Study reveals barriers to AI integration in manufacturing design

The integration of Artificial Intelligence (AI) into manufacturing processes has huge potential for improving productivity, efficiency, and safety. Machine learning models are already used to monitor equipment health and others predict supply-chain issues and consumer demand. However, research published in the International Journal of Mechatronics and Manufacturing Systems suggests that there remain barriers to the more widespread adoption of AI in production environments. In particular, there are obstacles to incorporating AI in the early design phase.

Yuji Yamamoto and Kristian Sandström of Mälardalen University in Eskilstuna, and Aranda Muñoz Álvaro of the Research Institutes of Sweden in Västerås, Sweden, explain how the early design phase is fundamental in determining how AI might ultimately be embedded into the manufacturing workflow.

They point out that it is during this period that engineers, data scientists, production staff, and other stakeholders have to align their goals with functionality and outcomes. However, this process can be stymied if there is a misalignment between the technical expertise of the data scientists and the practical knowledge of the manufacturing professionals.

Poor communication and unrealistic expectations then lead to the installation of an AI system that does not meet the operational needs of the factory floor.

One of the biggest problems the researchers found is that of cognitive overload, where those involved are overwhelmed by the complexity of the tasks at hand. The technical jargon of machine learning and AI, for example, is often inaccessible to those with expertise in production management but not in data science.

Conversely, data scientists may struggle to understand the intricacies of manufacturing operations, such as workflow design and the real-time adjustments needed to address unpredicted challenges. This knowledge gap between the two groups can lead to failure, especially if the AI system does not take into account the very dynamic nature of manufacturing.