Back to Agenda
FHWA Unknown Foundations Summit
NDT Tehcnologies and Unknown Foundations
Alan Rock, Summit Peak Technologies
Runing Ahang, Advanced Numerical Modelings
Phil Sirles, Zong Engineering
[Slides for this presentation were all illustrations, graphics, etc. text below is just comments]
Figure representing soil, rock and cement. Engineer gets depths, puts together model, figures out bearing capacity of foundation. Comes up with safety factor.
Velocity factors, determine bearing capacity of foundation. If can use image it's very helpful, need determine what you're looking for. A 2-d picture is not good enough. Radius of shaft helpful.
How we go provide engineer tool to get number what is bearing capacity of foundation? make a model of ground. If find right properties, then give better picture of what going on.
Can create model like what in ground condition. See depths of bedrock, see piles. Go from just image to numerical model. If produces signal and match field collection, then if accurate can load model until it breaks. Very simple model, if find properties in model that match properties in field. What if take same techniques right to field, instead take field to lab, take lab to field.
With UF don't have ideal case anymore. Soil not uniform. If Measure properties in field can take in variations, then load model and determine safety factor. Decide to support or replace. Engineer can determine what dealing with in field.
Tube difference Model
This is part of FHWA project. Want to show what we're doing and how apply to UF. Built model with 2 million ping pong balls. Have tube model and field condition model, measuring compression waves in water. Look at difference between field and computer model. See compression wave. Small difference but it comes in. if I have computer model and field conditions, then I'll see difference. Reverse time analysis.
Can detect anomalies in ground that see in field but don't have in model.
Rock vs. Soil surroundings
Not promoting any data, one better than another. Can process full wave-forms. Not analyzing by hand, back propagate in the model. can see if rock or soil. Shaft section.
Don't analyze difference in signals, put the different signals back in model and then find them in the model. pull these out, then reconstruct in high resolution where they are. Can see where no cracking is able to pull waves back. No change in any arrival time of signal.
Can't find strength of mats in ground. Know there's no rubber-y material in ground. Because can find cracking can find zero strength. That's where crack go fail.
Want produce a numerical model that can load, excavate, grout, how much difference is grout go make, can change model. a picture, the engineer is go have make conservative guesses. This gives more objective tool for engineers to evaluate foundations.
Goal is load capacity
Ground images indirect at best
Numerical modeling is anyone's guess—it's got a bad name. This isn't numerical modeling, objectively try guess numerical samples, but we want something we can't fiddle with.
Info is in the data. we can take data, build model the same way, and then make model produce same signals see coming out. Now it's able to give more accurate images. How entire signal behaves.
Reverse-time analysis has potential to extract info from entire signal; give better images; give better estimates of load capacity. Want to develop long-term corrosive cracking. Have potential to analyze foundation at certain point and time but can do longer term model to look at corrosion. Can estimate life expectancy of shaft if know how it is being flooded.
Q: Kornel: model has non-linear spring?
AR: In this case used linear.
Q: Can use different model to crack appropriate springs, let program figure out what is cracked?
AR: Yes, if he goes to any 2 different concrete shafts can notice different signals for concrete shafts.
Q: Let model do it on its own? Don't' intervene?
AR: Model does it.
Q: SS: Changing mass effect on ping pong ball
AR: Model still behaves the same. If I load this, what is bearing capacity? If I have xxx, be confident my model start behaving same way when I load it xxx. Has same behavior.
Q: SS: How long take to build model like this?
AR: Building one take s10-12 hours. Depends on where sensors are. Find what's down in ground.
Q: Khamis: Can look into foundations, see if concrete is deteriorated, what strength is.
AR: Yes but can't get unique model out of it. I can reconstruct what's in ground.
Back to Agenda