Zeszyty Problemowe – Maszyny Elektryczne Nr 84/2009 87

Marcel Janda, Miroslav SkalkaBrno University of Technology, Brno

OPTIMIZING OF COOLING IN ELECTRIC MACHINES

Abstract: This paper deals with thermal optimization of electric machines. There are described the connec-tions of finite element methods with optimizations algorithms. This connection is used for modifications of brushless motor construction. Solution of physical field model is realized by FEMM program which cooperate with optimizing program. Optimizing program was created in Java programming language. This program can be easily modified for various Finite Elements Methods solvers. Cooperation between optimizing program and FEMM is realized be LUA script. Four optimizing algorithms were chosen for using in program: Random find, Bees method, Hill climbing and Simulated annealing. Compare of these methods is in this paper. The part of this work is the complete thermal analysis of AC machine and modifications of cooling.

1. Introduction

This paper describe thermal optimization of parts of electric machines. This problem is very important in this time. Cooling of electric ma-chines can be solved by many methods. Four mathematical optimization methods are used in the paper. These methods can be used thanksdevelopments of computers. Computing power of modern computers can be use for calcula-tions of complex mathematic problems. Opti-mizations of cooling system can be realized by 2D or 3D models of electric machines linked with solution and optimalization software. Modern software can modify parameters of model constructions and parameters of cooling system. Designer of electric machines have various possibilities for solutions of physical field in motors and for modifications of con-struction in this time.

2. Optimization methods Optimization methods refers to the study of problems in which one seeks to maximize or minimize a function choosing the values of ariables from within an allowed set. We can use some of mathematics algorithms for solve of this problem. For example: Ant colony Beam search Bees algorithm Differential

evolution Dynamic relaxation Evolution strategy

Genetic algorithms Harmony search Hill climbing Random find Quantum annealing Simulated annealing Stochastic tunneling

Many algorithms are based on nature phenome-nons. Genetic algorithms is very popular in this time. Other methods are ignored but they are also successfully in solution of optimization problem. For example, algorithms of Bees me

thod or Simulated anneling is very simple for programming. Results of these methods are enough accurate.

3. Optimization of electric machinesThermal optimization of electric machines was realizated by special program. This program was created for cooperations with finite ele-ments methods. This program changes two parameters of 2D model. These parameters are number of cooling holes and their diameter. Fitness function is calculated from flux density in air gap and arithmetic mean of temperature in stator. Program find minimum of temperature by minimal change of flux density. Results of program are new parameters of 2D model of electric machine. This model can be use forsolve of electric machine’s cooling by other methods. For example: Finite Elements Method (FEMLAB, ANSYS), Thermal Circuit, etc. Results can be used to making 3D model of electric machines. User can use classical algo-rithm for solve of flow cooling medium.

4. Program for optimization of electric machinesOptimization program was created in Java pro-gramming language. This choice has variousbenefits. One benefit is possibility of using this program in Windows and Linux operating sys-tems.Program use four optimization algorithms:

Random find Bees method Simulated anneling Hill climbing

Random find is not an optimization algorithm, but it was used for test of program.

Zeszyty Problemowe – Maszyny Elektryczne Nr 84/200988

Optimization program cooperate with FEMM program. This program is used for solution of finite element method. 2D model of electric machines for FEMM was created in Autodesk AutoCAD. Cooperation between optimization program and FEMM is realized be LUA script. This scripting language is used for modify con-struction model and save results of simulations of physical fields.Objects structure of optimization program allow for many simple modifications of programstructure. One modifications of this program can be change of FEMM to Ansys or MATLAB programs. It can be used any solution program which contains a script language. Next advan-tage of this resolution is possibility using of this program for solution other problems which are different from thermal optimization.Optimization program is created for universal solve of optimization problems in all engineer-ing sections.

5. Comparation of optimizations methods

Comparate of optimizations method can be make by three parameters:

Time

There is comparation of four optimization me-thods in this paper. These methods are: Simu-lated annealing method, Bees method, Hill climbing method and Random find method. Very important parameter of optimalization is time, which need methods to solution of prob-lem. Random find method is fastest of usedmethods. Time is dependent on number of solu-tion steps. Random find method was set up on 50 steps.

Fig. 1. Compare of solution time for some optimization methods

Fitness functions

Fitness function is the parameter of optimaliza-tion methods. Optimization methods find solu-tion of function with minimal fitness function.

Minimal fitness functions had Simulated an-nealing method and Bees method. These me-thods had very good results by repeat start. Hill climbing method and Random find method had very different fitness function by repeat start.

Fig. 2. Compare of minimal fitness function for some optimization methods

Results

Optimization methods are used to find of cool-ing holes parameters. These parameters are number of holes and their size. Random find method and Hill climbing method were started two-time. Individual results are different. It shows bad accuracy of these methods. The best result have Simulated anneling method. This method has equally result by repeat start and result is really optimal.

Fig. 3. Compare of results for some optimiza-tion methods

6. ConclusionThere is compare of optimization methods. These methods were: Random find method, Hill climbing method, Bees method and Simulated anneling method. Compare parameters were: time of solution, results and fitness function. From results is evidently, that best method is Simulated annealing method. This method had very good results by repeat starts of optimaliza-tion program. Time of solution is too relatively good. Other methods had different results by

Zeszyty Problemowe – Maszyny Elektryczne Nr 84/2009 89

repeat start. Second good method is Bees me-thod. This method needs too much time to solu-tion and it can stop in local minimum of opti-malization function. Random find method is not classical optimalizatin method and it is the worst of all.

ACKNOWLEDGEMENTResearch described in this paper was financed bythe Ministry of Education of the Czech Republic, under project MSM0021630516, the project of the Ministry of Industry Trade No. F1-1M/199 and the project of the Grant agency CR No. GAČR 102/09/1875

7. Bibliography[1] Chapman J. S.: Electric machinery fundamen-tals Fourth Edition. International Edition 2005, ISBN 007-115155-9[2] Bartsch H. J.: Matematické vzorce. Mladá fronta, Praha, 1996, ISBN 80-204-0607-7[3] http:// www.wikipedia.com[4] http://www.bees-algorithm.com[5] Russell, Stuart J.; Norvig, Peter (2003), Artifi-cial Intelligence: A Modern Approach (2nd ed.),Upper Saddle River, NJ: Prentice Hall, pp. 111-114, ISBN 0-13-790395-2[6] S. Kirkpatrick and C. D. Gelatt, M. P. Vecchi:Optimization by Simulated Annealing, Science. Vol

220, Number 4598, pp. 671-680, 1983[7] Vítek O. Analýza magnetických obvodů EC motorů, Brno,Technická zpráva, Prosinec 2006[8] Moreno J. F., Martínek M. J. D., Hidalgo F. P., Ruiz J. R. S.: Thermal study of a three-phase induc-tion machine, To be published on ICEM 2002[9] Boglietti A., Cavagnino A., Staton D.A.: Thermal Sensitivity Analysis for TEFC Induction Motors, To be published on PEMD 2004[10] Krum, Al.: “Basic thermal analysis“ in Sergent J., Krum Al. Thermal management handbook for electronic assemblies, To be published in New York: McGrawHill, 1998, 5.1-5.22[11] Boldea, I., Nasar S. A.: The Induction Machine Handbook , Published 2002 CRC Press, ISBN 0849300045

Authors Marcel Janda, Ing.,e-mail: janda@feec.vutbr.czMiroslav Skalka, Ing.,e-mail: xskalk02@stud.feec.vutbr.czBrno University of Technology,Faculty of Electrical Engineering and Communication,Department of Power Electrical and Electronic Engineering,Technická 8, 61600 Brno, Czech Republic