PLC, HMI and SCADA occupy different layers of an industrial control system. A PLC reads field inputs, executes defined control logic and commands outputs. An HMI gives operators a local interface for status, commands, alarms and setpoints. SCADA provides supervisory visibility across controllers or areas and can add central alarms, trends, historian, events, reporting and user permissions.
The terms are related but not interchangeable. A clear architecture keeps time-critical control and required interlocks in the appropriate local layer, gives operators usable interfaces and sends only justified data and commands to supervisory systems.
Start at the field layer
Sensors, switches, transmitters, encoders, drives, valves, motors and protective devices create the real-world inputs and outputs. Their ranges, signal types, power supplies, isolation, fail states and calibration need definition before software design. A control system cannot correct an unreliable instrument or an undocumented field circuit.
The I/O list should name each point, engineering unit, normal state, alarm limits, scaling, update behavior and quality indication. The electrical drawings, process description and cause-and-effect requirements should agree with that list.
PLC: deterministic local control
A programmable logic controller executes sequences, permissives, interlocks, timers, calculations and equipment commands. It can coordinate conveyors, pumps, valves, drives and production steps. Required control should remain understandable, testable and maintainable by the site team.
PLC design should define scan and response needs, I/O capacity, spare capacity, program structure, manual/automatic modes, fault handling, restart behavior and backup. Protection and safety functions require their own approved architecture; they should not be casually mixed with convenience logic.
HMI: the operator’s working interface
An HMI presents the machine or process state to an operator. Useful screens show what is running, stopped, inhibited or in alarm; provide authorized commands; explain permissives; and guide recovery without overwhelming the operator. A screen full of animated equipment is not automatically usable.
Good HMI design establishes navigation, consistent symbols, alarm priority, units, setpoint limits, confirmation for consequential commands and role-based access. It should also state what happens when the HMI fails. The PLC and required equipment protection should not depend on a decorative screen remaining available.
SCADA: supervisory coordination and data
SCADA becomes useful when operations need visibility across multiple machines, lines, utilities or sites. It can consolidate alarms, trends, events and production or utility data; manage authorized supervisory commands; and provide historian and reporting functions. Its boundary should follow an operating requirement, not the assumption that every PLC needs a central server.
| Layer | Primary role | Typical users | Key design question |
|---|---|---|---|
| Field devices | Measure and act | Operators and maintenance | Are signals and fail states defined? |
| PLC | Execute local control logic | Controls engineers and technicians | Which functions must continue locally? |
| HMI | Local operator interaction | Machine or process operators | Can the operator diagnose and act safely? |
| SCADA | Supervise multiple assets and data | Control room, engineering and management | Which alarms, trends and commands belong centrally? |
| Historian/reporting | Retain and analyze time-based records | Engineering, quality and operations | Which data, resolution and retention are justified? |
PLC-only, PLC plus HMI, or SCADA?
A contained machine may need only a PLC and physical controls when operation is simple and data retention is limited. A PLC plus HMI suits equipment that needs recipes, guided operation, alarm acknowledgement or clearer diagnostics. SCADA is justified when the site needs central coordination, broader alarm management, historian functions, reporting or authorized supervisory control across several systems.
The decision should consider availability and support. Adding a supervisory layer creates servers, licenses, backups, user administration, network interfaces and maintenance responsibilities. Those additions can be valuable, but only when the operating case is clear.
Protocols are interfaces, not proof of architecture
OPC UA, Modbus and Profinet are general examples of industrial communication technologies. Selection follows the installed devices, supported functions, topology, timing, data volume, diagnostic needs, vendor support and approved security architecture. A protocol name in a proposal does not prove interoperability.
Before integration, prepare an interface matrix showing each device, port, protocol, address, data owner, required tags, command permissions, update rate and failure behavior. Validate a representative connection and document scaling, timestamps and quality flags. Gateways can bridge incompatible systems, but they also create another component to configure, secure, back up and support.
Resilience and communication loss
The design must say what happens when an HMI, SCADA server, network switch, controller or data link becomes unavailable. Required local sequences may continue, hold or move to a defined state depending on the process. Operators need clear indication of lost communications, and the system should avoid treating stale data as current.
Redundancy is not automatically required everywhere. It should be selected for the process availability target, failure consequence, recovery time and maintenance plan. Testing should include communication loss, power restoration, controller restart, backup restoration and local operation without the supervisory layer where that behavior is required.
Alarm, historian and user-management boundaries
Alarm design should define priority, cause, consequence, operator response and shelving or acknowledgement rules. Excessive low-value alarms can hide important events. Historian design should define which tags are stored, sample or exception rules, timestamp source, retention and access. User roles should follow operating responsibilities, with changes and remote actions logged where required.
Cybersecurity is a lifecycle responsibility shared across operations, engineering and IT. Asset inventory, network separation, restricted accounts, approved remote access, backups, update responsibilities and recovery procedures should be agreed for the actual site. No single product or protocol makes the architecture secure by itself.
Retrofit workflow for existing equipment
- Survey controllers, panels, field devices, networks and software versions.
- Collect current drawings, programs, backups and alarm lists.
- Identify unsupported components and undocumented interfaces.
- Define the operational problem and measurable acceptance criteria.
- Decide which functions remain local and which become supervisory.
- Develop the I/O, tag, interface and user-role matrices.
- Plan cutover, rollback and production-shutdown constraints.
- Factory-test sequences and representative interfaces.
- Site-test interlocks, alarms, failure modes and restoration.
- Deliver updated drawings, backups, training and support responsibilities.
Verified project evidence: Zhejiang
The Zhejiang smart logistics sorting line is a completed project with a published capacity of 20,000 bags/hour. Its verified public scope describes smart logistics, automated sorting, intelligent conveying and material handling.
That project evidence does not state that the site used OPC UA, Modbus, Profinet, AI, machine learning, predictive analytics or a specific cybersecurity architecture. This article therefore does not attribute any of those technologies to Zhejiang. Protocols and control features should be selected and verified for each project.
Buyer checklist
- Process description, operating modes and cause-and-effect requirements.
- Controller, HMI, drive, instrument and network inventory.
- Current programs, backups, drawings and licenses.
- I/O list, alarm list, historian/tag list and reporting needs.
- Local control, supervisory control and safety boundaries.
- User roles, remote-access policy and responsibility matrix.
- Failure behavior, redundancy need, backup and recovery tests.
- Shutdown window, cutover sequence and rollback plan.
- Training, documentation, spares and lifecycle support.
See PLC / SCADA & Industrial Control for the core service scope. Browse Insights, including factory retrofit versus replacement and how factory automation can affect energy and operating cost.
Functional design specification before programming
A functional design specification translates the process into control behavior. It should define operating modes, sequences, permissives, interlocks, alarms, setpoint ownership, restart rules and manual overrides. It should also identify which actions are local, which are supervisory and which require separate safety functions. Programming before this agreement often creates late changes and inconsistent operator expectations.
For brownfield work, the specification should record what the existing system already does, including undocumented operator practices. A retrofit that changes a familiar sequence needs training, cutover planning and a verified fallback.
Factory and site acceptance testing
Factory acceptance testing can exercise logic, screens, alarms and simulated interfaces before installation. Site acceptance testing confirms real wiring, instruments, drives, networks and process response. The test plan should include normal sequences, invalid commands, sensor faults, communication loss, power restoration and backup recovery.
Every test needs an expected result and evidence record. Open items should be assigned and retested. Final handover should include source files, compiled applications where applicable, device configurations, licenses, network drawings, I/O lists, alarm lists and backups in a form the owner can maintain.
Frequently asked questions
What does a PLC do?
A PLC reads defined inputs, executes control logic and commands outputs for equipment or a process.
What does an HMI do?
An HMI gives operators a local or machine-level view of status, commands, alarms, setpoints and procedures.
What does SCADA do?
SCADA supervises one or more control systems and can provide central visualization, alarms, trends, events, historian and reporting.
Can an HMI replace a PLC?
No. An HMI is an operator interface; required deterministic control and interlocks should reside in the approved control layer.
When can PLC-only control be sufficient?
For a contained machine or process where local controls, diagnostics and data needs are limited and explicitly defined.
What is a historian?
A historian stores time-based process values, states, alarms or events so authorized users can analyze operation and trace incidents.
How should communication protocols be selected?
Choose them from the actual device inventory, supported interfaces, performance, maintainability and approved network architecture.
What should happen if SCADA communication fails?
The approved local control and protection strategy should define continued operation, operator visibility, alarms and safe-state behavior.
Should safety functions depend on SCADA?
Safety-related functions should follow the approved safety architecture and should not be assumed to depend on a general supervisory workstation.
Does the Zhejiang project prove specific protocols, AI or cybersecurity architecture?
No. Its published evidence verifies application, capacity and completion status, not those specific technologies.