Industry 4.0 and smart manufacturing

In short

Industry 4.0 is the fourth industrial revolution: connecting machines, products and people through cyber-physical systems so that a factory senses its own state, decides and reconfigures itself with minimal human intervention. Smart manufacturing is the application side - using that connectivity plus analytics, digital twins and robotics to raise OEE, cut unplanned downtime and allow mass customisation down to a batch of one. It is examined in Mechatronics "Industry 4.0 / Automation" papers, appears in GATE Instrumentation and Production & Industrial sections as automation hierarchy, sensors, networks and reliability, and is a standing interview subject at Siemens, Bosch, ABB, Rockwell, Honeywell, Schneider, L&T, Tata Motors, Maruti, TCS/Infosys/LTTS digital-manufacturing practices and Tata Steel Kalinganagar (a WEF Global Lighthouse plant). The Government of India drives it through the SAMARTH Udyog Bharat 4.0 centres (SAMARTH stands for Smart Advanced Manufacturing and Rapid Transformation Hub — one Department of Heavy Industry initiative, not two), PLI schemes and the National Manufacturing Mission announced in the Union Budget 2025-26.

Explained

The four revolutions and what defines the fourth

• Industry 1.0 (c. 1784): water and steam power, mechanised looms. Industry 2.0 (c. 1870): electricity, assembly line, mass production. Industry 3.0 (c. 1969): electronics, PLCs, CNC, robots - automation of individual machines.
• Industry 4.0 (term coined at Hannover Messe, 2011): networking of those automated machines into cyber-physical systems (CPS) - a physical process with an embedded computing and communication layer that monitors and controls it through feedback.
• The jump from 3.0 to 4.0 is not more automation but horizontal and vertical integration: horizontal across the supply chain (supplier → plant → customer), vertical from sensor to ERP, plus end-to-end digital engineering across the product life cycle.
• Design principles (examined as a list): interoperability, information transparency (a virtual copy of the plant), technical assistance (operator support, AR), decentralised decisions, modularity, real-time capability, service orientation.
• Industry 5.0 (European Commission framing, 2021) adds human-centricity, sustainability and resilience - the robot collaborates with the human rather than replacing them; expect a one-mark comparison question.

The automation pyramid and ISA-95

1. Level 0 - field: sensors and actuators (proximity, encoder, load cell, RTD, thermocouple, VFD, valve). Smart sensors add self-diagnostics and digital output (IO-Link).
2. Level 1 - control: PLC, PAC, CNC, drives, safety relays. Scan-cycle determinism in milliseconds.
3. Level 2 - supervisory: SCADA and HMI - visualisation, alarms, historian, setpoints.
4. Level 3 - MES/MOM: manufacturing execution - work order dispatch, routing, genealogy and traceability, quality, OEE, labour and tool management. Horizon: shift to minutes.
5. Level 4 - ERP/business: SAP, Oracle - planning, purchase, costing, dispatch. Horizon: days to months.
6. ISA-95 / IEC 62264 defines these levels and the B2MML data exchange between Level 3 and 4; ISA-88 / IEC 61512 does the same for batch recipes.
7. Industry 4.0 flattens this pyramid: a field device can publish straight to a cloud analytics service, giving the "automation pillar" or RAMI 4.0 three-axis model (hierarchy levels × life-cycle & value stream × layers: asset, integration, communication, information, functional, business).

Core technologies (the nine pillars)

• Industrial IoT: instrumenting assets and publishing data; retrofit IoT gateways on legacy machines read from PLC registers or clamp-on current sensors.
• Cloud and edge/fog computing: cloud for storage, model training and fleet comparison; edge for low-latency control, bandwidth saving and data sovereignty - rule of thumb, anything needing a response under ~100 ms stays at the edge.
• Big data and analytics: descriptive (what happened) → diagnostic → predictive (when will it fail) → prescriptive (what to do).
• AI/ML on the shop floor: vision-based defect detection, anomaly detection on vibration, soft sensors, process-parameter optimisation, energy forecasting.
• Digital twin: a live, data-connected model of an asset, process or whole plant. Digital model (no automatic data flow) → digital shadow (one-way, physical to virtual) → digital twin (two-way, the model can act back). Types: product, production and performance twins.
• Additive manufacturing: tooling, jigs, spares on demand, lattice and topology-optimised parts; shortens prototype loops from weeks to days.
• Collaborative and mobile robots: cobots rated by ISO/TS 15066 power-and-force limits; AGVs follow a fixed guide path, AMRs localise and plan their own path (SLAM).
• Augmented and virtual reality: AR work instructions, remote expert support, VR operator training.
• Cybersecurity: OT security per IEC 62443 - zones and conduits, security levels SL1-SL4; the CIA priority inverts in OT to availability first, then integrity, then confidentiality.
• Add-ons often asked: blockchain for supply-chain provenance, 5G private networks for wireless control, simulation.

Communication, protocols and data models

• Fieldbus/industrial Ethernet: Modbus RTU/TCP (simple, register-based), PROFIBUS/PROFINET, EtherCAT, EtherNet/IP, CAN and CANopen, HART and IO-Link at sensor level.
• OPC UA (IEC 62541): the backbone standard of Industry 4.0 - platform-independent, service-oriented, carries an information model (semantics, not just bytes), secure by design (certificates, encryption); pub/sub mode plus OPC UA over TSN for deterministic traffic.
• MQTT: lightweight publish/subscribe broker-based transport for telemetry over unreliable links; Sparkplug B adds a payload and state-management spec. MQTT vs OPC UA is a favourite interview contrast: MQTT is a transport, OPC UA also standardises meaning.
• TSN (IEEE 802.1 set): time-aware scheduling and synchronisation (IEEE 1588 PTP) so control traffic and IT traffic share one Ethernet.
• MTConnect: read-only semantic standard for machine-tool data, widely used for CNC monitoring.
• Unified Namespace (UNS): a single real-time, hierarchically structured broker view of the whole plant - everything publishes, everything subscribes, which is what actually flattens the pyramid in practice.
• Interface layering: ISO/OSI reference model, and IEC 61131-3 for PLC programming languages (LD, FBD, SFC, ST, IL) plus IEC 61499 for distributed function blocks.

Smart manufacturing practice: metrics and methods

• OEE = Availability × Performance × Quality - the single most asked metric; world class is taken as ~85%, Indian discrete-shop reality is often 45-65%.
• The six big losses: breakdowns and setup/adjustment (availability); idling/minor stops and reduced speed (performance); process defects and start-up/yield losses (quality).
• Maintenance ladder: reactive → preventive (time/usage based) → condition-based (CBM, threshold on vibration, temperature, oil) → predictive (model forecasts remaining useful life) → prescriptive. Reliability terms: MTBF, MTTR, availability = MTBF/(MTBF+MTTR).
• Condition monitoring signals: vibration (ISO 10816 / ISO 20816 velocity severity zones A-D), motor current signature, acoustic emission, thermography, oil debris.
• Quality 4.0: inline machine vision, SPC with automatic control-chart alarms, automated gauging and CMM feedback, full digital traceability by serial/batch (mandatory in pharma and automotive recalls).
• Lean plus 4.0: 4.0 does not replace lean - digital kanban, e-andon, automated value-stream mapping; digitising waste only makes waste visible faster.
• Flexible and reconfigurable manufacturing systems (FMS/RMS): machine flexibility, routing flexibility, volume and product flexibility; mass customisation and lot-size-of-one are 4.0's commercial promise.
• Implementation path: (1) business-case and KPI baseline, (2) maturity assessment, (3) connect and collect (OT data layer), (4) visualise (dashboards, OEE), (5) analyse and predict, (6) close the loop/autonomy, (7) scale from pilot to plant - avoiding "pilot purgatory", where dozens of proofs of concept never scale.
• Maturity models: Acatech Industrie 4.0 Maturity Index (computerisation → connectivity → visibility → transparency → predictive capacity → adaptability); SIRI (Smart Industry Readiness Index, Singapore-origin, used by many Indian assessors); the Samarth Udyog centres offer free demo and assessment for MSMEs.

Indian context: policy, standards and constraints

• Samarth Udyog Bharat 4.0 - DHI's network of common engineering facility/demo centres (e.g. C4i4 Pune, I4.0 India at IISc Bengaluru, CMTI Bengaluru, IITD-AIA Foundation for Smart Manufacturing, PTC/IISc centres) for MSME awareness and pilots.
• Make in India, PLI schemes (auto, ACC batteries, electronics, drones), India Semiconductor Mission, and the National Manufacturing Mission announced in Budget 2025-26 to support clean-tech and deep-tech manufacturing.
• Standards bodies: BIS (LITD/ETD committees adopt IEC automation standards as IS/IEC), and Indian adoptions such as IS/ISO 9001 (QMS), IS/ISO 50001 (energy management), IS/ISO 45001 (OH&S), IS/IEC 27001 (information security). Quote specific automation IS numbers only after checking the BIS catalogue - many are direct adoptions and get revised.
• Law in force: the four Labour Codes (Code on Wages; Industrial Relations; Social Security; Occupational Safety, Health and Working Conditions) are in force from 21 November 2025 and replaced the Factories Act, Minimum Wages Act, Industrial Disputes Act and BOCW Act - so factory safety, hazardous-process and worker-consultation duties for an automated line now come from the OSH&WC Code and its rules, not the Factories Act, 1948.
• Data and cyber: the Digital Personal Data Protection Act, 2023 governs personal data (worker tracking, CCTV analytics); CERT-In directions require incident reporting within 6 hours and log retention - relevant when OT networks are breached. Rules under the DPDP Act were notified in 2025 with phased commencement, so check the current compliance date before quoting one.
• GST: capital goods and most automation equipment attract GST (slabs since 22 September 2025 are 5% / 18% / 40%, with 18% the usual rate for machinery and industrial electronics); input tax credit on plant and machinery makes the post-tax cost of a retrofit lower than the invoice value for a registered manufacturer.
• Real constraints in India: legacy machines with no digital port, power quality and unreliable connectivity, scarcity of OT-plus-IT skills, data-ownership disputes with OEMs, and MSME capital cost - which is why brownfield retrofit, not greenfield, is the dominant Indian project type.

Key points

• Industry 4.0 - networking of automated systems into cyber-physical systems with vertical, horizontal and end-to-end integration.
• Cyber-physical system - physical process + embedded computation + communication, closing the loop through a virtual representation.
• Nine pillars - IIoT, cloud, big data/analytics, autonomous robots, simulation/digital twin, additive manufacturing, AR, cybersecurity, system integration.
• Automation pyramid (ISA-95/IEC 62264) - Level 0 field, 1 control, 2 SCADA, 3 MES, 4 ERP; 4.0 flattens it.
• SCADA vs MES vs ERP - SCADA controls and visualises in seconds, MES executes and tracks orders in minutes, ERP plans in days.
• OPC UA (IEC 62541) - vendor-neutral, semantic, secure interoperability standard; with TSN it becomes deterministic.
• MQTT - broker-based lightweight pub/sub transport; Sparkplug B adds payload semantics.
• Digital twin - bidirectional live model; distinguish model (manual), shadow (one-way), twin (two-way).
• Edge vs cloud - edge for latency, bandwidth and privacy; cloud for scale, storage and training.
• OEE = A × P × Q - the headline smart-manufacturing KPI; six big losses map onto its three factors.
• Predictive maintenance - forecasts remaining useful life from condition data; sits above condition-based monitoring.
• Availability = MTBF/(MTBF + MTTR) - the reliability identity behind the availability factor of OEE.
• IEC 62443 - OT cybersecurity: zones, conduits, security levels; availability outranks confidentiality in OT.
• ISO/TS 15066 - collaborative-robot power and force limits; ISO 10218 covers industrial robot safety.
• RAMI 4.0 - three-axis reference architecture (hierarchy, life cycle, layers) plus the asset administration shell.
• Mass customisation - economic lot size of one, enabled by flexible/reconfigurable systems and digital routing.
• Indian push - Samarth Udyog Bharat 4.0 centres, PLI schemes, National Manufacturing Mission (Budget 2025-26); brownfield retrofit dominates.
• Labour Codes from 21 Nov 2025 - OSH&WC Code now governs factory safety duties for automated lines, replacing the Factories Act, 1948.

Formulas and examples

Example 1 - OEE of a CNC cell. A shift is 8 h with 30 min planned break; recorded downtime 45 min. Ideal cycle time 30 s/part. Total parts 700, rejects 21.

• Planned production time = 480 − 30 = 450 min; Run time = 450 − 45 = 405 min.
• A = 405/450 = 0.90.
• P = (30 s × 700)/(405 × 60 s) = 21,000/24,300 = 0.8642.
• Good count = 700 − 21 = 679 → Q = 679/700 = 0.97.
• OEE = 0.90 × 0.8642 × 0.97 = 0.7544 ≈ 75.4%.
• Cross-check: (30 × 679)/(450 × 60) = 20,370/27,000 = 0.7544 ✓.
• Loss: the 13.6% performance loss (minor stops/reduced speed) is the biggest single bucket - typically the first target for IIoT monitoring.

Example 2 - predictive maintenance business case. A press line suffers 10 unplanned breakdowns a year, average 6 h each; contribution lost ₹25,000 per hour. An IIoT vibration-and-current monitoring retrofit costs ₹12,00,000 plus ₹1,50,000 a year for cloud and support, and is expected to prevent 70% of breakdown hours.

• Present loss = 10 × 6 × 25,000 = ₹15,00,000/year.
• Saving = 0.70 × 15,00,000 = ₹10,50,000/year; net of operating cost = 10,50,000 − 1,50,000 = ₹9,00,000/year.
• Payback = 12,00,000 / 9,00,000 = 1.33 years ≈ 16 months.
• Simple annual ROI = 9,00,000/12,00,000 = 75%.

Example 3 - availability and reliability. A machine logs 2,400 operating hours with 8 failures; total repair time 40 h.

• MTBF = 2,400/8 = 300 h; MTTR = 40/8 = 5 h.
• Availability = 300/(300+5) = 0.9836 (98.4%).
• λ = 1/300 = 3.33 × 10⁻³ /h → probability of surviving a 100 h campaign R = e^−100/300 = e^−0.3333 = 0.7165 (71.7%).

Example 4 - edge vs cloud data sizing. A plant has 500 tags sampled at 10 Hz, 8 bytes per sample (timestamp + float, compressed).

• Raw rate = 500 × 10 × 8 = 40,000 B/s = 320 kbit/s; per day = 40,000 × 86,400 = 3.456 × 10⁹ B ≈ 3.46 GB/day (≈ 1.26 TB/year).
• Edge aggregation to 1-second averages cuts it 10× to ≈ 0.35 GB/day, but destroys the high-frequency content needed for bearing diagnostics - so keep raw data at the edge for vibration and send only features (RMS, kurtosis, band energies) upstream.

Example 5 - takt, WIP and cycle time. An assembly line runs 7.5 effective hours per shift, two shifts, to meet demand of 900 units/day.

• Takt = (2 × 7.5 × 3,600)/900 = 54,000/900 = 60 s/unit.
• If the measured throughput is 15 units/h per station chain and total lead time through the line is 4 h, WIP = 15 × 4 = 60 units on the line (Little's Law).
• Any station whose cycle time exceeds 60 s is the bottleneck; smart-manufacturing station-level cycle logging finds it automatically instead of by stopwatch study.

Where it is used

• Automotive and auto components: Maruti Suzuki, Tata Motors, Mahindra, Bosch and Hero plants run MES-driven traceability, AGV/AMR material movement, inline vision and torque-tool data logging for recall-proof genealogy.
• Steel and process industry: Tata Steel's Kalinganagar and Jamshedpur and JSW units use plant-wide digital twins, ML-based blast-furnace and energy optimisation; WEF Global Lighthouse recognition has gone to Indian plants in steel and consumer goods.
• Electronics, EV and battery manufacturing: PLI- and ISM-backed plants (cells, modules, SMT lines) built greenfield with full MES, SPC and traceability from day one.
• Pharma and FMCG: serialisation and track-and-trace, electronic batch records, GMP-compliant data integrity (21 CFR Part 11-style audit trails) and vision inspection on packing lines.
• Job roles: Industry 4.0/automation engineer, controls and PLC-SCADA engineer, MES or MOM consultant (SAP ME/MII, Siemens Opcenter, Rockwell FactoryTalk), IIoT solution architect, digital-twin/simulation engineer, OT cybersecurity analyst, reliability and condition-monitoring engineer, data engineer for manufacturing analytics - hiring at Siemens, Bosch, ABB, Rockwell, Honeywell, Schneider, Emerson, L&T Technology Services, Tata Elxsi, LTIMindtree, TCS and Cognizant.
• MSME retrofit and energy: clamp-on energy meters and IoT gateways for ISO 50001 energy management, machine-hour billing and PM-level production visibility in Pune, Coimbatore, Rajkot and Ludhiana clusters, often via Samarth Udyog centres.

Interview questions

• Q: How is Industry 4.0 different from Industry 3.0? A: 3.0 automated individual machines with electronics, PLCs and robots; 4.0 networks them into cyber-physical systems with vertical, horizontal and end-to-end integration, so decisions are taken on live data from the whole value chain.
• Q: What exactly is a cyber-physical system? A: A physical process tightly coupled to embedded computation and networking, so the system senses its own state, computes against a virtual representation and acts back on the physical world in closed loop.
• Q: Differentiate SCADA, MES and ERP. A: SCADA supervises and visualises the process in seconds, MES executes and tracks the production order on the shop floor in minutes (routing, traceability, OEE), and ERP plans materials, costs and dispatch over days - formalised as ISA-95 Levels 2, 3 and 4.
• Q: Why is OPC UA called the backbone of Industry 4.0 when MQTT is lighter? A: MQTT only transports bytes; OPC UA carries a standard information model so a consumer understands what a value means, with built-in security and, over TSN, determinism.
• Q: When would you compute at the edge instead of the cloud? A: When latency must be below roughly 100 ms, when bandwidth or connectivity is limited, when high-frequency raw data would be too costly to ship, or when data must not leave the plant for policy or regulatory reasons.
• Q: A digital twin versus a simulation model? A: A simulation is a standalone model run with assumed inputs; a digital twin is continuously fed by live asset data and can push decisions back, so it tracks the specific unit's current condition.
• Q: Plant OEE is 55%. How do you find out where it is lost? A: Break it into availability, performance and quality from time-stamped machine data, map losses onto the six big losses, and attack the largest - usually minor stops and changeover, not breakdowns.
• Q: Difference between condition-based and predictive maintenance? A: CBM triggers an action when a measured parameter crosses a threshold; predictive maintenance uses a model and trend to estimate remaining useful life and schedule the intervention before the threshold is reached.
• Q: Why does OT cybersecurity invert the usual CIA priority? A: Because a stopped or unsafe plant costs more than a leaked tag value - availability and integrity of control come first, and patching windows are rare, so IEC 62443 zones, conduits and network segmentation do the protecting.
• Q: How would you make a 20-year-old lathe with no network port "smart"? A: Retrofit non-invasively - current clamp or vibration sensor plus an IoT gateway, or read the existing PLC/relay signals through a digital-input module - to get run/idle/stop status and cycle counts first, then build OEE before attempting process-level data.
• Q: What is "pilot purgatory" and how do you avoid it? A: Dozens of proofs of concept that never scale; avoided by starting from a measured KPI baseline and business case, standardising the data layer and architecture before the use case, and planning rollout and ownership from the start.
• Q: Does Industry 4.0 make lean manufacturing obsolete? A: No - digitisation only makes waste visible faster; the process should be stabilised and the waste removed first, otherwise you automate and measure a bad process.
• Q: Which Indian law now governs safety on a highly automated line? A: The Occupational Safety, Health and Working Conditions Code, 2020 and its rules, in force from 21 November 2025, which replaced the Factories Act, 1948; machine guarding and cobot risk assessment also follow ISO 10218 and ISO/TS 15066.

Common mistakes

• Treating Industry 4.0 as simply "more robots" or "a dashboard" - the examinable core is integration, interoperability and closed-loop decisions, not hardware count.
• Mixing up the ISA-95 levels, especially claiming MES does real-time control or that SCADA schedules orders.
• Computing OEE wrongly: using total calendar time instead of planned production time (that is TEEP), or multiplying percentages already expressed as percentages and getting a figure above 100%.
• Forgetting unit consistency - ideal cycle time in seconds against run time in minutes is the most common OEE arithmetic slip.
• Calling any 3D CAD model a digital twin; without a live, automatic data link from the physical asset it is a model or at best a shadow.
• Assuming cloud-first: streaming raw 10 kHz vibration data to the cloud is unaffordable and unnecessary when features can be extracted at the edge.
• Quoting the Factories Act, 1948 or old IS numbers in answers - the Labour Codes replaced it from 21 November 2025, and many IS automation standards have been revised as IS/IEC adoptions.
• Ignoring change management and operator training in a project answer; most failed Indian 4.0 pilots fail on people, data ownership and connectivity, not algorithms.

Related topics

Mechatronic system design · Internet of Things basics · PLC programming and industrial automation · Sensors, transducers and signal conditioning · Robotics and automation · CNC machines and computer-aided manufacturing