Reinforced ANFIS-based Filter Replacement Prediction System with Multi Sensors for VOCs Emission Reduction in Urban Industrial Facilities

With the increase of air pollution, Volatile Organic Compounds (VOCs) emission control requirements and regulations for small air pollutant emitting facilities in urban centers are strengthening. This study proposes a multi sensor IoT network-based intelligent filter replacement prediction system for efficient operation of VOCs emission reduction facilities installed in automobile painting facilities. In the proposed system, several multi-sensor modules consisting of low-cost sensors are installed at the rear end of the adsorption tower of the prevention facility for full-time management of the facility, and to improve the sensor measurement accuracy, the measurement system is trained and optimized with Reinforced Adaptive Neural Fuzzy Inference System (RANFIS) model proposed in this study. This enables real-time monitoring by predicting VOCs emissions. Based on the predicted emissions, Decision Tree (DT) model is applied to predict the breakthrough rate of the filter material, activated carbon, and inform the filter replacement cycle for each facility manager. To verify the proposed system, eight sensor modules consisting of three types of sensors were attached to the exhaust vent of a real automobile paint booth VOCs prevention facility. To verify the accuracy of the sensors, the existing Adaptive Neural Fuzzy Inference System (ANFIS) model was applied for comparative verification. As a result, the RMSE values predicted by the RANFIS model for the three trained sensors are 14.757, 16.117, and 8.918, respectively, which are 73.6%, 82.4%, and 29.7% higher than the existing ANFIS model training. In addition, the DT model was applied based on the RANFIS results to predict the activated carbon replacement cycle, and the prediction accuracy was more than 80% for 80%, 70%, and 60% reduction efficiencies. Therefore, the proposed approach utilizing low-cost multi-sensors can be applied to continuously monitor the prevention facility and provide information on the activated carbon replacement cycle to managers, enabling efficient activated carbon filter management and pollution emission reduction in the prevention facility.