Predictive Modelling of Dust Accumulation in Air Conditioners using Support Vector Machines

Recent studies have shown increasing concerns about indoor air quality and air conditioning efficiency, particularly in urban areas. Dust accumulation in air conditioning (AC) units degrades air quality while simultaneously reducing system performance and energy efficiency. This research explores a machine learning approach to detect dust inside AC systems using multi-sensor data. Real-time measurements including temperature, humidity, airflow rate, power consumption, particulate matter (PM2.5), CO2 concentration, and filter usage hours were gathered from a prototype monitoring setup. The study employed a Support Vector Machine (SVM) classifier to differentiate between "Normal" and "Dusty" operational states. Sensor data underwent pre-processing and normalization to enhance model performance. Results indicate strong classification accuracy from the SVM model, demonstrating its effectiveness in identifying dust accumulation through sensor-driven patterns. The findings highlight SVM’s potential application in predictive maintenance for AC units, ultimately contributing to energy optimization and improved indoor air quality. The SVM model achieved a classification accuracy of 91%, effectively distinguishing between dusty and normal conditions. These results validate the model’s applicability in real-world environments for early dust detection and maintenance planning.