Continuous thermal pasteurisation of various bulk liquid media is an important step in the food and allied industries. The design of a continuous flow pasteuriser is typically predicated on mathematical models developed from experimental data - usually batch, bench - scale, ethods. Of particular interest is the effect of combined pasteuriser temperature ( T ) and liquid pH on inactivation and survivor of contaminants. However, bench - scale thermal survivor data may not adequately mirror those in a continuous flow pasteuriser. This research presents the development and experimental validation of rigorous models for thermal pasteurisation of bacteria as affected by combined process T - pH in both batch, bench - scale capillary studies ( static ) and in a pilot continuous flow pasteuriser ( dynamic ), within a defined liquid and range of exposure time, temperature and pH ( t - T - pH ). Five integrated stages in synthesis and model analysis were undertaken using stringent criteria for goodness of fit of an adequate model established. First, four published predictive models were assessed against published static data ( n [subscript T] = 248 ) for the thermal inactivation of Escherichia coli ( ATCC 25922 ) in a Carbopol ® 941 liquid food simulant in batch capillaries over a range of t - T - pH. The models tested were the Classical Arrhenius, Davey Linear - Arrhenius ( D - LA ), Square - Root ( Belehradek ) and a third - order Polynomial model ( nOP ). Analysis showed the D - LA model best satisfied the criteria for model selection and explained 96.0 % V in the thermal inactivation rate coefficient. Second, the D - LA model was assessed against limited, published dynamic data ( n [subscript T] = 109 ) for the same E. coli strain in identical food simulant. The model explained 60 % V in the thermal inactivation rate coefficient. On average, model predictions of survivor numbers from the dynamic data were less than that predicted from the static data, i.e. for a given ( t - T - pH ) more bacterial cells were apparently inactivated in the continuous flow pasteuriser than in bench - scale, batch capillary studies. Overall, however it was not clear from extensive analyses of available data whether there is a statistically significant difference in survivor numbers of viable E. coli between batch static and continuous flow dynamic data. Third, although the D - LA model best satisfied the criteria for goodness of fit of a model, it failed to accurately predict the observed tails in the static survivor data. New models ( KDT and a modified KDT ) were synthesised to predict tails and shoulders in survivor data. The modified KDT ( MKDT ) form gave improved predictive capability over the KDT model when assessed against published static survivor data for E. coli and L. monocytogenes ( n [subscript T] = 355 ) in the Carbopol food simulant. This model, however, could not be readily integrated with equations describing the performance of a continuous flow pasteuriser. Analyses indicated that a greater density of dynamic survivor data for E. coli was needed. Fourth, a pilot continuous flow pasteuriser was constructed and used to generate a greater density of dynamic survivor data of E. coli ( ATCC 25922 ) in a Carbopol ® 941 carrier liquid for rigorous comparison with predictions from the Lin ( 1976 ) isothermal continuous laminar flow process model. Direct steam injection heating was used. Extensive dye and digital - video studies, in a section of glass holding tube confirmed the practical implementation of the assumptions of laminar flow and rapid condensation of steam. Extensive practical experiments highlighted a non - isothermal condition along the holding tube. A highly linear dependence ( R ² > 0.90 ) of exposure temperature with holding tube length, i.e. exposure time, was demonstrated. This was accounted for using mathematical approaches and quantitatively incorporated into a D - LA model for the rate coefficient in an extended Lin process model. A block experimental design of 4 T ( 54, 56, 58, 60 ° C ) x 4 pH ( 4.5, 5.5, 6.5, 7.5 ) x 3 replicates with a total of ( n [subscript T] = 834 ) exposure times ( 16 - 198 s ) was carried out in the pilot continuous flow pasteuriser. Findings highlighted that greater numbers of E. coli were thermally inactivated in the flow pasteuriser than predicted. From a practical operating view, the predictions from the extended Lin model were therefore conservative - with reduced risk to public health. Highly significant differences in the rates of heat - up of bacteria in the pilot pasteuriser ( dynamic ) ( 0.0104 s ) compared with that in the batch ( static ) capillary tubes ( 1.6 s ) and, mode of heat transfer, together with partial effects of dispersion with increasing length of pasteuriser holding tube, are postulated to be the controlling process influences for the difference between the experimental survivor data and the extended Lin model predictions. The lack of agreement between the continuous pasteuriser data and predictions from the extended Lin model indicated that this model cannot be practically applied. A direct comparison of the experimentally derived dynamic survivor data from the pilot pasteuriser ( as ln N / N [subscript 0] ) was also made with both the published static and dynamic data at a number of defined t - T - pH. This comparison revealed that overall, more E. coli were inactivated in the pilot continuous flow pasteuriser than described by published batch static capillary and dynamic data. Importantly, these comparisons showed that batch thermal survivor data for E. coli do not adequately mirror those obtained in continuous flow systems. Fifth, in a search for an improved model for the inactivation data, the newly derived MKDT model was assessed against the experimental pilot pasteuriser data. This model was rejected, however, because it could not account satisfactorily for all tails in survivor curves. A Weibull form model with two coefficients ( a scale factor ( α ) and a shape factor ( β ) ) also did not adequately predict tailing and could not be reliably extrapolated with holding time. However, a modified Weibull form, also with two model coefficients ( β [subscript 0], β [subscript 1] ), did give an improved fit to available experimental data. This research highlighted statistically significant differences between the dynamic thermal survivor data for E. coli and standard bench - scale static capillary data for a defined liquid and range of t - T - pH. It is likely that findings from this study can be generalised. However, validation should be carried out for a range of common indicator micro - organisms in a range of liquid foods. / Thesis (Ph.D.)--School of Chemical Engineering, 2006.
| Identifer | oai:union.ndltd.org:ADTP/280231 |
| Date | January 2006 |
| Creators | Khoo, Khar Yean |
| Source Sets | Australiasian Digital Theses Program |
| Language | en_US |
| Detected Language | English |
Page generated in 0.0081 seconds