Smart Buildings, Industry, and Infrastructure Applications

Introduction

The Internet of Things (IoT) becomes most valuable when connected devices are integrated with physical processes, operational decisions, and measurable performance objectives. In a smart building, sensors and controllers coordinate lighting, heating, ventilation, air conditioning (HVAC), security, and energy storage. In industry, connected machines support automation, quality control, predictive maintenance, and flexible production. In infrastructure, IoT enables monitoring and optimization of water networks, farms, transport systems, and urban services.

These applications differ in their physical processes, safety requirements, communication needs, and economic objectives. A temperature sensor in an office may tolerate a delay of several seconds, whereas a protection signal in an industrial motor-control system may require deterministic millisecond-level performance. A successful design therefore begins with the application and its risks, rather than with a particular cloud platform or communication protocol.

This chapter develops architectures and engineering methods for designing IoT-enabled smart buildings, industrial systems, water and agricultural systems, and smart cities. It also introduces performance indicators, cybersecurity and governance requirements, and an illustrative integrated case.

Learning Objectives

After studying this chapter, a student should be able to:

1. Explain the architecture of smart building, industrial, infrastructure, and city IoT systems.
2. Design basic solutions for energy management, lighting, HVAC, automation, and predictive maintenance.
3. Select sensors, actuators, networks, computing locations, and data platforms according to application requirements.
4. Calculate and interpret important energy, reliability, and operational performance indicators.
5. Evaluate an IoT application using technical, economic, environmental, safety, ethical, and cybersecurity criteria.
6. Recommend appropriate implementation and governance practices.

1. Application-Oriented IoT Architecture

A practical architecture usually contains five functional layers:

The layers are logical rather than necessarily separate physical products. A modern gateway may provide protocol conversion, local control, cybersecurity functions, and temporary data storage. The architecture should preserve local operation if the external network or cloud service is unavailable. This principle is especially important for life safety, process safety, and production continuity.

1.1 Sensing, actuation, and control

Sensors convert physical variables into data. Typical variables include temperature, humidity, illuminance, occupancy, vibration, pressure, flow, current, voltage, power, air quality, and water level. Actuators apply decisions to the process: dampers adjust airflow, variable-frequency drives change motor speed, relays switch loads, valves regulate flow, and dimmers control lighting output.

A measurement is useful only when its accuracy, range, response time, installation position, and maintenance requirements are appropriate. For example, an occupancy sensor placed behind a partition may systematically undercount occupants. A current sensor installed on only one phase of a three-phase motor may provide an incomplete diagnosis.

A basic feedback loop compares a measured variable with a desired setpoint. If the measured room temperature is above its setpoint, the controller may increase cooling. In a simple proportional controller, the control action can be represented as:

where u(t) is the control signal, K_p is the proportional gain, and e(t) = r(t) - y(t) is the error between the reference r(t) and measured output y(t). This expression assumes a proportional controller and a control range in which the actuator is not saturated. Real HVAC and industrial systems often use proportional-integral-derivative (PID) control, supervisory rules, optimization, or model predictive control in addition to local safety interlocks.

1.2 Edge, fog, and cloud computing

Edge computing places processing near the equipment. It reduces latency, limits bandwidth use, and allows operation during loss of internet connectivity. Cloud computing provides scalable storage, fleet-wide analytics, remote access, and machine-learning services. A suitable design commonly uses both: safety and fast control at the edge, historical analysis and coordination in the cloud.

Data should not be transmitted merely because it can be transmitted. Sampling rate should reflect the dynamics of the process. A slowly changing building temperature may be sampled every one to five minutes, while vibration analysis may require much higher rates and local feature extraction. Excessive sampling increases network traffic, storage cost, and cybersecurity exposure without necessarily improving decisions.

2. Smart Buildings

A smart building integrates building management systems (BMS), energy meters, lighting, HVAC, access systems, fire systems, lifts, renewable generation, and sometimes electric-vehicle charging. Its objective is not simply minimum energy consumption. It must balance energy, comfort, indoor air quality, occupant productivity, resilience, safety, and operating cost.

2.1 Building energy management

Energy management begins with measurement at appropriate levels: whole-building supply, major floors or zones, HVAC plant, lighting, plug loads, and large equipment. Submetering makes inefficient systems visible and supports fair allocation of costs.

Building energy intensity is commonly calculated as:

where E_annual is annual energy consumption, usually in kWh/year, and A is conditioned or floor area, usually in m². The result is expressed in kWh/m²·year. The area definition and climate conditions must remain consistent when comparing buildings.

A load profile shows demand over time. Peak demand may determine network charges and transformer sizing even when total energy is moderate. Demand response can shift flexible loads such as chilled-water production, battery charging, or electric-vehicle charging away from system peaks. Critical loads must not be shed without an approved reliability and safety strategy.

A simple energy-saving estimate is:

where S is annual energy saving, and the two energy terms must refer to the same boundary and comparable operating conditions. A valid measurement and verification process adjusts for occupancy, weather, operating hours, and changes in building use.

2.2 Intelligent lighting

Smart lighting combines occupancy sensing, daylight harvesting, scheduling, dimming, and zone-level control. A daylight sensor measures available natural illumination, while an occupancy sensor determines whether lighting is needed. The design must avoid excessive switching, glare, flicker, and inadequate illumination.

Lighting power density is:

where P is installed lighting power in watts and A is floor area in m²; the unit is W/m². Lower LPD does not automatically imply better design because visual quality, task requirements, controls, and operating hours also matter.

A practical control sequence may maintain minimum task illumination, dim luminaires when daylight is adequate, and switch or dim unoccupied zones after a time delay. Emergency lighting and fire-related functions require separate compliance with applicable safety codes and should not depend solely on a cloud platform.

2.3 HVAC and indoor environmental quality

HVAC is often the largest building energy consumer. IoT improvements include zone temperature and humidity sensing, variable-air-volume control, occupancy-based ventilation, chiller optimization, filter monitoring, and fault detection.

The coefficient of performance (COP) of a cooling system is:

where Q_cooling is useful cooling capacity in watts and W_input is electrical input power in watts. COP is dimensionless. It should be calculated over a clearly defined boundary, such as the compressor alone or the complete cooling plant including pumps and fans.

Indoor air quality requires more than temperature control. Carbon dioxide concentration can serve as an indicator of ventilation adequacy in some occupied spaces, but it is not a complete measure of air quality. Sensors for particulate matter, humidity, volatile compounds, and outdoor conditions may also be required. Sensor calibration, placement, and ventilation-system response must be considered before using measurements for automated decisions.

3. Industrial Automation and Smart Manufacturing

Industrial IoT connects machines, programmable logic controllers (PLCs), supervisory control and data acquisition (SCADA), manufacturing execution systems (MES), enterprise resource planning (ERP), and maintenance platforms. The central design challenge is integrating operational technology (OT), which directly controls physical processes, with information technology (IT), which manages data and business processes.

3.1 Automation hierarchy

A typical hierarchy includes:

• Field devices: sensors, actuators, drives, and safety devices.
• Control level: PLCs, distributed control systems, and robot controllers.
• Supervisory level: SCADA, alarms, trends, and operator interfaces.
• Operations level: MES, production scheduling, quality management, and maintenance.
• Enterprise level: ERP, supply-chain, finance, and business analytics.

The Purdue-style separation of levels is useful for reasoning about access and data flows, although real installations may not follow it exactly. Safety instrumented systems and emergency shutdown systems should remain independent enough to achieve their required safety integrity. Adding connectivity must never bypass an existing safety function.

3.2 Overall equipment effectiveness

Overall equipment effectiveness (OEE) combines availability, performance, and quality:

The factors are dimensionless ratios between zero and one when expressed as fractions. Availability is commonly planned production time minus unplanned downtime, divided by planned production time. Performance compares actual production rate with an ideal rate for the operating time. Quality is good units divided by total units. The exact definitions must be agreed because planned maintenance, changeovers, minor stops, and startup losses may be classified differently.

For example, suppose a line has planned production time of 480 minutes, 40 minutes of unplanned downtime, an ideal rate of 2 units/minute, actual operating time of 440 minutes, 800 total units, and 760 good units.

Availability = 440/480 = 0.9167.

Performance = 800/(2 × 440) = 0.9091.

Quality = 760/800 = 0.9500.

Therefore:

OEE = 0.9167 × 0.9091 × 0.9500 ≈ 0.792, or 79.2%.

The result identifies losses but does not by itself explain their causes. Engineers should examine downtime codes, speed losses, reject causes, and the reliability of the data collection process.

3.3 Smart manufacturing

Smart manufacturing uses connected assets, digital work instructions, automated inspection, robotics, additive manufacturing, and production analytics. A digital twin is a maintained digital representation of a physical asset or process that uses current and historical data. It is useful for monitoring, simulation, optimization, and training, but it is not automatically accurate: model assumptions, sensor quality, and asset configuration must be maintained.

Machine vision can inspect dimensions, labels, surface defects, or assembly presence. AI models may classify images or detect unusual behavior, but model performance depends on representative training data, stable lighting, correct labels, and monitoring for data drift. Human review remains important for ambiguous or safety-critical decisions.

4. Predictive Maintenance and Asset Intelligence

Maintenance strategies include corrective maintenance after failure, preventive maintenance at scheduled intervals, condition-based maintenance using measured condition, and predictive maintenance using models to estimate failure risk or remaining useful life.

A basic reliability measure is mean time between failures (MTBF):

where T_operating is total operating time over the observation period and N_failures is the number of relevant failures. MTBF is meaningful only when the failure definition, operating population, and observation period are specified. It should not be used uncritically for repairable systems with changing operating conditions.

A predictive-maintenance workflow consists of:

1. Define the failure modes and business consequences.
2. Identify measurable precursors, such as vibration, temperature, current, pressure, or lubricant condition.
3. Establish a trustworthy baseline during healthy operation.
4. Extract features or use suitably validated models.
5. Set alert thresholds with acceptable false-alarm and missed-failure rates.
6. Link alerts to inspection, work orders, spare parts, and safety procedures.
7. Measure outcomes and retrain or recalibrate when conditions change.

For a rotating machine, vibration frequency components may indicate imbalance, misalignment, bearing defects, or looseness. Electrical motor-current signatures can provide additional evidence, but load variation and power-electronic drives complicate interpretation. A model that predicts an anomaly but does not lead to a practical maintenance action has limited operational value.

5. Water Systems and Agriculture

5.1 Smart water systems

IoT in water systems supports pressure management, leak detection, pump optimization, reservoir monitoring, water-quality surveillance, and demand forecasting. Sensors may measure flow, pressure, turbidity, conductivity, pH, chlorine residual, tank level, and pump status.

A simple water balance for a storage tank is:

where ∆V is the change in stored volume in m³, Q_in and Q_out are average inflow and outflow rates in m³/s, and ∆t is the time interval in seconds. The expression assumes that unmeasured leakage, evaporation, and other flows are either negligible or included in the flow terms. In a real system, a residual balance can indicate leakage or sensor error.

Water-quality data require strict calibration, sampling protocols, and chain-of-custody procedures where regulatory decisions are involved. Automated control of treatment processes must include fail-safe limits and independent verification.

5.2 Smart agriculture

Agricultural IoT combines soil-moisture sensing, weather stations, irrigation control, crop imaging, livestock monitoring, and farm-management software. A soil-moisture sensor should be calibrated for the soil type and installed at a depth relevant to the crop root zone. Irrigation decisions should consider crop stage, weather forecasts, soil water-holding capacity, and water restrictions rather than a single sensor reading.

Variable-rate irrigation and fertilizer application can reduce inputs, but poor spatial sampling may produce false precision. Connectivity challenges in rural areas make local control, store-and-forward operation, and low-power wide-area networks important design options.

6. Smart Cities and Infrastructure

A smart city is not merely a collection of connected devices. It is an institutional and technical system that coordinates transport, energy, water, waste, public safety, buildings, and citizen services.

Common applications include adaptive traffic signals, parking management, street-light control, public-transport information, flood monitoring, waste-route optimization, air-quality monitoring, electric-vehicle charging, and distributed-energy management. Interoperability is essential because city assets are often supplied by different vendors and remain in service for decades.

A useful city architecture includes local controllers, municipal communication networks, a shared data platform, geographic information systems, operational applications, and public-facing services. Data governance should define ownership, retention, access rights, quality responsibilities, and acceptable secondary uses.

Performance indicators should reflect public value, not only device counts. Examples include:

Indicators require baselines, defined boundaries, data-quality checks, and a stated reporting period. A city that installs thousands of sensors but cannot maintain them or act on their data has not achieved smart infrastructure.

7. Worked Application: Energy and Maintenance Upgrade in a University Building

The following is an illustrative case, not a report of a real institution.

Background

A four-storey university engineering building has a floor area of 8,000 m². It contains laboratories, offices, classrooms, and a server room. Annual electricity consumption is 1,600,000 kWh. The BMS records only central plant data, and several air-handling units operate at constant speed.

Problem

The facilities team reports high afternoon demand, complaints of overheating in some rooms, and repeated fan-bearing failures. The building has no reliable occupancy data and limited submetering.

Analysis

The initial energy intensity is:

EUI = 1,600,000 kWh/year ÷ 8,000 m² = 200 kWh/m²·year.

A survey identifies three major opportunities: occupancy-based scheduling, variable-speed fan control, and vibration monitoring on critical fans. The design team also finds that some temperature sensors are poorly located and that laboratory ventilation cannot be reduced below safety requirements.

Approach and solution

The proposed architecture includes room occupancy and temperature sensors, electrical submeters for HVAC and lighting, vibration sensors on twelve critical fan assemblies, and an edge gateway connected to the existing BMS. The gateway retains local schedules and safety interlocks. A cloud platform stores non-real-time data for dashboards and fleet analysis.

The control strategy schedules classrooms according to timetables, uses occupancy data to extend operation only when needed, resets supply-air temperature within approved limits, and commands variable-frequency drives subject to minimum ventilation requirements. The predictive-maintenance system compares vibration features with a healthy baseline and creates an inspection recommendation rather than directly stopping equipment.

Cybersecurity measures include network segmentation between building controls and office IT, unique accounts, multifactor authentication for remote access, signed firmware where supported, least-privilege permissions, backups, and an incident-response procedure. Occupancy data are aggregated for energy control and are not retained as personally identifiable movement histories.

Outcome estimate

Suppose the project is expected to reduce annual electricity use by 15% without reducing required laboratory ventilation.

Projected saving = 0.15 × 1,600,000 = 240,000 kWh/year.

Projected consumption = 1,600,000 − 240,000 = 1,360,000 kWh/year.

Projected EUI = 1,360,000 ÷ 8,000 = 170 kWh/m²·year.

These are estimates, not measured outcomes. Verification should compare post-installation consumption with a weather- and occupancy-adjusted baseline. Comfort complaints, laboratory safety, fan failure frequency, false alarms, and maintenance workload should also be tracked.

Learning points

The case demonstrates that sensors alone do not create value. Value comes from linking measurements to validated control sequences, maintenance actions, and performance verification. It also shows why application constraints matter: laboratory ventilation, fire systems, privacy, and cybersecurity may limit otherwise attractive control strategies.

8. Cybersecurity, Safety, Ethics, and Governance

Connected physical systems enlarge the attack surface. Threats include unauthorized access, ransomware, denial of service, manipulated sensor data, unsafe commands, insecure remote maintenance, and supply-chain compromise. A risk-based approach should identify assets, threats, vulnerabilities, consequences, and safeguards.

Important practices include asset inventories, secure configuration, network segmentation, authenticated communications, timely patching, vulnerability management, backups, logging, anomaly detection, and tested incident response. Legacy devices may not support modern security controls; compensating measures such as isolation, strict firewall rules, jump servers, and monitoring may be required.

Safety and cybersecurity are linked but distinct. A cybersecurity control that delays an emergency shutdown is unacceptable, while a safety interlock should not be replaced by an unvalidated analytics model. Safety functions must be engineered according to applicable electrical, machinery, building, process, and occupational regulations.

Ethical governance includes privacy, fairness, transparency, accessibility, and accountability. Occupancy analytics may reveal individual routines. Facial recognition and worker-monitoring systems create especially serious risks of surveillance and discrimination. Data minimization, purpose limitation, informed notice, retention limits, access controls, and human oversight are essential.

Procurement should specify open interfaces, data portability, update obligations, cybersecurity responsibilities, service-level agreements, and an exit plan. Vendor lock-in can make a technically successful pilot economically unsustainable over its lifetime.

9. Implementation and Evaluation Framework

A disciplined implementation proceeds through the following stages:

1. Define the operational problem and success criteria.
2. Establish the baseline and identify data gaps.
3. Map assets, processes, stakeholders, hazards, and dependencies.
4. Select sensors and controls based on accuracy, dynamics, environment, and maintainability.
5. Design communications and computing locations according to latency, availability, bandwidth, and security.
6. Pilot the solution in a representative but controlled area.
7. Validate data quality, control response, cybersecurity, and user acceptance.
8. Scale using documented standards, commissioning, training, and lifecycle budgets.
9. Verify benefits against the baseline and continuously improve.

Economic evaluation may use simple payback, life-cycle cost, or discounted cash flow. Simple payback is:

where C_initial is initial investment and B_annual is annual net benefit. This metric ignores discounting, degradation, residual value, and benefits after the payback period; it is therefore a screening measure rather than a complete investment analysis.

The most important professional lesson is to optimize the whole system. Lower energy use that causes equipment damage, poor indoor air quality, unsafe production, or excessive maintenance is not a successful IoT outcome. Engineering judgment must combine physical modeling, controls, data analysis, economics, human factors, and governance.

Contemporary Developments

Current developments include private industrial 5G, time-sensitive networking, interoperable semantic models, edge AI, digital twins, federated analytics, distributed energy resources, vehicle-to-building operation, and carbon-aware control. These technologies can improve responsiveness and coordination, but they do not remove the need for instrumentation, calibration, cybersecurity, standards, and competent operators.

AI is particularly useful for anomaly detection, demand forecasting, image inspection, and optimization under changing conditions. It should be evaluated using relevant engineering metrics, including false-alarm cost, missed-event cost, latency, robustness, explainability, and safe fallback behavior. A model should never be judged only by an accuracy percentage calculated on an unrepresentative test set.

Chapter Summary

Smart buildings, industrial systems, water networks, farms, and cities use IoT to connect measurement, control, analytics, and organizational decisions. Their architectures commonly combine field devices, local controllers, communication networks, edge computing, data platforms, and applications.

Building applications emphasize energy intensity, demand, lighting, HVAC performance, comfort, and indoor air quality. Industrial applications integrate PLCs, SCADA, MES, and enterprise systems while protecting safety and production continuity. OEE helps classify production losses, and predictive maintenance links condition data to practical maintenance actions. Water and agricultural systems require physical balances, calibrated sensing, and operation under uncertain connectivity. Smart-city programs must measure public outcomes rather than device counts.

Effective designs use application-specific performance indicators, validated baselines, local fail-safe control, cybersecurity, privacy protection, interoperability, and lifecycle planning. AI and digital twins can strengthen decisions, but their value depends on data quality, model validity, human oversight, and safe fallback procedures.

Key Terms

IoT: A network of physical objects capable of sensing, communication, computation, or actuation.

Smart building: A building that uses integrated sensing, control, communication, and analytics to improve operation.

BMS: Building management system used to monitor and control building services.

HVAC: Heating, ventilation, and air-conditioning system.

Edge computing: Processing performed near the physical source of data.

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

PLC: Programmable logic controller used for reliable industrial control.

MES: Manufacturing execution system connecting production operations with planning and quality activities.

OEE: Overall equipment effectiveness, calculated from availability, performance, and quality.

Predictive maintenance: Maintenance supported by condition data and predictions of developing faults.

Digital twin: A maintained digital representation of a physical asset or process.

EUI: Energy use intensity, commonly measured in kWh/m²·year.

COP: Ratio of useful cooling or heating output to input power.

Demand response: Adjustment of electricity consumption in response to system or price conditions.

Interoperability: Ability of systems or devices to exchange and use information effectively.

Operational technology: Hardware and software that monitors or controls physical processes.

Cyber-physical system: Integrated computational and physical system with sensing and actuation.

Data governance: Policies defining data ownership, quality, access, use, and retention.

Non-revenue water: Water supplied to a network but not billed because of leakage, metering, or other losses.

Measurement and verification: Systematic process for determining whether claimed performance savings occurred.

Review Questions

Short Answer Questions

1. What are the main functional layers of an application-oriented IoT architecture?
2. Why should safety-critical control remain available at the edge?
3. Define EUI and state its unit.
4. What three factors are multiplied to obtain OEE?
5. Why is sensor calibration important in water-quality and agricultural applications?

Long Answer Questions

1. Explain how an IoT-enabled HVAC system can reduce energy consumption while preserving comfort and indoor air quality.
2. Describe the integration challenges between industrial IT and OT systems.
3. Explain a complete predictive-maintenance workflow for a rotating machine.
4. Discuss cybersecurity, privacy, and safety requirements for a smart building.
5. Compare edge computing and cloud computing for industrial and city applications.

Analytical / Application Questions

1. A building uses 2,400,000 kWh/year over 12,000 m². A control project is expected to reduce consumption by 12%. Calculate the baseline and projected EUI, and identify two factors that must be controlled during measurement and verification.
2. An industrial line operates for 450 minutes out of 480 planned minutes, produces 810 units at an ideal rate of 2 units/minute, and produces 770 good units. Calculate availability, performance, quality, and OEE.
3. Design an IoT architecture for a municipal water reservoir. Specify sensors, local control functions, communications, cloud functions, cybersecurity measures, and at least four performance indicators.

Multiple Choice Questions with Answers

Question 1

Which function is most appropriately retained at the edge in a smart physical system?

A. Long-term fleet benchmarking

B. Emergency interlocking and local safety control

C. Annual sustainability reporting

D. Historical data visualization

Correct Answer: B. Emergency interlocking and local safety control

Question 2

A building consumes 900,000 kWh annually and has an area of 6,000 m². What is its EUI?

A. 15 kWh/m²·year

B. 150 kWh/m²·year

C. 540 kWh/m²·year

D. 5,400 kWh/m²·year

Correct Answer: B. 150 kWh/m²·year

Question 3

Which set correctly defines OEE?

A. Energy × Cost × Revenue

B. Availability × Performance × Quality

C. Reliability × Voltage × Current

D. Capacity × Pressure × Temperature

Correct Answer: B. Availability × Performance × Quality

Question 4

A cooling plant provides 500 kW of cooling while consuming 125 kW of electrical input. What is its COP?

A. 0.25

B. 2

C. 4

D. 625

Correct Answer: C. 4

Question 5

Which activity is the best first step in a predictive-maintenance program?

A. Install every available sensor

B. Define failure modes and their consequences

C. Replace all equipment immediately

D. Send all raw data to the cloud

Correct Answer: B. Define failure modes and their consequences

Question 6

In a water-tank balance, which quantity represents the change in stored volume?

A. Qin + Qout

B. Qin − Qout multiplied by the time interval

C. Qin ÷ Qout

D. Qout multiplied by pressure

Correct Answer: B. Qin − Qout multiplied by the time interval

Question 7

Which characteristic is most important when selecting a network for fast industrial control?

A. Advertising reach

B. Deterministic latency and reliability

C. Largest possible screen size

D. Highest historical storage capacity

Correct Answer: B. Deterministic latency and reliability

Question 8

Why should occupancy data in a smart building be minimized and aggregated where possible?

A. To increase personally identifiable tracking

B. To reduce privacy risk while retaining operational value

C. To eliminate the need for sensors

D. To prevent all energy measurements

Correct Answer: B. To reduce privacy risk while retaining operational value

Question 9

Which is a valid limitation of simple payback?

A. It includes all future discounting automatically

B. It ignores benefits and costs after the payback period

C. It requires no estimate of investment cost

D. It measures indoor air quality directly

Correct Answer: B. It ignores benefits and costs after the payback period

Question 10

Which outcome is the strongest indicator of a successful smart-city project?

A. Number of installed sensors alone

B. Number of vendor dashboards alone

C. Improved, verified public-service performance

D. Amount of data stored without analysis

Correct Answer: C. Improved, verified public-service performance