Measurement Device of Nondestructive Testing (NDT) of Metanil Yellow Dye Waste Concentration Using Artificial Neural Network Based on Microcontroller

This research applied the nondestructive testing (NDT) method of the emission gas produced by metanil yellow waste, carbon dioxide (CO
₂
), and hydrogen (H
₂
) gases with TiO
₂
as a catalyst material to detect the concentration of metanil yellow waste using artificial neural network based on a microcontroller. This device consists of Arduino UNO, LCD 16×2 I2C, MQ-8 H
₂
gas sensor, MQ-135 CO
₂
gas sensor, BH1750 light sensor, and DHT11 temperature sensor. The concentration of waste is 5–30 mg/L with variations range of 5 mg/L. Meanwhile, the neural network architecture was created using the MATLAB R2018a application. The architecture that owned 25 neurons was chosen due to its performance value with an error of
$3.98 \times {10}^{ - 25}$
and epoch 165. The result of artificial neural network (ANN) training on MATLAB of 4-25-1 architecture contains an error value of 0.024%. The weight and bias values in the training process are programming ANN to Arduino. The testing has an error value of 6.028% and a root-mean-square error of 0.552. Based on the error values, it shows that the training is reaching close to the actual value, it can be concluded that the device can work properly and is practical to use because it does not take long to observe.