Optimizing high-temperature and high-humidity dusty waste gas purification in tobacco curing machines using LSTM neural networks

The Flue Gas (FG) from the Drying Machine (DM) has the characteristics of high temperature, high humidity, and dust content. This characteristic leads to a higher concentration of process particulate matter in the directly treated and purified gas. Therefore, a purification method for the high temperature, high humidity, and dust-laden FG from cigarette factories is proposed, which quickly degrades organic matter while reducing pollutants generated during the condensation process. Derived from the working principle of the DM, a Spray Condenser (SC) is used for cooling and dehumidification, and a Long Short-Term Memory Neural Network (LSTM) is applied to forecast the Moisture Content (MC) at the tobacco outlet. Additionally, a Fusion Attention Temporal Convolutional Network (FATCN) is utilized to control the MC at the outlet, maintaining it within the set range, thereby reducing the Volatile Organic Compounds (VOCs) and deposits generated during the condensation process. Furthermore, Activated Carbon Adsorption (ACA) and thermal desorption with sliding arc discharge plasma purification technology are used to purify large particulate matter, some smoke dust, VOCs, and other harmful substances deeply in the FG. Results demonstrate that the proposed intelligent control–assisted purification system reduces exhaust gas temperature to 65–73 °C, humidity to 72–76%, achieves up to 95% particulate removal and 85% VOC removal, providing an efficient and scalable solution for high-temperature, high-humidity industrial waste gas purification.