Introduction
In the first two weeks of my internship at Abatek in collaboration with Daetwyler, I gained insight into the manufacturing process, from when a customer places an order to the final silicone product. This exploration helped me identify my focus area, which was the design tooling department.
AI-Driven Simulations for Design Optimization
This project started from an idea I saw in an R&D presentation. Although I didn’t have prior experience with this, my interest in AI, developed through a course I took, motivated me to explore its potential for design. Using a database with force and dimension data, I implemented Google’s Vertex AI, a machine learning tool, to analyze the correlation between these variables. The goal was to identify dimensions that satisfy specific force and stroke requirements.
To deepen the analysis, I applied chain regression, a method where I used one variable to predict the next. For example, I first used force to predict the angle, then used that angle to predict the length, and finally used both to estimate the thickness. This layered approach allowed me to map out how each dimension affects the overall force.
The two images below showcase this process. Both graphs displays the length of each part and the statistical correlation between each of them. These simulations could help refine design efficiency and accuracy for future projects.
Part 2: Chain Regression for Predicting Dimensions
After testing the AI-generated data, I found that the results were very similar to those from our current Excel-based program. We are now using these AI-generated dimensions to produce a dome and measure its force for accuracy. If this approach proves more precise, the AI tool could become a valuable asset for optimizing design. It could be applied to various tasks requiring quantization, such as shrinkage calculations or key noise reduction.
Conclusion
Looking back on my work with AI-driven simulations at Abatek, I can see how important this project has been for my understanding of how artificial intelligence can improve design processes. I started this project with a curiosity about how machine learning could make our designs more accurate and efficient. As I worked with Google’s Vertex AI and analyzed the data, I learned a lot about how different factors relate to each other and how these relationships can help guide design choices.
One of the most satisfying parts of this project was seeing that the AI-generated results were very close to those from our current Excel program. This confirmed my belief that AI tools can be really useful for improving design work. Using chain regression to predict dimensions helped me understand how different design elements are connected, making it easier to make better decisions.
This project also showed me the importance of being flexible in design and engineering. As we keep testing and improving the AI tool, I’m excited about its potential to help with other tasks, like shrinkage calculations and reducing noise in keys. Overall, this experience has not only improved my technical skills but also strengthened my problem-solving abilities and highlighted the value of teamwork in driving new ideas. I look forward to using what I've learned in future projects and exploring more ways technology can enhance design.
Hall of Projects
© 2023 All rights designed, developed and definitely owned