Artificial intelligence can improve factory performance, but only when it is guided by human expertise, according to new research from the University of Skövde.
Ehsan Mahmoodi, PhD Student in Informatics at the University of Skövde.
Manufacturers around the world are under pressure to produce more while reducing costs, energy consumption and waste.
In his doctoral research, Ehsan Mahmoodi at the School of Engineering has developed methods that combine computer simulation and artificial intelligence to improve production planning and design. Instead of testing changes directly in a factory, virtual models of production systems can be used to evaluate thousands of alternatives safely and efficiently.
"The goal is to help decision-makers find better solutions faster. Rather than relying on trial and error, we can use AI to explore many possible options and identify those that improve performance," says Ehsan Mahmoodi.
AI searches smarter, not harder
Factories often produce many different products using the same equipment, making planning highly complex. Managers must balance production speed, costs and quality while dealing with countless possible decisions regarding staffing, scheduling and resource allocation.
Ehsan’s research shows that optimization algorithms become significantly more effective when they are provided with relevant knowledge about the production system. Information about bottlenecks, resource constraints or problematic product combinations helps the AI focus on promising solutions instead of searching blindly.
Even relatively simple rules, such as avoiding the production of two problematic products one after another, can have a major impact on overall performance.
Human experience improves AI performance
One finding was surprising in Ehsan’s research:
"I expected more advanced algorithms to consistently perform better. Instead, we found that simple rules based on production experience could sometimes outperform much more sophisticated computational methods," says Ehsan Mahmoodi.
The results highlight the importance of combining artificial intelligence with human expertise. Knowledge gained through years of practical experience on the factory floor can substantially improve the quality of AI-generated decisions when incorporated into optimization tools.
Benefits for both industry and sustainability
The methods could help manufacturers increase output and improve delivery reliability without investing in new machinery. More efficient use of existing resources can also reduce energy consumption and material waste, contributing to sustainability goals.
The approach is not limited to manufacturing. Similar methods could potentially be applied in logistics, healthcare and maintenance planning—any area where computer simulations are used to support complex decisions.
Towards self-improving factories
Looking ahead, Ehsan plans to integrate the methods with digital twins; virtual representations of real production systems that are continuously updated with operational data. This would enable production systems to adapt to changing conditions in near real time.
The research will also be expanded to include environmental objectives such as energy consumption and carbon emissions alongside traditional production targets.
"The long-term vision is a practical decision-support tool that manufacturers of all sizes can use to continuously improve their operations," says Ehsan Mahmoodi.
