Cleaning In Place (CIP) is the process of cleaning product pipes and filling valves with a sanitizing product or water, which can be heated (hot CIP) or left at ambient temperature (cold CIP). Hot CIP is characterized by different consecutive phases (steps) corresponding to the evolution of the product temperature. This paper presents an unsupervised approach for detecting anomalous hot CIP and classifying the normal ones according to the number of steps. It is based on the combination of time series analysis representation with functional data clustering and does not require a training set. This study was conducted to find a valuable alternative to a supervised approach previously adopted, which reached remarkable results but required an effortful and time-consuming dataset preparation.
Tessoni et al. (Thu,) studied this question.