Differential Gear Design

[GKenworthy]

Hi, I'm designing a limited slip Salisbury (clutch-pack) differential and I'm having trouble with my gear calculations. It seems that the AGMA gear calculations that I'm taking from Mechanical Engineering Design (Shigley) are predicting that I need gears that are far larger than what I know is okay. I've got some rough dimensions for a similar differential that's capable of taking significantly more torque than I require and when I use its data I find that the AGMA method predicts that the gears should fail.

Does anyone have any suggestions on where I could look for some better or more accurate calculations I can try? Thanks very much!!

 

[diamondjim]

Garbage in, garbage out.
You probably are making some assumptions
which are not true in how you choose all
of the many variables in the calculations.

I would adjust the variables to agree with
the known and then compare your new gear
design to the known and factor it accordingly.

 

[sg20]

You need to define the maximum torque required and then chose your gears (using the appropriate pitch); AGMA is very good with gears but you also have to account for factors such as vibration, and gear cycles... Since you are doing clutch, life cycle is not a factor but vibration is. So you might want to give it bit more strength if you can.

 

[GKenworthy]

Thanks for the help, but I'm still not really sure how to continue. I do have an approximate input torque though I know there's some shock loading on the system. Our halfshafts are designed for 450Nm nominal (fatigue) and 800Nm shock under clutch drops/mis-shifts etc. I'm using this torque divided by two for the torque on each side gear then divided by the number of spyder gears for the torque at each gear mesh. That seems to be the proper method for differentials.

I have performed the calculations several times to ensure that I haven't made any errors. I have also tried using a calculator that I downloaded called MITCalc. The data published by the AGMA does not seem to provide a great deal of information about small gears - the size factor is not available for bevel gears of 10 teeth mating with 16 teeth. Consequently, I'm not fully sure my calculations are accurate, but it seems hard to believe that they're out by such a large margin.

My reference design is capable of withstanding 10x the torque in a racing application, yet my calculations say that it's only marginally handling my loads. In fact, they are gears from a late 80s Ford Escort re-worked into an LSD, which still should be putting near 5x the torque into them. This simply cannot be right.

diamondjim, are you suggesting that I essentially redesign my gears to have the same factor of safety (~0.4) that the reference parts have? It's hard to me to believe this is a good method, but it's the only thing I can think to do. Are there straight bending strength calculations I can do that do not account for fatigue?

 

[diamondjim]

I know there are material factors depending on the
materials and cleanliness of the steels as well
as manufacturing processes that can change the
values by at least 5 times the supposed normal
steels. I believe if you have 4 engineers go thru
the calculations, you will get 4 different answers
depending on their familiarity with all of the
variables. AGMA does not do a good job of listing
the sources of many of its factors in the equations.
Yes I am implying if you have known results from
an existing application is .40, I would assume your
new design would only have to be .40. Can you list
which pages of calculations you are using from
Shigley? He is a well known and has been a reliable
source.

 

[sg20]

This is the equation I used for bevel gear design, 8620 alloy steel.

Stress (bending) = Ft / (pi/P)*b*y
MS = (1600,000/stress)-1

Ft - tangential load
b - face width in inches
y - tooth form factor or geometry factor