Anomaly detection solutions: The dynamic loss approach in VAE for manufacturing and IoT environment

Anomaly detection is critical for enhancing operational efficiency, safety, and maintenance in industrial applications, particularly in the era of Industry 4.0 and IoT. While traditional anomaly detection approaches face limitations such as scalability issues, high false alarm rates, and reliance on skilled expertise, this study proposes a novel approach using a BiLSTM-Variational Autoencoder (BiLSTM-VAE) model with a dynamic loss function. The proposed model addresses key challenges, including data imbalance, interpretability issues, and computational complexity. By leveraging the bidirectional capability of BiLSTM in the encoder and decoder, the model captures comprehensive temporal dependencies, enabling more effective anomaly detection. The innovative dynamic loss function integrates a tempering index mechanism with tuneable parameters ( α and γ ), which assigns higher weights to underrepresented classes and down-weights easily classified samples. This improves reconstruction and enhances detection accuracy, particularly for minority class anomalies. Experimental evaluations on the SKAB and TEP datasets demonstrate the superiority of the proposed framework. The model achieved an accuracy of 98% and an F1 score of 96% for binary classification on the SKAB dataset and a multiclass classification accuracy of 92% with an F1 score of 85% on the TEP dataset. These results significantly outperform state-of-the-art models, including traditional VAE, LSTM, and transformer-based approaches. The proposed BiLSTM-VAE model not only demonstrates robust anomaly detection capabilities across diverse datasets but also effectively handles data imbalance and reduces false positives, making it a scalable and reliable solution for industrial anomaly detection in the context of Industry 4.0 and IoT environments. • Novel BiLSTM-VAE: Integrates BiLSTM and VAE to capture temporal dependencies in time-series data. • Dynamic Loss Function: Uses a tempering index with α and γ to address data imbalance. • Performance: Achieved an accuracy of 98% and 96% F1 on SKAB; 92% accuracy and an F1 score of 85% on TEP, surpassing traditional and state-of-the-art models. • Comparative: Outperforms established models such as LSTM, Autoencoders, and Transformer-based methods in terms of accuracy and robustness in handling imbalanced datasets. • Scalability: IoT-ready model for predictive maintenance; validated on SKAB and TEP.