Artificial Neural Network Prediction Model for Supercritical Water Gasification Data of Discarded Circuit Boards

Abstract

The gasification reaction is affected by the non-linear effects of temperature, pressure, feedstock concentration and other aspects due to the complexity of supercritical water. The conventional production forecasting methods have difficulties to predict conversion rates and yield of gas accurately. In this paper, an artificial neural network model is established to predict and analyse the conversion rates and the yields of gas products. The gasification experiment is carried out at different conditions. The data are taken as the learning samples, and 60 % of them are used for training and testing. Experimental data at different working conditions are randomly selected and inputted into the gasification artificial neural network model for prediction and analysis. The prediction accuracy of the carbon conversion rate of waste circuit boards is calculated whose coefficient of determination (R²) is 0.98 and root mean square error (RMSE) is 2.22. In addition, R² of hydrogen yield is 0.85, and the deviation RMSE is 1.16. The prediction results of the neural network model constructed in this paper are in good agreement with the experimental data. And the model could realize the accurate prediction of the conversion rates of both carbon and hydrogen, and gas yield of discarded circuit boards. It is found that a double hidden layer neural network model is suitable for the prediction and analysis of the conversion rates while the single hidden layer neural network model suits for the prediction of hydrogen yield. Results at different reaction conditions can be obtained by this model and a lot of resources are saved in industrial design.