Trains.com

Subscriber & Member Login

Login, or register today to interact in our online community, comment on articles, receive our newsletter, manage your account online and more!

model track geometry car ?

4000 views
49 replies
1 rating 2 rating 3 rating 4 rating 5 rating
  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Wednesday, July 28, 2021 5:19 PM

ATLANTIC CENTRAL
First, what do you mean by alignment of points? Length? Gap on open side? Gauge?

these are fasttrack turnouts.  35+ on the west end of a passenger station, many more on the east end which took very long to get working

there are inconsistencies in construction. 

  • solder blobs between the points and stock rail preventing tight closer. 
  • distance between the points is narrow requiring more switch machine travel. 
  • throwbar is tight requiring more switch machine force
  • turnout not spiked near throwbar
  • as well as other problems

as we've been installing switch machines, manual inspection, some filing, bending, spiking and resoldering fixed many issues allowing a car to properly track both turnout routes using a tortoise machine to switch the points.    expect more issues with trains

i think it would be helpfull to be able to detect a point that is not tight against the stock rail using a geometry car and am now becoming more curious what an accelerometer can do

 

please recognize this thread is about the a geometry car, not layout construction.   this is what has been done and is not going to be replaced (not my layout).   

of course you may not do it this way on your layout.

greg - Philadelphia & Reading / Reading

  • Member since
    August 2019
  • 39 posts
Posted by hgodling on Wednesday, July 28, 2021 7:01 PM

If you want to go the optical route, you want to be sure there is nothing between the camera and the track. Even clear pieces will distort the image and mess with the contract making things more difficult. 

There is a software called OpenCV that you may find helpful. I don't know if this runs on an Arduino. I know a few people that have had some success in object tracking in openCV. It has a bunch of libraries that may save you some time and keep you from reinventing the wheel. 

  • Member since
    January 2009
  • From: Maryland
  • 11,130 posts
Posted by ATLANTIC CENTRAL on Wednesday, July 28, 2021 7:20 PM

gregc

 

 
ATLANTIC CENTRAL
First, what do you mean by alignment of points? Length? Gap on open side? Gauge?

 

these are fasttrack turnouts.  35+ on the west end of a passenger station, many more on the east end which took very long to get working

there are inconsistencies in construction. 

  • solder blobs between the points and stock rail preventing tight closer. 
  • distance between the points is narrow requiring more switch machine travel. 
  • throwbar is tight requiring more switch machine force
  • turnout not spiked near throwbar
  • as well as other problems

as we've been installing switch machines, manual inspection, some filing, bending, spiking and resoldering fixed many issues allowing a car to properly track both turnout routes using a tortoise machine to switch the points.    expect more issues with trains

i think it would be helpfull to be able to detect a point that is not tight against the stock rail using a geometry car and am now becoming more curious what an accelerometer can do

 

please recognize this thread is about the a geometry car, not layout construction.   this is what has been done and is not going to be replaced (not my layout).   

of course you may not do it this way on your layout.

 

I will go back to sleep now and leave you to it. Seems to me you just need someone who knows how to lay track.

Sheldon

    

  • Member since
    May 2004
  • 6,893 posts
Posted by 7j43k on Wednesday, July 28, 2021 10:46 PM

gregc

 i think it would be helpfull to be able to detect a point that is not tight against the stock rail using a geometry car and am now becoming more curious what an accelerometer can do

 

I am not understanding why your eyeball is not telling you a point "is not tight against the stock rail".

If it cannot, perhaps you should find someone else to build turnouts.

If ALL of you are incompetent in this matter, you DO have a problem!

 

 

Ed

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 4:49 AM

7j43k
If ALL of you are incompetent in this matter, you DO have a problem!

don't understand how such comments help with the design of a geometry car ??

the east end trackwork (only partially shown in right hand photo) was similarly constructed and works.   eyeball inspections and testing with cars identified problems that were corrected.

there's little need for a geometry car if there are no problems.    with so many turnouts a geometry car can hopefully identify problems less noticable by eye/feel, without having to get on top of the layout. 

greg - Philadelphia & Reading / Reading

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 4:56 AM

hgodling
If you want to go the optical route, you want to be sure there is nothing between the camera and the track

no camera.   had suggested using reflection off top of rail to measure gauge using narrow slotted apertures.

wouldn't expect that approach to work with points.   thinking accelerometer may detect a wide point

greg - Philadelphia & Reading / Reading

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 7:22 AM

at some point, i'm imagining this car being part of a train during operation.   having a sounder (beeper) and leds visible from either side.

it would generate a beep when a defect is detected and turn on an LED for a short time (~5 sec).  different colored leds would indicate a verticle, horizontal, ... displacements using accelerometer.   

sensitivity would need to be made adjustable, presumably made more sensitive as more major problems are corrected

 

(is this pie in the sky?   the interlock project started pre-covid is finally getting installed)

greg - Philadelphia & Reading / Reading

  • Member since
    January 2009
  • From: Maryland
  • 11,130 posts
Posted by ATLANTIC CENTRAL on Thursday, July 29, 2021 8:02 AM

gregc

at some point, i'm imagining this car being part of a train during operation.   having a sounder (beeper) and leds visible from either side.

it would generate a beep when a defect is detected and turn on an LED for a short time (~5 sec).  different colored leds would indicate a verticle, horizontal, ... displacements using accelerometer.   

sensitivity would need to be made adjustable, presumably made more sensitive as more major problems are corrected

 

(is this pie in the sky?   the interlock project started pre-covid is finally getting installed)

 

The interlock project is interesting, and that is likely as good of an approach as any. But I could easily do that with about 30 ice cube relays @ $3 each - given we would both need the same buttons and lights, what did you spend on hardware?

In fact my staging yards will all have one button route control, and my interlockings will be similar requiring only one to three buttons to fully select routes.

Just my opinion, but I don't see any one railcar tracking in a way to accurately take these measurements.

But What do I know...

Sheldon

    

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 8:29 AM

ATLANTIC CENTRAL
But What do I know...

wow

greg - Philadelphia & Reading / Reading

  • Member since
    September 2003
  • 17,656 posts
Posted by Overmod on Thursday, July 29, 2021 9:05 AM

I have no question about using small high-resolution cameras or 'mouse sensors' in a track-geometry project, as an enormous amount of the work in making the devices and their interfaces is now at obsolescent-cheap levels of cost (thanks to many generations of feature-packed phones, high-resolution gaming and CAD pointing devices, structured light experimentation, etc.)

I don't think it would be rocket science to make something with two cameras with acceptable width of field and an appropriate light source (perhaps polyspectral) to allow small device aperture and hence long depth of field for resolution).  A similar arrangement with cameras in quadrature could remotely sense the end of the elongated indicator wire or pointer to get long baseline that was mentioned earlier, with far more precision than I think a commercial accelerometer core might practically furnish.

I do not know if weighting the car to give a better inertial reference frame is going to work at smaller scales -- it should for short track defects traversed relatively quickly, but it would have to be tried and tinkered with.  

  • Member since
    September 2003
  • 17,656 posts
Posted by Overmod on Thursday, July 29, 2021 9:15 AM

ATLANTIC CENTRAL
Just my opinion, but I don't see any one railcar tracking in a way to accurately take these measurements.

My two cents here is that perhaps one railcar shouldn't take all these measurements.  Have one car that tracks gauge, one that checks lateral transitions, one for vertical and cross-level anomalies, etc.  Run them separately and recursively... or as a geometry train of sorts.

  • Member since
    January 2009
  • From: Maryland
  • 11,130 posts
Posted by ATLANTIC CENTRAL on Thursday, July 29, 2021 10:04 AM

My appologies.

I stayed out of this topic because it seemed? I don't know a polite word? Not necessary......

I was only courious to understand the track problems you are facing.

You are a better man than me, I would just tell them/him "this is not the right way to do this, and the workmanship is sub par".

I build things for a living. After having the opportunities in life to be a "boss", an engineer, an "office worker", I found much more satisfaction (and just as much money) in actually creating the finished product.

Much of what I do is designed and built by my hands.

I have little tollerence for lack of skill, or poor workmanship.

I know this is just a hobby, but it requires skills......

Sheldon 

    

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 10:22 AM

Overmod
I don't think it would be rocket science to make something with two cameras

i don't think a camera is infeasible, but have little experience with image processing.

Overmod
My two cents here is that perhaps one railcar shouldn't take all these measurements.

i've suggested "different cars with different purposes".   i doubt all the electronics and any mechanical stuff could fit on a single car.  

a single car with accelerometer would handle 3 dimensions and code would be fairly cookbook, just a matter of recognizing a big enough jolt in any direction and turning on an LED.    still unsure about recognizing a lack of easement

Overmod
I do not know if weighting the car to give a better inertial reference frame

i wondered about this.  the accelerometer needs to move to register.   so the wheels need to have little lateral or vertical play and the platform needs to be light (but not too light)

greg - Philadelphia & Reading / Reading

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 10:25 AM

.

greg - Philadelphia & Reading / Reading

  • Member since
    September 2003
  • 17,656 posts
Posted by Overmod on Thursday, July 29, 2021 10:44 AM

gregc
a single car with accelerometer would handle 3 dimensions and code would be fairly cookbook, just a matter of recognizing a big enough jolt in any direction and turning on an LED. 

If that's all you're wanting to do, then a MEMS accelerometer (which used to be a bunch of little weighted piezo beams with their outputs averaged together so breaking a few wouldn't kill the output) ought to give you the trigger signal you want for that -- I don't have any data sheets for cores but you could ask suppliers for their OEM data.  The catch is whether at the same time the result will be sensitive enough to pick up anomalies without at the same time falsing when air blows on the car or a passing truck shakes the ground.  Suspect you'd have less trouble with your original idea of roller feelers and adjustable limit switches if you want a kind of go/no-go gage action for switch point transitions and suchlike.

  • Member since
    July 2009
  • From: somerset, nj
  • 3,760 posts
Posted by gregc on Thursday, July 29, 2021 11:00 AM

ordered I2C compatible MPU 6050

greg - Philadelphia & Reading / Reading

  • Member since
    May 2004
  • 6,893 posts
Posted by 7j43k on Thursday, July 29, 2021 11:20 AM

gregc

 

 
7j43k
If ALL of you are incompetent in this matter, you DO have a problem!

 

don't understand how such comments help with the design of a geometry car ??

 

 

You are quite correct that it doesn't help with the design.

It questions the NEED for such a device, at least one as complex as you want to build.

I can see that you are beyond this question, so any further comments from me will be on the design, only.

 

 

Ed

  • Member since
    September 2003
  • 17,656 posts
Posted by Overmod on Thursday, July 29, 2021 4:08 PM

gregc
ordered I2C compatible MPU 6050

That device supports InvenSense BlurFree, a technology about 15 years old that stabilizes camera capture (as in smartphones).  At the very least this should provide you error signals for inertial correction of some of the unwanted movement...

https://circuitcellar.com/wp-content/uploads/2012/06/CC2012050461.pdf

  • Member since
    January 2003
  • From: US
  • 85 posts
Posted by rbturner on Monday, August 2, 2021 7:26 PM

Here are some basic cars I made using Tichy trucks, Intermountain wheelsets and #148 Kadees.

The plastic was given to me so I have no details on it.

Randy
  • Member since
    September 2004
  • From: Dearborn Station
  • 21,736 posts
Posted by richhotrain on Tuesday, August 3, 2021 6:08 AM

The whole idea of a geometry car seems daunting because of the number of tests required. Why not start out with something simple like a car that measures track gauge in decimals. As it runs around the track, a green LED would be lit to show in gauge and a red LED to show out of gauge.

Rich

Alton Junction

Subscriber & Member Login

Login, or register today to interact in our online community, comment on articles, receive our newsletter, manage your account online and more!

Search the Community

ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT
Model Railroader Newsletter See all
Sign up for our FREE e-newsletter and get model railroad news in your inbox!