MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No.

2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

108

Design Optimization of 7.5 Kw, 4pole, 3-Phase, 50 Hz Induction Motor Employing Genetic Algorithm/Improved Genetic Algorithm using Sweep Frequency Response Analysis
Shivendra Prakash Verma Asstt. Prof. S.G.I.T., Ghaziabad

ABSTRACT Transformers are a critical part of an electrical utilitys asset base. Loss of a transformer in a utility, generation plant or process can cost many millions of dollars, depending on how long it is out-of-service. Online monitoring and diagnostics is a useful tool to help operators to manage their assets and make decisions on continuing operation, maintenance or replacement. To condition monitoring of larger power transformers, a practical method is to use the available bushing tap connection, which is called SFRA. Sweep Frequency Response Analysis (SFRA) is an established tool for determining the core deformations, bulk winding movement relative to each other, open circuits and short circuit turns, residual magnetism, deformation within the main transformer winding, axial shift of winding etc. In this paper, we will discuss a custom high-power signal generator that injects high-frequency signals on the bushing tap of the transformer under investigation, as well as a circuit to replace the bushing tap short and allow online operation of the system.

performance and optimized design depending on end user specifications. The effect of motor performance on surroundings and environment is of great concern also, e.g., the level of noise and acoustic comfort of train passengers and nearby residents has to be ensured. The motor starts making noise due to Maxwells air gap forces, which causes in creating stator vibrations in audible range. The problem of induction motor optimal design has received much attention since the beginning of computer sciences [4,13,16]. Many solutions and algorithms were designed based on following points. Actually many conventions are made on the basis of practical experiences with manipulations of Induction motor design, e.g., a small air gap improves the efficiency, but increases the magnetic sound level and decreases the overload capacity of the motor; increasing the stator length of yoke to diameter ratio generally lowers magnetic vibrations [14]; but it increases the material cost; and with a fixed motor size, it decreases the available out-put torque with increased rotor temperature. In literature so many single as well as multiple objective approaches for the electrical machine design has been proposed [13, 17]; using various methods and techniques. These multiple objective approaches may be dealt with several NLP methods [8, 9, 10, 11], as well as genetic algorithm [5, 6, 7, 12, 18] and improved genetic algorithm [15].

I. INTRODUCTION
Induction motors are being widely used for almost all practical purposes in industrial and domestic world, giving rise to the consumption of maximum share of the total electrical power generated. So, their efficiency, torque produced and cost becomes a major concern to be optimized. But the optimal design of I.M. is a problem due to the following reasons: 1. Induction motor involves so many variables, which nonlinearly affects the performance of machine. 2. Due to involvement of various conflicting parameters, objectives and their non-linear behavior during design and operation of the machine, it becomes quite necessary to balance between many variables, sacrificing one for the other to obtain the best possible

II. PROBLEM DEFINITION AND DESIGN APPROACHES

The generally used single phase equivalent circuit model of the motor is shown in Fig.1. The model offers reasonably good prediction accuracy with modest computational efforts, despite its shortcomings. This model is basically a per phase representation of a balanced poly-phase induction motor in the frequency domain, comprising six model parameters. The six parameters have their usual meaning [5].

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

109

value of objective function and x is design variable set. The penalty functions, P [gj (x), r], are expressed with respect to the type of inequality used. By means of exterior penalty function, constrained problems are converted to unconstrained problems by removing constraints. According to constraints, penalty function is defined as following [5]:
P(g j (x), r)
r [max (0, g j )]2 , j 1, 2, m 2 r [min (0, g j )] , j m, ....m

Fig. 1. Equivalent circuit Model of induction motor

(2)

When the constraint inequality is satisfied, the penalty function becomes inactive. The objective function emphasizes III. A N OVERVIEW OF OPTIMIZATION BY GENETIC the larger constraint violations and the optimization search tries to reduce these violations to zero, in the feasible region. ALGORITHM This would result in pushing the search into the feasible design In the most general sense, GA-based optimization is a region. All the constraints are satisfied within this region and stochastic search method that involves the random generation the optimization approach attempts to move the design into of potential design solutions and then systematically evaluates its best optimum solution [5]. and refines the solutions until a stopping criterion is met. There However, order of magnitudes of various constraints is much are three fundamental operators involved in the search process different from one another in case of electric motors. of a genetic algorithm: selection, crossover, and mutation. The genetic algorithm implementation steps are shown as Therefore, constraint functions need to be normalized with respect to the specified objective function to have meaningful follows: 1. Define parameters and objective functions (Initializing). convergence criteria. It is to be ensured that constraints with higher values do not dominate over others. The normalized 2. Generate first population in a random manner from constraint functions in the penalty function are developed as search space. shown in the following [5]. 3. Evaluate population by using objective functions and gj norm(x) = (bj ref - bj) / bj, j = 1, 2,n (3) arrange in order of merit. where bjref is the calculated value from the current generation/ 5. Start reproduction process (Selection, Crossover, evaluation whereas bj is the expert defined constant as shown in Table 1. The main purpose for defining the constraint bj is Mutation & elitism). to have the final design for practically be feasible and 6. Form new generation of offspring and treat as acceptable. In general, the constraints are decided upon with new population. To continue the optimization, return great care taking into consideration the availability of to step 3. materials, customers requirements and manufacturing To apply the GA approach, objective functions and standards. Table 1 is also referred to as the motor constraints have to be defined to evaluate how good each specifications and their constraint values. Constraint values motor design is. of variables can be expressed by following inequality [5] given below. 4. Test convergence. If satisfied then stop else continue.

IV. IMPLEMENTATION OF THE OPTIMAL DESIGN PROCEDURE

g1(x) = s - b1 0, g2(x) = Bsy b2 0, g3(x) = Bry b3 0, g4(x) = Bst - b4 0, g5(x) = Ff - b5 0,

The formulation of the objective function [5] is as per the following process:

In Eqn. 4, there are two different conditions of Where, F (x) is the objective function, motor material cost/ inequality constraint which are explained in following efficiency/torque, r is the penalty coefficient related to the sections.

F(x)  P[g j (x), r If F(x)  P[g j (x), r] ! 0 F(x) , If F(x)  P[g j (x), r] 0 (1) 0

g6(x) = Istart / I - b6 0 g7(x) = CosF - b7 0 g8(x) = Tp / T - b8 0 (4) g9(x) = Tst / T - b9 0 g10 (x) = h - b10 0

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

110

Table 1: Inequality Table For Motor Parameters Sr. no. Name Of Parameter 1 2 3 4 5 6 7 8 9 10 Slip, s Stator yoke flux density, Bsy Rotor yoke flux density, Bry Stator teeth flux density, Bst Stator slot filling factor, Ff Starting current to rated current ratio, Istart / I Power factor, cosF Pull-out torque to rated torque, Tp / T Starting torque to rated torque ratio Tst / T Efficiency (h) Inequality (Bj) (b1 = 0.05) (b2 = 0.6) (b3 = 0.6) (b4 = 1.7) (b5 = 0.90) (b6 = 7) (b7 = 0.8) (b8 = 1.5) (b9 = 0.6) (b10 = 0.85)

GA optimization algorithm was based on a roulette wheel selection, single point crossover, bit mutation and elitism.

4.1 FIRST CONDITION

It is not permitted that some constraints (as g j(x), j=7,10) are below the expert defined level. For example, high value of power factor is desired for good performance in induction motor. If the expert-defined constraint for power factor as shown in Table 1 was 0.8, then anything less than that would be a violation.

4.2 SECOND CONDITION

It is not permitted that some constraints (as g j(x), j=1,..6) are above the expert defined level. For example, stator yoke flux density may not exceed certain values on account of iron losses. If the expert-defined constraints for stator yoke flux density as shown in Table 1 was 0.6, then anything more than that would be a violation. The software developed for the design optimization of the induction motor was prepared using JAVA, which can analyze, optimize & evaluate design parameters and performance of motor. Parameters of the motor or materials used can be easily modified to investigate their effect on performance. Selection and optimization type (efficiency, torque, cost etc.) can be made depending on user. The

Fig. 2: Flow Chart For 3-phase Induction motor Design using nlp

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

111

The flow chart for design optimization process using NLP is given in Fig. 2 and the flow chart for design optimization process using GA is given in Figs. 3 & 4. Each block consists of a number of subroutines.

Min f (x) g (x) 0, j h (x) 0, j

i 1, 2, 3, ....., n i 1, 2, 3, ....., n

(5)

Where f(x) is the objective function of optimization, gi(x) and hi(x) are constraining functions, and x is the design variable set.

5.1 OBJECTIVE FUNCTIONS

In order to reduce the active (main) material cost and improve the efficiency of 3-phase induction motors, three different objective functions of optimization are defined separately as: [15]

f1 (x) CFe.pu Wfe  CCu.pu Wcu.s  CAl.pu WAl.r, f3 (x) T f 2 (x) K,

Fig. 3: Flow Chart For Optimization using GA/IGA

(6)

Where CFe.pu, CCu.pu, CAl.pu are the unit prices of magnetic material, copper wire and aluminum conductive bars respectively, W Fe, W Cu, and WAl are the masses of the corresponding materials, h is the motor efficiency and T is full load torque produced. = Pout-put / (Pout-put + Losses) (7) Where, Losses = Stator iron loss + Stator copper loss + Rotor copper loss + mechanical loss, & Pout-put = Pin-put Losses = 3 VL IL cos Losses (8)

5.2 DESIGN OPTIMIZATION VARIABLES

There is a lot of design variables involved in design of 3-phase induction motor and it is required to choose critically the suitable parameters as the optimization design variables as per the following carefully [15]:
X = [x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12] [L, D, lg , Z, Hs, Ws, lb, bb, ATm, Bav, Cu, Al]

(9)

Where L stator core length, D stator core inner diameter, l g length of air gap, Z number of stator conductors, Hs height of stator slot, Ws width of stator slot, lb height of rotor slot, bb width of rotor slot, ATm ampere conductors per meter, B av average flux density, dCu current density in stator winding conductor, dAl current density in rotor bars. Fig. 4: Flow Chart For 3-Phase Induction Motor Design USING GA/IGA

5.3 CONSTRAINING FUNCTIONS

The Genetic Algorithm usually adopts the sequential unconstrained minimization technique (SUMT) to solve the problems defined by Article [6.2]. Experiences and simulations have shown that if the penalty factor is selected too large, the The highly nonlinear constrained multivariable optimization CGA may cause premature convergence. On the other hand, if problem is very difficult to be solved by the conventional the penalty factor is selected too small, the CGA may yield a methods. The optimization problem for 3-phase induction highly computational burden. To overcome it, a new annealing motor design can be formulated as [15]: penalty function, which adopts self-adaptive annealing penalty

V. MODEL FOR 3-PHASE INDUCTION MOTOR DESIGN OPTIMIZATION

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

112

factors, is developed for Improved genetic algorithm (IGA) COST OF ALUMINUM: [15]: CAl.r = Le * Ab * S2 * PAl * CAl-pu f(x, s) = f(x) + p(x, s) (10) where, p(x, s) = s [Si |min(0, gi(x)| + S(x)|]

(15)

COST OF INSULATION: For a given voltage, Cost of Insulation material is where s is the penalty factor (s = 1 / T, Ti+1 = Tia, T is the proportional to Temperature rise of Machine and hence is value of the parameter under control, a is the proportionality proportional to total losses of the machine, and will be constant within the range of (0-1), p(x, s) the penalty function, considered in efficiency maximization. and f(x, s) the generalized objective function defined by the Total cost of the machine can be given by: IGA. To reflect the changing feature of the nature and imitate the parameter under control becoming more rigorous with the development of the generations, the fitness value of a string is defined as fit = fm - f(x, s) (11) CT = Cfe + Ccu.s + CAl.r + CInsulation + Cp (16)

6.2 EFFICIENCY AND TORQUE PRODUCED AT FULL LOAD

The function for minimizing the losses can be used as function for maximizing the efficiency. (17)

Where, fm is a maximum objective function in a generation. We know that, To avoid prematurity, another technique known as the multiEfficiency(h) a [Pout-put/ (Pout-put + Losses)] turn evolution strategy was used.

For maximum efficiency of the motor at full load, iron losses The IGA is applied for design optimization of the 3-phase can be taken as equal to P And friction and windage cu.stator. induction motor. In the process of optimization, the objective loss can be taken as 10% of full load rotor copper loss (P ). cu.rotor as well as constraining functions in Article (5,6.3) are So, now total losses can be given by: implemented as the following [15]: g2(x) = D - D 0 g1(x) = L - L 0, g (x) = l - l 0, g4(x) = Z - Z 0 g g 3 g6(x) = Ws Ws 0 g5(x) = Hs - Hs 0, g (x) = l l 0, g8(x) = bb bb 0 b b 7 g9(x)=ATmATm 0, g10(x)=BavBav 0 g (x)=d d 0, g12(x)=dAl dAl 0 11 Cu Cu Symbols have their meanings as given in Eqn 9. Total Losses = 2 Pcu.stator + 1.1 Pcu.rotor (19) Total Losses = 2 * 3 Iph2 Rswdg + 1.1 * Ib2 (Rb S2 + Rring) (20)

5.4 A PPLICATION OF IGA TO 3-PHASE I.M. DESIGN OPTIMIZATION

And Losses in the motor can be given by: Total Losses = Wi + Wcu + Friction and windage loss (18)

6.3 BREAKDOWN (PULL-OUT) TORQUE

(12) TBD a {1 / rotor reactance(Xr)} (21)

6.4 STARTING TORQUE

We can obtain better starting torque with R / X ratio near to 1 at standstill, i.e.

VI. OBJECTIVE FORMATIONS

The major performance parameters, which majorly affect Tst a Rr; If Rr / Xr = 1; Tst = TBD the machine performance, can be collected as: (1) Cost of machine, (2) Efficiency & Torque (FL) produced, (3) Break Better starting torque can also be obtained by deep bar or down torque, (4) Starting torque, (5) Power factor of the double cage rotor, as there is a limitation on rotor resistance machine, (6) Full load slip of machine, (7) Starting to rated due to increased losses with greater rotor resistance. current ratio, (8) Pullout to rated torque ratio, (9) Magnetic vibrations & noise.

6.5 POWER FACTOR

6.1 COST OF MACHINE

It involves following parameters: COST OF IRON: Cfe = L * OD2 * ks * Pfe * CFe-pu COST OF COPPER: Ccu.s = 3 * Lswdg * As * Pcu.d * CCu-pu (14) (13)

We know very well that power factor of the motor is better with smaller air gap length (lg ). Pf a (1 / lg) (22)

6.6 FULL LOAD SLIP

We know that slip, s = Pcu / Pg Or, s a Copper Losses (23) (24)

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

113

6.7 STARTING TO RATED CURRENT RATIO

Starting Current is low with high rotor resistance, so it can be understood that if R/X ratio is near to one for better starting torque, the starting current will be low and ratio of starting current to rated current will be better for such a motor, but at the cost of efficiency. 1 2 3 4 5

Table 3: Methods Used In Ga / Iga Sr.no. Parameter Population Size Cross over rate Mutation rate Elite count Stopping criteria Value / Criteria 400 0.8 0.01 2 IPerformance IndexI 0.15 && [1/{I Performance Index_1 Performance Index_0 I}] = null 10,00,000 (Max) Single point crossover with fixed cross over rate (0.8) Bit string mutation with fixed mutation rate (0.01) Fitness

6.8 PULLOUT TO RATED TORQUE RATIO

Pull out to rated torque ratio = (TPull-out / TRated) (1 / Xr * efficiency) (25)

6 7 8 9

6.9 MAGNETIC VIBRATIONS AND NOISE

We know that noise caused due to magnetic vibrations is a big problem, which is due to the zig-zag leakage reactance, which is inversely proportional to the air-gap length of the motor, i.e. Noise a Zig-Zag leakage reactance a [1/air-gap length(lg)] (26)

Number of generations Cross over method Mutation method Selection coefficient Fitness method

But on the contrary larger air gap give rise to the magnetizing 10 Genetic load reactance which causes lower power factor and hence inferior performance of the motor. So, one has to make balance These weights are decided for a motor running for an between the noise and power factor of the motor. Atta-Chakki under following conditions: 1. Continuous running for more than 20 hrs per day. VII. PERFORMANCE INDEX 2. The motor should be economic but due to its continuous running energy efficiency is more important. Now we can write performance index as: 3. The motor is to be used in a silent environment (say J = k1 CT + k2 (1 (1/h)) residential area). So, we have to consider weight for noise function also, but it is very less due to the fact + k3 [(Rr/Xr)1] that it conflicts with the power factor and hence the + k4 [1 (1 / lg)] efficiency of the motor. + k5 [1 (1 / Xr * efficiency)] + k6 (1 / lg ) (27) VIII. RESULTS Where, k1, k2....k6 are weight constants given to different functions, given in Table 2. Table 2: Weights For Performance Index Sr.no. Name of Function 1 2 3 4 5 6 Cost of the motor Total losses Starting torque PF (Length of air gap, lg) Pull out to rated torque ratio Magnetic Vibration and Noise Function Total Weight of Performance Index Weight Given to The Function 0.1 0.5 0.1 0.15 0.1 0.05 1 Results are summarized and analyzed in Tables 4, and Figs. 5, 6 & 7 given at the end.

IX. CONCLUSION
Result analysis of design, optimization and validation (by MATLAB/SIMULINK MODEL) gives the following conclusion: 1. The output of the GA program and its validation using SIMULINK / MATLAB shows that: The cost of motor decreases with increase in efficiency. Starting current of motor increases considerably but with better starting torque, which is required in the assumed case, as on load starting is required. Breakdown torque is also improved considerably and hence overload capacity also. The optimized motor design using GA/IGA reaches full speed faster having lesser transient time, lower slip & higher efficiency at full load and lower cost.

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

114

IX. TABLES
Table 4: Comparision of Performance By Output Of Simulink / Matlab Sr.no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Parameter / Machine No. Stator Voltage(Volt) Stator starting current(Amp) (DOL) Stator starting current(Amp) (Y-D) Stator steady state current(Amp) Rotor starting current(Amp) Rotor steady state current(Amp) Starting Torque(N-m) (DOL) Starting Torque(N-m) (Y-D) Break Down Torque(N-m) Load Torque(N-m) Time to reach full speed(Sec) Rotor Speed(rpm) Slip(%) Cost Function (a Cost in Rupees) 399.94 89.87 29.95 9.61 80.47 9.44 188.15 62.72 164.47 54.10 0.70 1472.00 1.86 Manual Amount 399.94 118.23 39.41 9.39 117.99 9.23 327.70 109.23 242.50 56.90 0.24 1475.00 1.67 1187.278 Nlp % Improv. 31.54 (incr.) 31.54 (incr.) 2.28 (decr.) 46.63 (incr.) 2.24 (decr.) 74.17 (incr.) 74.17 (incr.) 47.4 (incr.) 5.17 (incr.) 65.71 (decr.) 0.20 (incr.) 10.2 (decr.) Amount 399.94 148.99 49.66 9.33 148.14 9.19 474.36 158.12 355.50 57.20 0.15 1476.00 1.60 1110.611 Ga % Improv. 26.01 (incr.) 26.01 (incr.) 0.60 (decr.) 25.54 (incr.) 0.38 (decr.) 44.75 (incr.) 44.75 (incr.) 46.59 (incr.) 0.52 (incr.) 60 (decr.) 0.067 (incr.) 4.19 (decr.) 6.457 (decr.)

X. FIGURES

Fig. 5: Manual Design

Fig. 6: NLP Design

Fig. 7: GA/IGA Design

Charateristics Of 7.5 Kw Induction Motor (Transient Response Showing Variation Of Stator Voltage, Stator Current, Rotor Current, Torque, Speed, Speed Torque Curve Is Also Shown

REFERENCES
[1] [2] [3]

[4]

A. K. Sawhney (1998), A Course in Electrical Machine Design, Danpat Rai & Co. [5] K. Prasad (2006), Computer Aided Electrical Machine design, Satya Prakashan. S. J. Chapman (2002), Electric Machinery and Power System Fundamentals, McGraw-Hill, New York.

C. Singh and D. Sarkar (July 1992), Practical Considerations in the Optimization of Induction Motor design, IEEE Proceedings, vol. 139, no. 4. Cunkas, M., Akkaya, R. (2006), Design Optimization of Three-Phase Induction Motor By Genetic Algorithm And Comparison with Existing Motor, Mathematical and Computational Applications, Vol. 11, No. 3, pp. 193-203, Association for Scientific Research.

MIT International Journal of Electrical and Instrumentation Engineering Vol. 1, No. 2, Aug 2011, pp 108-115 ISSN 2230-7656 MIT Publications

115

[6]

[7]

[8]

[17] Mimi Belatel, Hocine Benalla, A Multiobjective Design Optimization of Induction Machine using CAD and ANNs, [11] F. Paracility, M. Villani (2004 June 16th 18th) Three phase Department of electrotechnic, university Mentouri, Ain Elinduction motor efficiency improvement with Die-cast copper Bey, Constantine, Algeria. rotor cage and premium steel, submitted for the SPEEDAM, [18] N. Bianchi, S. Bolognani (1998), Design optimization of Capri (Italy). electric motors by genetic algorithm, IEEE Proc. Electr. Power Appl., 145: 475-483. [12] F. Wurtz, M.Richomme, J. Bigeon, J.C. Sabonnadiere (Mar, 1997), A few results for using genetic algorithm in the design Springer.

[15] Lihan, Hui LI, Jingcan LI, Jianguo ZHU (26-29 September F. Paracility (2003), M. Villani; Design of high efficiency 2004), Optimization for induction motor design by improved industrial induction motors by innovative technologies and genetic-algorithm, Australian Universities Power new materials, Proj. of Department of Electrical Engg, Engineering Conference (AUPEC 2004), Brisbane, Australia. University of LAquila, supported by Indian ministry of [16] M. Poloujadoff, E. Christaki, C. Bergniann (Dec. 1995), Education, University and Research, December. Univariant Search: An opportunity to identify and solve conflict problems in optimization. IEEE Trans. On Energy [10] F. Parasiliti, and M. Villani (1999), Energy Efficiency Conversion, Vol. 9, no. 4. Improvements in Electric Motors and Drives, New-York: [9]

of electrical machines, IEEE Transactions on Mag., VOL. D.E. Goldberg (1989), Genetic algorithm in search, 33, No. 2, pp. 1982-1985. optimization and machine learning, Massachusetts: Addison[13] G. Liuzzi, S. Lucidi, F. Parasiliti, and M. Villani (May 2003), Wesley Publishing Company, Inc. Multiobjective Optimization Techniques for the Design of Darrell Whitley, A Genetic Algorithm Tutorial, Computer Induction Motors, IEEE Transactions on Magnetics, VOL. Science Department, Colorado State University Fort Collins, 39, No. 3. CO 80523 whitley@cs.colostate.edu. [14] J.Le Besnerais, A.Fasquelle (01-05 June 2008), V.Lanfranchi, F. Paracility (2003), M. Villani; Design of high efficiency M.Hacquet, and P.Brochet, Mixed-variable optimal design of induction motors including efficiency, noise and thermal induction motors with die-casting copper rotors, Energ y criteria, EngOpt 2008 International Conference on efficiency in Motor driven systems, Editors: F. Paracility, P. Engineering Optimization, Rio de Janeiro, Brazil. Bartoldi (book), Springer, ISBN 3-540-00666-4, pp.144-151.