ABSTRACT
Conventional single slope solar still integrated with a parabolic trough collector in addition to packaged glass ball layer was used in water desalination. Experimental work carried and data obtained were used in modeling the influential parameters affecting water desalination using principal component analysis to reduce parameters’ numbers. These parameters were then applied in constructing a Response Surface Model (RSM). System’s Performance has been predicted in terms of temperatures; saline water temperatures (Tw), glass cover temperatures (Tg), dry bulb temperature (Tdb), and wet bulb temperature (Twb) inside the conventional solar still. Along with the above temperatures, ambient air temperature (Ta), Oil inlet temperature (Toi) and solar intensity (I). The impact of these parameters can highly effect the water desalination yield. The RSM is developed to predict the impact of these parameters. Principal Component Analysis (PCA) reduced and categorized the number of effective parameters to three components containing the parameters (Tw, Twb, and Tdb), (I, Ta, and Tg) and (Ta, Toi, and I) respectively. These combinations were then tested using the two different RSM optimized models; Modified Reduced Quadratic and Two-Factor Interaction (2FI) model. These models were tested by monitoring nine statistical indices namely: RMSE, P-value, F-value, R2, R2adj, R2pred, BIC, PRESS, and AICc. Results obtained for this model showed minimum RMSE value and higher values of R2, R2adj and R2pred as well as lowest values for BIC and AICc. Concluding supremacy of (I, Ta, and Tg) Modified Reduced 2FI model over the others.
Nomenclature
AICcAkaike’s Information Criterion
ANNArtificial Neural network
BICBayesian Information Criterion
F-valueFisher–Snedecor distribution
Isolar intensity
PCAPrincipal Component Analysis
PRESSPredicted Residual Error Sum of Squares
PTaccompanying loading matrix
P-valueprobability value
R2Coefficient of determination
R2adjAdjusted Coefficient of determination
R2predPredicated Coefficient of determination
RMSERoot Mean Square Error
SSESum Square Error
SSTTotal Sum Square.
Tscore matrix
Taambient air temperature
Tdbdry bulb temperature
Tgglass cover temperature
ToiOil inlet temperature
Twsaline water temperature
Twbwet bulb temperature
Xindependent variable
λiEigenvalue associated with the eigenvector pi
Acknowledgments
The authors extend their appreciation to the Deanship of Scientific Research at King Khalid University for funding this work through Research Group Program under grant number (R.G.P.1/132/40).
Additional information
Funding
Notes on contributors
Osman Haitham
Haitham Osman is an Assistant Professor at King Khalid university, graduated from Newcastle University (UK). His research interests included but not limited to process modeling and optimization, advance process control, separation processes, and machine learning. He has published papers in journals and conferences in the area of process statistical modeling, numerical analysis, optimization, and artificial intelligence.
Madiouli Jamel
Jamel Madiouli is an Assistant Professor at Kairouan university. Graduated from Monastir University (Tunisia) in collaboration with RAPSODEE in EMAC, France, he undertook number of posts within Gafsa university, King Khalid University and finally Kairouan university.Dr Madiouli works in different areas within the Mechanical engineering arena, among them but not limited to drying, sintering, solar energy, desalination and shrinkage measurement during thermal processes
Shigidi Ihab
Ihab Shigidi is an Assistant Professor at King Khalid University, Graduated from Loughborough University, he undertook number of posts within King Khalid University.Dr Shigidi works in different areas within the Chemical engineering arena, among them but not limited to separation processes, desalination, transport phenomena and mathematical modelling.