Document Type : Research Paper

Authors

• Ghasem Zolfaghari ¹
• Sara Nezamparvar ¹
• Mahmood Amintoosi ²

¹ Department of Environmental Sciences and Engineering, Faculty of Environmental Sciences, Hakim Sabzevari University, Sabzevar, Iran.

² Department of Computer Sciences, Faculty of Mathematics, Hakim Sabzevari University, Sabzevar, Iran.

Abstract

The aim of this research is to evaluate the performance of the electrofilter at the Zaveh Cement plant and to provide an optimized model for predicting dust using MLP and RBF neural networks. Using 5048 data points, the impact of parameters such as pressure, temperature, and voltage was examined in MATLAB software. The MLP network was designed with four layers, including an input layer, a first hidden layer, a second hidden layer, and an output layer, and the output error from the execution (three times) was calculated. In the MLP, the Levenberg-Marquardt learning algorithm and the Hyperbolic tangent transfer function were employed with 1000 epoch. The MLP layer multiplier was selected for the RBF. To assess the performance of the electrofilter, humidity, speed, pressure, flow rate, and gas temperature were measured under isokinetic conditions using the KIMO HD device, and dust was measured with the Westech device. The MSE for MLP was 1.36 for training data and 2.78 for test data, while for RBF it was 1.39 for training data and 3.15 for test data (R²= 0.78 for MLP and 0.68 for RBF). The performance of the existing electrofilter did not decline significantly relative to the lifespan of the filters. The findings of this study suggest that if the neural network is properly trained, it can be an accurate and fast method for solving complex and time-consuming problems. It is recommended to investigate the effects of speed, humidity, and dew point on the output dust using the network.

Keywords

• Dust
• Electrofilter
• Sensitivity analysis
• MLP and RBF neural networks