Data-based model order reduction for online monitoring

Data-based model order reduction techniques allow compact modelling of mulit-phyics simulation models developed by IP-protected software and make thus the model exchange among the mictroelectronic supply chain feasible. Moreover, MOR helps realizing health assement of machines such as increasing the fault detection accuracy of a motor rig with a mounted sensor box. Due to data-based nature of the approaches, espacially nonlinear structural dynamics problems (creep deformations, large displacements, etc) can be catured well by the compact models.

When placing a smart field device (i.e. a measurement unit) on an electric motor, the response of the system caused by the operational healthy and faulty forces is affected by the transfer path between motor, mounting adapter, and the sensor on the PCB. Hence, the health assesment of the machine can be wrong. Therefore, a signal correction should be applied to the transfer path. To get the transfer path, a detailed mechanical simulation of the entire system is required. In general, such simulation models can have an extremely large number of DOF(degrees of freedom). To overvome this difficulty, reudced order models can be very usefuly.

Furthermore, for nonlinear strucutral models, data-based model order reduction can be beneficial. Typical examples are strucutres with large displacements as well non-elastic deformation such as creep.

For realization the signal correction of the electric motors an access to real data from the motors is needed

Siemens, Infineon, NXP, Siemens Industry Software

To generate data, licenses for Simcenter 3D is requited.