Request for help reducing friction and rolling noise of a cam follower 1

[John2004]

Hi everyone,

I would like to ask if anyone could please help me with the following problem.

I have a small conjugate dual-roller rib or blade cam that has an inner and outer cam-follower roller working on a corresponding inner and outer cam curve. The cam curve is basically a curved "rib" with a roller on each side of the rib. The rollers oscillate about a swing arm in response to cam oscillation.

The force of the small 3/16" OD rollers is pinching the cam rib and causing drag on the manually oscillated cam. The system also has roller noise that must be reduced as much as possible. The system is very sensitive to any friction between the cam & roller, or any drag or noise in the system.

I've tried to polish the cam, and add lube, but there is still too much friction. I could also try to polish the rollers.

Is there any type of hard coating I can apply to the cam and/or rollers that will help significantly ? Any type of special lube I could use ?

The maximum Hertz contact stress is probably around 180,000 PSI. Right now, the cam is made from hardened 4140 and the roller is hardened A2 Tool Steel.

Is there any type of non-metallic roller that I could use ?

I would appreciate any suggestions for alternate materials, coatings, lubes, or anything else I might try.

Thanks for your help.
John

 

[MikeHalloran]

It sounds like the rib is of constant thickness, and is not thinned to compensate for the angle of the surface relative to the rollers. I.e., one solution is to separate the rollers axially a bit, and thicken the cam where the pressure angle is low.

Another alternative is to put the rollers on a pivoting beam that can follow the cam, keeping a line between the rollers perpendicular to the cam surface.

I.e., I think it's a geometry problem, not a materials problem.



Mike Halloran
Pembroke Pines, FL, USA

 

[MikeHalloran]

Nowadays, small computers are so powerful and so cheap that you could just drive the cam followers with leadscrews, and throw away the cam, and maybe even save some money doing it.



Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

Thanks for your feedback. I'm afraid I'm stuck with using a cam on this design.

As far as this being a geometry problem, I'm not sure. I designed the cam with cam design software, and the inner and outer curve are mirror images or each other. The cam was CNC machined and then flame hardened.

One of the big problems on the design is the space constraint. The cam has to be so small that I need to use a small diameter cam follower roller in order to keep from undercutting the cam. The points on the cam near the dwells get too sharp with a larger diameter roller. It seemed even going to a 1/4" OD roller produced interference, but perhaps I could cheat a little. It does not look possible to use a larger roller than 1/4" OD and I would rather stay with a 3/16" OD roller so I can use the same cam and not redesign the curve for a 1/4" OD roller.

Presently, I'm making the rollers from 3/16" OD A2 drill rod. I cut the roller to length, drill a 2 mm hole in it, and then press a 2 mm OD hardened dowel through the on-center hole. The dowel extends from each end of the roller & the dowel "rotates with" the roller as it oscillates. The ends of the dowel are supported by plastic busings made by

http://www.igus.com.

As the roller oscillates, the dowel pin / shaft oscillates in the plastic bushings. It's a basic yoke type of roller mount.

Perhaps the plastic bushings are giving a little & I would be better off just letting the ends of the 2 mm OD dowel oscillate in the 4140 yoke, & lube it with some grease. It may be possible there is some deflection of the shaft causing problems, or perhaps heat treating the roller made it a little egg shaped. I did not think there would be any significant friction at these rollers, but I think the load is so high for such a small roller that this is causing the problem. The two rollers are pinching the cam rib, and causing drag. Springs return the cam to it's home position & drag is preventing this. Stronger springs mean more physical effort from the user. I need some way to reduce friction at the rollers as much as possible. Reducing the friction should also reduce the noise problem.

Other than this friction problem, the design works fine, which makes it frustrating the get everything together and run into an unforeseen problem like this.

Timken makes a 1/4" OD drawn cup needle roller. Perhaps I could install this over a 1/8" dowel and give that a try. The possible problem I see here is that drawn cup needle rollers rely on a press fit to properly size them, and make them truly round. It may also fit loose on the shaft unless I make a custom shaft.

http://www.ikont.com/

makes a miniature stud type cam follower (Part # CFS 2.5 V) with a 5 mm diameter ( 0.197" OD) which should be close enough. It can handle a 220 pound static load, the only problem is that the last time I checked, they cost about $50.00 a piece and the product requires two rollers per unit. I think it will be cost prohibitive unless the price comes down drastically for production quantities.

It seems a rolling element bearing is the best way to reduce friction, drag, & noise as much as possible, along with applying lube to the roller OD & making sure the surfaces are smooth. The problem is finding a bearing that has a diameter of 3/16" and that can carry a static load of 90 to 130 pounds.

A non-metallic self lubricating roller also seems appealing, but it's a matter of finding something that small that can handle the load.

If I could use springs to offset some of the roller force, what roller material would be the best alternative ?

Any other feedback would be appreciated.

Thanks again,
John

 

[MikeHalloran]

As you have pointed out, drawn cup needle bearings are not intended for service as cam followers. It has been my experience that they don't fail spectacularly when so abused, but they do impose an extra motion on the follower that is not in the cam, as the rollers pass the point of contact, and the shell at the point of contact is stressed alternately in compression and in bending. I.e., some systems can 'feel' the needles passing by. That happens even if you select them and only use shells that are round as delivered.

I'll accept that the cam design program was given the correct geometry, including the follower radius and trajectory. How much allowance did you leave in the design for operating clearance and manufacturing tolerances?










Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

>How much allowance did you leave in the design for >operating clearance and manufacturing tolerances?

The inner roller is on a spring loaded slider, so that there is no clearance between the rollers and the cam. Manufacturing inconsistencies in the cam profile can push the inner roller along the spring loaded slider by a few thousandths of an inch if needed. That part of the design seems to be working as intended. I designed the components this way because the application is sensitive to clearance between the rollers and the cam, and to allow the device to be less sensitive to manufacturing in installation inaccuracies.

The inner and outer rollers create opposing torque on the cam, and it's the spring loaded inner roller that returns the cam to it's home dwell position after the manually activated lever is released. I thought I could simply load the inner roller with the same force as the outer roller but it seems that due to roller friction, I need a little more force on the inner roller. I was surprised at how much of a difference just polishing the cam curve and adding some grease made.

It's so close to where I need to be. I think it's just a matter of finding the right roller approach at reasonable cost. I also have two opposing springs connected to the cam to hold it at the centered home position dwell. I found that using springs with a higher rate helps to overcome the roller friction, but I can't use springs that are too stiff as this will cause the lever to be too hard to activate. I think it's a matter of zeroing in on the right spring rate, but I think the first step is to get the cam follower roller friction as low as possible.

Thanks again,
John

 

[John2004]

After checking into this further, I think that this might be a case where the cam curves are not accurate enough and the cam is kind of binding between the two rollers. Can anyone please tell me what the tolerance of the cam curves should be, relative to the cam rotation axis, for a conjugate dual roller cam as shown in FIG. 2 at the following link ? ...

http://www.malcams.com/dynamicpage.asp?id=15&parentId=4

The cam was made from 4140 steel, CNC machined, and then the curve was flame induction hardened. I probably should have had the cam hardened and then ground to high accuracy.

Thanks
John

 

[John2004]

Note:

Regarding figure 2 at the link in my previous post, the very inner curve is not present in my cam. My cam just has the curved "rib" or blade that is in-between the two rollers.

 

[MikeHalloran]

I think the usual way to deal with a situation like this is to generate the cam with cutters that are a little larger than the actual rollers, or apply equivalent cutter compensation.

The amount is dependent on clearances and deflections and manufacturing tolerances, so it's more a design decision that you would make than a standard number we could tell you.

If you go back to the cam design program and upsize the rollers, does the program tell you the cam will bind, or raise a flag somehow?

Suppose you make another cam out of machinable wax, put it in the assembly, and run it through a revolution. How much wax will be shaved off?





Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

Thanks for your message.

I had thought that the cutter OD should be be the same size as the roller OD. The cam design program won't do dynamic analysis, so it won't tell if the cam will bind, it just outputs pressure angle, radius of curvature, follower versus cam displacement per desired increment, and polar coordinates for the curve, but the polar coordinates were not used as the CNC programming was done direct from a 3D CAD file.

The wax method you suggest may be a good idea. My feeling is that the curves would have to be machined or ground to at least within .0005" of each other.

Next time I think I will use pre-hardened 4140 that's about 32 RC and machine it without any further heat treat.

This cam was CNC machined, then the curve was flame induction hardened. I think it probably distorted the curves & possibly the bearing bore. I had toleranced each of the two curves at +/- .001" relative to the cam rotation axis, but I have no real way to check the part accuracy as I don't have that kind of inspection equipment. I now suspect it needed to be more accurate anyway.

The inner roller now has an adjustment screw to allow it to be adjusted into the inner cam curve & I changed the spring arrangement to reduce the follower loads onto the cam. My goal is to have the smallest amount of roller clearance possible, but still have smooth operation with no significant drag on the cam, i.e., no drag or binding that can be felt in the activation lever, or that will prevent the return springs from bringing the cam back to it's home position dwell.

Thanks again for the feedback.
Jonn

 

[MikeHalloran]

So ... where did the 3D CAD file come from?



Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

I created the curve with cam design software, then extruded that into a 3D file with AutoCAD. From that point, I think my machinist used his computer aided machining software to translate the AutoCAD file into an IGUS file that he could use for CNC programming.

I just talked to my machinist and he said he ground about .010" to .015" off the inner and outer curve lobe and it seemed to help quite a bit. He said he didn't know if they screwed up or if the curves from my CAD file were not accurate. I can send the CAD file off to a third party like Camco to verify. My feeling is that this is a part & assembly accuracy problem, but my machinist now feels it's something else.

The machinist is testing the accuracy of the curve against my cad file with a CMM machine, but he said he's having problems because there are so many calculation points along the cam curve. With the cam design software, I used 10 calculation points per each degree of cam rotation, which seemed reasonable for accuracy. The total oscillation sweep angle is about 33 degrees so there are about 330 calculation points along the curve, although the machinist said it looked like it had 1,000 points.

Any further feedback would be appreciated.

Thanks
John

 

[MikeHalloran]

It's starting to smell like the cam generation software was not made to understand the follower geometry properly.

You could verify some of it. I assume that at some point you get a file that represents the motion of the center of the follower roller. If you plot that in AutoCAD, and draw circles to represent the roller, then a curve connecting the roller tangents should approximate the cam surface, except at locations of high curvature, where the cam may not look like the follower center path.



Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

I rotated the cam and follower in AutoCAD for each degree of cam rotation. The follower rotation was based on the cam design software output, since it outputs cam versus follower displacement in degrees.

The curve seems accurate. There is a small place where the inner roller seems to protrude or extend into and beyond the curve by .001", but the polar coordinates are accurate. The .001" discrepancy could have something to do with the way AutoCAD handles curves or perhaps more calculation points were needed as this is what it looks like when I zoom in.

I could use the follower displacement output to draw followers around the base circle in AutoCAD, then draw a spline curve tangent to the rollers. This will take some time if I do it for 10 calculation points per degree, but should produce an accurate curve. Then, I can't extrude a spline into a 3D model, I would have to convert it to a polyline, which may reduce accuracy of the spline curve.

I wish I had a lisp file that could take a list of polar coordinates and draw a curve based on the coordinates and roller size or just design the cam curve via lisp in AutoCAD.

Thanks again,
John

 

[MikeHalloran]

You need to figure out where the .015" that the machinist ground off actually came from.



Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

When I went to the shop today & tested the unit myself, grinding off the .015" did not really improve things so I think it was a mistake to grind off the .015". That particular cam is now ruined. They were just using trial and error to get to the bottom of the problem, which is fine by me, I just want to get the thing working right.

I used a slightly skewed or asymetric parabolic curve for this cam and I could not find a commercial cam design software to design this type of curve so I used software from the book "Cams for industry" by John Reeve. I could not even find commercial cam design software that would design a standard symmetric parabolic curve for a swinging follower.

I think I will contact a cam manufacturing company and pay them to generate the curve and provide a CAD file. At least then I will know that the geometry of the CAD file is correct & I can go from there. I'm not sure what else to do.

Thanks
John

 

[John2004]

Hi everyone,

I had two independent parties verify that the cam profile is correct. The polar coordinates that were generated by the software are correct, however, they were not used for CNC programming.

My CAD file did not have 10 calculation points per degree as I thought the cam design software was generating. When I zoomed in very close in AutoCAD, I found that the points were about .03" apart and they were connected by straight lines and not arcs or curves. The polar coordinates generated by the software did have 10 calculation points per degree.

The CNC machine did not compensate and cut the cam exactly like the CAD file, with straight lines between the .030" spaced points. The machinist saw the flat spots and thought they were cutter marks, and polished them out by hand.

The only problem with the design at present is that there is a little drag on the cam around it's home position when you put a small preload on the rollers. I'm hoping that re-machining the cam accurately will take care of this problem.

The only way I know to get the curve in the CAD file smoother is to plot the polar coordinates in AutoCAD and then run a spline curve through the points, then extrude it into a 3D solid and send that to them for programming. However, they will probably translate that to an IGUS file and who knows what happens at that point.

Perhaps the best bet is to have them CNC program right from the polar coordinates. I may also get a quote from a company that specializes in cams and this may save some trouble in the long run.

Thanks again for your feedback guys.
John

 

[John2004]

Mike,

Regarding your previous comments about using drawn cup needle roller bearings as cam followers, did you use a caged roller bearings or full complement bearings ?

I think full complement bearings may work better since the bearing is "fully loaded" with needle rollers and this would offer more support to the outer ring or cup.

I'm going to give the full complement type a shot as I just can't find yoke mountable needle roller bearings that are this small, other than the .250" OD Timken drawn cup bearings.

John

 

[MikeHalloran]

A long time ago, I inherited responsibility for a clever assembly that used (full complement) drawn cup needle roller bearings as cam follower rollers, riding on molded plastic cams, driving piston pumps with pistons up to 7/16" diameter, driving water through a complicated plumbing system with an average bore of .025" or so.

The area ratio between the pump and the tubing greatly magnified imperfections in the cams and the dynamic deflection of the drawn cups. The difference was detectable with a good pressure transducer and oscilloscope, and in the performance of the end product.

Where it was a problem, one partial solution was to substitute a pair of instrument ball bearings for the needle bearings. The ball bearing races were a bit stiffer. You could still 'see' the balls passing by, but the pressure fluctuations were smaller.

Years later, I had occasion to design a similar assembly for a similar purpose, also having similar magnification by means of the fluid system. I used real cam followers, which worked much better. They were also considerably larger, so the base circle and the assembly were also much bigger.

Nowadays, I'd probably use brushless DC motors and high- resolution encoders driving leadscrews, and I'd substitute software for the cams. Cams haven't gotten more expensive, but electronic motion control has gotten much cheaper.





Mike Halloran
Pembroke Pines, FL, USA

 

[John2004]

Hi Mike,

One other question came to mind on this.

Should I design the cam curve to work with a roller that has a diameter about .015" larger than it's really going to be ?

Since my main problem is drag on the cam at one point, it seems to me that if I design the curve to work with a roller that's a little larger than it's really going to be, then the curved rib would be less likely to bind or drag in-between the two rollers. This may help account for manufacturing inaccuracies on the cam and roller, especially if I use the drawn cup needle rollers as cam followers.

I'm not sure what the ring OD tolerance or out of round tolerance would be for a .250" drawn cup needle roller prior to being pressed into a housing , I'm sure there's no official catalog tolerance since their designed to be pressed into a housing. I guess inspecting 50 or so is the only way to get an idea.

Perhaps I could get the same effect by using a cutter that is a little larger than the roller, although I may have more flexibility with the curve design since I can enter whatever roller size I want into the cam design software.

Thanks again Mike,
John