We all know that temperature affects materials but how does that ultimately affect our parts? On the surface it would appear a simple task to compensate for. All we should have to do is take the coefficient of thermal expansion (CTE), look at a thermometer and psyfer up the solution. Well, it doesn’t quite work out to be that simple. There are just a few more variables that we have to consider before we are able to make this calculation. Here are the variables that we really need to address:
- CTE of the part including different materials in the same part,
- CTE of the machine and feedback scales,
- Geometric changes in the machine with temperature changes,
- Heat added by the machining process, and
- Heat added or removed by cutting coolants and lubricants.
After we have all these variables considered, then we have to decide what temperature we want the part to be normalized to.
What’s the best way to control all these variables? Best practice is usually to eliminate them. Set the environmental temperature of the machine and the part to the desired normalized temperature; let the part and machine “soak” at that temp for 24 hours or so. Use a cutting coolant that has been conditioned to the same temperature as the normalized environment and then monitor these conditions and control these as needed. This is the perfect world but it isn’t the norm. What we normally see is a part designed to be dimensionally correct at 68F or 20C and a machining environment at 70F to 75F. Part CTE’s from 1 to 13PPM. Now we have to do some real psyfer’n.
Environmental Temperature Control
Environmental temperature control is the most common and is how CMS generally quotes the specifications of our machines.
The standard is +/- 1 degree C or about 3 degrees F. The CTE of steel is around 7.3 ppm/F. = .0073mm/m/1F = .022mm/m/3F = .22mm for a 10 meter machine in a 3 degree F temp spread. As you can see this is quite a bit of movement for a 30 some foot machine. Aircraft stringers, helicopter blades, flight controls, and many other parts and pieces are this big and must be machine to tighter tolerances than this. The tighter this can be controlled the better.
Just as a note, the Lawrence Livermore National Ignition Laboratory, who is the source for many high precision measurements, has a lab that is a 10 meter cube and is maintained to 0.1 degree F for their testing.
There are also standards, like the ASME 5.54, that have thermal drift as one of the standard measurements in machine tool analysis.
Also, the type of heat is also important. Radiant heat, convective heat, and conductive heat make materials react differently also. I won’t go into details here but heat from sunlight, ovens, fans, air conditioning vents, all affect the machine differently. Some manufacturers even go to the extent of running liquid coolant through the machine frames and drive components which we will discuss next.
Component Temperature Control
Another time honored method is to monitor and control component temperatures. Things such as gearboxes, motors, encoders, and ball screws also change with temperature. For many years, and even now, ball screws were drilled and coolant run through them to control their expansion. Ballnuts were, and are still, encased in coolant jacket housing, likewise with gearboxes and motors. All modern spindles are liquid cooled, or more correctly temperature controlled.
For large machines some of these methods can and are used; but for a long axis, 3 meters plus, ball screws are not a very good choice. Rack and pinions are normally used and are not of a one piece construction. This allows the gear rack to move with the machine structure and not have to two different alloys fighting each other.
Structural Temperature Monitor
This is a method that uses modern electronics to monitor the temperature of the structure and modify the distance traveled through an algorithm set in the unit. The way this works is there are several temperature sensors set along a machine’s axis. This data comes back to a control box that modifies the encoders’ pulse signal length to correct for the thermal movement in the structure of the machine. This unit is completely separate from the machines controller. The temperature sensors and the machine’s encoders both go through this box. The encoder signal is then modified and sent on its way back to the controller. This is a very good system but relies on two very important things to operate correctly:
(1) the temperature changes must happen very slowly and
(2) must be derived from a convective heat, that is air to material.
Added to the importance of this is that this system must be tested, tested, tested. It is not possible to just put in the CTE of steel and get good results. The reason for the convective heat and slow change requirement is the temperature sensors can only compensate for the entire length of the axis, averaging all of the sensors. We can place 100 sensors on the axis but if one area is hot and one cold the unit will only average the temperatures for the entire length give poor results in certain areas. CMS is currently building a machine with this system and we are hoping to offer it as a retrofit by the end of the year.
Real Time Positioning Control
This is currently the best way to handle temperature variations on long machines. Not surprisingly it is also by far the most expensive. This system uses a laser interferometer to measure the distance the machine travels. This is also not part of the machine’s controller. It is the same device that is used to calibrate machine tools. The difference is this laser is permanently mounted to the machine and is online, real time, all the time. In the case of a gantry machine with a master-slave axis it requires two systems to correct for that axis’ movement. It will also correct for temperature gradients and is self-calibrating. The cost of this system is around $100,000 per axis. CMS is also reviewing this system to possibly offer it in the future.
There is one other variable that must be considered before deciding which avenue to take: Geometric changes. All of the compensative systems mentioned above will correct the axis positioning of the machine At the servo motor, it will not 100% correct for geometric changes in the machine that affect the final position at the tool. Fortunately small temperature changes do not affect geometry, but large swings will. Let me give an example. A machine is 20 meters long. It is welded to the foundation that is 4 feet thick and is insulated and isolated from the surrounding concrete and has 1 inch steel plates embedded in it. In 20 meters we will have 10 “legs” the shoulders will be set on. On the shoulders we have rails that are welded on the shoulders and on the rails we have the bearings and the servo drive motors. Let’s say the air temperature changes 10 degrees from morning to noon. We could further complicate things by saying the sun shines in on only one of the shoulders from 10 till 2pm adding more radiant heat to one side. At this point we have one rail basking in the sun, the other rail taking off his morning jacket, but we have the foundation still buried deep in his cocoon. Now we have one wall expanding 1.5 mm, the sun drenched wall maybe 2 or 2.5 mm and a foundation that is completely unchanged. What happens? I have not been able to measure this on this long of a machine and at this level of temperature spread; but on the smaller ones I have measured, and this is what happens.
The base of the leg does not move but the top of the leg does. It causes small humps between the legs. These humps are very small but what affect it has at the spindle is amplified depending on the length of the Z axis. As the X axis rides up and over these humps it causes the Z axis to “pitch” forward and backward, much like the mast of a sailboat would bobbing up and down in the waves. This causes linear displacement errors that cannot be compensated out by any current known method. There is one, but it has only been used on the scale of computer processor chip size. The good thing is at smaller temperature changes the affect is near immeasurable.
I would say the moral of the story is, “There is not just ‘one’ thing that can correct for everything. Rather it must be the application of diligent environmental controls in conjunction with technology.”
The Part Itself as a Variable
As I mentioned earlier, we need to also consider the part. The CTE of the machine will have one value and the part might be considerably different. How do we compensate for this? The ideal situation would be to set the machine and the part at the desired temperature and maintain it very closely. But what if the nominal temperature is 68F and the machine and part are at 72F. There are several options at this point. If the machine is going to be carefully maintained at this temperature we could calibrate the machine to a correct dimensionality at 72F using the CTE of the material. This is a good solution for this situation. Now what if the environment changes to 77F? The machine’s CTE will make it expand by, let’s say, 30 micro-milli-inches and the part will expand 10 micro-milli-inches. What do we do now? We have a couple of options. We could recalibrate the machine to this new environment by using the CTE of the part; we could also scale the program of the part by subtracting the part’s movement from the machine’s movement and calculate the ratio of the overall size. Or we could put in place one of the active machine compensation devices we’ve been discussing.
Linear Scale Control Option
There is one other device that is simple, effective, and inexpensive and is currently in use at CMS to correct for all these crazy movements. Before I describe this wonderful cure, I must mention it is only available for smaller length axes up to 3 meters at this point. The device is a linear scale with the same CTE as the material being cut. As I mentioned earlier, our options available for control are the environment, the part, the machine, or the feedback device. In this case the scale would be made of a material with the same CTE of the part, while the machine would be allowed to expand and contract at any rate and the feedback device would only let the machine move the distance it reads. The other option we have here is to isolate the scale from the machine movement and control just the environment inside or around the scale. This is the approach CMS has taken. We use a steel tape scale that is mounted inside a housing that is mechanically isolated from the machine and we control the temperature inside the housing. This way the machine is always correctly calibrated regardless of external conditions. Now the part movement will still have to be considered, but this gives us one less variable to factor in.
I hope this gives you a little more information about the systems available and what your best options would be to assist in maintaining positioning accuracy.