Zusammenfassung
To increase the productivity or efficiency of manufacturing systems, theuse of model-based methods is one of the most promising approaches.Therefore, large varieties of models for different aspects of machinecomponents or entire machine tools are used for simulation throughouttheir life cycle phases. This includes models of dynamics with finiteelements or multi-body, geometrical data in CAD, electrical setups inECAD, models of wear or material flow or models of powerconsumption, to name a few. All those models have in common that ittakes a lot of time and resources to create, parameterize and maintainthem. Additionally, with the computational power of cloud computing, anew frontier has been set to record, store and analyze data efficiently.This computational power can be exploited for model identification,parameterization as well as learning for the mentioned multitude ofsimulation models used in the production field.As power consumption is one of the main drivers for production andoperational cost of a product, it is necessary to make operators ofmanufacturing units and machine tools aware of the current powerconsumption of their operating equipment. Moreover, model-basedapproaches for power consumption monitoring, prediction and even forthe power optimal operation of machine tools are being developed;however, they lack proper methods for the easy and quickparameterization of the applied consumption models. Most commonly,long power consumption measurements on site, usually whilstinterrupting the production process itself, are needed to create andparameterize these models. One spends a lot of effort before there is abenefit from model use.This paper presents an approach to power consumption model learning,driven by control signals of machine tools. It is based on the fact, thatthe machine control itself has the biggest influence on the powerconsumption behavior. The influence channels are the control signals tothe components.
Nutzer
