Industrial Communication, HMI and Supervisory Control

Introduction

Modern manufacturing systems depend on the reliable movement of information as much as on the movement of materials. Sensors measure temperature, pressure, position, speed, vibration, and quality variables. Controllers interpret these measurements and command actuators. Human–machine interfaces (HMIs) present operating information to people, while supervisory control and data acquisition (SCADA) systems collect plant-wide data, manage alarms, and support production decisions. Industrial communication networks connect all these functions.

Industrial communication differs from ordinary office networking because it must often provide deterministic timing, high availability, electrical noise immunity, long service life, safety, and predictable behavior during faults. A delayed email is inconvenient; a delayed emergency shutdown signal may be dangerous. Consequently, network selection and configuration are engineering decisions involving performance, interoperability, cybersecurity, maintainability, and lifecycle cost.

This chapter examines industrial network architectures and protocols, HMI and SCADA design, data acquisition, alarm management, historians, distributed control, interoperability, and dependable communication.

Learning Objectives

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

1. Distinguish field, control, supervisory, and enterprise network layers.
2. Explain the roles of PLCs, PACs, RTUs, DCS controllers, HMIs, SCADA servers, and historians.
3. Compare major industrial communication protocols and their uses.
4. Design basic tag structures, HMI screens, alarm systems, and data-acquisition paths.
5. Explain polling, event-driven communication, determinism, redundancy, and time synchronization.
6. Evaluate interoperability and cybersecurity risks in industrial networks.
7. Apply good engineering practice to a manufacturing automation case.

1. Industrial Communication Architecture

1.1 From signals to information

An automated machine commonly contains four functional stages:

• Measurement: sensors convert physical variables into electrical or digital signals.
• Control: a PLC, PAC, industrial computer, or DCS controller executes logic and control algorithms.
• Supervision: an HMI or SCADA system displays status, trends, alarms, and commands.
• Management: manufacturing execution systems (MES), enterprise resource planning (ERP), maintenance, and analytics systems use selected production data.

A useful architectural model is the Purdue Enterprise Reference Architecture. In simplified form, levels 0–2 perform physical control, level 3 coordinates site operations, and levels 4–5 support enterprise planning and external services. The model is not a mandatory physical arrangement; it is a way of separating functions and controlling information flows.

Figure: Functional layers of an industrial automation architecture. Information should cross layers through controlled interfaces rather than unrestricted connections.

1.2 Network topologies

A bus connects multiple devices to a shared communication medium. It uses little cable but requires careful termination and can be affected by a fault on the shared segment. A star connects devices to a central switch. It is easy to diagnose and expand, but the central switch is a possible single point of failure. A ring provides an alternate path when supported by a redundancy protocol. A tree combines these arrangements and is common in plants.

Topology must be selected from requirements rather than appearance. Important questions include:

• What maximum communication delay is acceptable?
• Must control continue if a cable or switch fails?
• How far apart are devices, and what electromagnetic interference is present?
• How many nodes and messages will be added in the future?
• Can technicians diagnose the network without stopping production?

1.3 Bandwidth, latency, jitter, and determinism

Bandwidth is the capacity of a link, usually expressed in bit/s. Latency is the time between transmission and reception. Jitter is variation in latency. A network can have high bandwidth but poor control performance if latency is unpredictable.

For a message containing L bits transmitted over a link of rate R bit/s, the serialization time is:

Here, tₛ is in seconds, L is the message length in bits, and R is the link rate in bit/s. This formula excludes propagation, processing, queuing, and protocol overhead. It is therefore only one component of total delay. Deterministic systems constrain both average delay and its upper bound.

2. Industrial Protocols and Interoperability

2.1 Ethernet and TCP/IP

Industrial Ethernet uses Ethernet physical and data-link technologies with industrial protocols above them. Ethernet provides frames and addressing but does not itself define the meaning of a temperature value or a motor command. TCP provides reliable, ordered delivery, whereas UDP provides datagrams without TCP's connection and retransmission mechanisms. TCP is useful for many supervisory applications; UDP or specialized real-time mechanisms may be selected where low overhead is important and the application handles loss appropriately.

The Internet Protocol (IP) provides logical addressing and routing. Industrial networks should not be treated as automatically safe merely because they use private IP addresses. Segmentation, access control, monitoring, and secure maintenance are still necessary.

2.2 Common industrial protocols

No protocol is universally best. Selection should consider installed equipment, real-time requirements, physical distance, certification, diagnostics, vendor support, cybersecurity, and total lifecycle cost.

2.3 Addressing and data representation

A tag such as Line1.PumpA.SpeedFeedback has meaning to a human, whereas a Modbus register or a raw byte sequence does not. Interoperability requires agreement about:

• device and network addresses;
• data type, such as Boolean, signed integer, floating point, or string;
• byte order and word order;
• engineering units and scaling;
• timestamp meaning and time zone;
• quality status, such as good, uncertain, or bad;
• command authority and permitted ranges.

For an analog input represented by a raw count x, linear scaling to engineering units can be written as:

where x is the measured raw value, xₘᵢₙ and xₘₐₓ are the raw-range limits, and y is the engineering value between yₘᵢₙ and yₘₐₓ. The relationship assumes linear sensor and input behavior and correct calibration. Units of y must be stated; for example, °C, bar, or mm.

Worked example. A 4–20 mA transmitter represents 0–100 °C. If the measured current is 12 mA, the fraction of span is (12−4)/(20−4)=0.5. The temperature is therefore 50 °C, assuming a linear calibrated transmitter and negligible loop error.

2.4 Interoperability versus integration

Interoperability means that independently produced systems can exchange and correctly interpret information. Merely connecting cables is not interoperability. A successful interface defines semantic meaning, data quality, update rate, command behavior, failure response, and ownership.

OPC UA is particularly valuable because it can expose structured objects, methods, metadata, and security mechanisms rather than only unnamed registers. However, an OPC UA server cannot correct poor tag names, incorrect units, or unsafe permissions. Interoperability remains an engineering and governance task.

3. Controllers and Distributed Control

A PLC executes a scan or scheduled task: inputs are acquired, logic is executed, and outputs are updated. A PAC generally provides broader computing, networking, and motion capabilities. An RTU is designed for geographically distributed monitoring and control, often with low power and intermittent communications. A DCS distributes control across controllers and emphasizes process control, operator stations, engineering tools, and high availability.

Distributed control reduces dependence on one central computer. If a supervisory server fails, local controllers should normally continue essential automatic control. This principle is called control autonomy. Supervisory commands should be validated by the controller, not trusted merely because they originate from an HMI.

A control architecture may include:

• local interlocks and permissives in the controller;
• safety instrumented or emergency systems independent of ordinary supervisory logic where required;
• redundant controllers or networks for critical processes;
• local manual stations for controlled recovery;
• supervisory commands such as start, stop, set-point, mode, and acknowledge.

The distinction between basic process control and functional safety is essential. A standard HMI alarm must not be assumed to perform a safety function unless the complete system has been designed, verified, validated, and maintained under applicable safety requirements.

4. HMI Engineering

4.1 Purpose and information hierarchy

An HMI should help an operator understand the process, detect abnormal conditions, diagnose causes, and take authorized action. It is not merely a colorful mimic diagram.

A practical hierarchy is:

1. Overview screen: plant or line status, major production indicators, active critical alarms.
2. Area screen: equipment relationships, flows, states, and abnormal conditions.
3. Unit screen: detailed equipment status, modes, interlocks, and commands.
4. Diagnostic screen: I/O quality, communication status, permissives, and maintenance data.
5. Trend and alarm screens: historical context and event investigation.

Normal operation should use restrained color. Bright colors should be reserved for abnormal or actionable conditions. Every command should show current state, command availability, mode, and feedback. A start button without confirmation that the motor actually started creates ambiguity.

4.2 HMI tag and command design

A robust tag includes a meaningful name, description, engineering unit, source, data type, update rate, alarm limits, quality, timestamp, and access classification. Commands should use a control handshake where practical: request, validation, execution, feedback, timeout, and fault indication.

For example, an HMI may issue PumpA.StartRequest. The PLC checks permissives, verifies that the pump is not in local mode, energizes the starter, and returns PumpA.RunningFeedback. If feedback does not arrive within a defined interval, the controller reports a start failure. The HMI should display the cause rather than simply changing the button color.

4.3 Human factors

HMI design should account for perception, workload, fatigue, language, accessibility, and emergency response. Good practice includes consistent symbols, readable text, clear units, stable screen layouts, alarm prioritization, and avoidance of excessive animation. Operators should be involved in testing because technically correct screens can still be operationally confusing.

5. SCADA and Data Acquisition

SCADA systems supervise geographically or functionally distributed assets. A typical system contains field instruments, PLCs or RTUs, communication networks, a data server, HMI clients, an alarm server, and a historian. Modern systems may use redundant servers and web-based clients, but the underlying responsibilities remain similar.

5.1 Polling and event-driven collection

In polling, a master or server asks devices for data at scheduled intervals. Polling is simple and predictable but may waste traffic when values do not change. Event-driven or unsolicited reporting sends data when a change or event occurs. It can reduce traffic and improve event awareness, but it requires buffering, sequence numbers, time synchronization, and recovery procedures.

A data-acquisition design should specify scan rate based on process dynamics. A rapidly changing variable may require a short interval; a tank level may not. Collecting every available value at the fastest possible rate increases storage and network load without necessarily improving decisions.

5.2 Historian functions

A historian stores time-stamped process values, events, alarms, and metadata. Many historians use compression or deadbanding. A deadband stores a new value only when the change exceeds a defined threshold. This reduces storage but can hide small changes if applied carelessly.

A historian should preserve:

• value and timestamp;
• quality and substitution status;
• source and engineering unit;
• configuration changes;
• alarm and operator events;
• time synchronization status.

The timestamp must be interpreted correctly. A server-receive timestamp is not equivalent to the instant at which a sensor measured the value. For sequence-of-events analysis, device-level timestamps and synchronized clocks are important.

6. Alarm Management

An alarm indicates an abnormal condition requiring operator awareness or action. An alarm is not merely a status message. A well-designed alarm has a defined consequence, priority, response, and return-to-normal behavior.

A useful alarm record includes tag, condition, priority, time raised, time acknowledged, time cleared, responsible area, operator guidance, and quality. Common priorities are advisory, low, medium, and high, but the exact scheme should be based on consequence and response time.

Alarm systems suffer from nuisance alarms, chattering alarms, alarm floods, stale alarms, and poorly documented limits. Techniques such as hysteresis, on-delay, off-delay, shelving under controlled policy, suppression by operating state, and rationalized priorities can improve performance. Suppression must not conceal a hazardous condition.

Alarm priority should be justified by risk, not by the visual importance of a tag. In a risk-based treatment, a simplified risk index may be expressed as:

where R is a relative risk index, P is the estimated likelihood, and C is consequence severity. This is not a universal safety equation; it is a qualitative or semi-quantitative screening model. Formal safety decisions require an appropriate hazard-analysis method, defined scales, and competent review.

7. Dependable Communication and Cybersecurity

7.1 Availability and redundancy

Availability is the proportion of required service time during which a system performs its intended function. For a repairable component with mean time between failures (MTBF) and mean time to repair (MTTR), a simple steady-state approximation is:

Both times must use the same units, and the approximation assumes statistically stable failure and repair behavior. It does not capture common-cause failures, planned maintenance, network dependencies, or imperfect repairs.

Redundancy may be applied to power supplies, network paths, switches, controllers, servers, and communication links. Redundancy is effective only when failure detection, switchover, configuration consistency, and testing are addressed. Two identical devices with the same uncorrected design defect do not provide meaningful independence.

7.2 Time synchronization

Accurate time supports event reconstruction, sequence-of-events analysis, production traceability, and coordinated control. Network Time Protocol may be adequate for many supervisory functions; Precision Time Protocol provides higher precision where supported by suitable hardware and network design. Time sources should be monitored, and devices should record synchronization quality.

7.3 Industrial cybersecurity

Industrial cybersecurity protects safety, availability, integrity, and confidentiality, with availability often receiving unusual emphasis. Important controls include:

• network zoning and conduits;
• firewalls and industrial demilitarized zones;
• least-privilege accounts;
• multifactor authentication for remote access where feasible;
• controlled vendor access;
• secure backups and tested restoration;
• patch and vulnerability management;
• application allowlisting where appropriate;
• logging, monitoring, and incident response;
• removal or disabling of unused services;
• documented asset inventories.

Security changes must be managed with production constraints. An untested patch can cause an outage, but indefinite postponement leaves vulnerabilities exposed. A risk-based maintenance process should include backups, rollback plans, testing, and authorization.

8. Applied Case: Packaging Line Communication Upgrade

Background

The following is an illustrative case. A food-packaging plant operates three filling and labeling lines. Each line contains PLC-controlled conveyors, drives, checkweighers, vision inspection, and reject stations. Operators report that screens are inconsistent, alarms are excessive, and production reports disagree with local counters.

Problem analysis

The existing system uses mixed serial links and unmanaged Ethernet switches. Some values are polled every second, while high-speed inspection results are transferred through undocumented data blocks. There is no common tag naming convention, clocks differ by several minutes, and a failed switch stops one complete line. Operators acknowledge alarms without guidance because many alarms are configured as simple Boolean changes.

Approach and solution

The engineering team first documented assets, data ownership, process criticality, and required update rates. Local PLCs retained sequencing, permissives, and fault handling. A segmented industrial Ethernet network was installed with managed switches and a redundant path for critical controllers. A supervisory zone was separated from the control zone by a firewall and controlled conduit.

A common tag model was established. Each tag included line, equipment, variable, unit, quality, and timestamp. For example, L02.Filler03.Temperature.PV identified the process value, while L02.Filler03.Temperature.Quality identified its quality state. OPC UA was used to expose structured data to the SCADA and reporting systems; existing field devices were retained through protocol gateways where replacement was uneconomic.

HMI screens were redesigned into overview, line, equipment, and diagnostic levels. Alarm workshops classified each alarm by consequence and operator response. High-rate inspection data remained in the controller or edge database at an appropriate sampling rate, while production summaries were sent to the historian.

Network and server health were monitored. Device clocks were synchronized, backups were tested, and vendor remote access required approval and time-limited credentials.

Outcome

The plant reduced duplicate tags and eliminated disagreements caused by inconsistent scaling. Operators could distinguish a process fault from a communication fault. Redundant network paths improved resilience, while local control continued during temporary SCADA unavailability. Alarm rationalization reduced nuisance alarms and made high-priority events more visible. The project also revealed that data governance—not bandwidth alone—was the principal cause of reporting inconsistency.

Learning points

1. Preserve local control autonomy; supervision should not be a single point of control failure.
2. Define information semantics before selecting an integration protocol.
3. Use alarm rationalization as an engineering activity involving operators.
4. Design redundancy, cybersecurity, and time synchronization together.
5. Treat historians and reports as governed information systems, not passive data sinks.

9. Implementation Method and Best Practices

A disciplined implementation normally proceeds through requirements, asset inventory, network design, protocol selection, tag modeling, HMI and alarm design, simulation or factory acceptance testing, site acceptance testing, commissioning, training, and lifecycle maintenance.

The design should maintain a communication matrix identifying every source, destination, protocol, direction, update rate, data type, criticality, and failure response. Test cases should include disconnected cables, stale data, invalid quality, duplicate addresses, controller restart, server failure, clock loss, alarm floods, and unauthorized commands.

Documentation should include network drawings, IP and device registers, switch configurations, controller backups, tag dictionaries, alarm rationalization records, user roles, software versions, and recovery procedures. Configuration management is essential because undocumented changes eventually become operational risk.

10. Contemporary Developments and Limitations

Edge computing places data processing near machines, reducing latency and bandwidth demand while enabling local analytics. Digital twins can combine control data, equipment models, and maintenance information, but their value depends on trustworthy data and validated models. AI-based anomaly detection may identify patterns that conventional thresholds miss; however, it can produce false positives, conceal changing process conditions, or be difficult to explain. AI should therefore support—not silently replace—validated control and human decision-making.

Time-sensitive networking aims to provide more predictable Ethernet communication through synchronization, scheduling, and traffic management. Wireless industrial networks can simplify installation and support mobile assets, but radio interference, battery life, roaming, and security must be engineered. Cloud connectivity can improve enterprise visibility while increasing dependency on gateways, identity management, and external availability. Neither AI, cloud computing, nor wireless communication removes the need for sound control fundamentals.

Ethical, Regulatory, and Professional Considerations

Automation professionals have duties to protect workers, the public, production assets, and the environment. They must not bypass interlocks merely to meet a production target, conceal data-quality problems, or grant unnecessary remote access. Safety functions require appropriate standards, competence, verification, validation, and change control. Production data may also contain personal information, operator activity records, or commercially sensitive information; collection and retention should be justified and protected.

Professional decisions should be auditable. When a system is uncertain—such as a sensor with bad quality or an unsynchronized clock—the interface should communicate that uncertainty rather than display a misleading precise value.

Chapter Summary

Industrial automation communication links field devices, controllers, supervisory systems, operations platforms, and enterprise applications. Network architecture must address topology, latency, jitter, determinism, availability, maintainability, and cybersecurity. Protocol selection depends on application requirements; no single protocol is universally superior.

PLCs, PACs, RTUs, and DCS controllers perform local control, while HMIs present information and authorized commands. SCADA systems supervise distributed assets, and historians preserve time-stamped values, events, and quality information. Interoperability requires agreement on semantics, data types, units, timestamps, quality, and command behavior. Alarm systems should identify actionable abnormal conditions and be rationalized by consequence and response.

Dependability is improved through autonomy, redundancy, time synchronization, tested recovery, segmentation, least privilege, and lifecycle documentation. Contemporary tools such as OPC UA, edge computing, AI, wireless systems, and time-sensitive networking offer useful capabilities but do not replace disciplined engineering. The central design principle is that communication should make control safer, clearer, more diagnosable, and more dependable.

Key Terms

Industrial Ethernet: Ethernet technology adapted for industrial environments and automation protocols.

PLC: Programmable logic controller used for industrial monitoring and control.

PAC: Programmable automation controller combining PLC functions with broader computing and networking capabilities.

RTU: Remote terminal unit used for distributed monitoring and control.

DCS: Distributed control system organized around networked controllers, operator stations, and engineering systems.

HMI: Human–machine interface through which operators observe and interact with equipment.

SCADA: Supervisory control and data acquisition system for monitoring and supervisory operation.

Historian: Database optimized for time-stamped process, event, and alarm data.

Tag: Named data point representing a measurement, state, command, or calculated value.

Determinism: Predictability of communication timing and response.

Latency: Time required for data to travel and be processed.

Jitter: Variation in communication delay.

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

Deadband: Threshold below which a change may not be recorded or reported.

Alarm rationalization: Structured evaluation of alarm purpose, priority, consequence, and response.

Redundancy: Use of alternate components or paths to maintain service after failure.

OPC UA: Secure, structured, platform-independent industrial interoperability framework.

Network segmentation: Division of a network into controlled zones and communication paths.

Quality status: Metadata indicating whether a data value is valid, uncertain, stale, or bad.

Control autonomy: Ability of local controllers to maintain essential operation without supervisory computers.

Review Questions

Short Answer Questions

1. What is the difference between latency and jitter?
2. Why should a tag include engineering units and quality status?
3. State two differences between polling and event-driven data acquisition.
4. What is the principal purpose of a historian?
5. Why should essential control remain in local controllers rather than in an HMI?

Long Answer Questions

1. Explain a layered industrial network architecture and the information flows between field, control, supervisory, operations, and enterprise levels.
2. Compare Modbus TCP, EtherNet/IP, PROFINET, OPC UA, and MQTT with respect to typical applications and limitations.
3. Describe the design of an effective HMI for a continuous manufacturing process.
4. Explain alarm rationalization, nuisance alarms, alarm floods, and the responsibilities of an operator-facing alarm system.
5. Discuss how redundancy, time synchronization, segmentation, access control, and recovery testing contribute to dependable communication.

Analytical / Application Questions

1. A plant's temperature transmitter produces 12 mA on a linear 4–20 mA, 0–100 °C range. Calculate the temperature and state the assumptions used.
2. Design a communication and HMI strategy for a packaging line containing PLCs, drives, vision inspection, a historian, and a remote maintenance connection. Justify protocol, topology, alarm, and cybersecurity choices.
3. A SCADA screen shows a motor as stopped, but the local starter indicates running. Analyze possible causes and propose a diagnostic sequence that distinguishes stale data, communication failure, incorrect feedback wiring, and actual motor failure.

Multiple Choice Questions with Answers

Question 1

Which device should normally retain essential sequencing and interlock logic when supervisory software fails?

A. Historian

B. Local PLC or controller

C. ERP server

D. HMI display

Correct Answer: B. Local PLC or controller

Question 2

A 4–20 mA transmitter represents 0–100 °C. What temperature corresponds to 12 mA under a linear assumption?

A. 25 °C

B. 40 °C

C. 50 °C

D. 75 °C

Correct Answer: C. 50 °C

Question 3

Which characteristic describes variation in communication delay?

A. Bandwidth

B. Jitter

C. Addressing

D. Scaling

Correct Answer: B. Jitter

Question 4

Which technology is especially associated with structured, platform-independent industrial interoperability?

A. OPC UA

B. Analog current loop only

C. Unmanaged Ethernet

D. Relay wiring

Correct Answer: A. OPC UA

Question 5

What is the main purpose of alarm rationalization?

A. To maximize the number of alarms displayed

B. To remove all operator decisions

C. To define actionable alarms, priorities, consequences, and responses

D. To replace safety systems with HMIs

Correct Answer: C. To define actionable alarms, priorities, consequences, and responses

Question 6

Which item is essential for correct interpretation of exchanged process data?

A. Screen color only

B. Data type, units, scaling, and quality meaning

C. A higher monitor resolution

D. A larger keyboard

Correct Answer: B. Data type, units, scaling, and quality meaning

Question 7

What does a historian primarily preserve?

A. Only user passwords

B. Time-stamped process values, events, and alarms

C. Only network cables

D. Mechanical spare parts

Correct Answer: B. Time-stamped process values, events, and alarms

Question 8

Which practice best supports industrial cybersecurity?

A. Unrestricted vendor access

B. Shared administrator passwords

C. Network segmentation and least privilege

D. Disabling all event logs

Correct Answer: C. Network segmentation and least privilege

Question 9

For a message of L bits transmitted at R bit/s, which expression gives serialization time?

A. L × R

B. R/L

C. L/R

D. L + R

Correct Answer: C. L/R

Question 10

Why is a start command with no running feedback poor HMI practice?

A. It uses too many colors

B. It cannot confirm that the commanded physical action occurred

C. It prevents all PLC scans

D. It automatically creates redundancy

Correct Answer: B. It cannot confirm that the commanded physical action occurred