Cluster analysis for identification of temperature breaks in a reefer container from IoT data

Temperature breaks are critical for deteriorating the quality of the perishable item during cold chain logistics in a reefer container. The breaches generally occur because of undesirable door opening and refrigeration shutdown or failure. Therefore, automated remote identification of such events from IoT data may enable appropriate action to avoid quality loss of the product. In this paper, we conduct few experiments in a reefer container emulating two temperature break events such as door opening, refrigeration shutdown and corresponding corrective events that results in moving the temperature towards the set point. The temperature data collected therefrom is used for automatic identification of the events. In particular, we cluster the subsequences from the time series using the similarity index computed from Dynamic Time Warping (DTW) algorithm and identify the patterns indicating the events related to temperature breaks. The result shows silhouette and adjusted rand index score of 62.5% and 74.5%, respectively. Thus, indicating correct grouping of the events.