Fouling and Cleaning of Plate Heat Exchangers for Milk Pasteurisation: A Moving Boundary Model

Abstract Plate heat exchangers (PHEs), widely used in the food industries, use large amounts of energy. Fouling reduces their thermal and hydrodynamic performance, and requires periodic cleaning (often after few hours), with productivity losses. Cleaning uses large amounts of water and produces wastes. Improving productivity, minimising energy, water and wastes, while ensuring effective pasteurization is of high interest and can be supported by simulation models. Current models describing the dynamic behaviour of PHEs with fouling typically rely on simplified assumptions, with limited predictive ability. More detailed models based on Computational Fluid Dynamics (CFD) are very computationally expensive for complete PHEs, not always better at fitting experimental data, and practically infeasible to use for operations optimisation and control. A 2D distributed, dynamic model described by Guan and Macchietto (2018) for a single PHE channel, is extended to enable the modelling of full PHEs with multiple plates arranged in any design configuration (e.g. with parallel, countercurrent, mixed parallelcountercurrent flow). The model is validated here for two such arrangements (A1 and A2) against experimental results. Four milk fouling models are assessed: two deposition mechanism (due to aggregate or denatured proteins), each with/without deposit re-entrainment. No single model was found capable of describing both arrangements. A1 was predicted best by models with aggregate proteins deposition, A2 with fouling by denatured proteins. A dynamic model for cleaning of fouled surfaces was integrated with the moving boundary model and validated against experimental data. This enabled for the first time in the literature the simulation of Cleaning-in-Place (CIP) procedures, establishing the duration of cleanings, and the seamless simulation of entire single and multiple heating-CIP cycles, with either pre-set or condition-based termination conditions for each phase. It is shown there is large scope for optimising the performance and productivity of overall heating-CIP cycles, and an optimal operation is demonstrated for arrangement A1.