Rear axle broke right inside cover

[Hugo Habicht]

Au contraire. Thats exactly the type of thing you can model with FEA.

Dan

Still, a human has to put in the load cases. This is where things can go wrong.

I found interesting what worst case load is; a friend of mine had to calculate a rear axle differential at university. Highest strain, as he explained to me, was when one wheel was spinning freely at full revs and then the vehicle falling down after a jump on very good grip asphalt. Not sure if this is applicable to tractors though

 

[McMXi]

Still, a human has to put in the load cases. This is where things can go wrong.

I found interesting what worst case load is; a friend of mine had to calculate a rear axle differential at university. Highest strain, as he explained to me, was when one wheel was spinning freely at full revs and then the vehicle falling down after a jump on very good grip asphalt. Not sure if this is applicable to tractors though

That was my real-world example which is a complex problem to solve due to myriad variables. You have to understand the types of forces, where they act, when they act, how they're affected by other forces and so on. Lots of assumptions have to be made and that's why at the end of the day it's empirical data from actual testing that is required. FEA is just anther tool in the tool box.

I use the built in solver in SW just about every week but it's basic at best, and I also have extensive experience with Ansys which is very powerful. But once again, the outputs are only as good as the inputs, and accounting for all of those without error or simplifying assumptions is the challenge.

 

[TheOldHokie]

Still, a human has to put in the load cases. This is where things can go wrong.

I found interesting what worst case load is; a friend of mine had to calculate a rear axle differential at university. Highest strain, as he explained to me, was when one wheel was spinning freely at full revs and then the vehicle falling down after a jump on very good grip asphalt. Not sure if this is applicable to tractors though

Any process including pen and paper is subject to human error and the same sort of misanalysis in your example. Modeling with FEA isnt infallible but it lets a designer do a far more comprehensive analysis. In the long run that produces more reliable outcomes

FEA is the mathematic process used by just about every automotive designer because it is predictive, accounts for more complex loading, and lets you get prototypes built and "real life" tested faster.

A quick internet search will return many videos showing how to model an automotive suspension using SolidWorks built in modeling engine. I imagine big time OEMs have even better modeling software and design engineers that are experts in applying FEA to their particular area of specialization.

Dan

 

[McMXi]

Any process including pen and paper is subject to human error and the same sort of misanalysis in your example. Modeling with FEA isnt infallible but it lets a designer do a far more comprehensive analysis. In the long run that produces more reliable outcomes

FEA is the mathematic process used by just about every automotive designer because it is predictive, accounts for more complex loading, and lets you get prototypes built and "real life" tested faster.

A quick internet search will return many videos showing how to model an automotive suspension using SolidWorks built in modeling engine. I imagine big time OEMs have even better modeling software and design engineers that are experts in applying FEA to their particular area of specialization.

Dan

Many of the bigger engineering companies such as Northrup Grumman use CATiA but not sure what is common in the automotive industries. I'll have to ask my boss what John Deere uses. He was in charge of hundreds of engineers before he left last year.

I'm well aware of what FEA is and how it works. Computational speed is why we have tools such as FEA. I've done calculations by hand to determine the plastic deformation at each node of a part and it's tedious and laborious process. The more nodes the more work. Obviously, with today's processors those calculations can be run millions of times a minute which is ideal for iterative or infinite series such as the Taylor series, Sine and Cosine functions etc. The software can spit out a solution once an acceptable delta (difference between current solution and previous solution) has been reached. Years ago I calculated the deformation of a rubber membrane stretched and sealed over the opening of a square hole. Symmetry helps a lot in those calculations, but I ended up creating a spreadsheet to calculate the deformation at each node for a given pressure and then created a plot showing the deformation. Very little of that was enjoyable with the exception of seeing the predicted shape and then validating the model via experimental process.

My model had nodes on a grid that were spaced 5mm apart, and the solution was limited to the resolution of the grid. Ansys or similar could have nodes every micron if necessary which would increase the accuracy of the predicted deformation. This is where FEA and computers shines.

Like I said, using FEA on a driveshaft on a tractor would be a waste of time and money which could be better invested somewhere else. The maximum stress in a rotating shaft occurs at the outer surface of the shaft. No surprises there so no need to create an elaborate mesh to figure stress profiles.

 

[torch]

And to the moon, appr. 300000km away, 24 bit mantissa floating point should be good enough to get you back to within a mile or so. Ever seen the Apollo computers?

Back in the mid-70's the family business was designing and installing fire protection systems. Grids of pipe, sometimes involving hundreds of nodes, required complex iterative and recursive calculations with a slide rule, but an experienced engineer could save a lot of time and effort by starting off somewhere in the middle -- an educated guess -- and refining from there with a slide rule (special Hazen-Williams scales -- not off-the-shelf), laboriously jotting results on a paper forerunner which lent it's name to the electronic "spreadsheet" of today. A good engineer's guess would often be within 5% of the final results.

We were approached by a salesman selling shared time on their engineering mainframe (NCR if memory serves, but that might have just been the hardware not the service). He boasted that their platform could handle any engineering problem because it put Man on the moon. After a few attempts to demonstrate, they had to give up and admit their system couldn't handle the complexity.

Flash forward 15 years: PCs were becoming commonplace and DOS based software could handle the complexity and produce acceptable results in less time. Nowadays it's even built right in to the CAD software with near-instant results on the fly as a designer alters a layout.

I still have my hydraulics slide-rule. And a copy of the early software. But I no longer have a computer that can actually run that software -- and to be honest, only a vague recollection of how to work the slide rule.

Back to the topic at hand: A solid bar is stronger than a pipe of the same diameter, but a pipe is stronger than a solid bar of the same weight. in other words, much of the strength is in the perimeter, not the core. Boring that hole and the size of the bore is a design choice. It only becomes a design flaw if it fails to meet the specifications.

Kubota engineers have no control over customer choices. Farmer John bouncing down the field with a 600 lb load on a tractor rated to carry, say, a 400lb attachment does not make axle failure a design flaw.

I'm not saying that's what happened, just pointing out the possibility. I think proving to a jury that none of the tractors in question were ever abused is a tough row to hoe (sorry, couldn't resist ).

 

[TheOldHokie]

Back in the mid-70's the family business was designing and installing fire protection systems. Grids of pipe, sometimes involving hundreds of nodes, required complex iterative and recursive calculations with a slide rule, but an experienced engineer could save a lot of time and effort by starting off somewhere in the middle -- an educated guess -- and refining from there with a slide rule (special Hazen-Williams scales -- not off-the-shelf), laboriously jotting results on a paper forerunner which lent it's name to the electronic "spreadsheet" of today. A good engineer's guess would often be within 5% of the final results.

We were approached by a salesman selling shared time on their engineering mainframe (NCR if memory serves, but that might have just been the hardware not the service). He boasted that their platform could handle any engineering problem because it put Man on the moon. After a few attempts to demonstrate, they had to give up and admit their system couldn't handle the complexity.

Flash forward 15 years: PCs were becoming commonplace and DOS based software could handle the complexity and produce acceptable results in less time. Nowadays it's even built right in to the CAD software with near-instant results on the fly as a designer alters a layout.

I still have my hydraulics slide-rule. And a copy of the early software. But I no longer have a computer that can actually run that software -- and to be honest, only a vague recollection of how to work the slide rule.

Back to the topic at hand: A solid bar is stronger than a pipe of the same diameter, but a pipe is stronger than a solid bar of the same weight. in other words, much of the strength is in the perimeter, not the core. Boring that hole and the size of the bore is a design choice. It only becomes a design flaw if it fails to meet the specifications.

Kubota engineers have no control over customer choices. Farmer John bouncing down the field with a 600 lb load on a tractor rated to carry, say, a 400lb attachment does not make axle failure a design flaw.

I'm not saying that's what happened, just pointing out the possibility. I think proving to a jury that none of the tractors in question were ever abused is a tough row to hoe (sorry, couldn't resist ).

There are a lot of unanswered questions and so far 8 failures out of how nany units in the field?

The judgement of two machine shops aside defects in material, workmanship, or misuse could easily account for that.and standard warranty terms covers the first two..

Dan

 

[hagrid]

This is why teh BX23S must remain upon the manicured lawn.

 

[TheOldHokie]

This is why teh BX23S must remain upon the manicured lawn.

Is that your official ID? We can't seem ti get something that basic from the gentleman that started this discussion

Dan

 

[McMXi]

A good engineer's guess would often be within 5% of the final results.

So +/- 5%?

So many real-world problems have numerical and empirical solutions rather than analytical solutions due to the complexity of many systems. In fact, I'd venture to guess that most real-world solutions are not analytical. Given that solutions are so often numerical, and given the time it takes to calculate those solutions, the pre-computer approach was somewhat Victorian and most things that lasted were overbuilt.

Today it's all about optimization and cost reduction.

 

[targabill]

Lots of interesting Mechanical info comments, Thanks, I'm an EE so not as up on stress testing a shaft, Axle in this case. I'm am quite mechanical on stuff like rebuilding engines etc. I even took to broken Axle out of my tractor waiting for a fix, probable welding it is my ONLY OPTION.

Just wondering how a designer would model the TORQUE on said axle.
I think about the force, Tractor in low gear, with a backhoe weighing maybe 3000#, newish tires for max traction in the dirt. Then you push the loader into a pile of dirt and catch an unseen rock or stump which stops the tractor dead in its tracks.. How would you measure such a torque? But I wouldn't expect my tractor to break down because of that ?? Am I expecting to much??

 

[torch]

How would you measure such a torque? But I wouldn't expect my tractor to break down because of that ?? Am I expecting to much??

I would start by assuming 100% available engine stall torque was being applied to the transmission then multiply by the total gear ratio. That's going to be the limiting factor.

And no, I don't think you are expecting too much. I also don't think the photos support the contention a twisting force caused the failure -- although certainly it could be a factor once the axle started to crack.