What is Industry 4.0, and how does it help improve decisions on the plant floor?

01 / Concept

Industry 4.0: information, models and connected systems

Industry 4.0 describes the integration of industrial equipment, processes and systems with digital technologies able to acquire, exchange, interpret and use information during manufacturing. A key foundation is cyber-physical systems: architectures linking physical variables, digital processing and, where appropriate, actions affecting the physical process.

Relevant technologies include sensors, the Industrial Internet of Things (IIoT), industrial communication, process models, simulation, digital twins, analytics, artificial intelligence and advanced control. Each addresses different needs and entails specific implementation requirements.

In the plant, the value emerges when measurements can be linked to products, orders, batches, equipment and operating conditions. This context helps operators identify deviations, engineers investigate causes and production managers verify the outcome of their decisions.

For further information on communication and integration, see OPC UA, MQTT and EUROMAP and our guide to industrial data architecture. The OPC UA specification provides mechanisms to represent and exchange industrial information using semantic models. [1]

02 / Historical background

Origins of Industry 4.0

The German term Industrie 4.0 came into wider use in 2011 as part of an industrial and technological initiative in Germany. In 2013, a working group associated with acatech published recommendations for implementation. The concept proposed using developments in electronics, software and communication networks to achieve more connected, flexible and coordinated manufacturing. [2]

The name refers to four broad stages: mechanisation using water and steam power; electrification and mass production; electronics and computer-based automation; and integration of physical and digital systems. These stages provide a historical overview, while technologies from all four coexist in today’s industry.

Integrating machines of different generations is particularly valuable for companies modernising production while retaining existing assets. This requires assessing signals, interfaces, access security and operating constraints for each machine.

03 / Engineering fundamentals

Phenomenological models and simulation: understanding process behaviour

Measurements reveal what happened. Phenomenological models, grounded in physical principles and process knowledge, help explain how a system evolves and estimate how it might respond to changing operating conditions.

Such models may use mass, energy and momentum balances, rheological relationships, heat transfer and mechanical constitutive equations. For example, a mould thermal model may relate heat supplied by the polymer, the heat capacity of mould components and heat removed by the cooling circuit. A conceptual balance is:

Illustrative expression for a lumped thermal system. Practical application requires defining control volumes, time-dependent heat sources, material properties, boundary conditions and appropriate discretisation.

In industrial engineering, simulation helps evaluate hypotheses before equipment is modified. Finite element analysis (FEA) examines stresses and deformations; computational fluid dynamics (CFD) investigates flow fields and heat transfer; injection-moulding simulation studies filling, packing and cooling within the limitations of the material models used.

These models require reliable inputs and experimental validation. Simulation results have engineering value when their assumptions, uncertainties and valid operating ranges are understood. A review of hybrid models explains how physics-based and data-driven approaches can complement one another to improve predictive capability and interpretability. [3]

04 / Digital twins

Digital twins: connecting the model with actual operation

A digital twin uses a digital representation of an asset or process linked to information from its physical counterpart to provide monitoring, estimation, prediction, virtual experimentation or decision-support services. The model, update frequency and interaction with the plant depend on the use case.

A filling simulation or CAD model can be part of a digital twin. Its ability to represent operation dynamically depends on the data and update mechanisms incorporated. The ISO 23247 establishes a reference framework for manufacturing digital twins. ISO 23247-5, addressing the digital thread across the product lifecycle, was published in June 2026. [4] [5]

In plastics processing, an integrated representation can relate mould geometry, material rheology, thermal conditions, machine parameters and quality results. Its practical value begins with an engineering question: which state must be estimated, which deviation predicted and which decision supported?

The literature on digital twins highlights challenges involving interoperability, model credibility, data availability and continuous verification. [6] See also our guide to digital twins in manufacturing.

05 / Variable estimation

Virtual sensors and hybrid models

In many applications, the variables relevant to quality control are difficult or costly to measure directly. The internal temperature of a part, the thermal state of a mould or certain rheological properties of the melt can be estimated using virtual sensors (soft sensors), which combine available measurements with physical, statistical or hybrid models.

A virtual sensor produces an estimate, so its performance must be assessed against reference measurements and under the operating conditions it is intended to cover. Process drift, formulation changes, equipment wear and signal availability affect its reliability.

A polymer extrusion study published in 2025 investigated predictive control of melt viscosity using a neural state-space model and soft-sensor feedback. The authors reported setpoint-tracking and disturbance-rejection results in simulation; industrial application requires additional verification. [7]

06 / Advanced control

Model-based closed-loop control

Closed-loop control continuously compares a measured or estimated variable with the desired behaviour and adjusts a manipulated variable to reduce deviation. PID controllers have used this principle for decades. Model-based approaches incorporate system dynamics to anticipate responses, estimate states or calculate constrained control actions.

In model predictive control (MPC), an algorithm uses an estimated state and dynamic model to calculate future control actions over a prediction horizon. It applies the first action and solves the optimisation again when new measurements arrive. Constraints may include temperature, pressure, speed, consumption or machine-specific limits.

Injection-moulding research has investigated model-based tracking of pressure profiles during filling and packing. In extrusion, strategies addressing nonlinear dynamics, feed disturbances and transport delays have been explored. [8] [7]

In an industrial cell, a project could investigate cooling-circuit supervision using a thermal model and temperature and flow measurements. The model would estimate thermal behaviour; an algorithm could recommend or, where technically authorised, adjust auxiliary-system setpoints within defined ranges. Validation would need to consider disturbances, response times, interactions with shared utilities and part quality.

07 / Automation and people

Adaptive control and adaptive automation

These terms sound similar but refer to different engineering mechanisms.

Adaptive control changes controller or model parameters when the estimated process characteristics change. It may be relevant when polymer viscosity, thermal conditions, mechanical condition or equipment dynamics vary. Adaptation requires stability conditions and strategies for detecting unreliable estimates.

Adaptive automation concerns how tasks are allocated between people and automatic systems in response to working conditions. A system might detect an anomaly, present evidence and suggest an intervention. In defined circumstances, routine actions may be automated while higher-risk decisions remain with authorised personnel.

A review by Bernabei and Costantino published in 2024 identifies opportunities for manufacturing from this dynamic allocation of functions, alongside challenges involving interfaces, automation levels and human factors. [10]

This connects the development of digital tools with plant experience. The people operating and maintaining equipment should help define alerts, intervention criteria, responsibilities and recovery procedures. Research on Operator 4.0 emphasizes learning and collaborative tools alongside the need to protect worker privacy. [11]

08 / Work on the plant floor

People’s knowledge remains essential

Experienced personnel recognise deviations that an isolated record can rarely explain: the sound of a transmission, a change in a part’s appearance or a machine’s behaviour during start-up. Combining these observations with production data and process models makes it easier to distinguish an occasional event from a problem requiring investigation.

Information serves different purposes according to responsibility. An operator can confirm an anomaly and document its context; maintenance reviews event sequences and conditions preceding failures; engineering compares hypotheses with measurements and simulations; quality teams relate products to recorded parameters; and plant management prioritises resources based on verified losses.

Researchers associated with NIST have studied an architecture integrating human knowledge with a digital twin. They propose combining human cognition, the physical environment and digital services to support decisions. [12]

Automation can change tasks, skills and staffing needs. The extent depends on the technologies adopted and how work is organised. Projects should therefore be designed with the people who will use the tools, with defined data-access criteria and training to operate, interpret and maintain the systems.

Industry 5.0 and a human-centred approach

In 2021, the European Commission proposed Industry 5.0 as an approach complementing digital transformation through three priorities: human wellbeing, sustainability and resilience. In manufacturing, this means considering technical outcomes, employment impacts, resource use and the adaptability of production systems together. [13]

09 / Manufacturing applications

Potential applications in plastics processing

Injection moulding: cycle stability and quality

Records of times, pressures, temperatures, positions and quality results can be used to investigate trends and anomalies. Filling and cooling simulations help examine expected system behaviour. In broader projects, these elements can feed hybrid models or experimental strategies for estimating and controlling defined variables.

See injection-moulding optimisation and monitoring and documentation of process signatures.

Extrusion: quality, material transport and dynamic response

Extrusion combines thermal zones, material transport, pressure generation and downstream stages with different delays. Dynamic models and virtual sensors can support analysis of temperature, estimated viscosity and product stability. Predictive control remains an active research area, particularly where delays and multivariable constraints are present.

Industrial cooling: interaction between machines and auxiliaries

Cooling conditions affect repeatability, cycle time and energy consumption. A study can combine thermal balances, flow and temperature measurements, production data and simulation to evaluate circuit changes. Advanced control depends on available actuators, the hydraulic architecture and the ability to intervene without adversely affecting other equipment.

Energy: measurement, modelling and optimisation

Measuring several circuits simultaneously separates machine and auxiliary consumption. Adding production information, operating states and process conditions makes it possible to calculate specific indicators and evaluate improvement scenarios. Model-based optimisation requires projected savings to be distinguished from measured results. See KERN ENERGY and ISO 50001 and energy measurement.

Maintenance and traceability

Actual operating hours, load conditions and event sequences help investigate failures. Degradation models or state estimation can be incorporated into specific projects once data and validation methods are available. A digital thread can also link product design, manufacturing, inspection and service-life information according to the implemented scope.

10 / Applied engineering

How these technologies relate to PM Tec Engineering and KERN IoP

Combining experience in process engineering, physical modelling, design and simulation with industrial sensing, data acquisition and historical records provides a basis for projects of increasing complexity. Available monitoring functions should be distinguished from developments requiring additional engineering and specific validation.

The systems KERN IoP | SMARTPRO and KERN IoP | MONITOR enable industrial information to be collected and reviewed according to the configured components. Modelling, simulation and analysis form part of PM Tec Engineering’s applied engineering work. Their integration is scoped according to objectives, available signals, controller compatibility and experimental evidence.

11 / Putting the principles into practice

How to start a technically sound project

The first decision is to select a specific problem: cycle variation, energy losses, quality deviations or recurring failures. Instrumentation is defined after identifying the variables and events that could explain the problem and the temporal precision required.

When developing a model-based application, we recommend documenting measured variables, states to be estimated, controllable inputs, known disturbances and operating constraints. The model is then developed or calibrated and compared against representative plant trials.

Intervention can progress from observation and recommendations to automatic control, subject to specific validation. Assessment includes data quality, response times, uncertainty, safety, access permissions, maintainability and behaviour when communication or sensors fail.

A solution becomes useful when it helps personnel understand the process, test improvements and retain the knowledge generated. See data-driven continuous improvement and verification of sustained improvements.

12 / Frequently asked questions

Frequently asked questions about Industry 4.0 and advanced control

Does a digital twin need artificial intelligence?

Its architecture may use models based on physical principles, statistics, rules or machine learning. Selection depends on the required fidelity, available data and the service the model will provide.

Is simulation equivalent to having a digital twin?

Simulation can be part of a digital twin. To provide functions linked to actual operation, the model needs connection, updating and evaluation mechanisms suited to its purpose.

How does advanced control differ from a monitoring dashboard?

A dashboard presents information for a person to analyse and decide. Advanced control calculates actions using validated logic and can send setpoints to authorised equipment. This second function requires additional assurances of integrity, stability and safety.

Can existing machinery be used?

Yes, where technically viable signals or measurement points exist. Closing the control loop automatically also requires verification of write permissions, execution times, limits and controller compatibility.

What role do plant personnel play?

Operators, maintenance personnel and engineers contribute knowledge of actual conditions, help validate hypotheses and define how results will be used. Training and the responsibilities of each role form part of system design.

Verified sources

References and standards consulted

1. OPC Foundation. OPC Unified Architecture, Part 1: Overview and Concepts. Specification.
2. acatech. (2013). Recommendations for implementing the strategic initiative INDUSTRIE 4.0. Institutional context.
3. Wang, J., Li, Y., Gao, R. X., & Zhang, F. (2022). Hybrid physics-based and data-driven models for smart manufacturing: Modelling, simulation, and explainability. Journal of Manufacturing Systems, 63, 381–391. DOI.
4. ISO. (2021). ISO 23247-1:2021, Automation systems and integration — Digital twin framework for manufacturing — Part 1: Overview and general principles. Standard.
5. ISO. (2026). ISO 23247-5:2026, Digital thread for digital twin. Standard.
6. Shao, G. (2024). Manufacturing Digital Twin Standards. NIST / EDTconf 2024. Publication.
7. Perera, Y. S., Li, J., & Abeykoon, C. (2025). Melt viscosity control in polymer extrusion using nonlinear model predictive control with neural state space modelling and soft sensor feedback. Journal of Process Control, 155, 103556. DOI.
8. Kariminejad, M., Tormey, D., & McAfee, M. (2024). Model-based pressure tracking using a feedback linearisation technique in thermoplastic injection moulding. Research work.
9. International Society of Automation (ISA). ISA/IEC 62443 Series of Standards. Series information.
10. Bernabei, M., & Costantino, F. (2024). Adaptive automation: Status of research and future challenges. Robotics and Computer-Integrated Manufacturing, 88, 102724. DOI.
11. Kaasinen, E., et al. (2020). Empowering and engaging industrial workers with Operator 4.0 solutions. Computers & Industrial Engineering, 139, 105678. DOI.
12. Kim, D., Bajestani, M., Shao, G., Jones, A., & Noh, S. (2023). Conceptual architecture of digital twins with human-in-the-loop-based smart manufacturing. ASME IMECE 2023. NIST.
13. European Commission. (2021). Industry 5.0: Towards a sustainable, human-centric and resilient European industry. Publication.