April 17, 2018
On an average day in India not so long ago, the circuit breakers on a single powerline got tripped. That caused the breakers on another line to go down. Then another. Then another. It happened again and again throughout their power grid, leaving more than 300 million people in the dark for 15 hours.
A few years later on a highway in China, construction and a spike in traffic created some congestion heading to Beijing. Fender benders followed. A few cars broke down. The situation descended into a major traffic jam that stretched for 100 kilometers and lasted for 10 days.
These scenarios were unrelated, but they had one key thing in common—small failures snowballed into full-blown catastrophes.
In his National Science Foundation CAREER Award-winning proposal, A Scalable Optimization-Based Framework for Modeling and Analysis of Cascading Failures, Assistant Professor Makan Fardad is tackling these cascading failures by developing a mathematical framework to expose the fragilities that exist within infrastructure networks so that they can be amended before causing large-scale failures.
With all our technology, how do minor faults turn into such big problems?
Normally these networks are stable. They can deal with most disturbances and shocks—even big ones. But, they are still vulnerable to some disturbances, shocks, and failures. If we know where to look, we can find fragilities that would allow even small shocks to destabilize the network, build momentum, and become massively amplified and propagated by the network’s natural dynamics.
In India’s blackout and China’s traffic jam, the initial failures most likely had natural causes, like weather or human error, which are unavoidable. We can only aim to identify networks’ weak spots and strengthen them before they create a cascading failure.
This is especially important today because technology has also democratized access to sensitive infrastructure and that can allow malicious groups to target our networks with the intention of doing harm, for example, through cyberattacks.
Is it fair to say that this “snowball effect” begins with a single “snowflake?”
Sometimes, yes. But, there generally is a combinatorial aspect to these problems. Often, it is multiple weak spots failing together that cause the larger system to fail. My research team is interested in finding the most consequential failures in the network.
While individual shocks may never initiate a cascade, if chosen wisely, multiple shocks together will push the network over the edge and past the tipping point. But it is not feasible to check all combinations of shock locations in a large network. There are just too many possibilities, so another part of my research is to devise computationally scalable optimization algorithms to tackle this.
What motivated you to work on this problem?
I find this type of unexpected behavior in systems incredibly intriguing. I also enjoy tackling these problems from a theoretical standpoint. Networks that demonstrate cascading behavior are generally mathematically challenging to analyze.
I first decided to study this field when I became fascinated by cascading behavior in social networks, such as the propagation of social contagion, the spread of rumors and misinformation, and the promotion of positive social change and collective action. This was before the days of the #MeToo movement, but I can think of no better example today.
Back then, I wondered how the self-immolation of a 26-year-old Tunisian street vendor sparked protests that led to a wave of uprisings that spread across 13 countries in North Africa and the Middle East to become the Arab Spring.
It’s amazing how a single event can snowball into something substantial.