Building a high-precision machine is complex and therefore expensive. A good CNC milling machine is expected to mill the desired shape to an accuracy of tenths or even hundredths of a millimeter. The larger the machine, the more difficult it is to be so precise throughout the entire working area of the machine.

There are milling machines that are 20 meters long, but even if they are only 1 meter long, one hundredth of a millimeter is one hundred thousandth of the entire length. The machine must be extremely stable and solid so that the forces generated during milling do not bend the machine. This is because even minimal bending can impair accuracy.

For comparison: if you hang 10 kg in the middle on a one meter long, 5 cm thick steel bar that rests on both ends, the bar will bend by a whole millimeter.

In addition to the stability of the machine, the machine itself – including the drive system, for example – must be manufactured very precisely, as any inaccuracy in the machine can lead to an inaccuracy in the workpiece being milled. Mechanical engineers have decades of experience in building precise machines.

But even the best mechanical engineer cannot change the laws of nature: every material expands when the temperature rises. The aforementioned one-meter-long steel rod is a quarter of a millimeter longer at 40 degrees Celsius than at 20 degrees Celsius. That’s huge if you want to mill to an accuracy of a hundredth of a millimeter.

Software compensates for errors

Software can compensate for these errors. CNC milling machines are controlled by a control system (in our case Siemens SINUMERIK ONE), which ensures that the machine’s milling tool mills at the exact coordinates required by the customer at any given moment. If the controller now knows that the machine is still cold and is therefore a quarter of a millimeter shorter than when it is warm, then it can simply add the expected error to the coordinates beforehand so that the machine – although it is cold and therefore shorter – still mills in the right place

For this purpose, the Siemens SINUMERIK ONE controller offers the option of entering a table (the technical term is “compensation matrix”) that specifies how much compensation is to be made in which situation.

Sounds simple. In practice, it is a little more complicated, as a modern CNC milling machine typically has five axes and each of these axis positions must be compensated and changes to one axis also have an effect on the other axes. This can be solved with math. However, it is more difficult to know which part of the machine is at which temperature at any time during production and by how much the length of each part changes at which temperature. The control system needs to know this in order to be able to calculate how much it needs to correct the coordinates so that it mills in the right place despite the machine heating up. If the machine has been switched off for a long time and is cold, then the whole machine has almost the same temperature. However, as soon as the machine is switched on and starts milling, it makes a difference whether it mills at the top left at the back for a long time or whether it mills at the bottom right at the front and thus generates heat elsewhere. The control system should therefore continuously adjust this compensation table during operation – even after the warm-up phase has been completed.

Research with mechanical engineering ecosystem

The Institute of Production Engineering and Photonic Technologies at the Vienna University of Technology, together with its ecosystem of leading mechanical engineers, is researching how the heating of the individual parts of the machine can be determined in a digital twin. Institute Director Friedrich Bleicher summarizes the goal: “Our aim is to be able to work just as precisely with a cold machine as with a machine that has already been warmed up.”

Our aim is to be able to work just as precisely with a cold machine as with a machine that has already been warmed up.

Friedrich Bleicher, Director of the Institute of Production Engineering and Photonic Technologies at the Vienna University of Technology

© TU Wien

In a research project, a large number of temperature sensors can be installed in the machine. In industrial practice, this is not economically feasible. “That’s why,” says Christoph Habersohn, head of department at the institute, “we use the data from the SINUMERIK controller together with existing temperature sensors to calculate the entire temperature distribution of the machine and the changes in length in a digital twin of the machine.” The digital twin can be created either with artificial intelligence or with a physical model.

© TU Wien

The thermal simulation in the digital twin of the machine is used to calculate the heat spread to the entire machine. The heat source is the engine – this is where the machine’s power is generated (time curve from left: shortly after starting, after 2 or 7 hours).

Matthias Kneissl, Head of the Siemens Machine Tool Systems Division in Austria and the CEE countries, adds: “The Siemens multi-error compensation technology – MEC – of the SINUMERIK ONE makes it possible to compensate for these length changes and thus avoid rejects during the ramp-up phase. This increases the competitiveness of our customers and also saves resources.”

The project is expected to improve the precision of the machine by a factor of ten from the temperature compensation in the warm-up phase alone.