The Rise of the Human Digital Twin in Smart Manufacturing

Digital twins have transformed the way factories manage their machines. A digital twin is a virtual version of a machine or process. Engineers can trial changes virtually before implementing them on the production line.  

Most systems closely monitor equipment and its performance. However, few systems keep track of the workers who operate these machines. This gap becomes more important as factories start using smarter tools and technologies.

A new idea called the Psychosocial Digital Twin, or PDT, is designed to help close this gap. The PDT creates a real-time model of a worker’s mental state. This model can help predict stress levels and workload during production.

Recent research has tested this approach with experienced factory workers. The results suggest that this idea could change the way managers design work in factories.

Why Worker Mental Load Matters

Operators in modern plants need to use more mental effort than before.  Workers monitor automated systems and watch large-screen displays. They often work alongside robots or use remote tools to get the job done. All these tasks can put a lot of mental strain on workers. Mental effort refers to the effort required to complete work.

When this psychological load gets too high, it becomes hard to stay focused. Mistakes become more common, and safety risks increase. If workers are overloaded for long periods, they can also become mentally fatigued.

How workers feel affects how well they perform and how engaged they are at work. Manufacturers must now balance meeting production goals with protecting their workers’ health. Supply chain shifts add more strain on production teams. Firms also struggle to keep skilled workers. All these problems are making factories reconsider how they manage their workforce.

Traditional dashboards do not show these human factors. They only display metrics such as output, downtime, and defect rates. They rarely give any insight into how workers feel while doing their jobs.

Extending Digital Twins to the Human Operator

Most digital twin research looks at machines and processes. These models are used to plan maintenance and improve production. Some studies also consider worker ergonomics, such as modeling physical strain during factory tasks. However, most research does not address mental workload. Meanwhile, cognitive research is finding new ways to track mental states. For example, changes in heart rate can show mental effort, and eye tracking can reveal fatigue or loss of focus.

Researchers sometimes combine different signals to predict when someone is overloaded. Machine learning models already use heart rate and skin signals for this purpose. However, few studies connect these signals to factory operations, and even fewer use them in digital twin systems.

The PDT concept connects these areas by combining factory data with information about workers’ states. This allows managers to simulate how changes impact both machines and people.

How the Psychosocial Digital Twin Works

The PDT collects multiple types of data simultaneously. It collects machine data from production systems, such as line speed, cycle time, and error counts. Environmental sensors monitor the area around the worker, measuring factors such as noise and lighting. Wearable sensors track signs of mental effort, such as heart rate variability and electrodermal activity, which can show stress. Eye tracking gives details about focus, fatigue, and changes in attention.

An artificial intelligence model processes these data streams collectively using a recurrent neural network trained on time-series data. This process generates a Cognitive Load Index (CLI), which provides a score ranging from low mental load to high stress.

Supervisors can view the Cognitive Load Index on a live dashboard and run simulation tests. For example, managers can check what might happen if they increase line speed by 15 percent or whether a worker’s stress levels could rise during a shift. The system predicts likely results, so supervisors can adjust plans before making real changes.

Testing the Idea in a Virtual Factory

Researchers tested the framework in a controlled experiment using a virtual factory created with the Unity engine. The simulation recreated a complex electronics assembly task. Seventy manufacturing operators took part, each with over fifteen years of experience. They wore wrist sensors and eye-tracking glasses to collect physiological and visual behavior data.

Participants were divided into two groups. In the first group, supervisors monitored production using only dashboard metrics. In the second group, supervisors used the PDT system, which let them see the Cognitive Load Index in real time and run simulations before making production changes.

Each worker took part in a 2.5-hour assembly session. During the shift, supervisors made several operational changes, including increasing line speeds and introducing new product variants. Researchers recorded all sensor data and performance metrics, and participants filled out workload surveys after the session.

What the Results Showed

The predictive model identified stress events with 87.4 percent accuracy. The PDT group saw significant improvements. High-stress events dropped by 42 percent, and average cognitive load decreased by 21.7 percent. Human task errors went down by 28 percent, and production performance became 38 percent more stable. Workers also reported feeling less workload in surveys, with NASA-TLX scores dropping by about 22 percent.

The system influenced how supervisors made decisions. Managers usually test ideas before putting them into practice. Simulations led supervisors to change or cancel 65 percent of tasks that caused high stress. Rather than making sudden speed increases, they introduced changes more gradually. Productivity improved during the trial, with researchers noting an 18 percent rise in output.

These findings support a central idea in Socio-Technical Systems theory: work systems achieve their best results when both human and technical elements work together.

Moving From Reactive to Predictive Ergonomics

Most ergonomics programs only act after problems show up. Companies usually deal with burnout or injuries once they happen. Process design often overlooks mental limits from the start. The PDT approach changes this way of thinking. Work designers now check how tasks affect people’s minds before launching them. This is like how engineers already test machines. With this approach, human performance becomes part of the planning process. 

This system also helps factories adapt. Production can react to more than just equipment problems. It can also adjust based on how workers are doing. These kinds of systems could be the next step in smart manufacturing. Even as automation grows, people are still at the heart of operations.

Industry Use Cases

Many industries could start using the PDT model right away. Automotive assembly lines often deal with many different types of vehicles. Simulations can help check workloads before a new model is launched. If stress levels are expected to go up, managers can divide tasks among workers. Using enhanced instructions can help lower mental effort.

Semiconductor clean rooms depend on operators who must stay highly focused. The PDT could flag overload during multi-machine monitoring. Supervisors can turn off non-essential alarms when running complex diagnostics.

Pharmaceutical packaging lines often run at high speeds for inspections. Simulations can show when faster speeds start to hurt quality checks. Lighting upgrades or interface changes might offset the added load.

Practical Challenges Ahead

Even though the results have been promising, there are still some challenges to overcome. The cost of hardware can be a significant hurdle. Investing in research-grade sensors and analytics tools is often necessary. Smaller companies may find these expenses difficult to manage. Offering these tools as a service could make it easier for companies to get started. Integrating new technology can also be challenging. Many factories continue to use older production systems. 

Standards such as OPC UA can help connect data across different platforms. However, some situations may still need custom middleware.

Building trust with workers is also a key concern. Constant monitoring can sometimes feel invasive to employees. Clear policies must limit how data is used. Data should support safety and work design only. It should never be used as a tool to monitor individual performance. Protecting privacy is essential for these technologies to be widely adopted. Experts also caution that workplace algorithms can be biased or misused.

The Future of Human-Aware Factories

The Psychosocial Digital Twin introduces a new approach to smart manufacturing. Instead of just modeling machines, digital twins now include the people who use them. This framework combines cognitive science, AI, and production data. It gives managers a better understanding of workers’ mental workload during operations. As a result, factories may be able to balance productivity with employee well-being, which are often seen as conflicting goals in traditional systems.

Researchers plan to test the system in real factories in the future. They may use long-term studies to track burnout, safety, and productivity. To protect privacy, they are considering methods such as federated learning, which allows models to be trained without collecting sensitive data in a single location.

Another development is the use of automated scheduling tools. These systems can adjust tasks in real time to help manage workloads. The industry is increasingly mindful of human needs.  While smart factories used to focus mainly on machines and automation, the next phase may give more attention to the people working alongside these systems.

If that shift succeeds, digital twins will not only mirror equipment. They will also reflect the human experience of work. 

—Authored by Prahlad Chowdhury, Managing Solution Architect, Fujitsu America Inc.