SOURCE:

Faculty: Engineering
Department: Mechanical Engineering

CONTRIBUTORS:

AZAKA, O. A
Enibe, S.O
Achebe, C.H

ABSTRACT:

A 2-Dimensional numerical model of coupled heat and mass transfer equations of
convective drying for rectangular and cylindrical Cassava slices was developed. The
experiment was conducted in the convective drier with controlled air temperature,
and a digital mass measuring scale. The developed unsteady-state partial differential
equations were solved by means of the finite element method by employing Galerkin
weighted residual approach. The state variables, surface temperature and moisture
content of the samples, as well as the shrinkage, and drying rate were determined
using the formulated model. The numerical model was used to predict the effect of
fixed and moving boundaries on the drying curve. The relative drying curve, drying
rate and surface temperature of the numerical simulation were in good agreement
with the experimental data analyzed. The results showed that the model best
described the drying process of cassava pellets under the conditions tested. The
diffusivity of the rectangular samples were determined to be in range of 1.4260 x 109
- 2.3304 x 109 m2
s-1 and described using a 3rd order polynomial
equation with the coefficient of
regression R² = 0.8989. The diffusivity plots clearly explains that the moisture
diffusivity of cassava pellets can be expressed as a function of temperature using a 3rd
order polynomial equation. The quadratic models which described the convective
mass transfer coefficients explain that convective mass transfer coefficient varies
with time. The R2 values and correlation coefficients (r) are mostly above 0.9
indicates that the convective mass transfer coefficient of the cassava samples is
adequately modeled as a quadratic function of drying time for each air temperature.
In finite elements analysis of the samples, solutions improved by mesh refinement.
Hence, the predictions generated for a 99-element model, a 367-element model, a
1395-element model and a 5447-element model show that the solutions progressively
fell within the experimental range of moisture content as number of elements
improved. The R2
, RMSE and r for the predictions are respectively 0.9340, 0.1924
and 0.9912 for the 1395-element model while the prediction accuracy improved to
0.9958, 0.0483 and 0.9984 for the 5447-element model. Marked improvement in
accuracy is recorded on mesh refinement. Using the 5447-element model, FE
predictions at the experimental intervals were compared with the measured values
and the corresponding statistical goodness of fit values. The R2 values were seen to
be above 90% where majority were up to 99% indicating high accuracy of FE