SCADA and HMI

Why it matters

A PLC controls a machine; SCADA lets people supervise a whole plant or a network spread over hundreds of kilometres — water pumping stations, power substations, pipelines, wind farms. The HMI is the operator's window into the process: how quickly an operator spots and handles an abnormal situation depends on how the screens, alarms and trends are designed. Mechatronics graduates frequently configure HMI screens, tag databases and alarm lists in their first jobs.

Key ideas

Where SCADA sits (automation pyramid, ISA-95 levels).

• Level 0: field devices — sensors, transmitters, valves, drives.
• Level 1: control — PLCs, RTUs, controllers executing fast control loops and interlocks.
• Level 2: supervisory — SCADA servers and HMIs: monitoring, operator commands, alarms, trends.
• Level 3: MES (production scheduling, quality, tracking). Level 4: ERP (business planning). The essential point: control and safety stay at Level 1. SCADA supervises — it sends set-points and commands, but the plant must remain safe if SCADA or the network is lost.

SCADA components.

• RTUs (remote terminal units) and PLCs at field sites: read I/O, run local logic, buffer data, communicate over long, slow or intermittent links.
• Communication: leased line, radio, cellular, fibre, satellite; protocols such as Modbus RTU/TCP, DNP3 and IEC 60870-5-101/104 (utilities), IEC 61850 (substations) and OPC UA (vendor-neutral, secure data exchange).
• Master station / SCADA server: polls or receives data, maintains the real-time tag database (one tag per measured or commanded variable, with value, timestamp and quality), and runs the alarm server.
• Historian: time-series database storing tag history for trends, reports and analysis; deadband and compression reduce storage.
• HMI/operator clients: graphics, faceplates, trends, alarm lists, reports.

Data acquisition modes. Polling: the master asks each station in turn — simple and deterministic, but the update time grows with the number of stations and the link speed. Report by exception: stations send only changed values (beyond a deadband), with periodic integrity polls — far more efficient on slow links (supported by DNP3, IEC 104, OPC UA subscriptions).

HMI.

• Can be a panel on the machine (local HMI, touch panel) or a PC-based SCADA client.
• Functions: process overview and detail screens, faceplates for motors/valves/PID loops, set-point entry with limits, alarm display and acknowledgement, trends, recipes, user log-in with access levels, audit trail.
• Good HMI design (ISA-101 "high-performance HMI"): grey, low-contrast backgrounds; colour reserved for abnormal conditions and alarms; a clear display hierarchy (plant overview → area → unit → detail); analog values shown with their normal ranges; consistent symbols. Red/green alone is a poor choice — about 8 % of men have red–green colour deficiency.

Alarm management (ISA-18.2, EEMUA 191). An alarm should require an operator action. Every alarm has a priority, a set-point, a deadband (hysteresis) and sometimes an on-delay to suppress chattering. Alarm floods — tens of alarms in minutes — hide the important one; guidance targets roughly one alarm per 10 minutes per operator in steady operation. Alarms are acknowledged by the operator and logged with timestamps.

SCADA vs DCS. SCADA: geographically distributed, event/data-acquisition oriented, tolerant of slow links, control mostly local in RTUs/PLCs. DCS (distributed control system): one plant site (refinery, power station), tightly integrated control, engineering and HMI from one vendor database, fast continuous control. The distinction is blurring.

Cybersecurity. Older protocols (Modbus, DNP3 without secure authentication) have no authentication. Defences per IEC 62443: network segmentation into zones and conduits, firewalls and a DMZ between plant and office networks, no direct internet exposure, user accounts and least privilege, patching, secure remote access, and monitoring. Stuxnet showed that control systems are real targets.

Formulas

t_poll = Σ_k (t_req,k + t_resp,k + t_turn,k)

• Polling cycle (s) over all stations k: request and response transmission times plus turnaround/processing delays.

t_frame = n_bytes × n_bits/char / baud

• Transmission time (s) of a serial frame; n_bits/char = 11 for Modbus RTU (start, 8 data, parity, stop).

Modbus RTU read-registers: request = 8 bytes, response = 5 + 2N bytes

• N = number of 16-bit registers read (address, function, byte count, data, 2-byte CRC).

Storage = N_tags × (bytes per sample) × (samples per second) × t

• Historian storage (bytes) over time t (s), before compression.

R_bits = N_points × bits per point / T_update

• Required raw data rate (bit/s).

Alarm active when x > SP; clears when x < SP − DB

• High-alarm hysteresis; SP = alarm set-point, DB = deadband (same units as x).

Worked examples

Example 1 (standard). A historian logs 2000 tags once per second, each sample (value, timestamp, quality) occupying 16 bytes. Find the daily storage, and the annual storage if deadband recording cuts the samples stored by 80 %.

1. Storage/day = 2000 × 16 × 1 × 86 400 = 2.765 × 10⁹ bytes ≈ 2.76 GB.
2. With 80 % reduction: 0.2 × 2.765 GB = 0.553 GB per day.
3. Per year: 0.553 × 365 = 202 GB ≈ 0.20 TB.

Answer: ≈ 2.76 GB/day raw; ≈ 0.55 GB/day and ≈ 0.20 TB/year with deadband.

Example 2 (GATE level). A SCADA master polls 20 RTUs on a 9600-baud Modbus RTU radio link (11 bits per character). Each poll reads 10 holding registers; each RTU's turnaround and radio key-up delay is 20 ms. Find the time per poll and the full scan time. A tank-level high alarm has set-point 4.5 m and deadband 0.2 m; at what level does it clear?

1. Request = 8 bytes; response = 5 + 2 × 10 = 25 bytes; total 33 bytes.
2. Bits = 33 × 11 = 363 bits → t = 363 / 9600 = 37.8 ms.
3. Per RTU = 37.8 + 20 = 57.8 ms.
4. t_poll = 20 × 57.8 ms = 1.156 s (the 3.5-character silent gaps, about 4 ms each, add a little more).
5. Alarm clears when the level falls below 4.5 − 0.2 = 4.3 m.

Answers: ≈ 57.8 ms per RTU, ≈ 1.16 s per full scan; alarm clears below 4.3 m. Report-by-exception or a faster link would be needed if the update must be well under a second.

Common mistakes

• Putting interlocks or safety functions in the SCADA/HMI layer instead of the PLC or safety system.
• Forgetting protocol overhead (address, function code, CRC, start/stop/parity bits) when estimating update times.
• Alarms without deadband or delay, so a noisy signal "chatters" and floods the operator.
• Using bright, colourful HMI screens where nothing stands out when something goes wrong.
• Treating SCADA networks as isolated and skipping security — most have some path to the outside.
• Confusing an HMI (operator interface) with SCADA (the whole supervisory system).

For GATE ME

Questions here are mainly conceptual MCQs — SCADA components and functions, levels of the automation hierarchy, polling vs report by exception, SCADA vs DCS, HMI features — with occasional simple numericals on polling cycles, data rates, number of samples logged, or communication time. Practise counting bytes and bits per frame and keeping track of units (bits vs bytes, ms vs s).

Quick check

1. At which automation level does the PLC run interlocks?
2. A tag is logged every 5 s. How many samples per hour?
3. How many bytes are in a Modbus RTU response reading 4 registers?
4. Name one advantage of report by exception over polling.
5. Why should an alarm have a deadband?

Answers: 1. Level 1 (control) 2. 720 3. 13 4. Much less traffic on slow links and faster reporting of changes 5. To stop it chattering on and off when the value hovers near the set-point