Autonomous process intelligence Teaching AI injection moulding expertise

Related Vendors

KAIST

CUMSA

BORIDE Engineered Abrasives

Renishaw plc

As skilled workers retire and factories become more international, manufacturers face the risk of losing critical process know-how. A new AI system developed at KAIST combines generative models and large language models to optimise injection moulding processes while transferring expert knowledge directly to operators in multiple languages.

Most of the plastic products we use are made through injection moulding, i.e. molten plastic is injected into a mould to mass-produce identical items. The method is highly efficient, but also sensitive: even minor changes in temperature, humidity or material behaviour can lead to defects. For decades, maintaining stable quality has therefore depended largely on the experience and intuition of highly skilled operators. As many of these experts approach retirement and manufacturing sites become increasingly international, concerns are growing that critical process knowledge could be lost.

Researchers in South Korea are now addressing this challenge with an AI-based approach. A team at the Korea Advanced Institute of Science and Technology (KAIST) has developed what it describes as the world’s first generative AI technology capable of autonomously optimising injection moulding processes, combined with a large language model (LLM)–based system designed to transfer manufacturing knowledge directly to the shopfloor. The work was carried out by a team led by Professor Seunghwa Ryu at the Department of Mechanical Engineering’s Innocore Prism-AI Center and has been published in the Journal of Manufacturing Systems.

Gallery

The first component of the research is a generative AI–based process inference technology that automatically determines optimal process settings in response to changes in environmental conditions or quality requirements. Until now, adjustments to factors such as temperature, humidity or target product quality have typically required repeated trial-and-error interventions by experienced workers.

To overcome this, the KAIST team implemented a diffusion model–based approach that works in reverse, inferring the process conditions needed to achieve a desired quality outcome. The model was trained using environmental data and process parameters collected over several months from a real injection moulding factory.

In parallel, the researchers developed a surrogate model that can stand in for actual production, allowing product quality to be predicted without running the real process. This approach reduced prediction errors to just 1.63 percent — well below the 23 to 44 percent error rates reported for established methods such as generative adversarial networks (GANs) and variational autoencoders (VAEs). Trials in which AI-generated process settings were applied in real production confirmed that the system could reliably produce parts meeting quality requirements, underlining its practical relevance for industrial use.

Natural-language queries meet process optimization

The second achievement is the IM-Chat, an LLM-based knowledge transfer system designed to address skilled worker retirement and multilingual work environments. IM-Chat is a multi-agent AI system that combines large language models (LLMs) with retrieval-augmented generation (RAG), serving as an AI assistant for manufacturing sites by providing appropriate solutions to problems encountered by novice or foreign workers.

When a worker asks a question in natural language, the AI understands it and, if necessary, automatically calls the generative process inference AI, simultaneously providing optimal process condition calculations along with relevant standards and background explanations.

For example, when asked, “What is the appropriate injection pressure when the factory humidity is 43.5 percent?”, the AI calculates the optimal condition and presents the supporting manual references as well. With support for multilingual interfaces, foreign workers can receive the same level of decision-making support.

This research is regarded as a core manufacturing AI transformation (AX) technology that can be extended beyond injection molding to molds, presses, extrusion, 3D printing, batteries, bio-manufacturing, and other industries.

In particular, the work is significant in that it presents a paradigm for autonomous manufacturing AI by integrating generative AI and LLM agents through a so-called tool-calling approach, in which the AI independently decides which functions or programs to invoke in a given situation.

Professor Seunghwa Ryu explained, “This is a case where we addressed fundamental problems in manufacturing in a data-driven way by combining AI that autonomously optimizes processes with LLMs that make on-site knowledge accessible to anyone,” adding, “We will continue expanding this approach to various manufacturing processes to accelerate intelligence and autonomy across the industry.”

Publications: Paper 1: “Development of an Injection Molding Production Condition Inference System Based on Diffusion Model,” DOI: https://doi.org/10.1016/j.jmsy.2025.01.008/ Paper 2: “IM-Chat: A multi-agent LLM framework integrating tool-calling and diffusion modeling for knowledge transfer in injection molding industry,” DOI: https://doi.org/10.1016/j.jmsy.2025.11.007

(ID:50669586)