Artificial intelligence in color classification of 3D-printed enhanced adsorbent in textile wastewater

The water quality of the wastewater treatment plant is generally observed by carrying out water quality testing and by visual observation by the wastewater engineer. However, to get the lab testing result is very time consuming and proper treatment cannot be implemented on time to treat the wastewater. To overcome this drawback, deep learning tools using convolutional neural network (CNN) and AlexNet were used with transfer learning to automatically detect, categorize, and quantify the acceptable and unacceptable color value of textile effluent. Both networks demonstrated excellent results in terms of their ability to accurately classify data, as well as their efficient training and testing time. For large data sets, CNN has the highest training accuracy while AlexNet has better performance in handling small data sets. Both CNN and AlexNet model has lower mean squared error (MSE) values of 0.169 and 0.151 with R2 values of 0.903 and 0.934 respectively, indicating the trained model has better performance and predictions closer to the true values. The mean absolute percentage error (MAPE) value for both models are below 10 %, further confirming that the models are highly accurate. This work can open an avenue for the application of deep learning tools for image processing in textile wastewater.