Automating transfer function estimation: LSRF method with coherence-based pre-filtering and weighting filters

The increasing individualization of products, coupled with a shortage of skilled workers in the fields of science, technology, engineering and mathematics (STEM), is a challenge for manufacturing companies worldwide. To achieve batch size 1 production without the expertise of longtime employees, machine tools must be able to automatically optimize machining processes in virtual space prior to execution. This requires simulation of the machining process. Therefore digital models are needed that represent machine tool and process behavior. However, due to a lack of skilled workers, these models must be built by the machine tools themselves. Based on a previous publication by Oexle et al., this paper presents a methodology for automating transfer function estimation using the least squares rational function (LSRF) method combined with coherence-based pre-filtering and the use of specific weighting filters. Fits  > 70 % could be achieved on measurement data sets with good coherence. However, for measurement data sets with poor coherence, good fits are possible but not guaranteed.