Overcoming waviness

Dear Doc: I cylindrical grind nitrided shafts with 20 spark-out revolutions. Sometimes I can spark-out forever and still never remove the waviness, and other times it takes only a few spark-out passes. I spark-out at 200 rpm with a constant wheel speed of 40 m/sec. (8,000 sfm) and a wheel diameter from 500mm to 350mm (19.7 " to 13.8 "). The level of waviness seems to depend on wheel diameter. Why is that?

The Doc Replies: No grinding wheel is perfectly true, and that imperfect roundness on the wheel inevitably gets put into the workpiece in the form of waves. The key is not to “catch the waves” when the workpiece makes successive spark-out revolutions.

Take the wheel rpm and divide it by the workpiece rpm. That’s the rpm ratio. If the rpm ratio is an integer value, such as eight, then the workpiece catches the wheel in the exact same wave and exacerbates the waviness (see Figure). If the rpm ratio is a fractional value, such as ¼, 1⁄3, ½, 2⁄3 or ¾, the situation is a little better as the previous wave is obliterated by the next wave, providing a series of scallops.

Courtesy of J. Badger

The height of waves on the workpiece depends on where the wheel hits during successive spark-out passes. 

The ideal situation is a long-decimal value, such as 7.84923785. Here, you never “catch a wave,” but just keep obliterating previous waves. I’ve seen unbalanced, out-of-true wheels generate little waviness because they found such a value.

In your situation, the wheel rpm ranges from 1,528 to 2,182 rpm. The rpm ratio is going from 7.63 to 10.91, passing through three integer values (eight, nine and 10) and numerous fractional values.

How do you make this happen on the machine? You can constantly monitor the rpm ratio to avoid integers and fractions, which is a hassle on an automated machine. You can run at a constant wheel rpm, which means the wheel surface speed will change throughout the life of the wheel. Or, the CNC programmer can choose a combination wheel and workpiece rpm that always gives a long-decimal value. That’s the best method.

Why don’t more CNC programmers program in long-decimal rpm ratios? It would take 10 minutes to do, but they typically don’t know about it. Too many of them have an attitude of “leave the grinding to the grinder.” That may change, but, in the meantime, you’ll have to sort it out yourself.

Dear Doc: I dress a form into my wheel with a plunge diamond roll. There’s a big argument in the shop whether it’s better to go unidirectional or antidirectional. I’ve been told it’s better to go unidirectional, but the machine operators say they won’t be able to hold surface finish when doing so. What’s the story?

The Doc Replies: I’ve convinced companies to switch from antidirectional to unidirectional, where the wheel and roll are moving in the same direction at the point of contact, or “crushing,” and seen problems disappear. Going antidirectional just dulls the wheel, causing burn and chatter problems.

If you can’t hold surface finish going unidirectional, go with a smaller grit and a 0.8 ratio for the surface speed of the diamond roll to the surface speed of the wheel. The exception is when you are removing a lot of material and do a rough-dress, rough-grind, finish-dress and finish-grind operation.