A Backpropagation Neural Network Algorithm in an Optical System for Detecting Borax and Formalin Contaminants in Food

Novianas, Hafis
Wirman, Shabri Putra
Fitrya, Neneng

Jurnal Fisika Unand
Vol 14 No 1 (2025)

The Bolin Detector is a device designed to detect borax and formalin contamination based on color differences. However, it has limitations in recognizing data based on contaminant levels. The system relies solely on threshold values for data classification, and several data points from samples exhibit overlapping values, making it difficult to differentiate between them. This research developed an Artificial Neural Network (ANN) to improve the performance of the Bolin Detector. The architecture used is backpropagation, with training methods including traingdx, traincgb, traincgf, and traincgp, as well as variations in the number of hidden layers and neurons. The results show that the ANN can recognize 100% of the training data and 97.83% of the testing data. The best accuracy was achieved using the traincgb method, with 85 neurons in the first hidden layer and 40 neurons in the second hidden layer.