The factory has its digital twin. The materials inside it need one too.
Siemens now trains factory inspection on a digital twin instead of waiting for prototypes. Materials R&D has the same problem, with one difference: its twin must predict what nobody has made yet.
Siemens’ partnership with NVIDIA tackles a problem we know well. Its teams in Erlangen used to wait weeks for a physical prototype of each new circuit board before they could train quality inspection. Now they train it on images from a digital twin. Siemens says this virtual ramp-up cuts engineering effort by up to 85% and time to market by up to 40%.
In materials R&D, the prototype is the sample. Each new lead-free piezoceramic has to be calcined, sintered, poled and measured, which takes days to weeks per composition. And the target is demanding: coupling within 10% of PZT-5H, a Curie point above 250 °C, no lead, limited bismuth and niobium, and it must sinter on the customer’s existing line.
Same problem, same answer
The physical world is slow and expensive to iterate on, so you want to make your mistakes in simulation first. The answer, in both cases, is a digital twin you can trust because real data keeps correcting it. Siemens builds its twin on real factory data. In our loop, every measurement, failures included, recalibrates the models.
One difference
A factory twin models machines that already exist. A materials twin has to predict something nobody has made yet, and standard simulation is weakest exactly where these ceramics are hard: temperature and disorder. So each of our predictions carries a calibrated uncertainty, and only the compositions with the best odds go into the furnace.
We are starting with lead-free piezoceramics, which manufacturers need before the EU exemption for lead ends in December 2027. If you make, use or invest in advanced materials, we would like to hear from you.

