Artificial intelligence-based classification of frying oil quality using gas sensor data

The chemical and physical degradation that frying oils undergo during use poses significant risks to human health and is associated with cardiovascular diseases, obesity, and certain types of cancer. In the literature, the determination of oil quality is mostly carried out through classical sampling methods based on chemical analyses, which are both time-consuming and costly. The increasing need for automation in the food industry indicates that sensor-based monitoring technologies and artificial intelligence–supported data analytics can offer faster, more accurate, and more sustainable solutions in quality control processes. This study aims to classify the quality levels of frying oils by using multidimensional data obtained from the BME688 gas sensor to monitor volatile compounds and degradation indicators formed during the frying process. The sensor data were preprocessed, various artificial neural network architectures were tested, and an accuracy rate of 99.14% was achieved with the model constructed using Heater Profile-411. The findings show that the BME688 sensor can reliably distinguish complex gas compositions and that, when integrated with artificial intelligence, it enables real-time evaluation of frying oil quality. The study aims to provide computer-based sustainable solutions for industrial processes.