10. Digital Twins, Simulation and Immersive Technologies

Introduction and Context

Modern manufacturing systems are increasingly designed, operated and improved through digital representations of physical assets and processes. A production line may be modelled before it is built, monitored through sensor data after commissioning, and supported through augmented-reality instructions during maintenance. These activities are related, but they are not identical. Simulation explores possible behaviour, a digital twin maintains a purposeful connection between a physical object and its digital representation, and immersive technologies enable people to interact with digital information in three-dimensional or context-sensitive ways.

For production managers, the significance of these technologies lies not in visual sophistication alone. Their value is determined by the quality of decisions they support: whether a new line can meet demand, whether a bottleneck can be removed, whether a robot program is safe, whether maintenance can be planned before failure, or whether an operator can be trained without interrupting production. The central managerial challenge is therefore to connect models, data, people and decisions into a reliable digital thread.

This chapter examines discrete-event simulation, virtual commissioning, digital threads, digital twins, augmented reality (AR) and virtual reality (VR). It considers their architectures, methods of implementation, applications, limitations, governance requirements and emerging developments.

Learning Objectives

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

1. Distinguish simulation models, digital threads and digital twins.

2. Formulate and assess discrete-event simulation models for manufacturing systems.

3. Explain the role of virtual commissioning in reducing implementation risk.

4. Design a high-level digital-thread and digital-twin architecture for a factory.

5. Evaluate AR and VR applications in design, training, maintenance and remote operations.

6. Interpret performance, validation and economic measures for these technologies.

7. Identify technical, organizational, ethical and cybersecurity risks.

8. Recommend an implementation roadmap aligned with production strategy.

From Models to Living Digital Representations

A model is a simplified representation of a system created for a purpose. A simulation model uses rules, variables and assumptions to imitate how a system behaves over time. A digital twin is more specific: it is a digital representation of a physical entity or process that is connected to the physical system through data and is used to monitor, analyse, predict or control it. The connection may be continuous or periodic, but it must be meaningful to the intended decision.

A digital thread is the traceable flow of information across the life cycle of a product, process or asset. It links requirements, design records, bills of material, process plans, manufacturing data, quality results, maintenance history and field feedback. The thread addresses continuity and traceability; the twin addresses representation and interaction with a particular physical system. A factory can possess a digital thread without having a real-time digital twin, and a twin may be poorly governed if it is not connected to authoritative life-cycle information.

These concepts can be viewed as layers. Product and process data provide the information foundation. Simulation uses that information to explore alternatives. Connectivity supplies operational data. The digital twin combines models and data for an ongoing purpose. Immersive interfaces make selected information understandable and actionable for people.

Discrete-Event Simulation for Production Decisions

Discrete-event simulation (DES) represents a system as a sequence of events occurring at particular points in time. An event changes the state of the system: a part arrives, a machine begins processing, a job finishes, a failure occurs, a worker becomes available, or a transport vehicle reaches a station. Between events, the state is assumed not to change in the relevant model dimensions.

DES is especially suitable for production systems involving queues, shared resources, variable processing times, failures, batch sizes and complex routing. Typical entities include parts, orders, pallets and vehicles. Resources include machines, operators, fixtures, buffers and inspection stations. Attributes describe entity-specific information such as product type, priority or due date. Events and routing logic determine how entities move through the system.

A basic production model may contain the following logic:

• an order arrives according to a defined pattern;
• the order waits for material and an available machine;
• processing time is sampled from a suitable distribution;
• the machine may fail or require setup;
• the order moves to inspection, rework or the next operation;
• completed orders leave the system.

The model can estimate throughput, cycle time, work-in-process, resource utilization, queue length, tardiness and service level. It can also compare alternatives such as additional capacity, different dispatching rules, buffer sizes, preventive-maintenance policies or product-mix changes.

Building and Validating a DES Model

A credible model begins with a decision question rather than with software selection. “Should a second machine be purchased?” is more useful than “Can the factory be simulated?” The modeller then defines system boundaries, performance measures, time horizon, assumptions and required data.

Data may include arrival rates, processing-time observations, setup times, breakdown distributions, repair times, routing probabilities and shift calendars. Data cleaning is essential because missing values, aggregated records and inconsistent timestamps can produce misleading results. Statistical distributions should be selected because they represent observed behaviour, not merely because they are convenient.

Verification asks whether the model has been built correctly. It includes logic reviews, conservation checks, extreme-condition tests and tracing individual entities. Validation asks whether the model adequately represents the real system for its intended use. Techniques include comparison with historical performance, expert review, sensitivity analysis and, where appropriate, formal statistical tests. A model is not “validated” in an absolute sense; it is judged adequate for a particular decision and level of risk.

Randomness creates further requirements. Multiple replications with different random-number seeds are normally needed. Warm-up periods may be required when the model begins empty but the real system operates continuously. Confidence intervals should accompany estimates, since a difference between two alternatives may be smaller than simulation variation.

Virtual Commissioning and the Transition to Operation

Commissioning traditionally involves testing a physical machine or line after installation. Errors discovered at this stage can cause schedule delays, rework and expensive production losses. Virtual commissioning moves part of this testing into a digital environment. A control program, such as a programmable logic controller (PLC) sequence, is connected to a simulated machine or process before the physical equipment is available.

Several levels are possible. In software-in-the-loop, control logic interacts with a software model. In hardware-in-the-loop, actual control hardware communicates with the simulated plant. A more advanced arrangement combines selected real equipment with simulated components. The objective is not to replace physical acceptance testing but to identify logic errors, unsafe sequences, timing problems, communication faults and abnormal operating conditions earlier.

Virtual commissioning commonly follows this sequence:

1. Create or obtain a geometric and behavioural model of the equipment.

2. Define sensors, actuators, I/O signals and communication interfaces.

3. Connect the control logic to the model.

4. Test normal cycles, product variants and fault conditions.

5. Record defects and refine the control program or equipment design.

6. Transfer verified logic and documentation to physical commissioning.

Its benefits include earlier debugging, reduced dependence on equipment availability, improved collaboration between mechanical, electrical and controls engineers, and better operator preparation. Its effectiveness depends on model fidelity, interface standards, version control and disciplined change management. If the simulated machine omits relevant timing, safety or mechanical constraints, successful virtual tests may create false confidence.

Digital Threads and Digital-Twin Architecture

A digital-thread architecture should establish how information is created, identified, exchanged, governed and reused. Product lifecycle management (PLM), manufacturing execution systems (MES), enterprise resource planning (ERP), supervisory control and data acquisition systems, industrial Internet of Things platforms and computerized maintenance-management systems may all contribute data. The architecture should not simply connect every system to every other system. It should define authoritative sources, common identifiers, event definitions, access rights and responsibilities.

A practical digital-twin architecture often contains five logical layers:

1. Physical layer: machines, products, tools, workers, sensors and production environments.

2. Connectivity layer: industrial networks, gateways, protocols, data acquisition and time synchronization.

3. Data layer: historians, event stores, contextualized data, master data and metadata.

4. Model and analytics layer: physics-based models, DES models, statistical models, condition-monitoring algorithms and optimization tools.

5. Application and interaction layer: dashboards, alerts, planning tools, AR devices, VR environments and control interfaces.

The twin’s purpose determines its design. A design twin may compare product configurations. A production twin may evaluate flow, capacity and quality. An asset twin may estimate equipment health. A performance twin may support energy or productivity optimization. These twins may be connected, but combining them increases complexity and governance demands.

A useful twin is not necessarily a complete virtual replica. It should contain the minimum sufficient data and model fidelity required for its decisions. For example, a maintenance twin may need vibration trends, load, temperature, operating hours and failure history, while a production-planning twin may require routing, demand, calendars and capacity constraints rather than high-frequency vibration data.

Immersive Technologies in Manufacturing

Virtual reality places the user in a computer-generated environment, usually through a head-mounted display and interaction devices. It is valuable when spatial understanding, rehearsal or safe exposure to hazards is important. Applications include assembly training, ergonomic assessment, factory-layout review, emergency drills and collaborative design reviews.

Augmented reality overlays digital information on the physical environment. Through smart glasses, tablets or projection systems, an operator may see work instructions, torque values, inspection criteria, asset identity or remote expert guidance while working on the actual equipment. Mixed-reality systems provide a stronger spatial registration between digital objects and physical surroundings.

The quality of an immersive application depends on more than visual realism. Important factors include registration accuracy, latency, field of view, readability, interaction effort, worker comfort, cognitive load and compatibility with the work task. A sophisticated interface can reduce performance if it forces users to search through irrelevant information or distracts them from hazards.

In design, VR permits engineers and operators to review reach, visibility, access and maintainability before construction. In training, it allows repeated practice of rare or dangerous tasks. In maintenance, AR can identify components, present sequential instructions and capture evidence of completion. In remote operations, a field worker can share a view with an expert who annotates the physical scene. Each application should be evaluated against a conventional alternative, such as classroom training, printed procedures or an ordinary video call.

A Decision Framework for Selecting the Technology

Technology selection should follow the decision problem. DES is appropriate when the issue concerns flows, queues, resources and variability. Virtual commissioning is appropriate when control logic and equipment behaviour must be tested before installation. A digital twin is appropriate when a continuing connection between an asset or process and its digital representation supports monitoring, prediction or control. AR is appropriate when workers need contextual information while interacting with the physical system. VR is appropriate when a safe, repeatable and spatially rich training or design environment is required.

A business case should examine:

• the decision or risk being addressed;
• the cost of data collection, integration, modelling and maintenance;
• expected effects on throughput, quality, safety, energy, downtime or learning time;
• model and data uncertainty;
• user adoption and workflow fit;
• cybersecurity and operational resilience;
• scalability across products, lines or sites.

Measures should include both technical and business outcomes. Technical indicators may include data latency, model error, availability, synchronization accuracy and false-alarm rate. Business indicators may include avoided downtime, reduced commissioning hours, improved first-pass yield, shorter training time, reduced engineering changes and return on investment. A pilot that produces an attractive visualization but does not change a decision is not a successful digital transformation.

Applied Scenario: Digital Twin for a High-Mix Assembly Line

Background

The following is an illustrative case. A manufacturer of industrial control cabinets operates a high-mix assembly line with frequent product changes. Management observes long and variable order lead times, although average machine utilization appears acceptable.

Situation and Problem

Orders pass through kitting, drilling, wiring, testing and packing. Components are often unavailable at the point of use, and urgent orders are inserted into the schedule. Supervisors disagree about whether the principal problem is insufficient capacity, poor material presentation or excessive changeover. The company plans a digital twin to support improvement without disrupting production.

Analysis

The project team first creates a DES model using historical order arrivals, routing, processing times, changeovers, replenishment delays and test failures. The model is verified through event tracing and validated against observed weekly throughput and average lead time. It shows that the apparent machine bottleneck is not continuously constrained. Instead, shortages in kitting and variable release of urgent work create queues and starve downstream operations.

A digital thread is then established using a common product and order identifier. Engineering bills of material, process plans, MES events, quality results and material transactions are linked. The twin receives production status at regular intervals and displays predicted completion dates, queue conditions and material risks. It does not automatically control the line; it supports supervisors in evaluating release and prioritization decisions.

Approach and Solution

Three alternatives are simulated: adding a drilling machine, introducing a kitting supermarket with defined replenishment rules, and applying a controlled order-release policy. The model indicates that the second and third alternatives together provide greater lead-time improvement than immediate equipment purchase. The company pilots the policy on one product family and uses AR instructions to show kit location, component identity and wiring sequence. A VR environment is used to train new assemblers on cabinet layout and safe work practices before they enter the line.

Outcome

The illustrative case demonstrates a reduction in avoidable waiting in the pilot area, but the more important result is managerial: investment decisions are based on system-level evidence rather than on utilization of a single machine. The twin is reviewed monthly because product designs, staffing and replenishment rules change. Its outputs are treated as decision support, not as unquestionable forecasts.

Learning Points

The case illustrates four principles. First, simulation should precede automation when the problem is not yet understood. Second, a digital twin requires identity and contextual data, not merely sensor connectivity. Third, immersive tools are most effective when embedded in a work method. Fourth, model maintenance is part of operational ownership, not a one-time engineering project.

Challenges, Risks and Professional Governance

Data quality is a central risk. Incorrect timestamps, missing sensor values, inconsistent asset names and undocumented manual workarounds can undermine a twin. Interoperability is another challenge: systems may use incompatible data structures, units, semantics and update frequencies. Version conflicts are especially dangerous when a simulation model, control program and physical configuration represent different revisions.

Model risk arises from simplification and uncertain assumptions. A prediction can appear precise while being based on weak evidence. Users should be shown confidence ranges, assumptions and data freshness where these affect decisions. Automated recommendations should have clear escalation rules and human accountability, particularly when they affect safety, maintenance authorization or product release.

Cybersecurity must be designed across the architecture. Connected machines expand the attack surface, and an immersive device may expose operational images or personal information. Controls include network segmentation, identity management, least-privilege access, secure update procedures, logging, backup and incident response. Safety systems should not be bypassed merely to obtain data or demonstrate a digital-twin capability.

Worker privacy and professional ethics require attention. Wearable devices and AR systems can record location, gaze, voice or task performance. Monitoring should have a legitimate purpose, transparent communication, proportionate data collection and defined retention. Technology should assist workers rather than become an undisclosed mechanism for punitive surveillance. Training data and algorithmic performance should also be assessed for bias across experience levels, languages, physical abilities and shifts.

Common implementation mistakes include beginning with a platform rather than a use case, attempting to model the entire factory at once, ignoring manual processes, failing to assign data ownership, and evaluating a pilot only by technical completion. Other errors include excessive visual detail, insufficient validation, lack of operator involvement and failure to budget for model updates.

Implementation Roadmap and Best Practices

A pragmatic implementation proceeds through stages. First, select a high-value use case with measurable performance and accessible data. Second, map the current process and identify authoritative data sources. Third, create a minimum viable model or twin with explicit assumptions. Fourth, verify and validate it with operators, engineers and managers. Fifth, integrate it into a real decision or workflow. Sixth, measure outcomes and expand only when the benefit is demonstrated.

Best practice is to establish a cross-functional product owner, a model owner and a data owner. Use standard asset identifiers, units of measure, naming conventions and interface contracts. Maintain a model registry containing version, purpose, assumptions, validation evidence and retirement status. Separate experimental models from models used for operational decisions. Train users to interpret uncertainty rather than presenting outputs as facts.

Where possible, design for reuse. Modular equipment models, standardized event definitions and open interfaces reduce the cost of extending a twin to another line. However, standardization should not eliminate necessary process knowledge. The aim is controlled reuse, not a generic model that fits no decision well.

Current Developments and Emerging Directions

Digital twins are increasingly combined with edge computing, cloud services, industrial data spaces and artificial intelligence. Edge processing can support low-latency monitoring while reducing the need to transmit all raw data. Machine-learning models can detect anomalies or estimate remaining useful life, although they require representative data and ongoing monitoring for drift. Generative AI may assist with natural-language access to maintenance records, simulation scenarios and work instructions, but its outputs require verification.

Interoperability initiatives are moving toward richer semantic models, standardized asset information and lifecycle traceability. Real-time 3D environments are also becoming more accessible, enabling collaboration among geographically distributed engineering and operations teams. The strategic direction is not simply toward more immersive or more autonomous systems. It is toward trusted, explainable and maintainable digital decision environments.

Chapter Summary

Simulation, digital threads, digital twins and immersive technologies address different but complementary manufacturing needs. Discrete-event simulation explores the behaviour of systems containing queues, resources, variability and competing priorities. Virtual commissioning tests control logic against simulated equipment before physical installation. A digital thread connects information across the product and production life cycle, while a digital twin maintains a purposeful relationship between a physical entity and its digital representation. AR supplies contextual information in the physical workplace, and VR provides a safe, repeatable three-dimensional environment for training, design and rehearsal.

Successful implementation depends on a clear decision purpose, trustworthy data, appropriate model fidelity, validation, interoperability, cybersecurity and user adoption. The strongest applications combine technical modelling with operational knowledge and governance. Digital technologies should therefore be judged by improved decisions and outcomes—not by connectivity, realism or novelty alone.

Key Terms

Augmented Reality: Digital information overlaid on a user’s view of the physical environment.

Digital Thread: A traceable flow of product, process and operational information across the life cycle.

Digital Twin: A connected digital representation of a physical entity or process used for monitoring, analysis, prediction or control.

Discrete-Event Simulation: Simulation in which system state changes at distinct events over time.

Entity: An item represented in a simulation, such as a part, order, pallet or vehicle.

Event: An occurrence that changes the state of a simulated system.

Hardware-in-the-Loop: Testing in which real control hardware interacts with a simulated plant or equipment model.

Immersive Technology: Technology that creates or overlays spatial digital experiences for human interaction.

Model Validation: Assessment of whether a model adequately represents a real system for its intended purpose.

Model Verification: Checking whether a model has been implemented according to its specification.

Mixed Reality: An environment in which physical and digital objects are spatially combined and interactively related.

Physical Layer: The machines, products, people, tools and environment represented by a digital system.

Queue: Waiting entities competing for a limited resource.

Simulation Replication: One independent run of a stochastic simulation using a specified random-number sequence.

Virtual Commissioning: Testing control logic with a simulated machine or process before or alongside physical commissioning.

Virtual Reality: A computer-generated environment that immerses the user in a simulated three-dimensional experience.

Warm-up Period: Initial simulation time excluded to reduce bias caused by starting conditions.

Work-in-Process: Items that have entered production but are not yet completed.

Review Questions

Short Answer Questions

1. What distinguishes a digital twin from a conventional simulation model?

2. Why are verification and validation separate activities in discrete-event simulation?

3. What is the purpose of virtual commissioning?

4. State two applications of augmented reality in manufacturing.

5. Why is a digital thread important for lifecycle traceability?

Long Answer Questions

1. Explain the construction, validation and interpretation of a discrete-event simulation model for a production line.

2. Discuss the logical layers of a digital-twin architecture and the role of each layer.

3. Compare virtual reality and augmented reality for manufacturing training and maintenance.

4. Explain how virtual commissioning can reduce project risk and identify its limitations.

5. Develop an implementation roadmap for a digital twin in a high-mix manufacturing plant.

Analytical / Application Questions

1. A factory has high equipment utilization but poor delivery performance. Explain how DES could determine whether the cause is capacity, batching, material shortages or release policy.

2. Design a minimum viable digital twin for a critical machining centre. Specify its data sources, model outputs, users and validation measures.

3. A company proposes AR glasses that continuously record operator location and task performance. Evaluate the proposal from operational, ethical, privacy and cybersecurity perspectives.

22. Multiple Choice Questions with Answers

Question 1

Which feature most clearly distinguishes a digital twin from a standalone simulation model?

A. It always uses a three-dimensional visual interface

B. It is connected to a physical entity or process for an ongoing purpose

C. It can only represent equipment rather than production systems

D. It eliminates the need for human decision-making

Correct Answer: B. It is connected to a physical entity or process for an ongoing purpose

Question 2

A production manager wants to compare buffer sizes and dispatching rules under variable arrivals and machine failures. Which method is most appropriate?

A. Discrete-event simulation

B. Virtual reality training

C. Augmented-reality overlay

D. Static product rendering

Correct Answer: A. Discrete-event simulation

Question 3

In simulation practice, verification primarily asks whether:

A. The real factory has achieved its target throughput

B. The model has been built according to its intended logic

C. The simulation has a visually realistic interface

D. The digital twin has eliminated process variability

Correct Answer: B. The model has been built according to its intended logic

Question 4

What is the main purpose of virtual commissioning?

A. To replace all physical acceptance tests

B. To test control logic against simulated equipment before physical installation

C. To provide operators with historical maintenance records

D. To create a marketing visualization of a factory

Correct Answer: B. To test control logic against simulated equipment before physical installation

Question 5

Which example best represents a digital thread?

A. A VR headset used for emergency training

B. A sensor measuring motor temperature

C. Linked requirements, design data, process plans, quality records and maintenance history

D. A simulation run using random processing times

Correct Answer: C. Linked requirements, design data, process plans, quality records and maintenance history

Question 6

A simulation begins empty, although the real factory operates continuously. What technique can reduce the resulting initial-condition bias?

A. Increasing the screen resolution

B. Using a warm-up period

C. Removing all random variables

D. Reducing the number of entities

Correct Answer: B. Using a warm-up period

Question 7

Which application is most naturally suited to augmented reality rather than virtual reality?

A. Practising hazardous maintenance without access to equipment

B. Reviewing a proposed factory layout before construction

C. Displaying torque instructions while an operator works on the actual machine

D. Simulating an alternative production schedule

Correct Answer: C. Displaying torque instructions while an operator works on the actual machine

Question 8

A digital twin provides highly precise forecasts, but its timestamps and asset identifiers are inconsistent. What is the most serious concern?

A. Insufficient visual realism

B. Data quality and contextualization risk

C. Excessive operator field of view

D. Too many simulation replications

Correct Answer: B. Data quality and contextualization risk

Question 9

Which is the strongest reason to select a minimum viable digital twin rather than model an entire factory immediately?

A. A smaller twin never requires validation

B. A focused twin aligns model fidelity and cost with a specific decision

C. A complete factory can never be represented digitally

D. Small twins do not require cybersecurity controls

Correct Answer: B. A focused twin aligns model fidelity and cost with a specific decision

Question 10

An employer uses AR devices to record worker movements. Which governance principle is most appropriate?

A. Collect all available data because storage is inexpensive

B. Use the data secretly to improve managerial control

C. Define a legitimate purpose, communicate it transparently and limit collection

D. Allow unrestricted access to all recorded information

Correct Answer: C. Define a legitimate purpose, communicate it transparently and limit collection