The CISDI Hot Rolled Strip Cooling Roller Health Monitoring System (HRSCR-HMS) is a platform widely deployed in hot rolled steel manufacturing, providing real-time monitoring, fault diagnosis, and health management for cooling rollers. Operating in industrial IoT environments with high data velocity and volume, the system faces challenges in scaling real-time fault diagnosis due to evolving fault patterns, severe data imbalance, and the need for both diagnostic accuracy and efficiency. Evaluations of state-of-the-art fault diagnosis (FD) methods, including online continual learning (OCL) algorithms like Camel, reveal their limitations in meeting the real-time adaptability and data processing demands of HRSCR-HMS. To address these challenges, we propose SRTFD, a scalable framework tailored for real-time fault diagnosis in industrial IoT systems. SRTFD processes high-velocity data streams using three core innovations: Retrospect Coreset Selection (RCS) for optimizing training efficiency by reducing redundant data, Global Balance Technique (GBT) for robust performance with imbalanced data streams, and Confidence and Uncertainty-driven Pseudo-label Learning (CUPL) for adaptive updates with unlabeled data. Experiments on industrial datasets demonstrate that SRTFD outperforms competing methods, addressing challenges of imbalance, redundancy, and scalability. Integration within HRSCR-HMS validates SRTFD as a practical, scalable solution aligned with the stringent demands of stream data processing in industrial IoT environments.

Online sentiment analysis has emerged as a crucial component in numerous datadriven applications, including social media monitoring, customer feedback analysis, and online reputation management.Despite their importance, current methodologies falter in effectively managing the continuously evolving nature of data streams, largely due to their reliance on substantial, pre-existing labelled datasets.This paper presents SentiStream, a novel co-training framework specifically designed for efficient sentiment analysis within dynamic data streams.Comprising unsupervised, semi-supervised, and stream merge modules, SentiStream guarantees constant adaptability to evolving data landscapes.This research delves into the continuous adaptation of language models for online sentiment analysis, focusing on real-world applications.Experimental evaluations using data streams derived from five benchmark sentiment analysis datasets confirm that our proposed methodology surpasses existing approaches in terms of both accuracy and computational efficiency 1 .