I have encountered engineers questioning survey accuracies in a recent project. We have been doing precise levelling, self-closed double run lines, each lines involve at least 2 benchmarks, consistent benchmarks and levelling routes for the monitoring points, INVAR staff, two-peg test before every field work with collimation error compensation applied (and check again to see if it works), only foresight and back-sight with no IS just for sure....The job's requirements are tight as the alarm will turn on if the settlement reached 3mm.
And then the engineers always question if the survey quality/accuracy is good enough as they see a couple of +/-0.5mm along the way. Apparently they don't know about random error theory and they think survey error is something wrong or mistakes committed by surveyors. I tried all I can to do the explanation, and I guess many of them fall into a fallacy: We don't know if the surveyor's right, but according to what we (engineers) know, this structure was built on foundation on a rock mountain so there is no reason for the movement, therefore it must be the error in surveying causing those strange readings.
Have you encountered these situations and how do u eventually work out a solution??ÿ
Chris
alarm will turn on if the settlement reached 3mm
Now that's a tight requirement.
An engineer should know that there is always some uncertainty in a measurement. If you repeat with consistent results they should believe your data.
You are expressing it with a confidence % aren't you? If it is a super critical measurement I would not be happy with 95% confidence (wrong 1 in 20 times) but would want 3-sigma or 99%.
If you are trying to detect 3mm of movement, you better be confident in your observations at the sub-mm level (95% confidence). That's a tough mark to hit...you better be setting control yourself and locking it down even tighter than that if you want to have any hope of meeting specs.
From the COE Deformation Monitoring Manual:
If you're not hitting that mark on every single monitoring session, you're going to see your results periodically (or regularly) trigger a warning. Especially if there is slight movement that is under the threshold.
Been there, seen that, learned from others' mistakes, won't touch monitoring unless and until I send the client a complete report on what it will take to meet their specs and then get approval to actually spend the time and money on it. Don't do it unless you run your preanalysis and prove out that you can hit specs. The liability isn't worth it.
Send him the registration information for Stats at the local community college...
Rock doesn't move? Tectonic plates don't shift?
local community college...
Do they have those in Hong Kong where this poster's profile says he is?
I'm curious where the thing is that you are monitoring and where your control is. If you are trying to detect 3mm, can you be sure your control has not moved 3mm? If you have truly taken enough measurements to show that statistically your data is correct, then you just present your evidence and the engineers have no leg to stand on. If you haven't taken enough, then there is not much argument you can make against what the engineers are saying. They might be right and you might be right but nobody knows.
While the original poster provided some details about his procedure, I wonder whether he applies the corrections used by the US NGS described (with links) here:
https://geodesy.noaa.gov/locus/equip.xhtml
See the LOCUS portion of the following presentation:
Geodetic leveling is very exacting. Working at the few millimeter level of precision requires consideration of elements not usually considered significant by surveyors.
Unfortunately in the first link there is a non-functioning link to a presentation titled “NGS precise leveling workshop presentation FTP site.” I have contacted the NGS webmaster to see about getting it fixed.
BTW, U.S. surveyors might be interested in the NGS training on geodetic leveling available at their I&M site. See: https://geodesy.noaa.gov/corbin/calendar.shtml
For the original poster: Do you solely provide the leveling data and associated files? Do the “engineers” perform the adjustments and analysis? Deformation analysis is a specialized field with some serious maths involved. Not knowing the extent and intent of you project, I defer further comment.
Good luck!
DMM
Rock doesn't move? Tectonic plates don't shift?
don't dare mention the glacial isostatic adjustment
...oops. sorry about that.
I'm curious where the thing is that you are monitoring and where your control is. If you are trying to detect 3mm, can you be sure your control has not moved 3mm? If you have truly taken enough measurements to show that statistically your data is correct, then you just present your evidence and the engineers have no leg to stand on. If you haven't taken enough, then there is not much argument you can make against what the engineers are saying. They might be right and you might be right but nobody knows.
3mm is pretty tight. If it is all relative (meaning that the whole thing can go up and down, but each pier needs to stay within 3mm of each other regarding settlement) then that simplifies things.
I have done monitoring where the oscillation in the measurements was greater than their spec. I explained that the error budget from the time and equipment they provided would not provide an error budget smaller than their spec. Often simply having a different crew running the equipment will create noticeable changes.
THEN I asked, "What is the real number, how much really matters?"
The senior engineer shook his head and grumbled something about the "auto optimizer button" or something in the design programs. If the original engineer designs something that cannot be built or measured, then that is their ignorance at work, not ours. The specs for deformation are created by someone with zero knowledge of measurement, in most cases.
But, to answer their question, you could run a baseline of the same length and elevation change as the project using the same techniques and equipment a few times (AND a different crew each run). Then run the results through something like StarNet for analysis and have some real answers.
And lastly, no one who has spent time outdoors and heard the sounds of the earth will imagine that a stone mountain is without any movement.
@bill93 Thanks for the reply I think I missed out some critical information in my original post.
The contract specs here says misclosure of levelling lines should be within 1 root k mm where k is the length of levelling route in km.
The benchmarks are gov controls, they are steel rods in reinforced concrete box on buildings corners. I don't have too much information about how the foundations of the buildings were created. Perhaps I should be looking at the old as-built drawings. BMs used are at least 100m away from the site where the engineers suggested that it should be the influenced zone within it. BMs are counter-checked and reviewed every month with other gov benchmarks further away from the site. The levelling route is starting from BM1 and go to BM2 and then to the monitoring points and then back to BM2 and finally back to BM1.
We use a digital level with 0.3mm sd per km double levelling with invar staff, if the misclosure exceeds 1mm, we redo the hole field work. Again there is no IS involved. As the monitoring readings are presented everyday, our resources only allow one levelling work per day. we will have repeated measurements (3 times) on each setup and if the staff reading deviates with more than 0.2mm spread we have that particular measurements redo.
For adjustment, the difference in provisional heights is corrected by the number of setup. Since we have fixed the levelling route and have equal leg length between foresight and backsight, the assumption was that collimation error is always eliminated and then we weight error in setup all the same. Adjustment by distance is also computed to counter check the adjustment by set-up, normally the readings don't differ by more than 0.5mm.
And we have monitored the points for almost a year, and the figure spread within +/- 1mm, and never touches the 3mm mark. Honestly I think this is what the resource allows. I understand that for the alarm being at 3mm, one might think well we should have a fluctuation within say 0.5mm only. But then as long as it is a daily monitoring to be reported before cob after the fieldwork, I cannot have more readings and decide which is the outliner and which should not be. And if I have a reading now differs from the previous 7 days, it can be (i) the benchmark actually moves; (ii) the monitoring point moves; (iii) observations are on both sides of the normal distribution curves (chances of this is slim as we are still bounded by the misclosure spec of 1mm of a 1km run).
And more importantly, as for the issue of confident level, I think we can only judge on the closing loop misclosure (e.g. instrument spec claims 0.3mm 1 km run for 1 sigma), and the spread of every staff reading. An impression I've got from those engineers was that they have already think the object being monitored there should not be moved (then what is the point of the monitoring works), they misplace the random error theory at the final product, the settlement value measured. The +/-1mm fluctuation is a result of random error in survey, the tiny little movement of benchmarks and monitoring points (even in the absence of vigorous site work), and we may never know the proportion between them.
I think there might be some hidden agenda in the engineer's mind but it's not my job to guess it, i want to know every thing in surveying possible I can possibly do, maybe not in this job, but in the future.
Cheers,
Chris
@rover83 Hmm we don't have this standard here but I think I can take it as a reference. But if they are well aware of that the alarm would not be 3mm...Anyway it is kinda hard to convey random error theory to them, where error is not mistake, hang on, we all learn statistics in all applied science fields, do we?
@dmyhill Yes but I'm not sure if ppl from other professions other than surveyors are really willing to listen, at least it is not happening here in this city. Giving a whole lecture starting from error theory to how to weight and do adjustment to preanalysis to postanalysis simply is way too clumsy in a busy workplace.
You've got an engineard running loose in the wild. That engineard needs to be corralled.
Clearly a beaver type obsessed with numbers.
Also called an anal retentive.
If it works for a government agency then the pointless costs created by its actions need to be highlighted somehow.
This probably won't help your situation but I only hire engineers, not the other way around, except one that needs me for corner searches, those scare him.
Don't let the engineer be the boss, be the engineer's boss, then everyone is happy.
don't dare mention the glacial isostatic adjustment
I once had a gawd-awful conference call with some Spanish engineers and surveyors at Ferrovial trying to justify why I didn't match the published elevation data from 60-year-old benchmarks.
"Well, 35.5 million years ago a bolide struck what is now Virginia, diverted what is now the Susquehanna River, and created the Chesapeake Bay...."
The Chesapeake Bay Bolide Impact: A New View of Coastal Plain Evolution - USGS Fact Sheet 049-98
@bill93 Hong Kong - good catch! The engineer is probably part of the CCP and thinks he can dictate what the rock will do.... Physics and such, be damned!
Clearly a beaver type obsessed with numbers
actually just the zero
the belief that zero is anything except a place holder on a number line is the issue.
there is no true zero
like Einstein said...It's all relative...
some Spanish engineers and surveyors
Weird. Pretty sure they have tectonic plate movement and subsidence in Spain, just like over here...
Concerning the original question, which I didn't actually answer in my first post:
Every session network/LS adjustment yields standard errors for each point in 3 directions. They always vary by a small amount, but we are trying to detect small amounts, so it's critical to evaluate whether they are significant from a statistical standpoint.
In my report spreadsheet I have columns for "Session Deltas" "Session Delta Confidence", and "Statistically Significant?"
The first is just the absolute deltas between sessions. The second is computed on a session-by-session basis. Take the square root of the sum of the squares of the current and previous session standard errors, and double it for ~95% confidence.
The "Statistically Significant?" column is just a true/false evaluation, and looks at whether the Session Deltas field is bigger than the Session Confidence field. If so, it's likely that there is movement.
Incorporating those additional columns in there makes it clear that there is variation in each session, and that we will always see some "movement" just based on random error, but there is a threshold below which we cannot say for sure that this movement is statistically significant.
That has gone a long way toward quieting the statistically challenged among our clients.
You still need to have rock-solid field procedures, QC/QA, and adjustment routines, because if you're sloppy it'll show up in the tables - and worse, it might show movement where there is none, or no movement when things are shifting.
still need to have rock-solid field procedures
no kidding.
like even better being certified for the quality of the data being collected like in the NGS etc type designation of 1st order 2nd order etc.
using a .5" total station with a 10' technician yields obvious results.
Happy Friday!