SOFT COMPUTING MODEL FOR THE FORECASTING OF MOISTURE RATIO OF MANIHOT SPP.pptx

SOFT COMPUTING MODEL

1. SOFT COMPUTING MODEL FOR THE FORECASTING OF MOISTURE RATIO OF MANIHOT SPP

2. Background of the study Problem statement Aim and objectives Significance of the study Literature review Materials and method Results and discussion Conclusion

3. BACKGROUND OF THE STUDY Definition of drying Drying is a process which involves the removal of moisture due to simultaneous heat and mass transfer. Prediction of drying parameters There are various methods for prediction of drying parameters of agricultural products. One simple way is to use available empirical correlations. However, Artificial Neural Network (ANN) can more accurately predict drying parameters.

4. Problem statement The shelf life of foods is determined by their moisture contents which can be computed using moisture ratio. Hence this research study on the development of an ANN model for the accurate prediction of moisture ratio of Cassava.

5. SIGNIFICANCE OF STUDY With the aid of this work, accurate prediction of the moisture content of cassava can be carried out with ease. This work will serve as a cost free means for simulating and controlling the cassava drying process.

6. LITERATURE REVIEW History of cassava Wild populations of M. esculenta subspecies flabellifolia , shown to be the progenitor of domesticated cassava, are centered in west-central Brazil, where it was likely first domesticated no more than 10,000 years. The oldest direct evidence of cassava cultivation comes from a 1,400-year-old Maya site, Joya de Cerén , in El Salvador . Types of cassava Sweet Cassava: Contains large quantities of cyanide compounds. Bitter Cassava: Contains much higher quantities of cyanide compounds.

7. Previous works on food drying Author Work Area of study Findings Thant et al., 2018 Studied the prediction of moisture ratio of paddy rice using ANN Food drying ANN was found to be a proper method for predicting moisture content of paddy. Beigi et al., 2019 Prediction of moisture content of vaccuum dried celeriac Food drying It was successfully proven that ANN could be used to predict moisture content during the drying

8. MATERIALS AND METHOD Materials and equipment The materials used for this research work were: Fresh cassava Water Oven Weighing balance Foil paper

9. METHOD Sample Preparation The freshly obtained cassava was washed, peeled by hand and sliced using a mechanical slicer into three different sizes of 2mm, 4mm and 6mm.

10. Drying procedure The slices were `dried using an oven (model DHG-9109) at 40 o C, 50 o C, 60 o C and 70 o C. The samples were removed and weighed using an electronic digital weighing balance. Weighing was done intermittently in intervals of 10 minutes until constant weight was achieved.

11. The nntool in the MATLAB software was used for the design and testing of various ANN models. The obtained laboratory data was randomly divided into 3 subsets for training (70%), validation (15%), and testing (15%). The networks performance was evaluated by correlation coefficient ( 𝑅 ) and mean square error (MSE). The best performing ANN configuration was then selected. ANN modeling

12. RESULTS AND DISCUSSION ANN model topology The selected ANN model consisted of 10 neurons in the hidden layer. Hence the network topology used was 3-10-1. 3 input neurons, 10 neurons in the hidden layer and 1 output neuron.

13. Input parameters Time, Temperature, slice thickness Output parameter Moisture Ratio (MR) Partition ratio 70-15-15 Network topology 3-10-1 Training algorithm Levenberg Marquardt Network type Feed forward back propagation Transfer function TANSIG R 0.99462 MSE 0.00089502 ANN configuration summary and performance

14. Assignment of weights and bias The assignment of weights and biases was automatically done by the neural network according to the linearly developed equation below: Where: nth sum of the weighted variables = weight assigned to drying time (t) = weight assigned to temperature (T) = weight assigned to slice thickness( th )  

15. Model transfer function The best performing neural network made use of the TANSIG transfer function in its single hidden layer as shown below  

16. Model equation The final model equation is given by: Where W o,i and b o are the weights and bias at the output layer (o) respectively.  

17. Model performance

18. CONCLUSION The model performed excellently well with an MSE value of 0.00089502 and R value of 0.99462. This demonstrated that ANNs can be used as an accurate and effective tool for predicting the moisture ratio of cassava from slice thickness, temperature and drying time input data.

19. THANK YOU