Faults Classification in Feedback Control Loops Using Bidirectional Encoder Representations From Transformers Large

In recent times, the efficiency and reliability for feedback control loops rely heavily on the prompt classification and detection of faults, which is crucial for optimistic performance of system. However, the existing approaches faced challenges in feedback control loops that include imbalanced data and handling of noisy data which creates ambiguity in fault patterns. Therefore, this research proposes Bidirectional Encoder Representations from Transformers Large (BERTL) for fault classification in feedback control loops. Firstly, the data is collected and processed for initialization of fault tolerance detection with the patterns exhibited. Next step involves the preprocessing methods utilization which includes galvanic isolation, and linearization that helps for enhancing the of fault classification assuring safety for accurate detection of faults. Followed by feature extraction with BERTL which contains 24 layers and 349 million parameters for effective classification of faults. Next step involves classification which is carried out by a fully connected dense layer, (Rectified Linear Unit) ReLu Activation Function and dropout layers for scaling better performance of model. From the results, the proposed model performed better than existing BERT for fault classification with 98.95% accuracy.