## Welcome

Tuesday, February 7th, 2023I jumped into this job at about 80% completion. Its a water crossing tunnel that goes down a shaft about 150', then flat for a few miles across at water body, then up a few hundred feet on the other side. It's a 30' diameter shaft that will hold 3 pipes that carry water. I have good, tight control at the surface of the shaft.

They are about to start placing the 3 pipes from the elbow in the tunnel and make their way up the shaft. They want me to "precisely" layout the elbow at the bottom of the shaft and then check for pipe extensions as they are welded and placed going up. I believe there are 6 lengths of pipe total that they will backfill with concrete after they've been placed. They haven't given me an actual horizontal tolerance. Vertically they've given me +/-10mm. As the pipes are placed there are two working decks that will concurrently rise with the pipe blocking my view between the bottom and top. They also make for a very unsteady platform.

The previous company had control at the top of the shaft facing down and would do a resection every time. I explained my concern with the geometry of that resection and stability of the work deck.

I threw out a few ideas but I'm hoping to hear what this community has to say? Does anyone have experience with this kind of work?

If you have tight control at the rim looking downward, with at least three points spaced evenly around, and you're observing angles & distances, I wouldn't have a problem with a resect from the middle of the shaft bottom.

I seem to remember running the numbers in StarNET for a similar scenario (slightly wider opening and slightly less deep, but still same situation) and they checked out. Ultimately we didn't have to do the work, but the math was solid. With good vertical on all three points, 10mm is attainable.

You're likely going to have more problems with the work platform stability than the resect, but as they move higher and higher it will be easier to sidestep that problem and set an instrument up top to look downward.

It's either that or they construct a platform at the top for you to drop a big old heavy plumb bob on piano wire.

If you have tight control at the rim looking downward, with at least three points spaced evenly around, and you're observing angles & distances, I wouldn't have a problem with a resect from the middle of the shaft bottom.

I seem to remember running the numbers in StarNET for a similar scenario (slightly wider opening and slightly less deep, but still same situation) and they checked out. Ultimately we didn't have to do the work, but the math was solid. With good vertical on all three points, 10mm is attainable.

You're likely going to have more problems with the work platform stability than the resect, but as they move higher and higher it will be easier to sidestep that problem and set an instrument up top to look downward.

It's either that or they construct a platform at the top for you to drop a big old heavy plumb bob on piano wire.

Maybe im looking at this wrong but the snip below shows the ideal geometry for a 3-point resection on the left, and the geometry I would get from this example on the right, to scale.

Not only is it not ideal but it would be hard to design worse geometry from 3 points. Based on the geometry I think I could be ok vertically but id be id be relying on under 4 degrees of separation on my resection.

I don't see how the math checks out, but I've been wrong many times before.

Yea I floated the exact situation you describe with the piano wire and plumb bob. It's tough with all the moving parts in and out of the shaft, lasers may be doable. Id still need to get my elevation down each time.

I've never done anything like this but could you mount a few 360's, peanuts or reflectors (if they're accurate) on the walls in a few places on the way down and traverse it for control at any elevation?

Something like this. This is to scale at 30'Dx150'

I don't see how the math checks out

Threw a quick StarNET job together over lunch. Here's the math:

Three control points spaced equally around a 30ft diameter circle at 150 ft elevation, each with a network accuracy of 0.005ft/1.5mm in the NEZ components.

Resect point in the center of the circle at zero elevation. Centering errors of 0.005ft.

One round of observations yielding mean angles with standard errors of 2" horizontal, 4" vertical, and 0.005ft slope distance.

Additional observations could be taken to improve things, but that's about 2mm horizontal by 2mm vertical error ellipse for the resect position. Obviously this is barring any blunders, systematic errors or outlier random errors, but it's definitely possible to do this and maintain 10mm accuracy for stakeout in both horizontal and vertical.

I've never done anything like this but could you mount a few 360's, peanuts or reflectors (if they're accurate) on the walls in a few places on the way down and traverse it for control at any elevation?

Something like this. This is to scale at 30'Dx150'

I was thinking this as well, not hard to throw in a anchor bolt and a threaded prism.

I don't see how the math checks out

Threw a quick StarNET job together over lunch. Here's the math:

Three control points spaced equally around a 30ft diameter circle at 150 ft elevation, each with a network accuracy of 0.005ft/1.5mm in the NEZ components.

Resect point in the center of the circle at zero elevation. Centering errors of 0.005ft.

One round of observations yielding mean angles with standard errors of 2" horizontal, 4" vertical, and 0.005ft slope distance.

Additional observations could be taken to improve things, but that's about 2mm horizontal by 2mm vertical error ellipse for the resect position. Obviously this is barring any blunders, systematic errors or outlier random errors, but it's definitely possible to do this and maintain 10mm accuracy for stakeout in both horizontal and vertical.

Would you mind sending me that job or posting it here so I can wrap my mind around it a bit better?

@rover83 I’m curious how the simulations compare to real world results? I’ve never known anyone that ran simulations like that so this is something new to me. 😮

I would just rent a laser tracker and spatial analyzer and never look back. No battling setting up tripods or dealing with steep angles. No worries about stability as i would set up semi permanent targets as i did each days work so even if they or wherever is convenient to mount the laser tracker it would be able to correct itself in real-time. At that distance you can measure the thread lengths . You can take and build the design into it from a solid works cad file or many other formats. As they build it you can correct up as you go. No pole rod to keep up with either. Just a shirt pocket with smr and maybe side puck. Not hard at all for you to have a demo and be able to run with confidence in about an hour on that truthfully. I don’t doubt you can do it with modern surveying equipment either. I have used piano wire in the past and plumb bobs. But I will have to admit for things like that the laser tracker is one sweet tool. I would say if it no more than a days work. The spatial analyzer folks might demo it for you and have an expert train you for a small fee. If it were close to me I would come show you just for a good lunch. Just wrap your head around xyz and transform whatever control exist and haul rump.

I’m curious how the simulations compare to real world results? I’ve never known anyone that ran simulations like that so this is something new to me.

Generally speaking, if the gear is calibrated, procedures are good, and the plan is followed, I have found that these numbers are solid. Done quite a few high-precision jobs (monitoring, dams/locks, heavy civil control, geodetic) where we used preanalysis to determine the optimal network configuration as well as the minimum number of observations needed.

I take the minimum number, bump it up by 30% or so to get plenty of redundancy, and so far I've always been able to hit specs.

I’m curious how the simulations compare to real world results? I’ve never known anyone that ran simulations like that so this is something new to me.

Generally speaking, if the gear is calibrated, procedures are good, and the plan is followed, I have found that these numbers are solid. Done quite a few high-precision jobs (monitoring, dams/locks, heavy civil control, geodetic) where we used preanalysis to determine the optimal network configuration as well as the minimum number of observations needed.

I take the minimum number, bump it up by 30% or so to get plenty of redundancy, and so far I've always been able to hit specs.

I didn't fully understand what was going on in that dat file. How did you get starting coordinates of your control and how did it process without any observation files? Something tells me you're doing something in Starnet I haven't seen yet and I'm very curious.

Something tells me you're doing something in Starnet I haven't seen yet and I'm very curious.

The PreAnalysis tool will take a set of a priori coordinates, a theoretical set of observations (using the questions marks in place of actual observation values), and use the input standard errors from the project and DAT lines to estimate the results.

If we know roughly where we are going to set points we can generate approximate coordinates for those points (we frequently use Google Earth or the Geolocation tab inside C3D and just pick a point from the aerials, or in your case I just created a circle at 30ft diameter and picked points around the edge) and then add the number of observations (1D, 2D, and/or 3D) we think we are going to need, and then run PreAnalysis to test what the expected results would be.

It's remarkably good at predicting likely results, but that shouldn't be surprising because it uses the same underlying statistical concepts (random error theory) that govern real-world observations and network adjustments.

Something tells me you're doing something in Starnet I haven't seen yet and I'm very curious.

The PreAnalysis tool will take a set of a priori coordinates, a theoretical set of observations (using the questions marks in place of actual observation values), and use the input standard errors from the project and DAT lines to estimate the results.

If we know roughly where we are going to set points we can generate approximate coordinates for those points (we frequently use Google Earth or the Geolocation tab inside C3D and just pick a point from the aerials, or in your case I just created a circle at 30ft diameter and picked points around the edge) and then add the number of observations (1D, 2D, and/or 3D) we think we are going to need, and then run PreAnalysis to test what the expected results would be.

It's remarkably good at predicting likely results, but that shouldn't be surprising because it uses the same underlying statistical concepts (random error theory) that govern real-world observations and network adjustments.

Thanks for the tip.

I had to fix a point and bearing otherwise starnet wouldn't process for me. I gave my other 2 CPs a standard error of 3mm and ran my project settings fairly tight. As good as I could realistically hope for. Even dropping the centering error to 0 (forced centering targets, unrealistic, even forced centering wouldn't be 0). Changing that actually didn't change the values at all, which I found weird.

I don't think Starnet is taking into account the geometry here, it's like its ignoring the elevation difference. When I alter the coordinates to be within 1 mm from each other my results are basically the same. Which in reality, is a basically one point resection. And the error should be massive.

Then when I swap axis and make it a horizontal solution I get more realistic numbers. Noting that now the error in the easting will represent my horizontal error and the error in northing will represent my vertical error (more or less)

And as a final confirmation I sucked the hz geometry in together and the error exploded as I'd guess it would.

This is what I came up with today for a proposal.