Industrial Communication and Networked Control

Introduction

Modern automated machines are not isolated collections of sensors, actuators, programmable logic controllers (PLCs), drives, robots, and supervisory computers. They are networked cyber-physical systems. Industrial communication networks transport measurements, commands, diagnostic information, configuration data, and time references between these devices. The quality of that communication directly affects machine productivity, safety, maintainability, and control performance.

A network used for industrial automation must be evaluated differently from an ordinary office network. Office applications generally tolerate variable delay and occasional retransmission. A motion-control system may require a command to arrive within a narrow time window, with predictable delay and very low jitter. Consequently, industrial networking is assessed using several related criteria: bandwidth, latency, determinism, synchronization, availability, interoperability, diagnostics, and security.

This chapter introduces major industrial communication technologies—Modbus, CAN, PROFINET, EtherNet/IP, EtherCAT, and OPC UA—and explains how they are integrated into networked control systems. The emphasis is on selecting and engineering technologies rather than memorizing protocol names.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the functions of industrial communication layers and network topologies.
2. Distinguish fieldbus, industrial Ethernet, and industrial information protocols.
3. Describe the operating principles and suitable applications of Modbus, CAN, PROFINET, EtherNet/IP, EtherCAT, and OPC UA.
4. Evaluate latency, jitter, determinism, bandwidth, interoperability, diagnostics, and availability.
5. Estimate communication timing and determine whether a network is suitable for a control task.
6. Design a basic layered architecture for a networked automated machine.
7. Identify practical, cybersecurity, safety, and governance issues in industrial networks.

1. Industrial Communication as a Control-System Function

A networked control loop contains a process, sensors, a controller, actuators, and communication links. A typical sequence is:

sensor measurement → input communication → controller computation → output communication → actuator response.

Communication introduces delay. If a sensor value is sampled at time t and used by the controller at a later time, the effective control action is based on older information. In a simple loop, the total communication contribution can be represented as:

Here, T_comm is communication-related delay, while T_sensor is sensor-interface delay, T_tx and T_rx are transmitting and receiving delays, T_queue is waiting time in devices, and T_switch is forwarding delay. All quantities are measured in seconds. The expression is an engineering decomposition, not a universal protocol equation; some terms may be combined or omitted depending on the system.

Latency is the time between sending information and its valid reception. Jitter is variation in latency from one transmission to another. Determinism means that timing behavior is bounded and predictable. A network can have low average latency but poor determinism if occasional congestion produces large delays.

For periodic communication with cycle time T_c, a basic timing requirement is:

T_comm,max is the worst-case communication time, T_compute is controller computation time, and T_c is the permitted control-cycle period. This condition is necessary but not sufficient for stability or adequate control quality. The plant dynamics, sampling rate, controller design, and synchronization error must also be considered.

Bandwidth and message load

The approximate bit rate required for periodic messages is:

N_i is the number of messages of class i transmitted per cycle, L_i is the message length in bits including relevant protocol overhead, and f_i is the transmission frequency in hertz. R_load is in bits per second. The estimate assumes that message classes are independent and that L_i includes the overhead being evaluated. It does not by itself account for arbitration, inter-frame gaps, retransmissions, or physical-layer encoding.

A useful approximate utilization measure is:

where R_link is nominal link bit rate. U is dimensionless. It is not a guaranteed maximum network utilization: protocol overhead, scheduling, collisions, priority rules, and safety margins must be included. High utilization usually increases queueing delay and reduces tolerance to bursts and faults.

2. Network Architecture and Industrial Layers

Industrial systems are commonly divided into levels:

The Purdue model is a useful conceptual guide, but real installations may combine or bypass levels. A sound design separates time-critical control traffic from less time-critical information traffic, even when both use Ethernet.

A protocol stack has a physical layer, data-link mechanisms, network and transport services where required, and application semantics. Industrial protocols also define device profiles, object models, cyclic data, acyclic parameters, diagnostics, and configuration procedures. Two devices may both use Ethernet and still be unable to interoperate if their application models differ.

3. Fieldbus and Industrial Ethernet Technologies

3.1 Modbus

Modbus is a simple request-response application protocol widely used for PLCs, instruments, meters, drives, and remote I/O. Modbus RTU commonly operates over serial links such as RS-485; Modbus TCP maps the Modbus application data into TCP/IP and Ethernet frames.

The protocol uses a client/server model, historically called master/slave. A client requests data or writes values, and a server responds. Common data areas include coils, discrete inputs, input registers, and holding registers. Values are often represented as 16-bit registers, so multi-register values require an agreed byte order and word order.

Modbus advantages include simplicity, low cost, broad availability, and ease of troubleshooting. Limitations include limited standardized semantics, relatively weak device-profile interoperability, request-response overhead, and no inherent guarantee of hard real-time behavior. Modbus is appropriate for many supervisory measurements and non-demanding device interfaces, but it is generally unsuitable for tightly synchronized multi-axis motion.

For RS-485 installations, correct termination, biasing, grounding, cable selection, node addressing, and line length are essential. A protocol analyzer or serial oscilloscope can distinguish wiring faults from addressing and register-mapping faults.

3.2 CAN and CANopen

The Controller Area Network (CAN) is a multi-master broadcast bus designed for robust distributed control. Messages are identified by an identifier rather than by a destination address. The identifier also determines arbitration priority: a dominant bit overrides a recessive bit, allowing the message with the highest priority to continue without destructive collision.

CAN provides error detection, acknowledgement, automatic retransmission, and fault confinement. Its strengths are robustness, compact frames, priority-based access, and extensive use in mobile machines and embedded systems. A disadvantage is that all nodes share bus capacity, and low-priority messages may experience delay when high-priority traffic is heavy.

CANopen adds a standardized application layer, including an object dictionary, process data objects (PDOs), service data objects (SDOs), network management, and device profiles. CANopen therefore provides more interoperability than raw CAN, although conformance and vendor-specific features still require verification.

For a fixed-priority CAN bus, exact worst-case response-time analysis can be complex. It must consider frame length, higher-priority interference, blocking by a lower-priority frame already in transmission, bit stuffing, and retransmission. Average bus utilization alone cannot prove that deadlines will be met.

3.3 PROFINET

PROFINET is an industrial Ethernet system commonly used with PLCs, remote I/O, drives, and diagnostic tools. It supports cyclic real-time communication, acyclic parameter and diagnostic communication, device discovery, configuration, and standardized engineering integration.

PROFINET Real Time (RT) traffic uses prioritized Ethernet mechanisms for ordinary industrial real-time exchange. PROFINET IRT (Isochronous Real Time) uses specialized scheduling and synchronization techniques for demanding motion applications. The required performance depends on controller class, devices, network topology, configuration, and cycle time; the label “PROFINET” alone does not specify a single deterministic performance level.

PROFINET systems use engineering descriptions such as GSDML files to describe device capabilities and modules. Diagnostics can identify device absence, channel faults, topology problems, parameter errors, and maintenance information. Its major practical strengths are integration with PLC engineering environments, rich diagnostics, and support for distributed I/O and motion.

3.4 EtherNet/IP

EtherNet/IP, administered by ODVA, uses standard Ethernet and the Common Industrial Protocol (CIP). CIP defines objects, services, device profiles, and communication models. EtherNet/IP supports explicit messaging for configuration and information exchange and implicit, often called I/O, messaging for cyclic data.

Common industrial mechanisms include producer-consumer communication, multicast or unicast transmission, device profiles, and CIP Safety or other extensions where applicable. Quality of Service and managed-switch configuration can help prioritize time-critical traffic. For more demanding coordinated motion, CIP Motion and appropriate time-synchronization mechanisms are used.

EtherNet/IP offers broad device availability and strong object-oriented semantics. However, ordinary switched Ethernet with EtherNet/IP should not automatically be assumed to provide hard real-time behavior. Performance depends on packet rates, switch architecture, traffic prioritization, controller capacity, and network design.

3.5 EtherCAT

EtherCAT is an Ethernet-based real-time system designed for high-performance cyclic I/O and motion. A master sends a frame through a sequence of slaves. Each slave reads and inserts its data while the frame passes through, rather than receiving, buffering, and retransmitting a separate frame in the conventional way.

This processing method reduces protocol overhead and supports tightly synchronized distributed devices. EtherCAT can use distributed clocks so that devices align their local time bases. It is particularly effective for servo drives, high-channel-count I/O, robotics, and synchronized measurement.

EtherCAT requires suitable EtherCAT devices and engineering tools; ordinary Ethernet devices cannot simply be inserted into the real-time process path as if they were EtherCAT slaves. Diagnostic information can identify link breaks, working-counter errors, device state, cable faults, and synchronization problems. Its high performance is accompanied by greater dependence on specialized device support and system configuration.

3.6 OPC UA

OPC Unified Architecture (OPC UA) is a platform-independent framework for secure industrial information exchange. It provides a rich information model, object and variable types, methods, events, historical access, security mechanisms, and client/server communication. OPC UA PubSub supports publish-subscribe exchange and can be mapped to transports such as UDP or time-sensitive networking in suitable implementations.

OPC UA is particularly valuable between control, supervisory, manufacturing, and enterprise systems. It can expose meaningful information such as a pump’s speed, operating state, alarm condition, engineering unit, and maintenance status rather than an unexplained register number. Security features include authentication, authorization, encryption, signing, certificates, and auditing, but these must be correctly configured.

OPC UA is not automatically a hard-real-time motion bus. Its suitability depends on the selected communication model, implementation, network, update rate, and timing requirements. In many architectures, a real-time field network performs closed-loop control while OPC UA supplies semantically rich monitoring and coordination data.

4. Comparative Evaluation

No technology is universally best. Selection should begin with control requirements, existing installed base, device availability, safety requirements, lifecycle support, and engineering competence.

5. Worked Timing Example

An illustrative machine has 24 sensors producing 32 bytes every 2 ms and 8 actuators receiving 16 bytes every 2 ms. Assume a simplified 20% allowance for protocol and framing overhead. Estimate the payload rate and approximate required rate including overhead.

Sensor payload per cycle is 24 × 32 = 768 bytes. Actuator payload is 8 × 16 = 128 bytes. Total payload is 896 bytes per 2 ms. The frequency is 1/0.002 = 500 Hz. Therefore:

Including the stated overhead allowance:

This is not a complete network design. It excludes Ethernet inter-frame gaps, preambles, acknowledgements, management traffic, retransmissions, switch processing, and safety margins. A 100 Mbit/s link may have sufficient nominal capacity, but the designer must still verify cycle-time behavior, worst-case delay, device limits, and fault response. If the measured worst-case communication delay is 0.65 ms and controller computation takes 0.40 ms, the condition for the 2 ms cycle is 1.05 ms ≤ 2 ms, so the basic timing budget is satisfied. Stability and application-specific control performance require further analysis.

6. Interoperability, Diagnostics, and Network Engineering

Interoperability has at least four dimensions:

1. Physical interoperability: connectors, cable, signaling, speed, grounding, and power are compatible.
2. Protocol interoperability: devices correctly implement the same communication rules.
3. Semantic interoperability: the meaning, units, scaling, byte order, and quality status of data agree.
4. Engineering interoperability: tools can configure, diagnose, update, and maintain the devices.

A network may pass physical tests while failing semantic interoperability. For example, two systems may both transmit a 16-bit value, but one may interpret it as signed temperature in tenths of a degree and the other as an unsigned speed value.

Diagnostics should be designed, not merely enabled. Useful diagnostic data includes device identity, link state, packet counters, error counters, missed cycles, watchdog status, time synchronization error, voltage and temperature, module status, and alarm history. A layered diagnostic procedure begins with power and wiring, then link status, addressing, protocol configuration, application mapping, timing, and controller logic.

Managed industrial switches can provide VLANs, Quality of Service, port mirroring, ring redundancy, access control, and event logging. Redundancy may use ring, line, parallel, or controller-redundant architectures. Redundancy improves availability but adds cost, configuration complexity, and possible common-mode failures.

7. Networked Control Design Method

A practical design sequence is:

1. Define every data item, its source, destination, size, update rate, deadline, priority, and safety classification.
2. Separate closed-loop control, supervisory, maintenance, and enterprise traffic.
3. Select the physical medium and topology according to distance, environment, mobility, and electromagnetic interference.
4. Select a protocol with suitable timing, device support, diagnostics, and lifecycle availability.
5. Calculate approximate load and construct a timing budget.
6. Configure priorities, synchronization, watchdogs, addressing, and fault behavior.
7. Test normal operation, overload, cable disconnection, device restart, clock error, and switch failure.
8. Document configuration, firmware versions, device descriptions, backups, and recovery procedures.
9. Apply cybersecurity controls and manage changes throughout the asset life cycle.

The fail-safe state must be defined explicitly. Loss of communication may require stopping motion, closing a valve, maintaining a safe pressure, or transferring control to a local controller. It is not acceptable to let each device choose an undocumented response.

8. Applied Case: Illustrative Packaging Cell

Background

Consider an illustrative packaging cell containing a PLC, four servo axes, distributed I/O, a barcode reader, a safety controller, an HMI, and a production database. The axes must coordinate at a 1 ms cycle. The barcode reader updates at 20 Hz, while production records are transferred every few seconds.

Problem

An initial design placed all traffic on a simple Ethernet network. Operators observed occasional motion faults during recipe downloads and large database transfers. The system had adequate average bandwidth, but the control traffic experienced variable delay and the diagnostic information was insufficient to identify the cause.

Analysis

Motion data requires deterministic cyclic exchange and time synchronization. Barcode data is important but not cycle-critical. Database traffic is bursty and should not compete directly with motion traffic. The original design therefore failed primarily in traffic classification and timing assurance, not nominal bandwidth.

Approach and solution

The redesigned architecture uses a motion-capable real-time industrial Ethernet network for servo and distributed-I/O traffic. The PLC and motion devices exchange cyclic data with configured priorities and synchronized clocks. The barcode reader uses a separate service connection. An OPC UA server publishes machine state, recipe information, alarms, and production counters to the HMI and manufacturing system. Managed switches separate control and information traffic using VLAN and priority policies. A firewall controls the boundary to the plant information network.

Watchdogs detect missed cycles, and the machine transitions to a defined safe state if critical communication is lost. Network diagnostics record port errors, cycle violations, device state changes, and time-synchronization status. A test plan intentionally disconnects cables, overloads the information network, restarts devices, and downloads a recipe during production simulation.

Outcome

The architecture prevents routine database bursts from determining servo timing. The most important improvement is not simply higher speed; it is separation of requirements, bounded timing, explicit fault behavior, and meaningful diagnostics. The final acceptance test demonstrates both normal performance and controlled degradation.

Learning points

• Average bandwidth is not a substitute for worst-case timing analysis.
• Real-time and information traffic should be separated logically or physically.
• A protocol choice must include device profiles, engineering tools, and maintenance capability.
• Diagnostics reduce mean time to repair only when alarms are actionable and recorded.
• Safety functions require appropriate safety-rated mechanisms and validation; ordinary network availability is not itself a safety function.

9. Security, Safety, and Governance

Industrial networks increasingly connect to enterprise systems, remote-service platforms, and cloud analytics. Connectivity increases value but also increases attack surface. Threats include unauthorized commands, credential theft, malware, insecure remote access, manipulated configuration, denial of service, and loss of data integrity.

Security should follow defense-in-depth principles: asset inventory, network segmentation, least privilege, strong authentication, certificate management, secure remote access, patch and vulnerability management, backups, logging, incident response, and tested recovery. Security controls must be coordinated with availability and safety requirements. A poorly planned security change can stop production; an absent control can permit dangerous manipulation.

Safety and cybersecurity are related but different. A safety system reduces risk from hazardous energy or malfunction; cybersecurity protects systems against unauthorized or malicious influence. Safety communication protocols and certified safety devices may be required for specific functions. Risk assessment, applicable machinery regulations, electrical standards, and organizational procedures must be consulted for the installation’s jurisdiction.

Ethically, engineers should avoid concealing diagnostic limitations, unsupported interoperability claims, or known vulnerabilities. Configuration files and operational data may contain sensitive production information. Access, retention, and sharing should follow organizational policy and applicable privacy and intellectual-property requirements.

10. Contemporary Developments and Limitations

Time-Sensitive Networking (TSN) extends IEEE Ethernet with mechanisms for time synchronization, traffic scheduling, and bounded transmission behavior. It can support convergence of operational and information traffic, but actual performance depends on compatible hardware, profiles, configuration, and application requirements.

OPC UA companion specifications improve semantic interoperability for domains such as robotics, machine tools, and energy. Edge computing can preprocess data close to machines, reducing unnecessary traffic and improving response to local events. Artificial intelligence can assist anomaly detection by analyzing network counters, timing deviations, and device health indicators. However, AI recommendations require validated data, explainable alarms, cybersecurity protection, and human responsibility for safety-critical decisions. AI should not be treated as a replacement for deterministic control logic unless the complete safety and control argument has been established.

Important limitations remain. Vendor ecosystems can create dependence on proprietary tools. Protocol conformance does not guarantee correct application integration. Ethernet speed does not guarantee determinism. Redundancy does not remove common-mode failures. Encryption can add processing and configuration overhead. Finally, a technically excellent network can fail operationally if maintenance personnel lack documentation and training.

Chapter Summary

Industrial communication networks connect sensors, controllers, actuators, supervisory systems, and enterprise applications. Their engineering quality depends on more than nominal data rate. Latency, jitter, determinism, synchronization, availability, diagnostics, interoperability, safety, and cybersecurity must be evaluated together.

Modbus provides simple and widely available register-based communication. CAN offers robust priority-based embedded networking, while CANopen adds device and application structures. PROFINET and EtherNet/IP provide industrial Ethernet ecosystems for cyclic I/O, diagnostics, and control. EtherCAT is optimized for tightly synchronized high-performance I/O and motion. OPC UA provides secure, semantic, platform-independent information exchange and is especially useful above the real-time control layer.

A successful design begins with requirements, separates traffic by timing and criticality, estimates load, verifies worst-case behavior, defines fault responses, and validates diagnostics. Protocol selection should reflect the entire life cycle, including devices, tools, skills, cybersecurity, safety, and maintainability.

Key Terms

Latency: Time between transmission of information and its valid reception.

Jitter: Variation in communication delay or cycle timing.

Determinism: Predictability and boundedness of timing behavior.

Fieldbus: Digital industrial communication network traditionally used for field devices and controllers.

Industrial Ethernet: Ethernet-based technology adapted for industrial control and diagnostics.

Modbus: Register-oriented industrial application protocol using serial or TCP/IP communication.

CAN: Multi-master controller network using identifier-based priority arbitration.

CANopen: Application-layer framework built on CAN.

PROFINET: Industrial Ethernet system supporting real-time automation communication.

EtherNet/IP: Ethernet implementation of the Common Industrial Protocol.

EtherCAT: Real-time Ethernet technology in which devices process frames as they pass through.

OPC UA: Secure, platform-independent framework for semantic industrial information exchange.

Cyclic communication: Repeated exchange at a configured period.

Acyclic communication: Event-driven or on-demand exchange, commonly used for parameters and diagnostics.

Producer-consumer model: Communication model in which data producers publish information that one or more consumers use.

Distributed clock: Local device clocks synchronized to support coordinated operation.

Interoperability: Ability of systems to exchange and correctly interpret information.

Watchdog: Mechanism that detects missing or late communication and initiates a defined response.

Time-Sensitive Networking: Ethernet standards and profiles supporting scheduled and time-bounded traffic.

Review Questions

Short Answer Questions

1. Define latency, jitter, and determinism in an industrial network.
2. What is the principal difference between cyclic and acyclic communication?
3. Identify two strengths and two limitations of Modbus.
4. Why is semantic interoperability important?
5. What functions are commonly provided by managed industrial switches?

Long Answer Questions

1. Compare CANopen, PROFINET, EtherNet/IP, and EtherCAT for automation applications.
2. Explain how a networked control loop is affected by communication delay and jitter.
3. Describe the role of OPC UA in a layered industrial automation architecture.
4. Develop a practical method for commissioning and diagnosing an industrial Ethernet network.
5. Discuss how safety and cybersecurity requirements differ and how they should be coordinated.

Analytical / Application Questions

1. A machine has a 4 ms control cycle, 1.1 ms worst-case communication delay, and 0.8 ms controller computation time. Determine whether the basic timing inequality is satisfied and identify additional evidence needed before accepting the design.
2. Select a suitable communication approach for a six-axis synchronized servo system, distributed I/O, and production historian. Justify the choice by separating timing and information requirements.
3. A network has adequate average bandwidth but produces occasional motion faults during large file transfers. Analyze likely causes and propose a verification and redesign plan.

Multiple Choice Questions with Answers

Question 1

Which characteristic most directly distinguishes a deterministic industrial network from an ordinary high-bandwidth network?

A. Predictable and bounded communication timing

B. Use of longer cable runs

C. A higher number of user accounts

D. Larger database capacity

Correct Answer: A. Predictable and bounded communication timing

Question 2

Which protocol is primarily known for simple register-based request-response communication?

A. EtherCAT

B. Modbus

C. OPC UA

D. PROFINET IRT

Correct Answer: B. Modbus

Question 3

In CAN arbitration, what generally determines which message continues transmission?

A. The message with the longest payload

B. The message sent by the physically closest node

C. The identifier-based priority

D. The message with the highest IP address

Correct Answer: C. The identifier-based priority

Question 4

Which technology is especially associated with processing an Ethernet frame as it passes through sequential devices?

A. Modbus TCP

B. CANopen

C. OPC UA

D. EtherCAT

Correct Answer: D. EtherCAT

Question 5

What is a major purpose of OPC UA information models?

A. To replace all physical wiring

B. To give industrial data meaningful and structured semantics

C. To guarantee hard-real-time servo control in every implementation

D. To eliminate the need for authentication

Correct Answer: B. To give industrial data meaningful and structured semantics

Question 6

A system sends 896 bytes every 2 ms. What is the payload bit rate?

A. 358.4 kbit/s

B. 896 kbit/s

C. 3.584 Mbit/s

D. 7.168 Mbit/s

Correct Answer: C. 3.584 Mbit/s

Question 7

Which statement about nominal Ethernet speed is correct?

A. It alone guarantees deterministic control timing

B. It removes the need for diagnostics

C. It guarantees interoperability between all Ethernet devices

D. It does not by itself guarantee bounded latency or determinism

Correct Answer: D. It does not by itself guarantee bounded latency or determinism

Question 8

Which communication type is commonly used for device parameters and diagnostics rather than fixed-period process data?

A. Acyclic communication

B. Carrier arbitration

C. Distributed clocking

D. Frame padding

Correct Answer: A. Acyclic communication

Question 9

Which is an example of semantic interoperability?

A. Two devices use identical connector colors

B. Two devices agree that a register represents speed in revolutions per minute

C. Two switches have the same enclosure size

D. Two cables have equal physical length

Correct Answer: B. Two devices agree that a register represents speed in revolutions per minute

Question 10

Why should production database traffic often be separated from servo traffic?

A. Database traffic always has a lower voltage

B. Servo traffic cannot use digital signals

C. Bursty information traffic can introduce timing variation into control communication

D. Servo systems do not require diagnostics

Correct Answer: C. Bursty information traffic can introduce timing variation into control communication