Deteksi Adulterasi Daging Babi pada Kornet Sapi Menggunakan Electronic Nose Berbasis Sensor MQ2 dan MQ9

Abstract

Cases of pork adulteration in processed meat products are increasing, particularly in beef corned, which is difficult to identify visually and may violate halal requirements. This study aimed to analyze the digital signal pattern of an electronic nose (E-nose) based on MQ2 and MQ9 sensors and to evaluate the accuracy of detecting pork adulteration in beef corned using supervised learning methods. The methodology includes preparing beef corned samples with varying compositions of beef and pork (0–50%), acquiring aroma data using the E-nose through flushing, sensing, and cleaning stages, and analyzing the data using statistical parameters (mean, standard deviation, maximum, minimum, skewness, and kurtosis). The data were then classified using SVM, CNN, and BiLSTM-CNN models and evaluated using a confusion matrix and training–validation loss curves. The results show that the sensing stage produces the most significant sensor response in distinguishing sample characteristics. The combination of mean– standard deviation–maximum–minimum parameters provides the best classification performance. The BiLSTM-CNN model achieved the highest accuracy (95–100%), followed by CNN, while SVM showed lower performance at low adulteration concentrations. Evaluation metrics indicate accuracy, precision, recall, and F1-score of 1.0, with no indication of overfitting. In conclusion, the E- nose system, based on MQ2 and MQ9 sensors combined with supervised learning, is effective and accurate in detecting pork adulteration in beef corned, particularly at higher levels of adulteration. Keywords: corned, electronic nose, pork adulteration, supervised learning