the door, and the refrigeration system of the trailer. Third, a correlation matrix of the temperature and the door state data obtained from the testbed is derived. In light of the correlation matrix, data are rearranged into seven groups. Then such seven groups are used as the features of interest to classify the state of the door and the state of the refrigeration system by supervised machine learning models, such as decision tree, SVM, weighted k-NN algorithm, and ensemble method. After comparisons of their estimation performances, the optimal feature groups, and the optimal classification models of door state classification and refrigeration system state classification are selected. It means that the door and the refrigeration system states required by the compact supervisory system could be estimated by internal temperature data of the refrigerated trailer. Lastly, the validity of the proposed supervisory system is investigated with a light refrigerated vehicle.