Heat transfer studies on canned particulate Newtonian fluids subjected to axial agitation processing.

摘要

Heat transfer to canned particulate laden Newtonian fluids was studied during free axial agitation thermal processing in a pilot STOCK retort which was modified to simulate the can motion in continuous turbo cookers. Evaluation of heat transfer coefficients (overall, U and fluid to particle, hfp) associated with canned liquid/particle mixtures, while they are subjected to free axial motion is difficult because of the problems involved with attaching temperature measuring devices to liquid and particles without affecting their normal motion. A new methodology was developed to evaluate U and hfp in Newtonian liquids. The methodology involved first correlating U and hfp as a function of input variables for cans in fixed axial mode of rotation in which both particle and fluid temperatures were measured using thin wire thermocouples. Subsequently, only liquid temperatures were measured in cans using wireless sensors in the free axial mode, and hfp values were empirically computed from the developed correlations and the measured temperatures. An L-16 orthogonal experimental design of experiment was carried out to select system and product parameters that significantly influence hfp and U for particles in the Newtonian liquid. With significant parameters selected, a response surface methodology and two full factorial experimental designs were used to relate U and hfp to process variables in each mode of rotation (fixed and free axial modes).;Dimensionless correlations were then developed using the evaluated data for heat transfer coefficients (U and hfp), in canned high viscosity Newtonian liquids (with and without particles) using stepwise multiple non-linear-regressions of significant dimensionless groups. In free axial mode, combining the natural and forced convection, Nu = A 1(GrxPr)A2+ A3(Re) A4 (Pr)A5 FrA 6 (rhop/rhop1)A 7 (e/100-e)A8 (dp/Dc) A9 (Kp/K1)A10 yielded a higher R2 (0.93) than using a pure forced convection model when particles were present in the can. Even in the absence of particles, and with the end-over-end mode of agitation where forced convection dominates, introducing natural convection term (GrxPr), improvedR2 from 0.81 to 0.97. Artificial neural network (ANN) models were also developed for heat transfer coefficient predictions and the trained models gave better predictions than dimensionless correlations. All ANN models developed could be implemented easily in a spreadsheet as either matrices or a set of equations.