Fouling modelling in a UHT unit based on plant data.

Abstract: Heat treatment of dairy products is essential for food safety, one way to perform this is ultra-high temperature (UHT) processing. During the process, macronutrients and minerals form fouling deposits which impairs heat transfer. Eventually the fouling needs to be removed chemically by alkaline and acidic detergents. Fouling buildup and removal are dependent on both internal parameters such as the physical properties of the product and external parameters such as time, flow rate, temperature, and concentration. Modelling fouling development and removal on an industrial scale is valuable from an economical and environmental perspective. By optimizing how long production and cleaning cycles should be the productivity can be increased and the chemical waste as well as the energy consumption can be decreased. Although many models for fouling have been developed, each model is highly dependent on the type of heat exchanger, the product, and the operational parameters. Furthermore, they do not account for upstream and downstream processes which can affect the operator’s decision. This thesis takes a data-driven approach to construct fouling development and removal models from 27 months’ worth of data that was retrieved from a specific heat exchanger. The raw dataset was divided into production and cleaning in place (CIP) cycles. Each cycle had specific requirements that had to be fulfilled e.g., for production they should be longer than 5 hours. CIP cycles were further divided into caustic and acidic cycles. However, due to deviations from standard operating procedures (SOP) and unreliable conductivity meters, caustic and acidic cycles could not be reliably differentiated, therefore 10 reliable CIP cycles were used instead. Regression models were developed for the production and cleaning cycles, where the fouling during production increased linearly with time while cleaning cycles had a more complicated relationship with fouling.