AdhesiveGPT (AdhGPT)
TWI Industrial Member Report 1204-2024
By Hasan Caglar, Kai Yang and Ewen Kellar
Industrial Need
In the current landscape, the process of selecting adhesives faces significant challenges. The absence of a centralised repository for comprehensive adhesive information, coupled with the time-consuming manual search and comparison of adhesives, introduces inefficiencies. Moreover, the potential for incomplete or outdated information and the lack of systematic data analysis may further complicate the task.
Recognising the pivotal role of adhesive selection across industries such as automotive, aerospace, oil and gas etc., the establishment of a dedicated database becomes imperative. The integration of an artificial intelligence (AI)-powered adhesive selection tool emerges as a transformative solution, offering multifaceted benefits. These include not only time and cost savings but also potentially enhancements in product performance, increased overall productivity, risk and maintenance reduction, stimulation of innovation, and the promotion of sustainable practices.
For TWI, harnessing the capabilities of such a tool becomes a strategic advantage. It positions the organisation to effectively support member companies in meeting their adhesive bonding requirements, fostering efficiency and innovation in their processes.
The potential impact of successful development in this field extends beyond specific industries, with virtually all sectors standing to gain. Whether in the realms of transport, aerospace, oil and gas, defence, leisure, or medical, the application of an advanced adhesive selection tool promises to bring about positive transformations and advancements.
Key Findings
- Data repository including over 1000 technical data sheets, safety data sheets and technical articles
- Generated a database of listed adhesives, encompassing adhesive type, properties, application areas, compatibility, performance data, safety ratings, and other pertinent attributes
- Trained a domain-specific Generative Pre-trained Transformer (GPT) model using the resources obtained from database
- A data library was created
- Created a user-friendly graphical user interface (GUI) to facilitate the input of necessary information by users
Impact
The project's key findings confirm success in meeting its objectives:
- Accurate and fast application specific adhesive selection and recommendations
- Reducing risk of incorrect adhesive selection
- Streamlined and efficient decision-making processes
- Making AdhesiveGPT accessible to experts and novices alike using user friendly GUI
- TWI can leverage this capability effectively to assist member companies with their adhesive bonding needs