Prompting a Bridge Collapse: Classic Textbooks vs GenAI

My To AI or Not to AI: Before You Prompt, Try Physics newsletter edition kicked off our series on domain specific generative AI. We recalled the shattering Memorex wine glass commercial of the 1970s to introduce the physical phenomenon of resonance. Prime-time viewers of the era knew and understood the phenomenon, which illustrates the importance of knowing the fundamentals.

👉 Prompting a generative AI chatbot to learn why the wine glasses broke would be equivalent to cutting butter with a chainsaw. Chatbots consume 10 times more energy than a simple search. 

However, we also showed the point where generative AI can augment a practitioner, such as an engineer. The genAI prompts have to be domain-specific. We then introduced a framework to construct domain-specific prompts for engineering workflows.  

👉  Engineering firms can incorporate generative AI into their engineering workflows to achieve faster, higher-quality results with fewer personnel.

In this edition of my newsletter, we will examine a real life case in civil engineering  to explore ways to incorporate domain-specific generative AI in engineering workflows. 

The Case of Galloping Gertie: The Tacoma Narrows Bridge

The resonance phenomenon depicted in the Memorex commercial of the shattering wine glasses can be very serious. The phenomenon can cause anomalous behavior and destruction of critical physical systems.  

In this context, the collapse of the Tacoma Narrows Bridge in Tacoma, Washington, in 1948 is noteworthy. The bridge, called “Galloping Gertie,” collapsed six months after opening.

The explanation of the collapse was initially thought to be due to wind conditions that induced resonance driven excessive vibrations. Lessons from this collapse are very useful for civil and mechanical engineering workflows.

Using Generative AI to Understand the Galloping Gertie

I used the domain framework I proposed in Prompt Engineering ≠ a Shortcut to Expertise. The salient elements of the domain framework are the fundamental courses in engineering such as solid mechanics, fluid mechanics, heat transfer, dynamics of structures, etc. Understanding the content from these courses enables one to construct domain specific genAI prompts. 

Therefore, to understand the root cause of the bridge collapse, I honed in on “dynamics of structures” in the framework. I recalled the undergraduate class taught by Professor Steidel at UC Berkeley. And, the authoritative book on dynamics of structures by Den Hartog that also has an explanation. 

I wanted to see if I can get to the level of understanding I have absorbed from the textbook. And, more importantly, if the explanations have been further refined since I took my course in 1982.

So, I tried Google Gemini. I was impressed. 

I decided to select “deep research” to generate a report with the question below.

The first research report entitled, The Collapse of the Tacoma Narrows Bridge: A Case Study in Aeroelasticity and Bridge Design Evolution was extremely thorough.  The explanation attributed the failure to aeroelastic flutter. 

Next, I wanted to go deeper into aeroelastic flutter, and look for an explanation of how to mitigate it. This is a phenomenon I had to take into consideration during my own work as a design engineer in hard disk drives. It also reminded me of airplane wing designs.

I posed the following five questions to Google Gemini and asked for deep research:

1. Why did the Tacoma Narrows Bridge collapse? 

2. What were the lessons learnt? 

3. How were the lessons incorporated in the replacement bridge?

4. Are there other systems that can be degraded by aeroelastic flutter?

5. How is aeroelastic flutter mitigated?

Gemini said the following and went to task.

I let Gemini do the research work as I continued doing mine. The second report took 25 minutes as it researched more than 100 websites. I got concerned about the Joules I was burning. However, in the interest of learning, I let it proceed to completion. The generated report lacked figures but was otherwise excellent.

Textbook Explanation 

I compared the detailed report to my understanding as taught by Den Hartog in the classic book Mechanical Vibrations.  I know this explanation off the top of my head. 

Below is the sketch recreated from Den Hartog’s book. As shown in the figure, the original bridge that collapsed had an “H beam” roadbed. Wind blowing across this bridge cross section caused eddy currents within the H beam shown in the figure. The resulting twisting or torsional oscillations aligned with the natural frequency of the bridge became severe, and the bridge eventually collapsed.  

Engineers extracted the lessons learnt and built the new bridge with a “box” construction for the roadbed. The new box design is stiffer and more resistive to torsional vibration. 

Stiffness to mass ratio determines the natural frequency, so higher stiffness shifts the natural frequency higher as well. The box design has holes to let the wind go through. These holes eliminated the edges that can trap the blowing wind, which would have created the dangerous eddy currents.

💡Key Takeaway: I love the simple explanation in the book. However, I also like the new content on aeroelastic flutter that Gemini provided. It was exactly what I was looking for to augment what I learnt in 1980. 

The Road Ahead

In this series of newsletters, we explore domain-specific generative AI by considering real-life examples. I’ll share another real-life case study related to data centers from my experience in the next one. These case studies will show you how to incorporate generative AI into engineering workflows. 

Given our example today, it looks like LLMs can be a great companion to traditional knowledge. But, I’m not getting rid of my classic textbooks yet.

About the Author

Former SVP, Chief Engineer, and Senior Fellow at HP, Chandrakant is a leader in AI, energy-efficient computing, and sustainability. He is an IEEE Fellow, ASME Fellow, member of the NAE, and the Silicon Valley Engineering Hall of Fame. Follow me on LinkedIn or email me at [email protected].

Want to stay ahead of the curve with insights into the newest advancements in Physical AI? Subscribe to Chandrakant’s newsletter at GenAI Works.

🚀 Boost your business with us—advertise where 10M+ AI leaders engage

🌟 Sign up for the first AI Hub in the world.

📲 Our Socials