Undergraduate Admissions

Follow us!

What on earth is he doing?

Why is Dano trying to knock down Millikan library with a sledgehammer?

This summer I’m working on a distributed seismic network. It will hopefully be a faster network, with thousands of nodes, which gives us the potential for early warning.

Everybody knows that big important organizations like the USGS (United States Geological Survey) have seismic sensors to detect earthquakes. These sensors are very high quality and are very expensive ($10,000) to install. Our plan is to use really cheap sensors and hook them up to regular people’s computers. That way, you have many more sensors, which makes for a more robust network.

When the system is working, each individual computer will keep track of recent accelerations experienced by that system. If there’s shaking above some threshold, then that client computer will send a message to the server saying “Hey, I’m getting shaking.”

Then, on the server side, we use Bayesian analysis to determine if there was an earthquake and, if so, where it was. Basically, you consider all the possible places an earthquake could have happened by dividing your map into a grid. For each place, you calculate a probability that an earthquake happened there, based on the messages the server got and the probability of getting exactly those messages if an earthquake was happening. If an earthquake is happening, one of the grid points will jump up to 90% or above very quickly. It sounds computationally intensive, but with parallel computing across multiple processors it’s very manageable.

Hopefully our system will be able to go from earthquake to early warning in six seconds. That would give about twenty seconds of early warning. It doesn’t sound like a lot of time, but automated systems can really help. One of the big problems with earthquakes is that the doors on fire stations get twisted so that they can’t open. Then, the fire engines and ambulances can’t respond to the earthquake until they tear down their own door. An automated system could open up the door every time we think there might be an earthquake, just in case.

As part of this, we have to validate our sensors. If our sensors can’t detect seismic activity, then the whole project is a bust. So, I have to compare our sensors to the professional sensors. Luckily, since the USGS has an office on campus, there are four professional sensors here. I asked one of the guys who helps maintain it what I should do and he pointed out that one of the sensors is in the basement of Millikan library and that it’s unlocked. He said that it would be no problem for me to set up my sensor next to the professional one. So, I got a little Eee PC and hooked up my sensor right next to his. 

The idea was to wait for significant seismic activity, then compare the readings from both devices. We got tired of waiting, though, so one of my professors suggested that I “help it out” by creating some seismic activity. So, I got to use a sledgehammer! I used a piece of plywood to pad it so that I wouldn’t break up the concrete, but I just wailed on the floor for a full minute. One of the librarians came over and asked me some very pointed questions about what I was doing. It was obvious that she was uncomfortable with me taking a sledgehammer to her building! Unfortunately, I had turned my data logger off right before doing it, so I had to go back later that day and try again. This time, it was late in the evening, so no librarians were disturbed. Today I’ll analyze the data and see how good our sensors are.

Dan Obenshain