John Wanjiku, Technical Marketing Engineer, Mentor Infolytica, a Siemens Business
The trend in CAE platforms is towards system-level (model-based) design with the aim of replicating the actual products. There is also an effort to ensure that these platforms are tailored for design engineers to ease the learning and application curves. As actuators become more electrified, system-level design becomes indispensable as it analyzes how a drive not only satisfies the load requirements but also noise and emission standards. Integration is also part of the design process, rather than at later stages, reducing development time and costs.
However, the system-level approach requires fast and accurate models. Although multiphysics (2D and 3D) co-simulation is accurate, it is slow, requires different solvers and is expensive (skilled analysts, hardware and licenses). 1D models (e.g. analytical models, tables, response surface models etc.) solves most of these challenges at the cost of accuracy. This article discusses the need, merits and applications of 1D motor models with an accuracy of FEA for high-performance drives that cannot be met by analytical models.
System-Level Electric Drive Design
Multiphysics is currently difficult to implement owing to the drawbacks highlighted previously. Conventionally, each drive component (e.g. motor or converter) and one physics (electromagnetic or thermal etc) are analyzed one at a time. The dependency on other subsystems and physics is loosely tied. Integration and system-level performance are done at later stages. Hence, design iterations and extended lead-times are inevitable, driving up costs.
A hybrid of multiphysics and 1D approach is practical. At the system-level, 1D models are used to set the design criteria and account for subsystem interaction. At the subsystem level, the analysis is via a hybrid of multiphysics and 1D models, while the interaction with the rest of the system is via 1D or fast multiphysics models. To generate these 1D models, low-penalty assumptions and empirical calibration, are necessary.
MotorSolve, a fully automated electromagnetics (EM) and thermal FEA package can generate 1D motor models with an accuracy of FEA. It has a user-friendly GUI and an automated model, meshing, solver, results and model export processing capability, which has increased the fraction of the analysis time, permitting rapid motor design.
System-Level Electric Motor 1D Models
High power density motors are prone to saturation and a variation of inductance which affects performance. These effects are difficult to model, modify and generalize analytically, especially for complex optimized rotor geometries (e.g. flux bridges and barriers in IPMSMs). The formulation also requires EM analysts. Co-simulation is possible at the cost of solution time. Tabulated motor parameters (e.g. flux-linkage) extracted using transient solvers have also been used.
MotorSolve innovates this 1D motor model generation process illustrated in Figure 1 by reducing the number of FEA operations, without sacrificing the FEA accuracy.
First, the supply, machine and thermal settings are specified. Then the rotor and stator are designed using customizable templates. The drive-type is also selected. The machine electromagnetic and thermal analysis is done via performance charts for supply, drive and machine performance. Fields distribution visualizations allow localised field analysis (eg. flux density distribution, current density, losses, demagnetization prediction etc). The machine is then calibrated using experimental data and exported to various 1D model formats.
Benefits of 1D Motor Models
1. Shareable between firms by concealing important design information (e.g. materials, stator, rotor and winding designs). Reduces the need for NDAs, saving on costs and lead-times.
2. Cost effective. The models have the accuracy of FEA (material nonlinearity, harmonics and geometric dependencies), without the penalty of computation and FEA costs (license and skilled EM analysts).
3. Suited for design engineers. MotorSolve’s user-friendly GUI and automated processes reduce the learning and development times.
4. Suited for system-level transient analysis. The models are several orders of magnitude faster than multiphysics models, with the precision of FEA.
Motor Emulation Using 1D Models
1. Testing and verification of real-time controllers in software (SIL), hardware (HIL) and power-in-the-loop (PIL), without the need of a physical motor, or when it is being prototyped.
2. Emulation of full power electric machines in PILs. The 1D motor models generate signals that are enforced through power amplifiers. This is a versatile and cost-effective way of testing power converters without investing in test benches and motors.
3. Testing for compatibility and interchangeability of subsystems at the system-level, e.g. performance changes owing to different motor types, thermal circuit designs etc.
4. Transient analysis (e.g. start-up, response, faults analysis etc) since there are no physical limitations such as power in SIL design and testing.
System-level 1D motor models are therefore essential in the emulation of electric machines in the design of more electrified transport, robots, wind generators, other servo and variable speed drives.
Example of Typical Applications of 1D Motor Models
BMW i3 and Prius 2010 motors are used to show the application flexibility of MotorSolve 1D models. The motors’ data is courtesy of Oak Ridge National Laboratory (ORNL), and information available in the public domain. The models were processed as illustrated in Figure 1, and imported in the respective system-level modellers for torque and thermal analysis as seen in Figure 2 (a) and (b), respectively. A 2D FEA BMW i3 model was also exported to MagNet for torque comparison.
The torque waveforms of the BMW i3 motor shown in Figure 2 (a) have the same average value, but there are minor differences in the torque ripple. The 1D motor model used in Simulink considered skewing, while in MagNet it was neglected (although it can be taken into account) to save on solution time. Mesh noise effect is also noticeable on the MagNet waveform. A very small time-step is also necessary for MagNet to sample the torque ripple effectively.
Figure 2 (b) shows the evolution of a Prius motor’s temperature for different locations of the motor cooling circuit in the overall PHEV thermal circuit in FloMASTER, under an SFTP US06 drive cycle.
These two examples show the versatility of the applications of 1D motor models generated by MotorSolve.
This blog highlighted the trend of system-level analysis approach, where the product performance is wholly analyzed and integration issues are resolved early; reducing the time to market. This approach requires not only fast, but accurate 1D motor models. MotorSolve meets this need for electric motors as it generates 1D models with an accuracy of FEA, which are well suited for high-performance drives such as in transportation, robotics, servo and variable speed drives applications. The application flexibility of MotorSolve 1D models was demonstrated using BMW i3 and Prius 2010 motors for torque and thermal analysis.
More information on how these models were generated, analyzed and calibrated in MotorSolve will be presented at the Motor & Drive Systems 2018 conference in Orlando.