top of page
  • Writer's pictureDaniel Lim

Learning AI/ML/DL for Data-driven design

In the past, I ventured into the world of programming and artificial intelligence (AI) through various online and on-site courses, even enrolling in academic programs like UC Berkeley's Introduction to Artificial Intelligence (CS188). However, my pursuit lacked a clear purpose, leaving me wondering why I was learning these subjects.

It wasn't until I delved into simulation-based and data-driven mechanical design that I began to grasp the immense potential of integrating artificial intelligence (AI), machine learning (ML), and deep learning (DL) into my ongoing research projects. Two of my primary research endeavors illustrate this point vividly:

1) The data-driven design of multifunctional stealth metamaterials link

2) The data-driven design of upper limb exoskeletons link

Both of these areas heavily rely on human intuition, demanding extensive hours and repetitive iterations in the design process. For instance, in the multifunctional stealth metamaterial design, we discovered how slight alterations in the unit cell structure could significantly impact both mechanical properties and radar-absorbing capabilities. Achieving the optimal design for a specific application (e.g. airfoil) required thousands of simulations. Incorporating AI/ML/DL into the process promises to automate these tasks, reducing the time and computing resources required for optimization.

In the case of the upper limb exoskeleton, the challenge lies in customizing the exoskeleton to fit various hand shapes, a task that has kept the cost of partial prosthetic hands over $5000. Despite the benefits of 3D printing that reduce fabrication time and cost, the time-consuming customization process remains a bottleneck. Imagine f we could streamline this process by allowing human designers to focus on whether the AI-generated model suits the user or not.

Recognizing the immense potential for progress, I've come to realize that incorporating AI/ML/DL into my research toolkit is not optional but essential. I'm taking my first steps in this direction by enrolling in online courses through Coursera. I plan to document my learning journey as a trailblazer and share my experiences with others in the future.

bottom of page