The sensor networks described here could have a dual purpose, both for inspection as well as for longer-term infrastructure maintenance. I keep advocating for infrastructure spending as a way to both jump-start the economy and also to embed high-ROI electronics (like these sensor nets) within said projects. Whether it will happen is another story. (Unfortunately, it won't.)
While it is undoubtedly true that the deflection of a bridge under a specific load does increase as strength elements fail, it seems like it will be quite a challenge to find the actual correlation of that data with real failure.
I would offer instead a means of change detection that would need fewer sensors, a much less complicated algorithm, and has already been demonstrated as working, although on a much smaller scale. That method is to monitor the bridge resonance, both frequency and decay rate. The immediate benefit is that fewer and easier to calibrate sensors could be utilized, which the fact is that accurately calibrating a strain gage installed application is complex and requires both skill and patience. (Just ask any of the load cell manufactrers.) An accelerometer for the lower frequency , lower amplitude ranges is much simpler to calibrate.
While a change in resonance would not point to the exact location of a fault, it should be a very reliable indicator that a failure has ocurred.
I did not see or perhaps I missed a mention of an ultrasonic crack detection sensor. I would think putting a few of those in strategic locations on the structure might be helpful but perhaps they are too expensive.
So what are the warning signs that a bridge is about to reach the end of its safe life? Extra deflections past some norm? Cracks that begin to spread more rapidly? Perhaps unusual noises when a load is imposed? Any other ideas?
What is the cost associated with these sensors and this kind of monitoring? I am sure it is less than the cost of a catastrophe.
Hey folks: Glad to see a few interesting comments. But you need to know the subject matter is hardly new. My firm has been doing this type work on bridges for over 10 years with excellent results. Makes me wonder why the author or the PhD student mentioned thinks this technology is some new, whiz bang development. It's a simple fusion of sensors, wireless and the Internet - reasonably well understood technologies. The biggest issue facing our industry is when the FHWA or AASHTO will start demanding its use instead of relying solely on visual inspection, which is highly variable and quite subjective.
Also, there is no reason to expect a "cloud" algorithm will ever resolve all the sensor data in a magic puff. Determining condition or remaining life span for a bridge is not a subject that can be treated with precision mathematically without human intervention. It still takes a lot of engineering judgment and must be left in the hands of highly experienced bridge engineers with insurance coverage and frankly, something to lose.
I would like to remind all of you that I-35W was given a "clean bill of health" by the University of Minnesota a few years before it collapsed. Expecting a University PhD student to make these determinations and then using that information for bridge management is the height of foolishness. So, while the article is interesting in context, I hope I've added some much needed realism to the comment stream.
On www.aositilt.com you can find numerous high resolution tilt sensors and signal conditioners. For an analog output a set of sx-003D-NULL high resolution sensor and an EZ-TILT-3000 module were used to provide clients angular resolution of about 0.001 arcdeg. If a small system like that is installed in a flex joint of the bridge, then angular changes of lets say 0.05 for 3000lb are easily detected and can be differentiated from other diflections. This is a very low cost method to count trucks vs cars crossing the bridge.
It is true that you can find other suitable sensors and modules on that site, but this one worked great. the only two additional items needed were regulated power source and a DAQ.
You are correct. our client did exactly that. The data was used to count trucks and cars based on the diflection. then it was used to monitor slow degradation of the curve and try and predict service of the bridge.
This seems like a really compelling example for sensor networks and one that could really deliver huge benefits. What about using that sensor-collected data and feeding it into FEA or another type of simulation analysis tool to pinpoint and perhaps correct problem areas and leverage that data for future bridge designs?
Doug: I have to laugh at your example of a talking GPS. Thing is, it's not so far-fetched.
It is clearly very interesting. We supplied many tilt sensors to be embedded into small bridges to monitor diflection and to differentiate between small and large vehicles. Very interesting stuff. Keep the good work.
Next Chuck, you're going to be telling us that the data from the strain gauge sensors can be transmitted to Blutetooth in our cars and we can know before we approach a bridge that it may be unsafe for the size of load we're carrying. A voice will announce: "Buirdge may collapse, Seek alternate route." Cool stuff.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This discussion will examine what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.