Cleaning within the food industry is a complex process which is necessary for both product sterility and also for efficient processing conditions. At present, cleaning is expensive, mainly due to the cost of chemicals and plant downtime. Some attempts have been made to optimise the process, but these were inadequate due to the lack of understanding of the fundamental mechanisms. The primary aim of this work is to develop a better understanding of those mechanisms. This dissertation studies the chemical and physical factors affecting the removal of whey protein deposits from stainless steel tubular heat exchanger surfaces. The cleaning process involves diffusion of the cleaning solution into the deposit matrix, reaction to form a swollen gel and dissolution of the reaction products away from the surface. As well as these chemical effects, the fluid imparts shear stresses on the deposit surface. Overall mass removal during cleaning was measured using a protein assay. At the same time a micro-foil heat flux sensor (MHFS) measured heat recovery over a section of the surface. The data showed the unexpected result that mass removal and heat recovery feature different sequences. This difference was shown to be due to deposit swell, providing a greater thermal resistance, whilst deposit removal still occurred from the surface. The MHFS was also used to measure the thermal conductivity of the deposit (0.28 W/mK), which was found to be significantly lower than water. A kinetic analysis of the data suggested the process was mass transfer controlled .. A novel approach to cleaning using pulsed flow was investigated. The aim of this work was to increase cleaning rates and hence reduce cleaning times. A cleaning rate enhancement of up to 70% was found by applying a reverse flow pulse to the system of frequency 2 Hz and displaced volume 2 ml. The reasons for this enhancement were partly due to increased mass transfer rates but mainly due to the higher shear forces acting on the deposit surface. Reverse flow pulses as opposed to non-reverse flow pulses had a significant effect on the enhancement. The change in the deposit microstructure was studied during cleaning, using a range of surface analysis techniques: SEM and freeze drying, X-ray elemental mapping, CSEM, SEM and glutaraldehyde fixation. These techniques showed that the cleaning solution diffused into the deposit structure within 3 seconds. In addition, the deposit morphology changed from an aggregate structure to a fused structure after 3 seconds and then changed further, after 4 minutes, to a structure resembling a fine stranded gel. On the basis of these results, a physical model was proposed to predict both heat recovery and mass removal during cleaning. The model initially assumed that the deposit consisted of a series of nodes, all of the same height. The observed swelling of the deposit was modelled as an increase in node height and removal was modelled stochastically as a decrease in node height. The removal mechanisms included uniform (assumed to be due to mass transfer or reaction limitation) and random removal from random nodes. A sensitivity analysis was carried out to investigate the balance between these effects. The best fit mass removal curve was achieved by assuming that 99% of the deposit was removed randomly. The heat recovery curve was found to fit by: (i) the deposit swelling and affecting the superficial velocity, resulting in a higher Do and (ii) removal occurring from inside the deposit structure, resulting in no change in the deposit thickness but in an increase in the deposit voidage.