A neural network based model to analyze rice parboiling process with small dataset.

Area of Science:

• Agricultural Engineering
• Food Science and Technology

Background:

• Optimizing parboiling processes is crucial for rice quality.
• Developing accurate predictive models for milling quality with limited data presents challenges.

Purpose of the Study:

• To develop and validate models for predicting the milling quality of Tarom parboiled rice.
• To compare multivariate regression and artificial neural networks (ANNs) for parboiling process modeling.

Main Methods:

• Parboiled rice samples were prepared varying soaking temperatures, steaming times, and final moisture contents.
• Milling quality characteristics (milling recovery, head rice yield, degree of milling, whiteness) were measured.
• Multivariate regression and ANNs, with K-fold cross-validation, were employed for modeling.

Main Results:

• The artificial neural network model demonstrated higher accuracy in modeling the parboiling process compared to multivariate regression.
• An ANN with one hidden layer, Tansig transfer function, and 18 hidden neurons was identified as the optimal model.
• The developed ANN model proved reliable for selecting optimal parboiling parameters using small experimental datasets.

Conclusions:

• Artificial neural networks provide a promising and accurate method for modeling complex parboiling processes.
• ANNs can effectively predict rice milling quality, enabling efficient optimization of processing parameters.
• This approach facilitates reliable parboiling process control even with limited experimental data.