Johannes Ziske

A Simple Phenomenological Model for Magnetic Shape Memory Actuators

This paper presents a new phenomenological model for magnetic shape memory (MSM) alloy actuators. The model was implemented as a lumped element for multi-domain network models using the Modelica language. These network models are rapidly computed and are therefore well suited for MSM-based actuator design and optimization. The proposed MSM model accounts for the 2-D hysteresis of the magnetic field-induced strain as a function of both the applied magnetic flux density and the compressive stress. An extended Tellinen hysteresis formulation was utilized to compute the mechanical strain of the MSM material from measured upper and lower limiting hysteresis surfaces. Two alternative approaches for the computation of the lumped element have been implemented. The first method uses hyperbolic shape functions to approximate the limiting hysteresis surfaces and offers a good balance of simulation accuracy, numerical stability, computational speed, and ease of parameter identification. The second method uses 2-D lookup tables for direct interpolation of the measured limiting hysteresis surfaces, which leads to higher accuracy. Finally, a test case having simultaneously varying compressive stress and magnetic flux density was utilized to experimentally validate both methods. Sufficient agreement between the simulated and measured strain of the sample was observed.

Novel Electrodynamic Feed Units for Small Machine Tools and Automation

Short travel ranges up to approx. 25 mm as in future small machine tools enable linear direct drives with simple single-phase design. Especially designs with moving magnet(s) and an iron core stator winding allow for large actuator constants, i.e. high forces at little losses and small volume. Different types of those compact, dynamic and cost-effective linear axes and tables as well as a novel planar direct drive have been developed, built and tested. They feature integrated ball or flexure guides, integrated incremental or absolute position sensors with resolutions from 0.16 to 1.25 \(\upmu \)m, embedded flatness-based position control, sensorless force control and various control interfaces. Selected prototypes and their features are presented.

Phenomenological Models of Solid State Actuators for Network Based System Modelling

Smart materials, such as thermal or magnetic shape memory alloys, provide interesting characteristics for new solid state actuators. However, their behavior is highly nonlinear and determined by strong hysteresis effects. This complex behavior must be adequately considered in simulation models which can be applied for efficient actuator design and optimization. We present a new phenomenological lumped element model for magnetic shape memory alloys (MSM). The model takes into account the two-dimensional hysteresis of the magnetic field induced strain as a function of both the compressive stress and the magnetic flux density. It is implemented in Modelica. The model bases on measured limiting hysteresis surfaces which are material specific. An extended Tellinen hysteresis modeling approach is used to calculate inner hysteresis trajectories in between the limiting surfaces. The developed model provides sufficient accuracy with low computational effort compared to finite element models. Thus, it is well suited for system design and optimization based on network models. This is demonstrated with exemplary models of MSM based actuators. System models and simulation results are shown and evaluated for different topologies.Copyright

A New Method for Coupling Transient Network Models and Stationary Finite-Element Models

In many cases, the accuracy of transient multi-domain network models can be improved by coupling to distributed models, e.g. finite-element (FE) models, which compute for specific element parameters, flow or potential variables of the network model. Two opposing methods are known. The first is direct simulator coupling. It requires solving of the distributed model in each iteration step of the network model simulation. The second is the uncoupled calculation of characteristic maps from stationary distributed models which are then used in the transient model in form of look-up tables. Since the course of the base parameters of the characteristic maps is unknown before the transient simulation runs the stationary distributed model has to be solved for all grid points of the spanned parameter space. Both methods lead to an inefficient high number of necessary calculations of the distributed model which usually determines the computing costs. We present a new approach which significantly reduces the number of necessary computations. The main idea is combining both methods and successively computing grid points of the characteristic maps depending on the current need while solving the transient model. This is demonstrated for the example of an electromagnetic actuator. In the presented example, the number of FE model calculations was reduced to a tenth.Copyright