A Complete Solution for Electromechanical System Design Dr. Weizhong Fang, Ansoft Shanghai, China Dr.

Uwe Knorr, Ansoft Corporation, USA Scott Stanton, Ansoft Corporation, USA Dingsheng Lin, Ansoft Corporation, USA

Abstract With increased requirements for higher energy efficiency of electrical drive systems and an increased use of electrical drives in transportation applications a more accurate prediction of the behavior of electrical machines is neces sary. The design of electrical drive systems involves many aspects; from very detailed electromagnetic, thermal and semiconductor level analyses to system level mechanical and control design tasks. Often these designs are performed by individual design specialist/groups without an integral approach taking the various interactions between the system components into account. This paper will describe a complete design environment for electrical drive systems from component to system level design providing an integrated toolset for various drive design tasks. Integrating the design flow The design of drive systems involves a wide variety of analyses and design methods. They reach from finite element analysis for electromagnetic and thermal design problems over detailed electrical simulations including dynamic and thermal semiconductor behavior to rather abstract and higher level control design tasks. For each of these levels specific design packages are available using different numerical methods. Which method is used depends on the design phase and the simulation target as well as on the availability of models and comput ation power. Today’s technology provides unprecedented computation power already on PCs. This allows the engineer to re-evaluate traditional design methodologies and to go to more complete design strategies. For the design of drive systems frequently stand-alone tools are used. These point tools have their individual strengths but usually miss interaction or even interfaces for data exchange. However often results of a specific design stage are required in another design stage as input. Depending on the implemented rules in a design group there are different ap-

proaches for the design of products. Usually the first step is a selection of an initial design based on technological specifications and customer requirements. In terms of a drive system the selection of the electrical machine type and the specification of the dimensions, winding data, materials etc. would be a typical approach. At this design stage lots of design variants have to be compared and analyzed. Therefore analytical solutions in combination with design databases can be used efficiently. Numerical methods are still to time consuming to narrow the design space. After the initial design an implementation specification is available, covering most of the design requirements and specifications. From here the design paths can vary depending on the number of design issues covered and design teams involved. Usually the design path splits into different physical domains. In terms of the drive system the motor design is transferred to the electrical machine design specialist. At the same time system designers can start to evaluate circuit topologies and control strategies. For both levels analytical methods can deliver specific data sets. While in the machine design geometry and material or winding schemes are dominant, in the design of the power electronics and control system lumped parameter models are usually used. This splitting of design flows allows concurrent developments and saves valuable design time. Each of the two design paths evolve separately and run several design cycles. Each of them may include optimizations.
Parameter Sets
Model Generation

Initial Design

Circuit, Control, System Simulation

Geometry & Material

Model Generation Model Generation Model Order Reduction

Parameter Sets, Lookup Tables

FEA-based analyses

Parametric Solution

State Space Equivalent Circuit Generation

In the following sections solutions of Ansoft´s Electromechanical System Design Suite are demonstrated for the achievement of different design goals during the design process. Switched reluctance drive systems receive continuous attention due to the robustness and low manufacturing costs. the Finite Element Analysis (FEA) based tool for highly precise electro-magnetic design. 3phase SRM motor and drive system used in this paper has no real industrial background. 3 the lamination proposal as it is generated by the software. The software package supports this by an intuitive and easy to use graphical user interface. The analysis is performed using analytical methods and therefore only lasts several seconds. As an example for the design of a drive system a switched reluctance motor system is used. The 6/4-pole. Inside the kernel three different simulation engines are Due to the analytical methods used in RMxprt the analysis runs only several seconds. stator and rating information can be entered in easy to use dialogs (Fig. 2 Design Results . 3 Design Results . Fig. 4). These data help engineers to evaluate the quality. Both geometry and performance data are available as a result. Extensive data evaluation capabilities and automatic report generation save valuable time while analyzing the data and creating pres entations. Design space reduction First step in a design is the selection of an appropriate design for the given specifications.Performance Data Fig. it allows the engineer to modify designs quick and . current at different positions. Since the analysis is relatively fast. Simultaneously a system level model is created for a direct integration into SIMPLORER – the power electronics and drive simulator. System level design path SIMPLORER – the system level design tool of the drive design suite – was specifically developed for power electronics and drive system simulation. Fig. However several issues. easy in order to investigate different solutions in an extremely short time. SIMPLORER is based on a unique simulator coupling technology (Fig. RMxprt can also be linked to Optimetrics – a powerful optimization tool. RMxprt is a tool that helps designers to create a whole machine design out of a number of technological specifications. Modern power electronics seem to be a possible solution to reduce torque ripple by using smart control algorithms. 1 Design Data Input The advantage of an integrated design environment comes to full effect when the next design stages have to be involved. 1). Geometry and material data can directly be generated Maxwell. It was selected due to its generic design avoiding the use of proprietary design data.Lamination Fig. Fig. Several features support an easy and fast model generation and provide exceptional simulation speed even for large-scale problems. After selecting the required motor type the technical data for rotor. such as noise caused by torque ripple prevented a wide range introduction of this motor type so far. 2 shows the flux linkage vs.

The position dependent control of the switches is modeled based on state machines. Another application could be the measurements of characteristic values online while a simulation is running. there are no parameters to be modified. each where the according phase is turned on. Using different accuracy levels the user can save considerable time during the simulation process. the symbol automatically shows the geometry of the motor. It allows analyzing complex control structures. specifically for power electronics applications developed tool. By placing an RMxprt symbol and assigning the according file name automatically an icon with the geometry of the motor is generated as well as all electrical and mechanical pins. Generally models are provided in multiple abstraction levels. The Block Diagram simulator is based on a dis tributed integration method. To see the effects of the switching behavior of power semiconductors dynamic IGBT or MOS models can replace the idealized static model. dynamic and dynamicthermal models for detailed electrical analysis. SIMPLORER´s open Simulator Coupling Technology provides a way to co-simulate with external tools. The position information is derived from the ac- . One switching cycle can be subdivided into 6 states. which is represented by the transitions between the states. mechanics or hydraulics on a more abstract level. The driver circuit is modeled us ing SIMPLORER´s static IGBT models. especially when it comes to the generation of pulse patterns for the power electronic semiconductors. This A_p A_m B_p B_m C_p C_m Torque Position + Torque Speed A + V position Fig. Since the model was specifically generated for an electrical machine defined by given geometric dimensions and material data. Ideal and static power switch models can be used advantageously for system design tasks.running concurrently. They can be used very easy for the modeling of controls of power electronic systems. dynamic and dynamic-thermal components are available. For each component several accuracy levels are available. Depending on the simulation target the user can select the according modeling level. These values can be used as cost functions for the integrated optimizer. link allows including the FEA machine model directly into the systems simulator and running it simultaneously. Typical applications are vector control units for drive systems. 5 shows a part of the schematic with the SR motor symbol and the driver bridge circuit. Thereafter the driver circuitry and the control are setup with SIMPLORER generic components. A state remains active until the position of the rotor has passed the off-angle (a_of). The software provides several power electronics and drives specific components such as power semiconductors and electrical machines. Fig. 5 RMxprt Model in SIMPLORER The SR motor model created during the analysis with RMxprt can be easily loaded into the SIMPLORER schematic. When the model file is imported into SIMPLORER. 4 Simulator Coupling Technology IA IB A + IC A The Circuit simulator is on non-SPICE based. All pins for electrical and mechanical connections are generated. State Machines extend the modeling capabilities towards discontinuous processes. In terms of power semiconductors ideal switches as well as static. Each engine was specifically designed for a physical domain that is frequently part of a drive system. D2 IGBT1 IGBT3 D4 IGBT5 D6 C/C++ Simulation Data Bus Simulink Vsrc Maxwell Circuit Simulator Block Diagram Simulator MathCad D1 IGBT2 D3 IGBT4 D5 IGBT6 State Machine Simulator + + A Fig. A future implementation using that technology will be the direct integration of the Maxwell2D transient finite element solver.

00e-004 1.60m 0.00e-004 6. The torque waveform shows some little oscillations in the rising slope of the signal. Fig. 7 shows the motor torque and Fig. No user interaction is necessary.00e-004 8. This saves significant computation time without loosing accuracy. Since all geometry and material data are available it is easy to generate a Finite Element model. 7 Motor Torque .80m 0. All data from RMxprt is already assigned and ready to use. Fig.50 IA.10m 0. 8 the three phase currents at constant speed.I IC.00e-003 t RMxprt automatically computes the minimum field domain needed for the finite element model.30 Torque.I 5. 8 Phase Currents PBoff SET: VSB:=0 PBon SET: VSB:=1 SET: pos1:=pos1+a_on FEA based machine design refinement The SIMPLORER system level model is generated automatically during the analysis with RMxprt.00m 0 0 1. The design of the drive system using the RMxprt models provides results in a very short time with a acceptable accuracy for many design purposes.25 0.30m 0. rp-pos1 >= a_off rp-pos1 >= a_on Fig.I 0.10 50.00e-004 4.00e-004 7.50 PCon SET: VSC:=1 SET: pos1:=pos1+a_on rp-pos1 >= a_off PCoff SET: VSC:=0 0 0. rp-pos1 >= a_off rp-pos1 >= a_on Phase Currents 7. The results correspond to the results of the analytic solution with RMxprt.50m 0.00 0 PAon SET: VSA:=1 SET: pos1:=pos1+a_on rp-pos1 >= a_on PAoff SET: VSA:=0 -2. that with an FEA based equivalent circuit model these short disturbances are eliminated. where all data has to be entered manually.00m t Fig. 6 Position Dependent Control Fig. 6). On the push of a button all necessary data stored as ready to open Maxwell project. This seamless integration saves valuable design time compared to disconnected software packages. Fig.00e-004 2. 9 Geometry in the 2D modeler 0.20m 0. Motor Torque 0.I IB. In each of the states the control signals for the phases are set accordingly (Fig.00e-004 3. These short changes are cause by the analytic nature of the solution when the stator and the rotor tooth are starting to overlap. 10 shows the result of Maxwell´s automatic meshing process. 9 shows the geometry in the modeler of Maxwell2D.tual rotor position using SIMPLORER´s integrated expression evaluator.40m 0.90m 1.15 0.00e-004 5.70m 0. Later it is shown.20 Fig.

Fig. . 14 represents flux linkage versus time. sensors or electrical machines. 14 Flux Linkage versus Time More accurate data for the system level simulation After the detailed analysis of the design some parameters of the electrical machine may have changed due to geometric and/or material and other modifications. A better and easier solution is the automatic generation of system level models based on a state space methodology. Fig. This technology can be used advantageously for several electromagnetic systems. Especially for Switched Reluctance (SR) machines. 11 Field Plot Fig. The traditional method is to manually extract lumped parameters for conventional Park-transformation models. The external circuit can be defined using the built in schematic editor. 12 a color shade representation of the magnetic filed density. such as solenoids. 10 Finite Element Mesh The Maxwell 2D transient solver analyzes the machine considering eddy current effects. 11 presents a filed plot and Fig. 13 Transient Torque Waveform Fig. saturation. To include the results of a FEA analysis there are different possibilities. motion effects and losses. this approach is a useful alternative to create fast and accurate system level models. The following pictures show some results of the transient simulation. This process however is time consuming and requires linearization that might lead to inaccurate results under certain circumstances. Fig. where a closed formulation based on a set of differential equations is not possible. Fig. All mechanical properties are defined in easy to use dialogs. In the result of the analysis the user gets electromagnetic field information as well as transient wave forms. Fig. 13 shows the transient torque waveform. Fig. Obviously an updated simulation on the system side will be necessary. 12 B Color Shade Representation Motor designers can derive a variety of design information from these graphical displays.

Multidimensional lookup tables embedded into the equivalent circuit represent the nonlinear behavior of the system.00e-004 4. If eddy currents are essential. Users can easily transfer data from one design level to another and model extraction is supported. Fig.00m 0 1. With the presented design suite these requirements are fulfilled. 16 show.00e-004 5.10 50.I 0. The simulation results in Fig.00e-004 8.00e-004 2.00m 0 -50. 16 Equivalent Circuit Model . However the achieved accuracy is sufficient for most detailed system type analyses. that the torque waveform now does not have the short disturbance in the rising slope. Even if the FEA based model usually has a very high number of equations (usually in the order > 10000) with the computation power available today it is possible to run these problems together.00e-004 7. The generation of the model is fairly fast because the parametric solution is based on multiple static finite element computations. The equivalent circuit model generated can be introduced into a SIMPLORER schematic the same easy way an RMxprt model is loaded.35 Torque.00e-004 3. The key for that solution is SIMPLORER´s simulator coupling technology. 17 Motor Torque Conclusion The design of electrical machines involves a wide variety of design methods and design data. the most accurate chance to integrate FEA results into a system level simulation is a direct integration of the finite element solver into the system simulator. In order to handle this complexity design tools must provide a tight integration and seamless data exchange capabilities.Based on the geometry and material definitions from the transient analysis. Motor Torque 0. where the flux linkage for various rotor positions and current distributions in the winding is determined and recorded Maxwell2D´s equivalent circuit generator derives a new system level model.00e-004 6. These mass elements have all the major mechanical non-linear effects already built in.30 0. 15 shows the same schematic with the imported model.00e-004 1.20 0.15 0. This of cause implies that the system level model will not contain eddy current effects. The transient FEA solver and SIMPLORER ex change simulation data at each simulation time step. D2 IGBT1 IGBT3 D4 IGBT5 D6 The mechanical load is modeled using a rotating mass element from the SIMPLORER mechanical library. a 2 or more dimensional parameter sweep model can be generated easily. Vsrc D1 IGBT2 D3 IGBT4 D5 IGBT6 + IA A + IB A + IC A N_1 _2 N N_3 N_4 N_5 _6 N MchRMas1 J Torque A + N_7 N_8 GAIN Angle position Fig.00e-003t Fig.25 0. among others stick friction and backlash. From the results of this parametric FEA analysis.