CG position from 4 corner load cells

[trainguy]

All,

We are attempting to use 4 load cells at corners of a railcar to determine its CG.

Using 3 points instead (which I'd rather do because it's statically determinate) is not feasible.

A discussion on thread404-117973 touched on this. Does anyone have a repeatable method of establishing the CG position (in plan) using the 4 load cell readings? We even have the capability of raising and lowering 1 of the corners slightly and taking new readings. These additional ones could provide the missing equation!

Any ideas?

tg

 

[JStephen]

I don't see the issue, actually. If you're using four supports and trying to calculate the load at each, it's statically indeterminate and all. But if you're using load cells to determine the loads, that's not an issue, you know what the loads are. Pick any reference line, sum of loads x distance divided by sum of loads is the CG in that direction. How repeatable it is would depend on how accurate the measurements are.

 

[3DDave]

It should be a summation of moments/summation of readings solution.

Statically determinant only means that the loads internal to the car can be calculated.

One item that is critical is to eliminate cross-coupling between horizontal loads and vertical loads. Load cells should be interally balanced to eliminate moment inputs and off-axis components from being resolved to on-axis values. Retesting with the load cells turned 90 and 180 should verify that if there is any question.

 

[rb1957]

how do you find the CG of four things (areas, forces) ? ... sum(Ax)/sum(A) ... no?

Quando Omni Flunkus Moritati

 

[rb1957]

btw, being statically indeterminate isn't an issue here. ok, if you want to make it an issue ('cause it might affect the 2nd decimal place) you could verify that the scales have the same stiffness, or you could re-weigh with the scales in different positions, you could accurately measure the rotation of the item (if all scales have the same stiffness, then redundancy is not an issue).

Quando Omni Flunkus Moritati

 

[trainguy]

What we've found many times in the past was that because of various amounts of carbody initial twist, or other reasons unknown to me, we could have calculated lateral CG position vary by as much as 1 to 1.5 inches (on a 10 ft width) from 1 weighing operation to the next (which occurs 2 minutes later). The scales all have the same stiffness, but the railcar doesn't.

We are attempting to narrow this down to within 1/4 inch.

Railcar is approx 10 ft wide by 89 ft long by approx 14ft high, and we're trying to nail down lateral CG pos'n. Load cells essentially at corners.

It may be that load cells are not precise enough for what we're looking for.

tg

 

[rb1957]

fair enough ... we are making assumptions about stiffness of scales, of the thing being weighed, about the accuracy required

Quando Omni Flunkus Moritati

 

[IRstuff]

How level is the structure that the scales sit on? How accurately can you even position the car itself; can you even repeat its position on the scales to 1/8 inch?

Fundamentally, though, in response space, the CG sits on a flattish hump, i.e., minor perturbations around the CG are not necessarily that detectable. Your requirement of 1/4 inch requires better than 1/8 inch accuracy, and out of 10 ft, that's a 0.1% accuracy, which doesn't sound like something that's doable on a rail car. Are the wheels even that well balanced?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[rb1957]

"We even have the capability of raising and lowering 1 of the corners slightly and taking new readings." ... so you could do a statically determinate balance (like you wanted to do in the first place) if the stiffness of the thing being weighed is significant, you'll see the thing being weighed creak and groan as it adjusts to the new reaction set.

Quando Omni Flunkus Moritati

 

[JStephen]

If you total the loads from the 4 load cells and compare them at the two different stations, that'd give you an idea of the accuracy.
If you've done this more than once, and get a shift from one station to the next, is it somewhat repeatable? or random? If just one load cell were off, it'd throw the calculated CG off, but would be similar each time you measured identical items.

 

[3DDave]

Stiffness of the rail car affects what loads are seen at the various load cells, but it won't directly affect the measurement of the CG. For example, if the CG was exactly centered and the railcar was dead soft, then all four load cells would read the same and the calculations would find the CG is in the middle. If the car was perfectly rigid and only two opposite corner wheels touched the track, half the weight is on those two load cells and the others read zero. The calculation again puts the CG in the middle.

If, however, the car teeters on those two wheels and, because the CG has a vertical component, the CG shifts, then the load cells will get the new distribution of loads, and the calculation will show the side-to-side location of the CG has changed, because it has.

There is also the potential for the car to change shape over time due to uneven heating from solar load. You may wish to account for this as well.

 

[3DDave]

... half the weight is on each of those two load cells ...

 

[BrianE22]

Do you have spherical buttons on the top of the load cells (even better - top and bottom of each cell)?

 

[Compositepro]

I don't see what the issue is here. As some have alluded the load cells must be designed so they only support and measure vertical loads directly in-line with the axis of the load cell. Then calculation of the CG in the plane of the cells is easy. The vertical location of CG would require tilting the car.

 

[rb1957]

because the CG is higher than the support plane, if the structure is very flexible then the position of the CG could move slightly.

if the car is rigid then the height of the CG doesn't change the location of the CG.

if the car is flexible then the height of the CG will work it's way into the results and add an unknown into the calcs ... the scales will tell you the point the CG is acting through, the line that it is on.

i guess that's a quesion to tg ... do you want all three CG co-ords or only two ?

Quando Omni Flunkus Moritati

 

[DerbyLoco]

Are you checking the shell off its bogies? I assume a large vehicle like this would have at least 4 axles. We have endless trouble getting consistant wheel loadings under a single two axle bogie on a test rig to within +/-5% and wheel loading on a whole vehicle are a constant nightmare. The centre of gravity usually is calculated from mass reports and the fun starts when it is way off centre!
Why bogies in UK - trucks in USA?

 

[rb1957]

"Why bogies in UK - trucks in USA?" ... different sides of the "pond"? two cultures separated by a common language?

why lorries in UK and trucks in the US? ... this could go on forever!

Quando Omni Flunkus Moritati

 

[dhengr]

Trainguy:
Are the trucks still on the car, how have you gagged them up to the body, or are you just working with the car body? What kind of a car is it, along with a few more dimensions and the lt.wt.? A flat car won’t have many vertical component issues, but a high cube box or passenger car will have some vertical CG issues Where are you jacking or weighing it, at the jacking pads, at the side bearings, where and how? Could you put the car on a knife edges (round bars) on or near the center of each center plate, and move it until it is balanced. Then use a load cell under each center plate to find the longitudinal CG, or balance it on the side sills or center sill near the center. Maybe you should level the car as best you can on four jacks, then shim the load cells under their weighing points, so they all have 100lbs. on their dials; that’s zeroing them out, and should be subtracted out. Then lower the car onto the load cells. I would check a few of the critical dimensions on the car, such as: does the center line match the centers of center pls., of side bearings, of jacking pads, of side sills or sides, any other heavy components, for two reasons; so you have a sense of any obvious off center wts. and secondly so you know where you are weighing w.r.t. the centerline. If I could get within .25",with some repeatability, I think I would be pretty happy. Purchasing just told me that the guys who welded the left side of the underframe consistently used about 28 lbs. more welding consumables one each car, go figure.

 

[MintJulep]

You are attempting to measure the CG location of a rail car to within 1/4" accuracy?

Good god man, why?

You won't be able to manufacture them sufficiently repeatably to keep the CG location of each one that close.

The amount of grease and crud that accumulates on the underside will change the position by more than that.

It will never be loaded sufficiently uniformly for it to matter.

Measure as best you can as you have in the past.

Preferably, repeat on a few cars.

Calculate the average location of multiple readings.

Draw a circle around that average that is large enough to contain all of the individual readings, plus a little more.

For whatever balance calculations you are trying to do, calculate them for the average location and for a point on that circle that is the worst case for the calculation at hand.

Report nominal and "not expected to exceed" values.

 

[racookpe1978]

Concur with the above.

The "real world" railcar is suspended BY SPRINGS in each of four corners between 2 sets of wheels on 4x axles. Those springs will settle and move as the car moves, and as the car is loaded and unloaded, weighed on the load collars and re-loaded on the springs. Not much, true. But what 2-3 percent?

Let's assume that you can reliably and consistently put your load back on the "perfect" car within 1/16 of an inch left/right and fore/aft every time.
Let's assume your load itself is consistently built every time within 1/16 of ITS center of gravity and at the same mass every time and at the same vertical center of gravity every time.
Let's assume your railcar is reliably and accurately and consistently re-built every time with its center of gravity within 1/16 of an inch every time with the same mass everytime.

Even with those three (almost impossible) assumptions, just placing the load "within spec" +1/16 on the car with a load CG change of 1/16 inch + a rail car CG change of 1/16 of an inch would throw the final CG 1/16 + 1/16 + 1/16 = out-of-measurement tolerance.