Proceedings of the 2008 International Conference on Electrical Machines

Paper ID1133

Advanced Parametric Environment for Electrical Machines Design Optimization

2

Konstantinos G. Papadopoulos and 1Christos Mademlis Alexandros M. Michaelides, 2Christopher P. Riley, 2 Nick Robertson and 2Isabel Coenen Aristotle University of Thessaloniki, Department of Electrical and Computer Engineering, Thessaloniki, 54124, Greece Tel & Fax: +30 2310 996234, e-mail: mademlis@eng.auth.gr Vector Fields, 24 Bankside, Kidlington, Oxford OX5 1JE, UK Tel: +44 (0)1865 370151, fax: +44 (0)1865 370277, e-mail: info@vectorfields.co.uk
2 1

Abstract- The paper describes a template-style front-end to a generic electromagnetic modeling tool, for the analysis and optimization of Electrical Machines. A two and three-dimensional FEA model for a generator and motor can be created in minutes, using templates with 'fill in the blanks' style screens. Accurate virtual prototypes can then be produced to help engineers provide answers on the performance of specific machine designs rapidly, and perform searching 'what-if?' investigations to identify the design characteristics of the perfect machine. Optimization tools are also available within the Environment, enabling engineers to find the 'best' solution automatically. Equally important is that the Environment is structured to allow creation and analysis of customised geometries, including special proprietary features.

TABLE I OFFERED MACHINE TYPES

2d-version
Induction Machine Synchronous Machine Switched Reluctance Machine Permanent Magnet DC Machine (rotor armature) Brushless PM Machine (many variants)

3d-version
Induction Machine Synchronous Machine Switched Reluctance Machine Permanent Magnet DC Machine (rotor armature) Brushless PM Machine (many variants) Axial Flux PM Machine

I.

INTRODUCTION developing an object oriented build up design environment [4] and with sensitivity analysis [5]. The present approach aims to develop a design environment for two and three dimensional analysis of electric motors and generators that could fulfill the needs of both the experienced and less experienced designer. The user provides the necessary geometric and electrical data for the machine through friendly dialog windows. The software builds the resulting machine model, performs the necessary solutions and provides simulation results at selected operating conditions. Variation of the given design parameters allows different scenarios to be tested and through an iteration process the user could arrive at an optimal machine design. Alternatively, the parametric model can be used to drive an optimization tool within the software, setting specific objective functions for the software to achieve. II. IMPLEMENTATION AND SIMULATION EXAMPLES The Electrical Machines Environment is an add-on toolbox available with the established commercial packages, Opera-2d and Opera-3d. Within the Environment, a FEA model for a generator or motor can be created in minutes using templates with 'fill in the blanks' style screens. Templates have been designed for most common electrical machine types, as listed in Table I. As with analytic computer programs, these templates represent the most characteristic geometries used in rotating

Many engineers designing rotating electrical machines currently employ analytic computer programs as the starting point for new designs. Such software solves electromagnetic equations for specific geometries, and is typically inexpensive and very quick to run. However, analytic solutions can compromise accuracy and, more importantly, are closedsystems that cannot be modified except by the originators. Analytic programs compute an average result for the overall geometry and only approximating. The alternative is a CAE tool employing, for example, Finite Element Analysis (FEA). These programs typically offer flexible GUIs, allowing users to simulate any design concept with supreme precision and accuracy. Wider analysis options are also on offer; for example, FEA programs can accurately compute eddy currents and naturally evaluate motional effects. However, the time required for analysis using FEM software, with its three step approach of pre-processing, solving and post-processing is unfavorable. While solution times have steadily decreased over the years owing to steady technological advances in computers, significant effort is still required by the user at the pre-processing stage, that is, building the geometry and setting the right conditions for solution. Thus, several works have been presented for improving the design environment enhancing the electromagnetic analysis [1], [2], adapting the dimensional model of the electromagnetic devices [3] and

978-1-4244-1736-0/08/$25.00 2008 IEEE

Proceedings of the 2008 International Conference on Electrical Machines

(a)

Fig. 3. Graph of electromagnetic torque versus rotor angle of a 4-phase, 8/6 SRM (typical simulation results)

(b) Fig. 1. Dialog window requesting information for: (a) the stator and (b) the rotor of the SRM

Figs. 1(a) and (b) show one such example for the definition of a 4-pahse, 8/6 Switched Reluctance Motor (SRM). All machine geometry information including teeth numbers, lengths, and angles are parameterised providing geometric flexibility. The program builds the machine geometry based on these parameters (Fig. 2). If the user is satisfied with the geometry created, they may proceed to analysis. Analysis data, specific to each type of machine is subsequently entered, as well as solution details, including mesh density and the required resolution in the results. The program proceeds with solutions to multiple cases and machine specific post-processing. One such example of results, the SRM electromagnetic torque Vs mechanical rotor angle curve is shown in Fig. 3. All output data is stored into named folders so that users are able to recover and further examine results. As an additional example, sequential dialog windows for the definition of the brushless PM synchronous machine rotor are illustrated in Figs. 4 and 5, respectively. A sample result of the model solution is shown in Fig. 6 representing the graph of static torque versus rotor angle on a 3-phase, 8-pole surface mount magnet PM synchronous motor.

Fig. 2. 4-phase, 8/6 SRM 3d-model

machinery. One important feature of the Environment is that templates are built using generic scripting and parameterisation techniques and the underlying code can easily be modified by users, providing the freedom to create and analyse customized geometries, including special proprietary features such as profiled stator teeth in SRMs or flux weakening features in PM machines.

(a)

Proceedings of the 2008 International Conference on Electrical Machines

Fig. 6. Graph of torque versus rotor angle of the PM synchronous motor (simulation results)

(b)

(c) Fig. 4. Dialog window requesting information for: (a) the PM motor type, (b) shape of the magnets and (c) dimensions for the magnets and retaining can of the rotor Fig. 7. Dialog window informing the user of insufficient room for stator tooth construction.

Fig. 5 PM synchronous motor 3d-model (3-phase, 8-poles, surface mount magnet type PM synchronous machine)

III. MANIPULATING DESIGN CONSTRAINTS

The structure of the Environment is open to the user. The user is able to examine the logical organisation of the models

and analysis settings and change or add specific features. Addition of features can range from the addition of minor geometrical features, winding arrangements, complete stator or rotor structures or alternative analysis and post-processing requests The design of the machine is subject to constraints which are activated during the model definition. These are geometrical constraints and are derived from the technical drawing. A set of algebraic expressions have been assigned for each design parameter so that the respective design constraint is implemented. When the input value of a geometric parameter is out of the range specified to each parameter & model, the software responds with an error message and prompts the user to alter the input value through a technical drawing. Such constraints simplify the desired parameterization within a machine model and avoid the cost of aimless designing iterations. The use of variables and expressions in the design constraints allows changes to the geometric dimensions to be made quickly.

Proceedings of the 2008 International Conference on Electrical Machines ter space. The optimiser begins by submitting a range of designs across the input parameter space to the Opera batch processor, to gain a diffuse knowledge of the relationship with the objective space. The searching algorithm then begins to home in on regions of interest where minima occurs. However, exploratory models are also built in sparse regions of input space to reduce the likelihood of missing other small but potentially deep minima. A balance is therefore maintained between the two to prevent effort seeking tiny improvements on potentially false minima (see sample objective function evolution and Pareto space of Figs. 9 and 10). The optimiser search algorithm analyses the stochastic s properties of the input space and utilises a Kriging-assisted surrogate method to predict the shape of its solution surface and thus determine the position of the next model with the highest likelihood of improvement. Where multiple objective functions are specified, solutions are ranked according to their location between Pareto surfaces in the objective space, [6],[7].

All constraints can be adjusted/altered by the user, who can also provide additional constraints pertinent to the particular electric machine variant designed. In similar fashion, postprocessing can also be modified or added-to matching the expectations of the user. IV. OPTIMISATION Once the user has produced a design using the Electrical Machines Environment they can chose to optimise it automatically using the general purpose Opera Optimiser. The optimisation process takes the original geometry, adjusts it automatically, solves the model using finite elements, checks the results for improvements and carefully selects a new geometry with a high likelihood of further improvements to the design. During a simple interactive set-up procedure (Fig. 8) the user is able to select important input parameters from the design

Fig. 9. The Evolution with iteration of: the two normalised objective functions (left); the normalised Fourier harmonic constraint, A5 < A3 (right). Fig. 8. Optimiser dialog window displaying the constraints tab.

environment; these will be adjusted as the optimiser creates new geometries in its search for a global minimum. A postprocessing analysis with resulting parameters can be created to allow the optimiser to define the quality of the generated model. Input parameters can be assigned upper and lower limits, to prevent the construction of infeasible models and to define the size and shape of the input parameter space. However, due to the automatic geometry checking within the Machines Environment the optimiser will not construct geometrically bad models. These models are not simply ignored however; the optimiser realises the implications upon this region of the input parameter space. Constraints can be imposed onto the optimisation by creating functions of the input and output variables. Analysed model geometries can then be seen to satisfy the constraints in graphical form as a function of the interaction number. Again, the optimiser does not simply discard models which do not satisfy the constraints; it realises the implications on the input parame-

Fig. 10. The location of the iteration inside the objective function space showing the nine first rank Pareto solutions.

Proceedings of the 2008 International Conference on Electrical Machines Synchronous machine example in OPERA-2d In order to demonstrate the optimisation of an electrical machine, a synchronous machine with thirty six stator teeth and an asymmetric six-bar, four-pole rotor (as shown in Fig. 11) was constructed in Opera-2D using the Electrical Machines Environment.

Fig. 11. The example synchronous machine before optimisation.

The objectives of the optimisation were to minimise undesirable normalised Fourier harmonics of the radial magnetic field component on a 1800 arc along the gap region. High order harmonics are produced by both the rotor bars and the stator teeth, while lower order harmonics are generated by the rotor shape. Thus, the A3 and A17 harmonics were selected as objectives to be minimised. The harmonics were normalised to the primary harmonic of the original model to maintain consistency. Four critical input parameters were selected as optimisation variables: the asymmetric radius of curvature of the rotor end; the width of the rotor end; the stator tooth width; and the inner stator coil width. Intelligent limits were chosen on the input parameters to define the size of the four-dimensional input space. Constraints were also imposed on numerous noneobjective Fourier harmonics so that they maintain their relative

Fig. 13. Radial B field component along a 180 degree arc inside the gap region demonstrating the reduction in high order harmonics from the original design (top) and a Pareto solution (bottom).

Fig. 12. The change in rotor and stator tooth geometry between the original (left) and a Pareto solution (right).

relationship to the objective harmonics found in the original model. Thus, preventing their growth is a response to the minimisation of the objective harmonics. Fig. 9 shows the objective functions and one of the constraints development as the optimisation progresses. The optimisation process converged to nine Pareto rank one solutions after 117 iterations; it took approximately twelve hours on a relatively cheap dual processor desktop PC with 2GB of memory. The majority of the time was spent, not in solving the finite element models since each of these took only a few minutes, but in the optimiser Kriging algorithm bes tween iterations; due to the large four dimensional input space and subsequent matrix inversions. The evolution of the objective functions and constraints through the optimiser iterations can be displayed graphically s (Fig. 9), as can the location of models within the input and objective parameter spaces. Fig 10 displays the model locations inside objective space and distinguishes between feasible, un-

Proceedings of the 2008 International Conference on Electrical Machines feasible and Pareto solutions. The resulting geometric changes to the machine are displayed in Fig. 12. Examination of the nine first rank Pareto solutions shows that the seventeenth order harmonic has been reduced to between a third and a half of its original value depending on the model. The constraints imposed on other harmonics resulted in them being reduced also. The third order harmonic was seen to be reduce by approximately ten percent from its original value in most of the solutions. This implies that the rotor input parameters selected do not provide sufficient control of this harmonic and that an intelligent replacement should be selected; thus, allowing the optimisation process to be repeated. Fig. 13 reproduces the magnetic wave form inside the gap region and demonstrates the improvement of a Pareto solution over the original design due to it containing smaller high order Fourier harmonics. The optimisation of this synchronous machine can be viewed as a demonstration of the type of route now available to machine designers striving for the ultimate system design and that further examinations are required with the aim of improving the purity of Fourier terms further. Switched Reluctance machine example in OPERA-3d In this example, the width of the stator and rotor teeth of a 1phase 4/4 switched reluctance machine, shown in Fig. 14, were optimized to provide a sinusoidal change of stator flux linkage against angle, a feature that has been associated with smoother torque variation and lower acoustic noise.

Fig. 15. Optimisation Results (initial R2 = 0.894, final R2 = 0.981)

V. CONCLUSIONS This approach to design can deliver significant advantages in today's market environment. The accuracy of FEA simulations, combined with the easy to interpret delivery of results, gives designers the means to rapidly make informed decisions whether the need is simply to make the most cost-effective solution for a given application, or to come up with something new. Currently, there's enormous pressure to improve energy efficiency for instance. FEA allows searching 'what-if?' investigations to be performed rapidly, identifying the design characteristics of the right machine with great accuracy. Preliminary design studies can be performed in minutes. Optimisation tools are also available within the Environment, enabling engineers to find the 'best' solution automatically. VI. REFERENCES
[1] C. F. Parker, J. K. Sykulski, S. C. Taylor, and C. S. Biddlecombe, Parametric Environment for EM computer aided design, IEEE Trans. Magnetics, vol. 32, no. 3, pp. 1433-1437, May 1996. F. Deng and N.A. Demerdash, Comprehensive salient-pole synchronous machine parametric design analysis using time-step finite element-state space modeling technique IEEE Trans Energy Conversion, vol. 13, no. , 3, pp. 221-229, Sept. 1998. R. Rong and D.A. Lowther, Adapting design using dimensional models of electromagnetic devices IEEE Trans. Magnetics, vol. 32, no. 3, pp. , 1437-1440, May 1996. M.B Norton, P.J. Leonard, object oriented approach to parameterAn ized electrical machine design IEEE Trans Magnetics, vol. 36, no. 4, , pp. 1687-1691, July 2000. P.J. Weicker and D.A. Lowther, sensitivity-driven parametric elecA tromagnetic design environment IEEE Trans. Magnetics, vol. 42, no. 4, , pp. 1199-1202, April 2006. G.I. Hawe and J.K. Sykulski hybrid one-then-two stage algorithm for A computationally expensive electromagnetic design optimization.COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering, 26 (2). pp. 236-246, (2007). G.I. Hawe and J.K. Sykulski, Considerations of Accuracy and Uncertainty with Kriging Surrogate Models in Single-Objective Electromagnetic Design Optimization.IET Science, Measurement & Technology, 1 (1). pp. 37-47, (2007

[2]

[3] Fig. 14. The 1-phase 4/4 Switched reluctance motor topology

[4]

In examining this problem, the Optimiser needed to operate on a series of (instead of a single) OPERA-3d solutions. Flux linkage was measured at ten rotor positions B() , = 0,5,10, ,45 Flux values are normalized, so they vary between 0 and 1. A goodness of fit parameter, R2 was then constructed, which compared the normalized B() to the standard cos() function. R2 was therefore the objective function to be maximized, with an ideal value of 1. Fig. 15 illustrates the end result of the Optimisation, always bearing in mind that the Optimiser was working within the user-specified variable constraints.

[5]

[6]

[7]