Source-linked AI summary
Industrial Internet of Things Intelligence Empowering Smart Manufacturing: A Literature Review
Yujiao Hu, Qingmin Jia, Yuao Yao, Yong Lee, Mengjie Lee, Chenyi Wang, Xiaomao Zhou, Renchao Xie, F. Richard Yu
TL;DR
Manufacturing transformation creates a need for software and coordinated IIoT intelligence beyond existing equipment and networking foundations. The paper synthesizes academic, government, industry, and lighthouse-factory evidence, proposes a five-layer architecture, and reviews technologies, impacts, ethics, and open challenges. Reported factory cases illustrate practical gains, including 10%-20% higher on-time-in-full performance from AI scheduling and a 52% assembly-efficiency improvement from collaborative robots.
Problem
Manufacturing has equipment and networking foundations but lacks software for managing industrial data and production processes.
Method
The paper provides a comprehensive overview using academic, government, industry, and lighthouse-factory sources, alongside Chinese enterprise practices and a five-layer IIoT intelligence architecture.
Results
The review identifies seven enabling technologies, analyzes ethical and environmental impacts, and reports lighthouse-factory gains including 10%-20% higher on-time-in-full performance and 52% higher assembly efficiency.
Takeaways & Limitations
IIoT intelligence is organized across equipment, networking, software, modeling, and analysis and optimization layers, with future work targeting digital twins, deterministic response, resource deployment, and industrial generative AI.
Abstract
from arXiv · showhide
The fiercely competitive business environment and increasingly personalized customization needs are driving the digital transformation and upgrading of the manufacturing industry. IIoT intelligence, which can provide innovative and efficient solutions for various aspects of the manufacturing value chain, illuminates the path of transformation for the manufacturing industry. It's time to provide a systematic vision of IIoT intelligence. However, existing surveys often focus on specific areas of IIoT intelligence, leading researchers and readers to have biases in their understanding of IIoT intelligence, that is, believing that research in one direction is the most important for the development of IIoT intelligence, while ignoring contributions from other directions. Therefore, this paper provides a comprehensive overview of IIoT intelligence. We first conduct an in-depth analysis of the inevitability of manufacturing transformation and study the successful experiences from the practices of Chinese enterprises. Then we give our definition of IIoT intelligence and demonstrate the value of IIoT intelligence for industries in fucntions, operations, deployments, and application. Afterwards, we propose a hierarchical development architecture for IIoT intelligence, which consists of five layers. The practical values of technical upgrades at each layer are illustrated by a close look on lighthouse factories. Following that, we identify seven kinds of technologies that accelerate the transformation of manufacturing, and clarify their contributions. The ethical implications and environmental impacts of adopting IIoT intelligence in manufacturing are analyzed as well. Finally, we explore the open challenges and development trends from four aspects to inspire future researches.
I. INTRODUCTION
The paper argues that fragmented surveys have obscured a system-level understanding of IIoT intelligence. It therefore synthesizes transformation pressures, industrial practice, a five-layer architecture, enabling technologies, and future challenges.
- Manufacturing transformation is driven by stagnant productivity, personalized customization, delayed innovation, shrinking humanpower, and demand for comfortable one-stop service.
- Existing surveys focus on specific IIoT domains, limiting clarification of how advanced technologies contribute across an overall intelligence architecture.
- The review broadens its evidence base beyond academic papers to government documents, industry reports, lighthouse factories, and official customer stories.
- IIoT intelligence is defined as value-chain techniques, methods, products, and platforms that build autonomous capabilities across manufacturing activities.
- The proposed architecture has five layers: equipment, networking, software, modeling, and analysis and optimization, whose interactions enhance manufacturing intelligence and digitization.
- Seven technology categories accelerate IIoT intelligence: industrial robots, machine vision, networking, digital twins, deep learning, smart hardware, and cloud/edge computing.
- Future challenges and trends are organized around digital control, deterministic response, cost-friendly operation and deployment, and IIoT intelligence proliferation.
II. IN-DEPTH PERSPECTIVES ON SMART MANUFACTURING
Smart manufacturing transformation responds to labor, customization, productivity, and competitive pressures, while advanced technologies and policy support provide enabling conditions. Chinese enterprise practice emphasizes production, customer, innovation, digital, and organizational capabilities.
- A. Background: The paper frames intelligent transformation as necessary because manufacturing faces social pressures and technical developments that challenge existing production systems.
- A. Background: 2014–2019 Chinese manufacturing saw falling workforce levels alongside significantly rising workforce costs.
- A. Background: Cloud computing, 5G, industrial data analysis, and IIoT support manufacturing upgrading, while governments promote transformation through policy initiatives.
- A. Background: Personalized customization and patient experience increase manufacturing-system complexity across the end-to-end product life cycle.
- B. Secrets to smart manufacturing: Foxconn’s transformation approach emphasizes effective production architecture, customer-centric value chains, innovation business models, digital capabilities, and proactive technical and organizational capabilities.
A. Understanding IIoT intelligence
IIoT intelligence spans the manufacturing value chain, combining digital connection and perception, intelligent analysis and cognition, and real-time decision-making. Its iterative capabilities support industrial applications and transformation across product development, quality control, and other value-chain activities.
- IIoT intelligence comprises techniques, methods, products, and platforms deployed throughout the value chain for digital connection, perception, intelligent analysis, cognition, and real-time decision-making.
- Its value-chain scope includes R&D, production, operation and maintenance, marketing, management, and services, where intelligent technologies combine with industrial scenarios, mechanisms, and knowledge.
- IIoT intelligence iteratively collects value-chain data, learns knowledge, verifies optimized models and algorithms, scales solutions, and begins a new improvement cycle with newly generated data.
- Applications: IIoT intelligence supports generative design and collaborative innovation, which can speed product design and reduce design costs.
- Applications: Automated visual inspection uses cameras, sensors, and AI-based machine vision to identify components, detect defects, assess assembly quality, and prompt corrective actions.
- Applications: Digitizing physical experiments enables lower-cost digital trial and error, fault prediction, and improvement recommendations during product R&D.
3) Organizational behaviour impacts – increase labour efficiency through remote monitoring and control:
IIoT intelligence shifts manufacturing toward connected, remotely monitored, and analytically optimized operations. Its five-layer architecture links equipment, networks, software, modeling, and optimization through bidirectional information flows, supporting labor efficiency and operational improvements.
- Organizational behaviour impacts: Remote monitoring and control connect equipment, workers, and intermediate products, improving safety through unmanned operations and reducing exposure to hazardous sites.
- Organizational behaviour impacts: The shift from one-man-one-control to one-man-multicontrol improves Overall Equipment Effectiveness (OEE) and labor efficiency.
- Operation: Cloud manufacturing creates intelligent factory networks by encapsulating distributed manufacturing resources as cloud services and managing them centrally to encourage collaboration.
- Deployment: ACG Capsules uses color-matching AI and a digital twin to optimize production scheduling, realizing a 10%-20% on-time-in-full increase.
- Application scenarios: Johnson & Johnson Consumer Health in India uses IIoT intelligence for predictive maintenance, improving asset reliability and reducing unplanned machine downtime by 50%.
- Architecture: The hierarchical architecture extends the classic equipment-and-network structure with software, modeling, and analysis-and-optimization layers to accommodate newer forms of intelligence.
- Analysis and optimization: Advanced analytics and machine learning process sensor data to forecast maintenance needs, optimize production schedules, and improve production decision-making.
A. Equipment layer
The equipment layer establishes the physical and computational foundation for IIoT intelligence. Industrial robots and smart sensors address automation, perception, safety, labor pressures, and the computing resources required for industrial data.
- Labor shortages, high labor costs, and dangerous, dirty, dull, or delicate tasks motivate the adoption of industrial robots in manufacturing.
- Smart sensors improve perception of production environments and enable high-precision process control, while digital transformation creates demand for extensive storage and analysis resources.
- Equipment-layer resources include production equipment—industrial robots and sensors—and computation devices providing cloud, edge, or fog computing and storage.
- Industrial sensors convert perceived physical quantities into electrical signals and support manufacturing processes through modalities including air quality, distance, image, motion, pressure, and speed sensing.
- Industrial robots are automated, programmable systems capable of movement on three or more axes and assist tasks such as welding, assembly, packaging, inspection, and testing.
- Cloud computing offers Internet-delivered servers, storage, databases, and software as an economical and flexible way to deploy enterprise computation resources.
2) Connotation:
The IIoT intelligence connotation is organized around networking, industrial software, and modeling methods that connect physical and cyber spaces and support optimization. These layers address information isolation, data management, process representation, and industrial decision support.
- Networking: The networking layer includes methods connecting human users, sensors, robots, cloud, edge, fog, and other smart devices.
- Networking: Advanced networks such as SDN, deterministic networks, time-sensitive networks, 5G, and network slicing pursue bandwidth, speed, steady transmission, and massive connectivity.
- Industrial software: Human, sensor, and robot data continuously moves through IIoT networks, but the architecture requires industrial software to manage data and production processes.
- Industrial software: Industrial software integrates industrial knowledge with information technologies and comprises R&D, production control, information management, and embedded software.
- Industrial software: Industrial software is evolving toward systematic platforms, cloud-based deployment, and lightweight development as industry environments become more competitive and complex.
- Modeling: The modeling layer represents physical processes digitally and transfers decisions from digital space to the physical world for applications including remote control, predictive maintenance, and process optimization.
- Modeling: Digital twins connect cyber and physical spaces through data, connection, model, interaction, and application, supporting description, analysis, diagnosis, prediction, prescription, and cognition across process lifecycles.
E. Analysis and optimization layer
The analysis and optimization layer applies intelligent algorithms to industrial problems that remain unresolved by cyber-physical connectivity alone, including planning, scheduling, topology optimization, and prognostics. The paper illustrates IIoT intelligence through lighthouse-factory transformations and reported operational, financial, and environmental returns.
- E. Analysis and optimization layer: The analysis and optimization layer addresses manufacturing process planning, scheduling, topology optimization, and industrial prognostics.These problems require algorithms beyond the control channel created by the modeling layer.
- Manufacturing process planning and scheduling: Manufacturing process planning selects and sequences operations, while scheduling assigns resources over time against criteria such as delay, throughput, or cost.The combined problem is typically NP-hard, with computational complexity increasing substantially as scale grows.
- Topology optimization: Topology optimization places material within a prescribed design domain to obtain the best structural performance through automated conceptual design.It occurs at the initial stage of product manufacturing and replaces conventional trial-and-error design.
- Industrial prognostics: Industrial prognostics estimates and anticipates events involving industrial assets and production processes through data-based workflows.The surveyed approaches include descriptive prognostics and physics-based, data-driven, and hybrid methods.
- Lighthouse-factory evidence: By 2023, the Global Lighthouse Network included 54 end-to-end lighthouses, including 17 sustainability lighthouses.The paper uses lighthouse factories and cross-company collaborations to examine IIoT intelligence in manufacturing transformation.
VI. TECHNOLOGY STUDIES ON IIOT INTELLIGENCE
The paper surveys technologies that support IIoT intelligence through intelligent equipment, machine vision, industrial networking, and related digital capabilities. Examples from manufacturing applications show improvements in assembly efficiency, delivery timeliness, safety, and network reliability.
- Industrial robots: Industrial robots act as programmable, automatically controlled, multipurpose equipment and have played a significant positive role in industrial processes.The paper notes substantial remaining room for further development.
- Industrial robots: Collaborative robots improved assembly efficiency by 52% at Haier’s Qingdao refrigerator factory, while robotics increased on-time delivery by 11% at Sany Heavy Industry.The examples cover flexible manufacturing and logistics execution, respectively.
- Machine vision systems: Machine vision reconstructs understandable models of the real world from sensed images and supports industrial tasks requiring precision, speed, robustness, and usability.Its online visual feedback can improve machining, welding, injection molding, and other processes.
- Machine vision systems: Unilever’s machine-vision supervision platform decreased unsafe behaviours by 78%, illustrating a safety application of online visual feedback.Machine vision is also used for product measurement, robotic guidance, and quality improvement.
- Networking: 5G connects industrial machines, vehicles, robots, and software through ultra-reliable low-latency communication for real-time control and management.Network slicing supports massive connectivity and service isolation in industrial applications.
D. Digital twin connects cyber-physical spaces
Digital twins connect cyber and physical spaces by modeling manufacturing systems and supporting interoperability, while deep learning supplies data-driven intelligence for industrial tasks. The section also highlights unresolved interoperability challenges, annotation costs, and difficulties in representing complex environments.
- Digital twin: Digital-twin research in manufacturing focuses on accurate modeling and interoperability across structure, attributes, behavior, processes, and heterogeneous data.Modeling approaches include data-driven, knowledge-based, and hybrid methods.
- Digital twin: Interoperability enables mutual execution between physical and digital spaces, but current research remains at an early stage.Open issues include mutual intelligibility, dynamic updating, data heterogeneity, and submodel heterogeneity.
- Digital-twin applications: Digital twins improved on-time delivery in full by 13% at ACG Capsules and produced a 159% profitability enhancement at Aramco.The examples apply digital twins to production planning and scheduling and to energy-consumption reduction, respectively.
- Deep learning: Deep learning automatically extracts multilevel data representations through multilayer computational models and backpropagation without relying on prior data processing.The paper links these methods to scheduling, planning, computer vision, and other IIoT applications.
- Deep learning: Supervised machine vision improves detected-edge quality compared with traditional detectors, whereas unsupervised approaches address industrial settings where annotation is expensive.Unsupervised defect detection reconstructs defect-free references and uses residuals to identify defects.
- Deep reinforcement learning: Deep reinforcement learning remains challenged by complex visual observations and dynamics, with reported performance below optimal state-representation performance.Industrial applications are nevertheless expanding as more cases are reported.
- Deep learning applications: Ingrasys’s AI demand-forecasting model delivered 27% more accuracy, illustrating deep learning’s reported application across manufacturing processes.The paper also describes automatic optimization of performance and drawing parameters using past strategies.
F. Smart hardware builds intelligence foundation
Smart hardware supplies sensing and computing capabilities for IIoT intelligence, while cloud and edge deployment reshape manufacturing resources, business models, and operational responsiveness. Cloud approaches emphasize sharing and remote access; edge approaches emphasize low latency, bandwidth efficiency, and autonomous operation.
- Smart hardware: Smart hardware advances through smart sensors and increasingly powerful computing devices, strengthening the foundation for IIoT intelligence.The paper identifies breakthroughs in sensor technology and computing capability.
- Cloud manufacturing: Cloud manufacturing provides manufacturing resources, information, and capabilities as internet services in a flexible, scalable, and collaborative environment.Shared equipment, software tools, and computational power can improve resource utilization and production decisions.
- Cloud-based deployment: Cloud-based deployment lets managers and clients access applications and data remotely instead of relying on on-premises hardware and infrastructure.Its pay-as-you-go model reduces upfront physical-infrastructure investment, particularly for enterprises lacking capital for extensive IT infrastructure.
- Edge-based deployment: Edge-based deployment moves processing toward local devices or edge servers when real-time processing, low latency, and reliable connectivity are important.It reduces raw-data transmission and can continue critical functions without continuous cloud connectivity.
- Industrial applications: Cloud-company collaborations apply IIoT intelligence to research and development, supply chains, manufacturing scale, and workforce tools.The paper uses STMicroelectronics and other industrial collaborations as practical examples of deployment optimization and business novelty.
VII. OPEN CHALLENGES AND FUTURE TRENDS
The paper frames future manufacturing around four capabilities—digital control, deterministic response, cost-friendly operation and deployment, and IIoT intelligence proliferation—and identifies unresolved technical challenges for achieving them.
- VII. OPEN CHALLENGES AND FUTURE TRENDS: Future manufacturing requires digital control, deterministic response, cost-friendly operation and deployment, and IIoT intelligence proliferation.These keywords answer the paper’s question about capabilities required by future manufacturing.
- A. Digital control: Current IIoT architectures support digital control through sensors, data analysis, PLCs, DCS, and remote monitoring, but remain short of real-time visibility and immediate command execution.The desired capability requires interaction across IIoT layers and cooperation among robots, sensors, computation, and network resources.
- A. Digital control: Digital twins could connect cyberspace and physical space, but require software-model interoperability, standardized industrial communication protocols, high-quality data, and continuously updated datasets.Perfect datasets covering every modeled-object condition are impractical, making update mechanisms necessary for sustained performance.
- B. Deterministic response: TSN provides deterministic network traffic, yet best-effort end-edge-cloud networks leave the response times of time-critical IIoT computing tasks uncertain.Open questions concern scheduling computation and network resources and sharing networks between communication traffic and computing tasks.
- B. Deterministic response: Researchers have begun exploring computing-and-network convergence paradigms, but deterministic responses for scalable IIoT computing tasks remain insufficiently studied.The paper identifies this issue as a future research direction.
C. Cost-friendly operation and deployment
The paper treats cost-friendly operation and deployment as complementary manufacturing priorities: lowering operating costs through resource strategies while meeting time requirements with economical computation and storage deployment.
- C. Cost-friendly operation and deployment: Cost-friendly operation optimizes costs, improves efficiency, and supports financial sustainability, while cloud manufacturing enables on-demand manufacturing.The paper argues that resource pricing and supply strategies deserve more attention alongside platform-side scheduling.
- C. Cost-friendly operation and deployment: Cloud- and edge-based deployment can reduce costs for industrial data storage and analysis, but best-effort network latency may prevent time-critical tasks from meeting deadlines.The paper therefore identifies cost-friendly deployment as an open challenge rather than a solved capability.
- C. Cost-friendly operation and deployment: Enterprises without intelligent-transformation foundations need cost-effective deployment solutions that account for personalized computing-server and network-device costs.The paper separately asks how to optimize existing deployments for enterprises with established transformation foundations.
- C. Cost-friendly operation and deployment: Enterprises with established transformation foundations need deployment structures for computing servers and network devices optimized around personalized transformation needs.This question complements the challenge of designing initial deployments for less-prepared enterprises.
1) Job displacement and workforce transition:
IIoT intelligence has both beneficial and adverse manufacturing impacts: it can improve resource efficiency and reduce waste, while automation, infrastructure growth, and unequal access create social and environmental concerns.
- 1) Job displacement and workforce transition: IIoT-enabled automation may displace workers performing some manual tasks, creating concerns about worker well-being and workforce transition.The paper recommends communication, early timeline disclosure, and diversified skills training during transformation.
- 2) Privacy and data security: IIoT intelligence raises privacy concerns because manufacturing companies collect and analyze extensive data about employees and stakeholders.The proposed response is transparency about data collection, processing, and usage, supported by clearly communicated privacy policies.
- 3) Social and economic equality: Large corporations may have greater ability to adopt IIoT solutions, potentially disadvantaging smaller manufacturers and less developed regions in productivity, competitiveness, and innovation.The paper expects providers to offer flexible and scalable solutions tailored to different users.
- Environmental impacts: IIoT intelligence can improve energy, water, and raw-material monitoring, reduce emissions, and lower waste from defects, production errors, and scrap.These benefits are associated with process optimization, predictive maintenance, and quality control.
- Environmental impacts: More IIoT devices, sensors, data centers, networks, and industrial large-model applications may increase electronic waste, energy consumption, raw-material extraction, and manufacturing impacts.The paper suggests durable, repairable, recyclable devices and more sustainable supply chains as mitigation strategies.
F. Richard Yu
Richard Yu is a professor at Carleton University whose research spans wireless cyber–physical systems, connected and autonomous vehicles, security, distributed ledger technology, and deep learning.
- F. Richard Yu: His research interests include wireless cyber–physical systems, connected and autonomous vehicles, security, distributed ledger technology, and deep learning.