Machine Learning tool to prevent and control Bag Filter clogging

Filter clogging represents a crucial issue in industrial filtration systems, and it can lead to various problems, such as reduced filtration efficiency, increased maintenance costs, and potential production disruptions. This paper presents a Machine Learning (ML) integrated solution, developed using Python, within a digital environment system created using LabVIEW to prevent and control filter clogging. The primary objective is to optimize the monitoring, control, and maintenance of a Bag Filter pilot plant. The digital environment incorporates a data comparison tool that trigger a compressor for filter cleaning. A Machine Learning algorithm has been integrated into the digital environment, where a comparison within real data acquired and the predicted one were performed to validate the selected algorithms and evaluate the Data Comparison tool. This initial phase's outcomes were used to develop a control system using LabVIEW software. The software compares estimated pressure drop values, reflecting optimal working conditions, with directly recorded pressure drop values. If the difference exceeds 10% of the optimal pressure drop, the controller activates a compressed air cleaning system, removing excess dust from the filter sleeves using an air jet. Additionally, when the estimated and acquired values diverge consistently over a 60-minute period, a user warning is triggered for filter sleeve replacement. Several tests were conducted at different air inlet velocities to assess the error between acquired and predicted data. The results indicated that the ML algorithm's error distribution remained below 5% for approximately 80% of the cases.