Applied machine vision and artificial neural network for modeling and controlling of the grape drying process

• Applied machine vision (MV) for measuring of shrinkage accrued during grape drying process.
• Applied artificial neural network (ANN) for predictive modeling of grape drying process.
• Coupling of ANN and MV for online predicting of moisture content and control of grape drying process.

This paper presents a new method for predictive modeling of grape drying process for on-line monitoring and controlling of this process. The shrinkage during drying plays an important role in determining the accuracy of the drying model. Machine vision (MV) was used to measure grapes shrinkage during drying process to produce raisins. An artificial neural network (ANN) was developed to predictive model of the grape drying in a hot air dryer. ANN inputs were air drying temperature, velocity, shrinkage and moisture content at time and output was moisture content at time t + Δt. The results showed that the ANN had better performance than MLR. The best ANN was obtained by three layers (4 inputs, 5 nodes in hidden layer and 1 output) with 0.00004 MSE and 0.99947 R2 for training and 0.00003 MSE and 0.99952 R2 for testing data. This ANN model could predict the moisture content of grapes at time t + Δt by knowing the input data at time t. Also, this ANN model and MV were coupled for on-line control of the grape drying process.