Towards General Industrial Intelligence: A Survey of Continual Large Models in Industrial IoT

TL;DR

This survey articulates how the Industrial Internet of Things (IIoT) can evolve from a data pipeline into an intelligent infrastructure that fully supports continual, large-scale models across four lifecycle stages: data foundation, training, connectivity, and evolution. It categorizes industrial large models into language-, vision-, time-series-, and multimodal-based systems and explains their roles in classification, detection, planning, and industrial Q&A. The paper details data-reservoir strategies (sensing, data freshness, augmentation, generation), training-ground considerations (bandwidth, latency, security, fine-tuning vs prompting), and connectivity mechanisms (modularity, routing, merging, edge–cloud collaboration) that enable emergent intelligence in IIoT contexts. It also discusses continual learning paradigms, traditional CL methods, and novel approaches tailored for large pre-trained models within IIoT, offering theoretical insights and practical guidance for deploying General Industrial Intelligence (GII). The work highlights data generation, efficient training, adaptive routing/merging, and continual evolution as key avenues for advancing industrial AI in real-world, resource-constrained environments, with broad implications for industry 5.0 objectives and human–machine collaboration.

Abstract

Industrial AI is transitioning from traditional deep learning models to large-scale transformer-based architectures, with the Industrial Internet of Things (IIoT) playing a pivotal role. IIoT evolves from a simple data pipeline to an intelligent infrastructure, enabling and enhancing these advanced AI systems. This survey explores the integration of IIoT with large models (LMs) and their potential applications in industrial environments. We focus on four primary types of industrial LMs: language-based, vision-based, time-series, and multimodal models. The lifecycle of LMs is segmented into four critical phases: data foundation, model training, model connectivity, and continuous evolution. First, we analyze how IIoT provides abundant and diverse data resources, supporting the training and fine-tuning of LMs. Second, we discuss how IIoT offers an efficient training infrastructure in low-latency and bandwidth-optimized environments. Third, we highlight the deployment advantages of LMs within IIoT, emphasizing IIoT's role as a connectivity nexus fostering emergent intelligence through modular design, dynamic routing, and model merging to enhance system scalability and adaptability. Finally, we demonstrate how IIoT supports continual learning mechanisms, enabling LMs to adapt to dynamic industrial conditions and ensure long-term effectiveness. This paper underscores IIoT's critical role in the evolution of industrial intelligence with large models, offering a theoretical framework and actionable insights for future research.

Figures (28)

• Figure 1: Layering the general industrial intelligence ecosystem.
• Figure 2: Industrial AI is transitioning from traditional deep learning models to large-scale transformer-based architectures, while the IIoT is also evolving in this process, advancing from a simple data pipeline to an intelligent infrastructure. Our key insight is that IIoT plays a critical role in supporting the full lifecycle of large models, including data foundation, model training, model connectivity, and continuous model evolution.
• Figure 3: A roadmap for general industrial intelligence with four generations of models based on task solving capabilities.
• Figure 4: (A) Actual bearing fault testbed. (B) indicates four common types of bearing faults1, i.e., (a) Rolling body wear, (b) Inner race wear, (c) Outer race crack, (d) Outer race wear.
• Figure 5: EMG/E-skin based human activity recognition and gesture recognition.