Processing and control strategies for the improvement of spray-dried coconut milk quality

Spray drying of liquid containing hydrophobic material such as coconut milk fats depends entirely on its feed formulation and process operating conditions. Good encapsulation of fats can only be achieved by spray drying of a stable emulsion. However, an excessive amount of additives is usually used to stabilize the emulsion without considering the natural quality of coconut milk. The main goal of this research is to propose process and control strategies for the improvement of the quality of spray-dried coconut milk by minimizing the use of additives. In order to achieve this goal, the effect of feed formulation and inlet drying temperature on the spray dried powder is evaluated. It also focuses on maintaining product quality by the application of a nonlinear model-based inferential control strategy. Different sonication amplitude setting (60, 80 and 100%) and sodium caseinate (SC) concentration (0 to 2% w/w) were used in the emulsification process to evaluate the effects on the degree of stability of the emulsion. The effect of sodium caseinate concentrations and drying temperatures (140, 160 and 180ºC) to spray dried coconut milk was assessed based on the physical and functional properties of the powder. The one-dimensional model with the integration of reaction engineering approach (REA) model was used to predict the dynamic behaviour that relates operating temperatures to the moisture content of powder produced. The empirical model, i.e., nonlinear autoregressive with exogenous input (NARX) model and neural network (NN) model, were used in the inferential control system as system identification and soft sensor estimator, respectively. High stability and good properties of coconut milk emulsion were achieved with the addition of SC concertration ≥ 1% w/w and ultrasonic amplitude setting of ≥80% The emulsification process significantly reduced the effective average particle size to lower than half (<5 μm) of its original size. Increasing the SC concentration (≥ 1% w/w) reduced both the creaming index and free fat content to 0% and <10%, respectively without significant change in apparent viscosity. This indicates that the emulsion has achieved high stability. High ultrasonic amplitude setting (≥80%) leads to the formation of stabilized emulsion with submicron range size (<1 μm) of droplets. For spray dried coconut milk, good physical and functional properties of powder were achieved with the addition of SC concentration ≥1 % w/w and inlet temperature of ≤160°C. Higher the SC concentration (≥%1 w/w) produced smaller-sized powder particles (<30 μm) which leads to low particle density and bulk density of the powder. High stability emulsion with 1% w/w of SC was unstable during the atomisation process due to re-coalescence of fat as the size of droplet increased to >2μm after the spray drying process. Adding SC to emulsion reduced the moisture content of powder to less than 5% without significant change due to SC concentration. The lowest moisture content (<4%) was obtained at the inlet temperature of 180°C. The highest free fat content, insolubility and droplet size were obtained at the inlet temperature of 180°C regardless of SC concentration. The presence of fleck was also noticed in the powder. The reaction engineering approach (REA) model is used to represent the drying kinetics of a single droplet of coconut milk. Integration of the REA model into the one-dimensional model enables accurate prediction of dynamic moisture content of spray dried coconut milk. The developed REA model accurately predicts the drying behavior of the coconut milk droplet with R2 value of 0.9786 obtained during model validation. Integration of REA model into one-dimensional model leads to a high accuracy model to predict the variables of the spray drying process with the mean absolute percentage error (MAPE) was found to be 17.1% for moisture content and 6.2% for outlet temperature. Good control performance was achieved by the nonlinear inferential control system in controlling the moisture content of coconut milk powder. Minimal offset (<0.0003 kg/kg) of the responses at various set points were obtained which indicates the accuracy of the neural network estimator. Less overshoots were obtained by TL PI inferential control during setpoint tracking. On the other hand, good control performance was obtained by ZN PI inferential control during disturbance rejection.

Text FK 2021 88 - IR.pdf Download (2MB)

Actions (login required)

View Item