High-speed gearboxes are critical core components in the energy industry, including fluid and oil applications. Because they must be custom-designed for specific installation environments and surrounding conditions, standardization is difficult. This ...
High-speed gearboxes are critical core components in the energy industry, including fluid and oil applications. Because they must be custom-designed for specific installation environments and surrounding conditions, standardization is difficult. This requires professional engineers with years of design experience and know-how to specify the size, component specifications, and other design parameters. This process is time-consuming and often leads to over-design for risk avoidance. While programs for optimal gear design are being introduced, they require expert-level gear engineering knowledge. Recently, a shortage of specialized technical personnel has become a common issue across industries. In the gearbox design field, there is a significant shortage of experienced designers capable of considering all relevant elements, such as gears and bearings.
This research explores how AI, widely utilized across all fields, can be applied to the design domain. It specifically targets the high-speed gearbox sector, attempting to practically implement AI-driven design solutions. This research serves two purposes: finding methods to integrate AI into the high-speed gearbox design process and implementing practical solutions for the high-speed gearbox sector, where the shortage of design personnel is severe. Since AI relies on data-driven learning and inference, data is paramount. To secure data, design parameter-related data was generated based on bending stress and contact stress—the fundamental design formulas for gear tooth profiles. This data was then used for training and inference via an AI Large Language Model (LLM). This method of extracting design-related data was termed an AI-Powered design approach. We verified whether junior engineers could utilize it in the design assistance phase and whether expert designers could expand its application to the design verification domain. Most companies, treating data as an asset, are actively seeking investment and application for AI utilization. While attempts to leverage AI technology in the manufacturing industry are ongoing and spreading, there remains a sense of uncertainty about where to begin.
This research explores and implements a solution enabling even novice designers in manufacturing to leverage the design capabilities of expert designers based on AI-Powered designer experience data. The paper implements a data generation approach targeting high-speed gearbox design, where the AI model learns and infers using an AI LLM. The key design parameter learned and inferred by AI is the center distance of the high-speed gearbox, which is central to layout sizing design. Compared to the traditional iterative trial-and-error approach for optimal sizing design, the AI-powered high-speed gearbox design method utilizing AI LLM proved highly effective. Data for training was generated according to the existing design methodology, and the interrelationships among the acquired data were organized in tabular form. This method applied the AI LLM RAG technique to utilize the generated data as external data. The results of inferring the design center distance of the high-speed gearbox using this method confirmed a significant reduction in design time. It is anticipated that applying this approach to similar design areas could assist experienced design engineers with expertise and know-how or, when necessary, enable junior designers to substitute for senior designers.