Time Series Abnormal Vacuum Pattern Detection and Classification in Refrigerator Cycle using Convolutional Autoencoder and Multi-Layer Perceptron

In this paper, a CAE-MLP-based AI model is proposed to detect and classify various abnormal patterns occurring in the vacuum process of refrigerator cycles. Traditional rule-based diagnostic methods have limitations in effectively detecting new abnormal patterns or irregular process delays. To address these challenges, this study employs a Convolutional Autoencoder (CAE) to learn the patterns of normal data and detect abnormal data based on reconstruction error. Additionally, a Multi-Layer Perceptron (MLP) is used to classify the detected abnormal patterns into specific failure types. The proposed CAE-MLP model demonstrated high performance even with imbalanced data. By integrating a rule-based method, the model achieved an F1 Score of 1.0 for anomaly detection and 100% accuracy in classifying failure patterns. This methodology not only enhances the quality control of the refrigerator manufacturing process but also suggests scalability and applicability of anomaly detection and classification systems in vacuum processes across various industries.