Baghouse design for milk powder collection.

This thesis aims to improve the understanding of the factors that determine the performance of baghouses used for milk powder collection. The research focuses specifically on the similarities and differences between milk powder collection and other common baghouse applications. The thesis also aims to demonstrate the value of recent developments in computational fluid dynamics in developing predictive models of baghouse performance. It is hoped that the findings of the thesis may find application in the New Zealand dairy industry, where such baghouses are commonly used to collect milk powder after spray drying. The effect of operating temperature and humidity on the performance of baghouses was investigated by examining both the forward filtration process and pulse cleaning process. Forward filtration was examined in a series of bench scale experiments, then scaled up to the pilot scale to confirm the findings. The effect of humidity on the pulsing performance was then investigated at the pilot scale. The importance of pulse system design was investigated at the pilot scale in a separate set of experiments. Pulse nozzle position, pulse pressure, and pulse duration were varied and the effect on the baghouse pressure differentials was measured. A computational fluid dynamics (CFD) filter model designed for membrane filtration was adapted with some success to simulate a milk powder baghouse. The model was successful in predicting the length of the low pressure zone at the top of the bag, and the general trends in overpressure associated with changes to the pulse system geometry. The model was not successful in predicting the acceleration of the filter bag during the pulse. The model was used to simulate both forward filtration and pulsing, to extend the results of the experimental investigation. The effects of changes in the pulse nozzle height, pulse nozzle diameter, and pulse pressure were simulated, as well as the effect of gravitational settling during forward filtration, to extend the results of the previous experiments. There is a clear opportunity remaining for further work to extend the basic model developed here and to adapt the model to simulate large industrial baghouses. Experiments on the bench scale and pilot scale indicated that increased cohesive forces between particles improve the performance of milk powder baghouses by lowering the resistance of the filter cake during forward filtration and aiding cake removal during pulse cleaning. Under the conditions typical of industrial milk powder baghouses, cohesive forces are governed primarily by liquid bridging between particles, due to melted fat (particularly at high temperatures) and softened lactose (at high humidity levels). As a range of milk powders with different compositions are produced commercially, the relative importance of lactose-based and fat-based cohesion differs between powder types. Cohesion promotes the formation of porous structures in the filter cake, improving the cake permeability. In skim milk powder (SMP), particle cohesion is dominated by softened lactose, and is highly moisture dependent. In the bench scale experiments conducted here, increasing the relative humidity from 6% to 17% decreased the specific cake resistance from 1.69x10⁹ m.kg¯¹ to 8.23X10⁸ m.kg¯¹, and decreased the proportion of powder adhering to the filter from 14% of the total supplied powder to 3%. The combination of these effects decreased the total resistance over the filter from 1.09X10⁹m¯¹ to 1.89X10⁸; m¯¹, an 83% reduction. The low deposition at high humidity suggested that the porous cake structure formed at high humidity levels was fragile, so that deposited particles were prone to subsequent dislodgement, especially in areas where the shear velocity near the filter surface was high. In pilot scale experiments, the porous cake structure formed at high humidity was more easily removed from the filter bag, resulting in more effective pulse cleaning. It was concluded that particle cohesion promoted cake filtration over depth filtration, as particles tended to adhere to the cake surface immediately upon contact. As depth filtered particles are more difficult to remove, the shift toward cake filtration at high humidity improved the pulse cleaning performance. A high-fat milk protein concentrate (MPC) powder was also filtered on the bench scale apparatus. Particle cohesion in the MPC powder was dominated by liquid fat, and showed a clear dependence on temperature but not on humidity. Increasing the temperature from 30°C to 90°C caused the specific cake resistance of the MPC to decrease from 1.06x10⁸ m⁻¹ to 3.94x10⁷m⁻¹, a 63% decrease. The deposition of MPC powder was unaffected by either temperature or humidity. Gravitational settling of particles in large baghouses was found to produce significant variations in the properties of the filter cake throughout the baghouse. Experimental results with the pilot scale baghouse found a strong decreasing trend in the particle size with increasing height in the baghouse, with the mean particle size decreasing from 117 μm at the bottom of the baghouse to only 31 μm near the top of the filter bag. The filter cake thickness also decreased sharply with height. Results from the CFD simulations indicated that in the pilot scale baghouse particles larger than 120 μm in diameter tend to fall out of the air flow and collect in the bottom of the baghouse, instead of depositing on the filter. While industrial baghouses tend to have a higher elutriation velocity than the pilot scale baghouse used in this study, the large size of industrial baghouses provides ample opportunity for particles to segregate on the basis of size. In addition, bench scale results indicated that high air velocities near the filter surface may cause particles to rebound from the filter. This may occur in industrial baghouses in the region near the inlet, where the air velocity is highest. The reverse pressure differential induced in the filter bag by a cleaning pulse was found to increase with distance from the cell plate. Positioning the nozzle too close to the bag opening created a low pressure zone just beneath the cell plate, where the pressure remained lower inside the bag than outside throughout the pulse. This may lead to poor cleaning at the top of the bag. In the pilot scale baghouse, positioning the nozzle at least 0.7 m from the bag opening eliminated the low pressure zone. The optimum distance of 0.7 m is is dependent on the nozzle type and bag diameter, but can be directly applied to recent industrial baghouse designs in the NZ dairy industry, which have the same nozzle type and bag diameter as the pilot scale baghouse. The design of the pulse cleaning system is important in achieving good baghouse performance. Increasing the pulse tank pressure on the pilot scale baghouse from 3.5 bar to 6.5 bar caused a 30% reduction in the forward pressure differential after the pulse, while decreasing the pulse pressure below 3.5 bar caused the pressure differentials to increase indefinitely. Altering the nozzle position had no effect on the overall pressure differentials, but did alter the local acceleration at different points on the filter bag during a pulse. CFD simulations indicated that decreasing the distance between the nozzle and the bag opening from 0.7 m to 0.1 m increased the overpressure at the bottom of the bag from 770 Pa to 3500 Pa, but this was offset by the appearance of the low pressure zone at the top of the bag as mentioned above. CFD simulations indicated that the diameter of the pulse nozzle altered both the mean bag overpressure generated by the pulse, and the distribution of the overpressure over the bag surface, with the low pressure zone at the top of the bag becoming longer at large nozzle diameters. The pulse duration was found to be unimportant, with experiments on the pilot scale baghouse finding that this had no effect on either the overall baghouse pressure differentials or the length of the low pressure zone at the top of the bag. The project has extended the understanding of milk powder baghouse performance by relating the moisture-dependent properties of lactose and the temperature-dependent melting of dairy fats to baghouse performance. The project has also provided a useful design tool in the form of the CFD model. The project demonstrates an opportunity for further CFD research into baghouse design, as the basic model developed here could now be modified to directly simulate large industrial baghouses. It is hoped that the results from this thesis will find application in the New Zealand Dairy Industry