PaperlayerView exact revision on arXiv

Source-linked AI summary

Data Management in Industry 4.0: State of the Art and Open Challenges

Theofanis P. Raptis, Andrea Passarella, Marco Conti

arXiv:1902.06141v2cs.NI

TL;DR

Networked Industry 4.0 environments generate diverse, large-scale industrial data, while existing surveys do not holistically cover its management across heterogeneous deployments. This article synthesizes practical use cases and multidisciplinary literature into an architectural analysis and taxonomy spanning field deployments to cloud services, and identifies open research challenges.

  • Problem

    Existing surveys do not holistically cover diverse data-management aspects across heterogeneous networked industrial deployments.

  • Method

    The article analyzes practical I4.0 use cases, architectural designs, and multidisciplinary literature to classify data properties, management philosophies, technologies, and services.

  • Results

    The survey produces a taxonomy of I4.0 data-enabling technologies and data-centric services spanning field-level deployments through cloud-level systems, and identifies open research challenges.

  • Takeaways & Limitations

    The synthesis supports understanding current data management in networked industrial environments and selecting future research opportunities.

Abstract

from arXiv · show

Information and communication technologies are permeating all aspects of industrial and manufacturing systems, expediting the generation of large volumes of industrial data. This article surveys the recent literature on data management as it applies to networked industrial environments and identifies several open research challenges for the future. As a first step, we extract important data properties (volume, variety, traffic, criticality) and identify the corresponding data enabling technologies of diverse fundamental industrial use cases, based on practical applications. Secondly, we provide a detailed outline of recent industrial architectural designs with respect to their data management philosophy (data presence, data coordination, data computation) and the extent of their distributiveness. Then, we conduct a holistic survey of the recent literature from which we derive a taxonomy of the latest advances on industrial data enabling technologies and data centric services, spanning all the way from the field level deep in the physical deployments, up to the cloud and applications level. Finally, motivated by the rich conclusions of this critical analysis, we identify interesting open challenges for future research. The concepts presented in this article thematically cover the largest part of the industrial automation pyramid layers. Our approach is multidisciplinary, as the selected publications were drawn from two fields; the communications, networking and computation field as well as the industrial, manufacturing and automation field. The article can help the readers to deeply understand how data management is currently applied in networked industrial environments, and select interesting open research opportunities to pursue.

I. INTRODUCTION

Industry 4.0 is transforming manufacturing through interconnected technologies, services, and data-intensive networked environments. This survey organizes the resulting data-management landscape across use cases, architectures, technologies, services, and future challenges.

  • Industry 4.0 combines interoperability, virtualization, decentralization, distributed control, and communication to support smart factories and networked industrial environments.
  • Emerging enablers include reconfigurable assembly lines, industrial Internet of Things, industrial cyber-physical systems, and data-centric services such as analytics and machine learning.These services address industrial data volumes, traffic, mappings, and conversions across formats.
  • The survey extracts data properties from practical I4.0 use cases and identifies the technological enablers needed to realize them.The extracted properties guide the subsequent analysis.
  • The literature review covers communications, networking, computation, industrial, manufacturing, and automation research, emphasizing work directly applied to industrial environments.Related work focused on non-industrial environments was purposefully excluded.
  • The article presents a holistic taxonomy spanning field-level deployments to cloud-level technologies and services, then identifies open research challenges for networked industrial data management.It claims that no previous work provided this practical, broad survey of data properties, management, technologies, and services.
  • It reviews recent architectural designs according to data presence, coordination, computation, and the extent of distributiveness.

II. COMPARISON WITH EXISTING RELATED SURVEY ARTICLES

The survey positions itself as a holistic review of industrial data management, bridging data management and industrial networking while distinguishing its scope from focused prior surveys.

  • Survey scope: The article surveys industrial data management across networked industrial environments rather than focusing on a single technology, service, or application area.It emphasizes the intersection of communications, networking, computation, industrial, manufacturing, and automation research.
  • Survey contribution: The paper claims to systematically extract, dissect, categorize, and combine data-management topics with industrial networking in one survey.Its stated contribution is to bridge two complementary fields that are rarely treated together holistically.
  • Prior survey coverage: Existing surveys address narrower areas, including petrochemical IIoT, data-driven manufacturing, control, cloud manufacturing, virtualization, and big data analytics.These studies contribute focused coverage of particular technologies, services, application domains, or control problems.

E. IIoT technologies

The surveyed IIoT-related literature spans architectures, enabling technologies, applications, challenges, wireless networking, scheduling, control, and product-service systems. The paper situates these topics within a broader data-management review grounded in use cases and extracted data properties.

  • IIoT technologies: IIoT surveys cover industrial networking, intelligent sensing, cloud computing, big data, smart control, security management, applications, and challenges.The reviewed scope extends across multiple Industrial Internet layers and application domains.
  • Industrial wireless technologies: Related work also addresses cognitive radio, industrial wireless sensor-network quality of service, spectrum access, interference management, sensing, handoff, and network security.These surveys concern communication reliability and real-time operation in industrial wireless environments.
  • Applications and control: Higher-level studies examine multi-factory scheduling, proximity between suppliers and construction sites, and networked-control limitations such as packet dropouts, delays, and quantization.The reviewed applications connect industrial coordination with scheduling, synchronization, and control constraints.
  • Product-service systems: Product-service-system surveys cover requirements management, product defects, and product modeling for collaborative production and consumption.Product-service systems are treated as an application near the top of the Industry 4.0 automation pyramid.
  • Data-management perspective: The paper complements these focused studies with practical extraction of data properties and technological enablers from fundamental Industry 4.0 use cases.The extracted properties guide the subsequent analysis of industrial data-management technologies.

A. Use cases necessitating high data efficiency

High-data-efficiency use cases combine demanding data requirements with industrial monitoring, control, identification, and coordination needs. Their challenges arise from dense sensing, distributed processing, timing constraints, traffic, and criticality.

  • Petrochemical plants: Petrochemical plants use dense wireless sensing, RFID, and IIoT integration for continuous monitoring, diagnosis, maintenance prediction, and failure prediction.Thousands of sensors can generate increased wireless traffic while supplying small-volume measurements with varied readings.
  • Automotive assembly: Automotive assembly requires end-to-end timing validation because distributed electronic control units implement critical assembly functions.Data is mainly distributed through wired deterministic networks, allowing traffic regulation in an offline centralized manner.
  • Shipbuilding: Shipbuilding applies data processing to fault detection and diagnosis across complex construction processes and vessel operation.The use case involves one-off manufacturing, complex planning, backorders, and overloaded capacity between consecutive processes.
  • Asset monitoring: Mass-production monitoring combines real-time asset location with contextual information such as machine power usage and vibration.Associating location with context can identify components machined by worn or damaged tools.
  • Customized assembly: Customized assembly uses controller data fusion and part-identification data, with high criticality because assembly must remain quick and accurate.Different controllers exchange data for fusion, whereas custom-part identification requires smaller identification data.
  • Use-case comparison: Table II organizes extracted data properties across recent Industry 4.0 use cases.The table is presented as a synthesis of data properties reported in recent works.

B. Other usecases

These use cases span monitoring, healthcare, and process control, with data characteristics varying by application layer and operational objective.

  • B. Other usecases: Container-terminal crane scheduling targets an energy-saving and service-efficiency trade-off while supporting rapid handling for mega-vessels.The passage identifies crane scheduling as critical because handling cranes mainly contribute to terminal energy consumption and service efficiency.
  • B. Other usecases: Refrigerated-warehouse optimization uses small volumes of temperature sensor data sent periodically to a central control station for long-term planning.Changing temperature set points can reduce product quality and increase costs, motivating comparisons involving electricity, maintenance, and energy consumption.
  • B. Other usecases: Industrial healthcare monitoring combines heterogeneous services, using limited sensors that generate small data volumes for long-term or real-time optimization.The passage frames safety monitoring, smart factories, and automated healthcare as recent industrial functions and service combinations.
  • B. Other usecases: Production-process control spans shop-floor vibration control, PLC design, and application-layer economic optimization, with data volume and traffic depending on the integration layer.The passage explicitly contrasts small or large data volumes and low or high network traffic across layers.
  • B. Other usecases: The architectural review extracts data-management properties and supported technological enablers to identify recent Industry 4.0 design trends.The review focuses on data presence, coordination, and computation, with Table III presenting the extracted management information.
  • B. Other usecases: Distributed computation places tasks across networked devices, enabling concurrency but introducing no global clock and independent device failures.Compared with concentrated computation, distributed computation uses multiple generally less powerful computers that coordinate by passing data.

A. Architectures focusing on assembly line and industrial robots

Architectures for assembly lines, robots, and cloud manufacturing range from centralized designs to decentralized or service-oriented arrangements, with data presence and coordination varying accordingly.

  • A. Architectures focusing on assembly line and industrial robots: Product-family architecture uses formal computational models and centralized methods, leaving limited room for ubiquitous data presence and coordination.The design addresses academic and industrial requirements for product-family architecture through a formal computer-assisted approach.
  • A. Architectures focusing on assembly line and industrial robots: Interoperable end-to-end manufacturing localizes data management conceptually even when data reside across different factories and manufacturing partners.The architecture supports communication and data exchange throughout the manufacturing life cycle, from supplier search through execution and monitoring.
  • A. Architectures focusing on assembly line and industrial robots: Cloud manufacturing research proposes decentralization through autonomous work systems acting as service providers rather than relying solely on a centralized cloud platform.The design permits data generation from various sources, including third-party online knowledge clouds.
  • A. Architectures focusing on assembly line and industrial robots: Manufacturing-service composition uses regulated coordination and computation in a highly centralized service-supporting system.Its data resources come from manufacturing, laboratory, and management sources and serve service requestors.
  • A. Architectures focusing on assembly line and industrial robots: Service-oriented manufacturing frameworks support asynchronous concurrent computation, while other PLC-oriented designs concentrate data presence, coordination, and computation.The service-oriented framework runs concurrent software behaviors across multiple machines; the IEC 61131-3 and IEC 61499 integration remains fundamentally concentrated.
  • A. Architectures focusing on assembly line and industrial robots: A five-layer computer-integrated manufacturing architecture separates physical, functional, managerial, informational, and control concerns, but keeps intra-layer coordination and computation focused centrally.The architecture is hierarchical, with each layer treated as a separate entity.

B. Architectures focusing on IIoT / ICPS, and WSAN

IIoT, ICPS, and WSAN architectures combine pervasive data generation with hierarchical or centralized coordination, while distributing computation across devices, subnetworks, or layers in selected designs.

  • B. Architectures focusing on IIoT / ICPS, and WSAN: Hybrid wireless architecture combines distributed communication and data entities with hierarchical coordination, while devices perform local computations to offload managers.The design is explicitly multi-tier and uses local computation to reduce the burden on local and global managers.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: The three-layer IIoT architecture connects IIoT nodes, gateways, and control, focusing on energy consumed by large numbers of deployed nodes.Its broader deployment includes RESTful service-hosted networks, a cloud server, and user applications.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Deterministic IIoT architecture uses simple packet switches configured by an SDN control plane, with pervasive data presence but centralized coordination and scheduling.The design seeks convergence between deterministic industrial networks and best-effort IIoT while supporting low latency and jitter.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Closed-loop wireless networked control architecture links plant sensors and actuators, controller nodes, and an intermediate network through wireless communication.The setting has ubiquitous data presence across the plant, controller, and network components.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Service-oriented control designs combine distributed computation with per-layer centralized coordination for data from ubiquitous WSAN sources.The coordination may also be viewed as decentralized under an alternative interpretation of the design.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Hierarchical WSAN and smart-factory architectures support pervasive data from stationary and mobile sources and decentralized computation through subnetworks.Examples of sources include automated guided vehicles, mobile workers’ devices, and WSANs.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Plant-wide monitoring decomposes processing into blocks and integrates block data through a centrally located decision-fusion algorithm.Plant-wide data have larger volume, multiple types, differing sampling rates, and potentially low collection density.
  • B. Architectures focusing on IIoT / ICPS, and WSAN: Future architectural insights consider ubiquitous data presence across integration layers while retaining centralized coordination and computation for ultra-high reliability.The discussion focuses on TSN and 5G designs.

A. Data enabling industrial technologies

The paper surveys industrial data-enabling technologies and data-centric services across networked industrial environments, from physical deployments and IIoT devices to cloud and applications.

  • A. Data enabling industrial technologies: Industrial networked environments couple physical processes with IIoT computational processes that receive data, calculate outputs, and apply them to the physical plant.The cyber part also provides and uses data-accessing and data-processing services.
  • A. Data enabling industrial technologies: The survey organizes Industry 4.0 building blocks into data-enabling industrial technologies and data-centric industrial services.Its coverage extends from field-level physical deployments to the cloud level.
  • A. Data enabling industrial technologies: The paper positions fault management, clustering analytics, reusable software, reactive test generation, modular reconfiguration, and predictive maintenance as important ICPS or IIoT operations.These operations rely on industrial data exchange, coordination, collaboration, and networked configuration.
  • A. Data enabling industrial technologies: The taxonomy includes AR/VR, camera and vision systems, anomaly detection, fault diagnosis, and multi-agent systems among data-centric services.These service categories are listed as part of the paper’s industrial data-management coverage.
  • A. Data enabling industrial technologies: Additional data-centric services include decision making, job scheduling, machine learning, big-data analytics, ontologies and semantics, human-in-the-loop systems, and energy management.The listed categories span operational decision support, analytics, semantic integration, human participation, and energy-related applications.
  • A. Data enabling industrial technologies: Recent IIoT research addresses topology optimization, packet scheduling, massive M2M communication, real-time RFID monitoring, data access control, and distributed data exchange.The examples include scalable supply-chain access control and a ZeroMQ-based industrial data-exchange mechanism.

2) WSAN:

WSANs collect and relay industrial-environment data, but harsh conditions make reliable real-time communication difficult. Related industrial systems use networked data for control, robot localization, navigation, collaboration, and flexible assembly.

  • 2) WSAN:: WSANs use dispersed sensors and actuators to monitor industrial conditions and deliver data centrally through single-hop or multi-hop communication.Industrial environments expose these networks to dust, heat, water, electromagnetic interference, and wireless-device interference.
  • 2) WSAN:: Communication improvements target reliable real-time WSAN operation through slot assignment, channel switching, synchronization, link-quality estimation, and cooperative relaying.These mechanisms address reliability, spectrum sharing, timing, link assessment, and secure data management.
  • 2) WSAN:: NCSs close control loops through communication networks, exchanging control and feedback data between plants and controllers.Network delays and data dropouts can prevent NCSs from satisfying performance requirements.
  • 4) Industrial Robots:: Industrial robot research covers stationary and mobile robots, including data-driven tracking control, localization, navigation, consensus, and collaborative coverage.Mobile-robot localization combines heterogeneous sensor or ambient data, while distributed exchange supports multi-robot coordination.
  • 5) Assembly Line:: Assembly processes combine resource identification, recognition, data collection, transmission, mining, and feedback control to support flexible production.Flexibility enables products with multiple styles, models, options, or configurations to share a production line.

5) Assembly Line:

Assembly-line data management spans sensing, models, communication, real-time operations, and feedback control. Industrial communication technologies are selected and adapted to satisfy application-specific reliability, latency, and throughput requirements.

  • 5) Assembly Line:: Modern assembly lines use sensor acquisition, CAD/CAM models, and simulation systems to increase data availability and support flexible manufacturing.Sensor systems generate large amounts of small-volume data, whereas CAD/CAM systems generate considerable large-volume data.
  • 5) Assembly Line:: Real-time manufacturing operations monitor machine variables, extract features from sensor streams, redefine products from 3D data, and assess production exceptions.These operations use heterogeneous data from factory machines, sensor streams, and CAD systems.
  • 5) Assembly Line:: Assembly-line research addresses uncertainty and diversity in remanufacturing data, complex-product assembly quality, and theoretical models of machine reliability and process dynamics.Proposed frameworks include probabilistic Boolean networks, timed event graphs, and data-driven autoregressive models.
  • 5) Assembly Line:: Industrial M2M communication focuses on device-to-device links, channels, transmissions, and one-hop exchanges across wired and wireless technologies.Research spans circuit and network models, antennas, filtering, multiplexing, interference management, and QoS guarantees.
  • 5) Assembly Line:: Communication configuration directly affects industrial data-management metrics such as data loss and delay, making communication design important for resource-intensive applications.The literature includes self-triggered sampling for NCSs and fast network-joining methods.
  • 5) Assembly Line:: IEEE 802.15.4e extends 802.15.4 for industrial requirements involving reliability, bounded latency, and interference protection.Other technologies covered include IEEE 802.11 WLAN, CAN, OPC-UA, EtherCAT, ISA100.11a, and WirelessHART.

B. Data centric industrial services

Data-centric industrial services apply sensing, computation, and learning to maintenance, control, anomaly detection, and fault diagnosis. The literature also highlights practical constraints from data volume, centralization, assumptions, memory, and bandwidth.

  • AR / VR:: AR and VR maintenance services centrally process large video volumes and use context databases to connect sensed physical information with spatially arranged maintenance content.Context sensing interprets raw sensor and camera data into low-level contexts such as marker IDs and transformation matrices.
  • Data centric industrial services: Camera and vision technologies support pattern recognition, fault estimation, template matching, and data-driven control tuning for industrial systems.The cited control approach uses iterative tuning and available data rather than a typical model-based design.
  • Data centric industrial services: Systems health management uses sensor data to assess system health, diagnose anomalies, and predict remaining useful performance over an asset’s life.One prognostics approach maps raw vibration data into monotonic features with early trends that can be predicted.
  • Fault diagnosis:: Online anomaly detection remains difficult because existing approaches may be centralized and complicated or restricted by strict assumptions, while industrial WSANs generate high capture rates and total volumes.These constraints complicate application to practical large-scale networked industrial systems.
  • Fault diagnosis:: Fault diagnosis methods address fault location, isolation, reconstruction, and reliability, but implementations may require prior data knowledge, additional memory, or sufficient acquisition bandwidth.These requirements are especially relevant to high-frequency signal detection and time-critical industrial networks.
  • 6) Multi-Agent Systems:: Multi-agent production systems aim to provide modular, flexible, robust, and adaptive operation, but classical systems’ static data hierarchies make modification difficult.A cited platform uses a central repository containing engineering data and information from line-design projects.

6) Multi-Agent Systems:

Multi-agent and machine-learning services use distributed data exchange, sensing, and computation to support industrial coordination and decision-making. Recent big-data approaches increasingly place ownership and processing closer to industrial edges while retaining centralized alternatives.

  • 6) Multi-Agent Systems:: Distributed industrial controllers and multi-agent methods coordinate agent data exchanges through decentralized control, optimization, and cultural algorithms.These approaches address industrial process integration and distributed decision-making under varying capacities.
  • 6) Multi-Agent Systems:: IIoT sensing can support automated decision-making inside and outside the shop floor by collecting data for machine coordination and employee-activity evaluation.The cited systems combine IIoT networks with information processing and data-based feedback or coordination.
  • 6) Multi-Agent Systems:: Manufacturing scheduling research spans single-machine, multiple-machine, multi-assembly-line, and inter-factory problems, including decentralized scheduling and operation control.Examples address multi-robot cells, multiple part types, serial batching, and related machine constraints.
  • 6) Multi-Agent Systems:: Industrial machine-learning services apply data-driven methods to QoS prediction, traffic classification, and other functions across enabling technologies.Deep learning can predict numerous industrial parameters and attributes, although efficient training is nontrivial.
  • 6) Multi-Agent Systems:: Big-data analytics evaluate industrial data across installation layers to study user preferences, technological-enabler behavior, and operational issues.The literature presents both generic edge-deployment frameworks and centralized manufacturing prediction toolboxes.
  • 6) Multi-Agent Systems:: Recent industrial approaches push computation decentralization toward the edge, particularly regarding data ownership and wireless network capacity.Examples include real-time WSAN data gathering, edge analytics, and investigations linking data processing with energy consumption.

11) Ontologies / Semantics:

Ontology services represent and formally define industrial data, entities, and their relationships, supporting automation and integration across industrial processes.

  • Ontologies / Semantics:: Ontology services formally represent, name, and define categories, properties, and relations among data and entities in industrial processes.
  • Ontologies / Semantics:: These services can automate tasks throughout industrial-system life cycles, from design through commissioning and operation.
  • Ontologies / Semantics:: Industrial standards such as IEC 61850 and IEC 61499 frequently support ontology services.
  • Ontologies / Semantics:: Semantic links enable automated integration and distributed updating in cloud manufacturing resource-service clouds.
  • Ontologies / Semantics:: Ontology services also support production-network systems, business integration, CAD assembly-model retrieval, and visual exploration.

14) Energy Management:

Energy management spans IIoT, WSANs, robotic cells, and assembly lines, with data supporting monitoring and optimization while deployment constraints remain.

  • Energy Management:: Industrial low-power WSAN protocols enable energy-conscious operation, but energy consumption still limits ubiquitous deployment of perpetual unattended devices.
  • Energy Management:: Real-time and historical usage data can help identify whether WSAN components are functioning properly.
  • Energy Management:: Energy optimization targets entire robotic cells and assembly lines through holistic minimization, dynamic low-power reconfiguration, and machine-consumption reduction.
  • Energy Management:: No significant contributions were found on energy-management issues for the data-enabling technology of networked control systems.

B. Data distribution in local and mobile clouds

The survey identifies centralized cloud collection as constrained by bandwidth, data-control, and cost concerns, motivating local and mobile distribution alongside distributed security.

  • B. Data distribution in local and mobile clouds: Cloud-manufacturing data collection commonly assumes network infrastructure can deliver all data to a cloud back end for processing and value extraction.
  • B. Data distribution in local and mobile clouds: Exclusive reliance on global clouds may provide insufficient bandwidth, reduce stakeholders’ control over data, and increase storage and computation costs.
  • B. Data distribution in local and mobile clouds: Local devices can distribute storage and computation into local or mobile clouds, using global clouds when global information or additional capacity is needed.
  • B. Data distribution in local and mobile clouds: Security mechanisms are well represented for IIoT/ICPS, WSANs, NCS, and M2M communication, but notably absent for assembly-line and industrial-robot enablers.
  • B. Data distribution in local and mobile clouds: Centralized security can introduce data loss and detection delay, whereas distributed solutions are described as more agile and robust for large-scale deployments.
  • B. Data distribution in local and mobile clouds: Architectural trends combine centralized data management for assembly lines and robots with decentralization for IIoT and WSANs.
Loading 1902.06141v2…