Você está na página 1de 14

Lumped parameter thermal model for electrical machines of TEFC design

P.H. Mellor
D. Roberts D.R. Turner

Indexing terms: TEFC designs, Thermal models, Electrical machines

Abstract: A lumped-parameter thermal model is described which provides both a steady-state and transient solution to the temperatures within an electrical machine of the TEFC design. The model is sufficiently complex to identify the temperatures at most locations in the machine, including the peak temperatures in the endwinding and the surface temperatures of the rotor. It is formulated out of purely dimensional information and constant thermal coefficients and is therefore easily adapted to a range of frame sizes. The thermal behaviour of the TEFC machine is accurately described by the solution of just eight linear differential equations. The model is therefore suitable for application to online temperature estimation for protection and duty-cycle evaluation. The application of the thermal model to a medium (75 kW) and two small (5.5 kW) induction motors is described in detail. The model performance is confirmed by experimental temperature data obtained from varying load tests on each of the three induction motors.
1
Introduction

The highest volume of AC machines manufactured are of the enclosed TEFC design. With increasing material costs there has been a tendency to reduce the frame size of these machines for a given output at the expense of both poorer efficiency and a lower tolerance to thermal overload. In addition, machines are now being designed with materials that are highly temperature sensitive, for example the neodymium-iron-boron magnets used in high-field synchronous machines. Because of these and other constraints, the margin for error in the thermal design of TEFC machines is often small, and there is an increased need for accurate analytical thermal models. The thermal models of low- to medium-rated TEFC machines need not be excessively complex because of the basic nature of their construction and cooling. Thermal failure in these machines is most likely to occur in either
Paper 8118B (Pl), first received 9th October 1990 and in revised form 29th January 1991 Dr. Roberts and Dr. Turner are, and Dr. Mellor was formerly, with the Department of Electrical Engineering and Electronics, University of Liverpool, Brownlow Hill, PO Box 147, Liverpool L69 3BX, United Kingdom Dr. Mellor is now with the Department of Electronic and Electrical Engineering, University of Sheffield, PO Box 600,Mappin Street, Sheffield S1 4DU, United Kingdom

the stator or rotor windings. Thus, to ensure a satisfactory lifetime, the peak slot or endwinding temperatures must remain below the designated insulation class limit, whereas the rotor temperatures must be limited to prevent mechanical distortion or fatigue. A thermal model should therefore be sufficiently detailed to distinguish between these stator and rotor components. Ideally the model should provide a transient solution so that the effects of rapidly varying loads or short periods of overload can be assessed. The lumped-parameter thermal model described in the paper was found to meet these requirements. The motor is divided geometrically into a number of lumped components, each component having a bulk thermal storage and heat generation and interconnections to neighbouring components through a linear mesh of thermal impedances. The lumped parameters are derived from entirely dimensional information, the thermal properties of the materials used in the design, and constant heat transfer coefficients. As a result the model is easily adapted to a range of frame sues. Although the calculation of the parameter values can be complex, the resulting set of thermal equations, which fully describe the machine, are relatively easy to compute. The model is therefore suitable for online temperature monitoring for many applications including motor protection [l, 23 and duty-cycle evaluation [3]. The lumped parameter model has been applied specifically to three separate TEFC induction motors. The nameplate details of these machines are given in Table 1. The first two have ratings of 5.5 kW, but are from different manufacturers, and the third is a medium-rated 75 kW machine. The three motors are of similar construction, the only major difference being the presence of rotor cooling holes in the first 5.5 kW motor (A in Table 1). Each has an aluminium frame, shrink fitted onto the core pack during manufacture, and an extruded aluminium rotor winding. The stators were all mesh wound with a Class F insulation system.
Table 1 : Induction motor nameplate details Motor Framesize Power, kW Voltage, V Current, A Poles
A B

132s 132s 250M 5.5 5.5 75 415 415 415 10.9 11.2 133 4 4 4

Lumped-parameterthermal model

The lumped-parameter thermal model is sufficiently detailed to include all the major components and heat
205

I E E PROCEEDINGS-B, Vol. 138, No. 5, SEPTEMBER 1991

transfer mechanisms within the machine without being over complex. Previous work on these models by many authors has demonstrated how heat transfer between bulk components can be successfully determined in the steady state. In particular Perez and Kassakian [ ] mod4 elled each component in terms of a thermal node which approximated to the mean temperature within the component. Any heat generation due to the losses in the component was introduced as a point source at the node. The paper extends the work to the full transient case by including thermal storage as an additional heat transfer at the node, related to the mean temperature and the components total thermal capacity. The geometry of a TEFC induction motor can be subdivided into the 10 components shown in Fig. 1, where

(iv) The thermal capacity and heat generation are uniformly distributed. On using these assumptions, the solution of the heat conduction equations in each of the axial and radial directions produces the two separate three-terminal networks shown in Fig. 2B. In each network, two of the terminals

Fig. 2A

General cylindrical component

L
Induction motor construction 6 Endwinding 7 Endcap air 3 Stator teeth 8 Rotor winding 4 Stator winding 9 Rotor iron 5 Air gap 10 Shaft
1 Frame 2 Stator iron

Fig. 1

symmetry is assumed about the shaft and a radial plane through the centre of the machine. The influence of the asymmetrical temperature distribution that exists in most machines, for example as a result of the external fan being mounted at one end, is therefore taken to be small. The solid components of the frame (I), stator (2, 3), windings (4, 6 ) and rotor (8-10) are all modelled as a network of thermal resistances based on a general cylindrical lumped component. Two further components of negligible thermal capacity represent the air gap (5) and the endcap air (7). The 10 components are either interconnected directly or, where appropriate, with additional thermal resistances to account for the convective heat transfer across the cooling air paths within the machine. An extra thermal resistance is also included to allow for the poor thermal contact between the rough stator laminations and the frame. Any heat transfer due to radiation from the internal surfaces is neglected.
2.1 General cylindrical component

Fig. 2B

Independent axial and radial thermal networks

represent the appropriate surface temperatures of the component, whereas the third represents the mean temperature 0, of the component at which any internal heat generation U or thermal storage C is introduced. The central node of each network would give the mean temperature of the component if there were no internal heat generation or storage. The superposition of internal heat generation results in a mean temperature that is lower than the temperature given by the central node, which is reflected in the networks by the negative values of the interconnecting resistances R,, , R,, . The values of the thermal resistances in each network come directly from the independent solutions of the heat conduction equation in the axial and radial directions [3]. These are given in terms of the dimensions of the cylinder and the axial and radial thermal conductivites k, , k, by the following expressions

The solid components of the induction motor are based on the general cylindrical component shown in Fig. 2A. To obtain simple, but physically significant, expressions for the network of thermal resistances that describe the heat conduction across the general component, the following assumptions are made (1) The heat flow in the radial and axial directions are independent. (ii) A single mean temperature defines the heat flow both in the radial and axial directions. (iii) There is no circumferential heat flow.
206

R,a

-L 67rka(rf - r:)

(3)

(4)
IEE PROCEEDINGS-B, Vol. 138, N o . 5, SEPTEMBER 1991

cooling air is modelled in the usual manner by a single thermal resistance R , , which has a value equal to the reciprocal of the surface area A, in contact with the cooling air times a boundary film coefficient h R
=-

1 hA,

The total thermal capacitance of the cylinder is found from the material density p, the specific heat cp and the motor dimensions as

C = pcpa(rf - r i ) L (7) The two one-dimensional networks are combined simply by connecting the points of mean temperature together. If it is assumed that the temperatures in the cylinder are symmetrical about a central radial plane so that the face temperatures 8, and 8, are equal, it is only necessary to model half the cylinder with correspondingly only half the heat generation and thermal capacitance. The thermal network is then reduced to the form shown in Fig. 2C, which consists of two internal nodes and four thermal resistances, R , , R , , R , and R , , given by L R , = R , , + 2R,, = 6ak,jr: - r:)

The four sets of film coefficients used to describe the convective heat transfer from the various surfaces within the induction motor are listed below. At each surface, two values of film coefficient are required, one for when the machine is stationary, when the external and internal fans are ineffective, and the other for when the machine is rotating. These two cases are denoted by the subscripts s and r (i) hls, h,, = heat transfer between frame and external air (ii) h,,, h,, = heat transfer between stator or rotor and air gap (iii) h,,, h,, = heat transfer between stator iron, rotor, endwindings or endcaps and endcap air (iv) h,,, h,, = heat transfer between rotor cooling holes and circulating endcap air. Only the coefficients h,, and h,, are to be found directly from test. The motor is run at constant load until thermal equilibrium is reached; hl, is then determined from the surface-ambient temperature gradient and the total machine loss. Similarly h,, is found from a lowvoltage locked-rotor test, where, under thermal equilibrium, the heat dissipated from the motor surface is equal to the total electrical power input. About the air gap, the stator and rotor act as two concentric cylinders rotating relative to each other, as a result, work [ S , 61 concerned with such geometries is used to determine the air gap film coefficients h,, and h,, . In a small to medium induction motor, any heat emitted from the rotor surface would be transferred directly across the air gap to the stator. There would be little or no axial heat flow from the air gap into the adjoining endcap air, and hence this effect could be neglected. The air gap film coefficients may be defined in terms of a dimensionless Nusselt number N N u ,the air gap width I , and the thermal conductivity of air kair$0 that

R , = 2R3,

This general network makes it possible to allow for different thermal conductivities in the axial and radial directions, which enables, for example, the thermal effects of

h,

=7
bg

kair

RC

92

Fig. 2C

Combined thermal network for symmetric component

The value of the Nusselt number for the convective heat transfer between two rotating smooth cylinders is given by Taylor [SI. In an actual machine, however, there will be a greater heat transfer across the air gap than is described by the smooth cylinder equations, because of the additional fluid disturbances caused by the winding slots. Experimental results obtained by Gazley [6] suggest that the slot effects will cause approximately a 10% increase in the heat transfer. The Nusselt numbers for small air gap machines are thus obtained from the modified expressions of Taylor NN,
= 2.2

the stator and rotor laminations to be considered. The general cylinder can be used to model a solid rod, such as the shaft, by evaluating the above expressions as the radius r , tends to zero and removing the node which corresponds to the central temperature e,.

NT,

< 41
41 < N To d 100

(14) (15)

N N , = 0.23N$:3Ng;27

2 2 Convective heat transfer


The convective heat transfer between the exposed surfaces of the solid components and the internal or external
IEE PROCEEDINGS-B, Vol. 138, NO. 5, SEPTEMBER 1991

The dimensionless Taylor N T , and Prandtl N,, numbers are defined from the air gap dimensions and fluid constants using the expressions given by Taylor [SI. The fluid constant values, which are temperature dependent, are taken at the expected full load air gap temperature. The critical value of 41 from the Taylor number refers to
207

a change from laminar flow, which is normal in small air gapped machines, to turbulent flow. When the machine is stationary the heat transfer across the air gap will be by conduction only, which corresponds to a Nusselt number of 2.0. In the endcap region, a single film coefficient is used to model the heat transfer to and from all the surfaces in contact with the circulating endcap air. Clearly this is a simplification, as separate film coefficients would apply to each of the different surfaces in the path of the cooling air. However, the complex analysis that is required to evaluate a film coefficient for each case cannot be justified, especially when considering the accuracy to which they could be obtained. The values of the endcap film coefficients h,, , h,, are found from the experimental work of Luke [7], on the dissipation of heat from endwindings by forced ventilation. The results of this study can be approximately linearised for small cooling air velocities U to (16) The constant term in this expression represents the heat transfer by natural convection alone and is therefore the value of the stationary film coefficient h,,, The cooling air velocity c' may be estimated from the product of the rotor angular velocity w , , the log-mean radius rm of the endring fins and the efficiency 9 with which the find fan the internal air, using
c' =

h,, = 15.5(0.29~ 1) W" + ,'

7.5 mjs

(17) Because of the lack of available information on the radial air velocity, a rather arbitrary value of 50% was assumed for the fan efficiency. The final film coefficients of interest are those for any cooling holes in the rotor. The stationary value h,, is assumed to be equal to that of the endcap air. The rotational coefficient h,, is obtained from a theoretical analysis presented by Mori and Nakayama [SI, which is concerned with the heat transfer of a pipe rotating about a parallel axis. In this work the Nusselt number for the pipe heat transfer is expressed in terms of the Reynolds number N R , and Kayleigh number N , , of the centrifugal field of air within the pipe as
U =

rmwrq

2.3 Stator- frame thermal contact resistance The stator core to frame interface thermal resistance requires careful consideration because of its position on the main heat flow path of the stator losses to the ambient air. It cannot be assumed that a thermal short circuit exists between the frame and the stator back iron because of the rough surface of the stator laminations at the interface. The thermal resistance of the interface is defined in a similar manner to eqn. 12 in terms of a single thermal contact coefficient h, and the contact surface area. The value of this coefficient is a function of the pressure which is created when the frame is shrunk onto the core during manufacture and the roughness of the stator surface. The contact pressure may be estimated from the original dimensions of the frame and stator bore prior to the shrink fit and the physical properties of the core and frame materials. A more detailed explanation of this calculation is given by Roberts [3]. At low contact pressures, the thermal contact coefficient can be found directly from the experimental curves of Brunot and Buckland [SI, which show the effect of pressure on the thermal contact of aluminium shims sandwiched between two laminated magnetic steel blocks, These curves, however, only extend to pressures of 1.4 x lo6 Pa (200 lb/ in') which are far below those normally expected in practice. The thermal contact coefficient at higher pressures may be extrapolated by applying a semi-empirical equation presented by Shlykov and Ganin [lo] to the experimental points of Brunot. The resultant curve for pressures up to 15 x lo6 Pa (2200 lb/in2) is presented in Fig. 3, where the contact coefficient values are plotted on a reciprocal scale. The Figure is thus representative of the variation of contact thermal resistance with contact pressure.
x103

(1 8) The definitions of the dimensionless Reynolds and Rayleigh numbers given in Reference 8 require, in addition to relevant dimensions and air fluid contants, estimations of the airflow velocity and temperature gradient along the cooling hole. In the calculation of h,, for the 5.5 kW induction motor designed with rotor cooling holes, low values were assumed for the temperature gradient and the airflow velocity. The internal film coefficients for the three induction motors modelled are listed in Table 2.

N,,

0.83(NR, N,,)'.'

5ot
c o n t a c t pressure, N / m 2

Fig. 3

Frame-core contact coefficient against contact pressure

Table 2 : Internal film coefficients


Film coefficients W / m 2K
Induction motor

x Bruno1 [U] experimental points 0 Shlykov [IO] equation Values used in thermal models. 0 5.5 k W motor A 0 5.5 kW motor B 0 75 k W motor C

A, 5 5 kW

B, 5 5 kW C, 75 kW
1993 2192 15 5 38 6 00 00 55 3 1028 155 65 1 00 00

61 8 67 9 15 5 38 9 15 5 18 8 208

A substantial variation in the frame contact resistance existed between the three machines tested, and this threw doubt on the validity of the use of the extrapolated curve. In the thermal model of the 75 kW machine, the contact coefficient was obtained directly from the curve and, through lack of detailed data from test, its validity could only be justified by the final accuracy of the model. In the
I t E P R O C E E D l h G S - B . V o l 138, N o 5 , S E P T t M B E R I991

first 5.5 kW machine (A) the core had been previously removed for test purposes and then re-inserted into the frame with a loose interference fit. As a result the contact pressure could not be substantiated and a measured value was used for the contact resistance, however, this value would seem to agree with the curve, given the likely low value of contact pressure. Finally, detailed steadystate tests were performed with the second 5.5 kW machine to find the true contact resistance; this value, however, was approximately half the value suggested by the curve. The thermal contact coefficients used in the formulation of the thermal models of the three machines are plotted for comparison in Fig. 3 at the respective estimates of contact pressure. In view of its position in the main heat flow path, the stator-frame contact would be expected to have a significant influence on the final outcome of the model. However, a sensitivity analysis shows that even a large variation in the value of the thermal contact resistance has surprisingly little effect on the estimates of winding temperatures obtained from the model. Details of this analysis are given in Section 3.5, which also shows that the model is similarly insensitive to the values of the convective heat transfer coefficients.

centre. Allowance is made for the laminations in a similar manner to that made for the stator back iron.

2.4.4Slot winding: The sections of the winding that lie


in the slots are modelled as solid cylindrical rods, made up of an array of conductors and insulation. In determining the axial and radial conductivities, it is assumed that only the copper conductors transfer heat axially along the slot, but radially the winding acts as a homogeneous solid which has a conductivity that is 2.5 times that of the insulation alone. The value of 2.5 is taken from Reference 11 and was derived from measurements on actual impregnated windings and will include to a certain extent the effects of voids in the insulation. The radius of the rod is chosen to give a cross-section equivalent to that of the available area in the slot assuming a 100% slot fill. The slot liner is modelled separately as a polyamide strip of thickness consistent with the class of insulation used in the machine.
ambient

2.4 Details of model components


Appendix 7 gives details of the geometries and the corresponding formulas for each of the ten networks of thermal resistances and capacitances which are used to model the induction motor. These networks combine to form the total mesh for the entire machine given in Fig. 4. In the Figure, the primed nodes 2, 3, 8 and 9 refer to the central nodes of the four components, which were based on the general cylindrical network given in Fig. 2C, and an additional junction 4 between the stator teeth, stator winding and stator back iron. A brief description is given below of the assumptions and simplifications made in applying the general conduction and convection models to the networks of each component, where, in most cases, because of symmetry, only half the machine is considered.

2.4.1 Frame: The frame component includes the entire


ribbed cooling structure and the endcaps. The whole structure is assumed to be at a uniform temperature and to dissipate heat externally through a single frame to ambient convective thermal resistance. It receives heat from the stator across the frame-core contact resistance and by convection from the endcap air.
I

2.4.2 Stator back iron: The stator back iron component


consists of the stator lamination pack minus the teeth which are modelled separately. The general cylinder form is modified to account for anisotropy due to the laminations by introducing a stacking factor in the radial direction and using a value lower than that of mild steel for the axial conductivity obtained from data published in Reference 11.

Fig. 4

Total induction motor thermal network


ambient

U = nodal heat generation c = thermal capacitance to

2.4.3 Stator teeth: An expanded version of the general


cylindrical component is applied to the stator teeth by modelling them as a collection of cylindrical segments connected thermally in parallel. The equivalent arc $e of the tooth segment is calculated to give a segment crosssectional area equal to that of the actual tooth profile. The circumferential heat flow from the slot windings is modelled by an additional resistance between the slot faces to the point of mean temperature at the tooth
IEE PROCEEDINGS-B, Vol. 138, N o . 5. SEPTEMBER 1991

2.4.5 Endwinding: The endwinding is modelled as a


homogeneous toroid structure, representing the circumferential mesh of conductors and insulation, and legs which are short cylindrical extensions of the stator slot windings. In a similar manner to the slot winding, the radial heat flow is considered to be that of homogeneous rods. An axial transfer of heat is assumed to occur from the mean temperature point in the toroid to the stator
209

slot winding along the copper conductors of the legs. Where the external toroid radius is not given explicitly by the design drawings, it is estimated as the mean radius of the stator slots. The equivalent radius of the toroid crosssection is found from the winding mass that remains after the subtraction of the copper in the slot section and legs. The endwinding model is weighted to estimate the peak hot-spot temperature rather than the mean as this is important in the evaluation of insulation aging and failure. A theoretical solution of the heat conduction equation for the radial temperature distribution in a cylinder gives a ratio of peak to mean temperatures of 1.5 : 1. This factor of 1.5 is used to weight the thermal resistances of the endwinding model to obtain the peak winding temperature. The thermal capacitance is reduced by the same amount to preserve the correct level of stored energy.

where Ci is node thermal capacitance, Oi is node temperature, Rji is thermal resistance between adjoining nodes i, j , and ui is heat generation at node i. Seven of the nodes in the network have no thermal capacitance and heat generation associated with them. On eliminating these mathematically, the system can be reduced to just eight linear differential equations, expressed in matrix form as

2.4.6Air gap: The air gap forms a connection between


the stator teeth, the part of the stator winding exposed in the slot openings and the rotor surface. The corresponding thermal resistances are found from the contact areas of these solids and the air gap film coefficient.

where [C] is a column matrix of thermal capacities, [F] is a square matrix of internodal conductances, and [U] is a column matrix of thermal generators. The above equation applies to the induction motor when it is rotating. At standstill, a separate conductance matrix [YJ applies because of the change in value of the convective elements of the branch thermal impedances. With no supply and hence zero heat generation, the standstill equation becomes

2.4.7Endcap air: The circulating air in the endcap is


assumed to have a uniform temperature. A single film coefficient is used to describe the convective heat transfer from all surfaces apart from the rotor cooling holes, where a separate value is used. The contact area offered by the endwinding toroidal model is increased by 50% to allow for surface irregularities and the greater area of the flatter structure of a true endwinding.

The above differential equations are best solved for varying load and loss using a quasistatic approach where it is assumed that the loss remains constant over short discrete time intervals. Discrete solutions are formulated over the kth time interval from the original equations by using eigenvector techniques [l] to yield equations in the iterative form

Cek+11

2.4.8 Rotor winding: The rotor winding is modelled as a


continuous aluminium cylinder surrounding the iron core, with a volume equal to that of the cage bars. Joined to the end of the winding is a disc of equivalent volume to the total end-disc including the fan. This approach is used for simplicity and because a good thermal contact exists between the extruded aluminium bars and the rotor laminations.

CAr1[Iek1 + c'kl Lek + 1 1 = CA,] Lek]


=

(22) (23)

Both the above solutions are referred to the ambient temperature, which must be added as an offset to obtain the absolute motor temperatures. Allowance for the variation of winding resistances with temperature is included in the running equation by recalculating the generation factors at each time step, so that

2.4.9Rotor iron: Account is taken of the laminations in


the rotor iron, as was done for the stator iron. If axial cooling holes are present, the thermal resistances of the general cylindrical form are increased in proportion to the reduction in cross-sectional area.
2.4.10 Shaft: The shaft is modelled as a cylindrical rod with no internal heat generation. The axial heat conduction is modelled as three sections, one that lies under the rotor iron, a second that lies under the bearing, and a third that acts as a thermal connection between the mean temperatures of the first two. The mean temperature of the entire shaft is taken to be the temperature at the centre of the third section. A good thermal contact is assumed to exist between the shaft and the frame across the bearings, any shaft external to the bearings is therefore considered to act as part of the frame.

Cek+ll

cArI[ekl + (['I + CalCBkl)[ukl

(24)

Here the sparse matrix [a] contains the temperature coefficients of resistance for the conductors present in the three winding components.

2.6 Distribution of motor loss


The heat generators that form the input to the thermal equations are derived from the motor electrical and internal windage loss. The core (iron) loss is distributed among the stator and rotor components in such a way that the majority of the loss occurs in the components that model the stator teeth and the combined rotor bars and teeth; nodes 3 and 8. The energy used to drive the internal fan on the rotor is ultimately dissipated as heat and appears in the model as if it were rotor core loss. Bearing losses are ignored but could, if necessary, be included as a heat generation in the shaft. The 12R (copper) loss is divided between the slot and the rotor endwindings, nodes 4 and 6, in proportion to the quantity of conductor in each component. In certain applications of the thermal model, such as duty cycle evaluation or online temperature prediction, it may be convenient to estimate the loss directly from the motor terminal voltage and current. This can be achieved through the use of the electrical equivalent circuit. A previous analysis [l] of the equivalent circuit of Say [12]
1EE PROCEEDINGS-B, Vol. 138, N O . 5 , SEPTEMBER 1991

2.5 Solution of thermal equations


The complete thermal network of the induction motor given in Fig. 4 has a total of 15 nodes. The individual node temperatures are defined by a set of 15 heat balance equations of the form

210

shown in Fig. 5 gives the following expressions, assuming a balanced three-phase supply, for the core loss @ ,stator Z2R loss W, and rotor Z2R loss W , per phase in terms of the motor phase voltage V, and current I ,

w.=
'

--

[cR,

cX,(X,+2x,)
-

-[I

x , + 2x,

cxm

+,I;

The results of the varying load tests are presented in Figs. 7-10, where it can be seen there is a good correlation between the simulated and measured temperatures. The results for the endwinding are often identical, which indicates that the endwinding simulations are in fact a little low. This is because the endwinding temperatures in the model are weighted to give the peak hot-spot temperature, whereas it is unlikely that the corresponding test thermometers were located exactly at the hottest point.
3.1 Location of thermometers The platinum resistance thermometers were located in each of three induction motors at the positions shown in Fig. 6, which corresponded as far as possible to the points of mean temperatures of the accessible model components. In most cases, the temperature of an individual component was taken from an average of two or more thermometers positioned at both the drive and the nondrive ends of the machine.
1

w,=

-R2

X,(X,

+ 2x,) v; +

c2X, R, x , 2x, 1;

The parameters in the above equations are in the usual notation, and so RI, R,, x,, X, and R, refer to the

r-3pIR2,s
+ cx2

Fig. 5 Induction motor electrical equivalent circuit zI = R, jxl


z2 = R , z, = R ,
x, = XI

+ + jx2 +jX,

drive end (DE) Fig. 6 Location of test thermometers 1 Ambient 2 Frame 3, 4 Stator iron 5, 6 Stator teeth 7, 8 Slot winding 9, 10 Endwinding

non-drive end
(NDE)

I I

stator and rotor resistances, the total leakage reactance, the magnetisation reactance and the core loss resistance, respectively, all referred to the stator. The constant c, which has a value slightly greater than unity, compensates for the error introduced by moving the magnetisation branch forward of the stator impedance. Similarly, the negative current factor in the core loss expression describes the reduction in excitation caused by the voltage drop across the stator resistance.
3
Performance of thermal model

The thermal models of the three induction motors were assessed by subjecting each motor to a randomly varying load which included periods of overload and standstill. The temperatures within the motors were monitored using small platinum resistance thermometers and were electronically recorded at 1 min intervals. These were compared with the simulations of the respective thermal models which were solved using the iterative procedure described by eqns. 23 and 24. The constant matrices of the iterative solution were formulated using a time step of 1 min, a time which is small compared with the thermal time constants of the machine. This also allowed the simulated temperatures to be correlated directly with the test data taken at the same interval. The heat generation at each time step was calculated from the corresponding test values of terminal voltage and current and the motor electrical equivalent circuit. The equivalent circuit parameters for each induction motor had previously been obtained from the conventional no-load and locked-rotor tests. The rotor resistance values obtained at a frequency of 50 Hz from the latter test were adjusted to account for the low frequencies present in the rotor during normal operation. These parameters were validated by confirming that the losses estimated from the equivalent circuit were in good agreement with those obtained from additional load tests on a dynamometer.
IEE PROCEEDINGS-B, Vol. 138, NO. 5, SEPTEMBER 1991

The stator iron and teeth thermometers were inserted into small radial holes drilled to a depth which would align the thermometers with the radial centre of the component, at an axial distance of one-sixth of the core length from the motor centre. The number and circumferential position of the winding thermometers was governed solely by the availability of suitable spaces, and it was not always possible to locate them at the optimal positions. The stator slot winding thermometers were inserted into convenient spaces between the array of winding conductors and the slot liner. The endwinding thermometers were located as closely as possible to the radial centre of the endwinding. Thermal cement was employed throughout to ensure a good thermal contact between each thermometer and its surroundings.

3.2 5.5 k W motor, A


The results of the varying load test performed on the first 5.5 kW induction motor are shown in Figs. 7a-e. The large temperature drop between the stator iron and the frame, as much as 20C, is attributed to the particularly low frame-core contact pressure present in this machine. The frame temperature rise that occurs shortly after the supply is switched off is a feature of all three induction motors and is a result of a rapid redistribution of the stored thermal energy as both the internal heat generation and forced ventilation are removed. Because the thermal model has eight separate time constants, it is able to follow this transient well. The temperature dependence of the winding heat generators was included in the model simulations for all
21 1

140 120 loo[

80
60-

Ob

2
0

three motors. To demonstrate how this feature improves the model performance, the endwinding temperatures were recalculated assuming constant stator and rotor winding resistances, set at temperatures of 60 and 80C, respectively. This result is presented in Fig. 8. Although the constant temperature simulations are not poor, the thermal tracking is improved by the inclusion of temperature dependence in the winding heat generators.

q
100
60
40

IO0
."..___.,

20
1

0
140 100

2
b

20 161

I.
1

t1me.h Fig. 8 5.5 kW motor A simulated and measured endwinding temperatures without thermal dependence of heat generation - - - - - - - simulated __ measured rated current
~~

3.3 5.5kW motor, B


The results for the second 5.5 kW motor are shown in Figs. 9a-e. For this machine, an experimentally measured value was used for the frame-core contact resistance, it is therefore not unexpected that the simulated values of the stator iron temperatures lie close to those measured. If the theoretical value obtained from Fig. 3 had been used, the simulated endwinding temperature at full load would have been approximately 7C higher, an increase of 3%.
60
40

20
0
1

d
140r

@ -

100

40 20

2 e

time,h Fig. 7 5.5 kW motor A simulated and measured temperatures Frame e Endwinding b Stator iron - - - -. simulated . c Stator teeth measured d Stator winding rated current
U
~ ~ ~

3.4 75 k W motor, C The load test on the 75 kW induction motor was performed at the manufacturer's works. Experimental results were not available for either the stator teeth or the stator winding, because of the difficulty of drilling holes accurately into the stator teeth and the high slot fill of the windings. The results for the remaining stator components are presented in Figs. loa-c. Additional information was available for the 75 kW motor from steady-state tests which had previously been performed by the manufacturer. A large number of thermocouples were embedded throughout the machine including four locations in the rotor winding and iron. Over 100 steady-state temperatures were recorded while the motor was running at full load. Many of these data could be averaged to give an indication of the mean temperatures in most of the model thermal components. A comparison of these results with the model predictions is given in Table 3 where it can be seen that there is, in general, a good agreement, although the model tends to overestimate temperatures, particularly in the rotor. The estimated steady-state heat flow distribution in the 75 kW motor at full load is given in Fig. 11. The results from this analysis are similar to those expected in practice. The major heat flow path is through the stator core, representing approximately 70% of the loss. The majority of the remaining heat is extracted through the
I E E P R O C E E D I N G S - B , Vol. 138, N o . 5 , S E P T E M B E R 1991

212

0 U

$
c

:"
160 100
80

endcap air and is made up of approximately a third of the endwinding loss and half the total rotor loss. The heat loss along the shaft via the bearings is estimated to be small. The thermal paths that are critical to the
140 1201
100

ob

3
0

a;

Y L 3

100

e f

Q a l

0' 0
160
U
0

3
b

140 I201 a#- 100

!100 . l J

2
n

a
C

? !
U 3

3
time. h

6
01

'

'

4
C

'

'

'

U
0

Q 160

- - - - - - - - ,,
I

. -

g E,

60 20 2
3 d 4

40

'0

6
a

time, h d Fig. 10 75 kW motor C simulated and measured temperatures Frame b Stator iron c Endwinding - - - - - - - simulated measured rated current
~ ~~

160r

steady-state model can be identified from Fig. 11, which suggests that it may be possible to simplify the model by removing nodes 2, 9 and 10. This approach would not be valid for the transient thermal model, because all these nodes contain thermal capacity.
Table 3: Comparison of steady state temperatures in 75 kW motor at full load Model comDonent time, h Fig. 9 5.5 kW motor B simulated Frame e Endwinding b Stator iron - - - - - - - simulated c Stator teeth ___ measured d Stator winding - - rated current
a

Measured temDerature 52.6 (average of 36) 64.0 (average of 1 1 ) 84.8 (average of 24) 103.5 (hottest) 149.0 (average of 2) 141.7 (average of 2)

Predicted Percentage temDerature difference 55.5 70.4 90.4 102.1 168.0 166.5 +6 +10 +7 -1 +13 +18 213

Frame Stator iron Slot winding Endwinding Rotor winding Rotor iron

IEE PROCEEDINGS-B, Vol. 138, N o . 5, S E P T E M B E R 1991

3.5 Sensitivity analysis The sensitivity of the thermal model to the constant coefficients which describe the heat transfers by convection and across the frame-stator interface was investigated by
ambient

4g,0

temperatures. Variations in the other parameters have little effect on the modelled winding temperatures. The sensitivity of the 75 kW motor thermal model to the seven thermal conductivity values that describe the conductive heat transfer within the lumped components was also investigated. The endwinding temperature was most sensitive to the shaft conductivity, however a 20% change in the shaft conductivity caused a change of just 1.6"C. As a result, it is concluded that the model is highly insensitive to these parameters.

Conclusions

Fig. 11 Heatjlow distribution in 75 kW motor C ut full loud heat flow as percentage of total loss

considering the effect of varying these parameters on the estimates of steady-state temperature. The results for the 75 kW induction motor at full load with each of the heat transfer coefficients increased by 20% in turn are summarised in Table 4. Not surprisingly, the model is most sensitive to the frame-ambient heat transfer; however even a large variation in this parameter results in a change of only a few degrees in the estimated winding
Table 4 : Sensitivity of thermal model to heat transfer coefficients : changes, in degrees Celsius, in estimated steadystate temperatures in 75kW induction motor at full load when each heat transfer coefficient is increased in turn by

20%
Model component Heat transfer coefficient Frameambient Stator-frame h, h, Frame Stator iron Stator teeth Slot winding Endwinding Rotor winding Rotor iron Shaft
214

Air gap h ,

Endcap air h,,

-5.9 -5.9 -5.9 -5.9 -5.9 -5.9 -5.9 -5.9

0.0 -1.7 -1.7 -1.5 -1.3 -1 .o -0.9 -0.5

0.0 +0.3 i0.4 i0.4 -0.1 -6.7 -6.6 -3.5

0.0 -0.5 -0.9 -0.7 -2.2 -4.2 -6.1 -3.2

A lumped-parameter thermal model is shown to be a highly effective method of estimating both steady-state and transient temperatures in TEFC machines. The model is formulated from purely dimensional data, physical constants and published heat-transfer characteristics and is thus readily adapted to a wide range of machine sizes and ratings. The only parameter that needs to be obtained by test is the frame-ambient thermal resistance, the rest of the model is created directly from the design drawings. The model predictions are surprisingly insensitive to any one individual value of the many thermal constants that describe the heat conduction and convection within the machine. The final accuracy of the model is attributed to the collected values of these thermal constants as a whole and the methods used to calculate the model parameters from the dimensions. The solution of the lumped-parameter model reduces to a set of just eight linear differential equations and is therefore easy to compute. The transient thermal behaviour of the machine is thus represented by eight thermal time constants, which provide sufficient detail to track winding temperatures accurately during sudden changes in load. In simplifying the model further, there would be a danger of losing this detail for a relatively small further reduction in computation, although clearly a much cruder model could be used if only a steady-state analysis were required. The full model is sufficiently simple to be suitable for programming into a microprocessor based device [I, 23 which is able to solve the motor temperatures in real time from external inputs of the motor supply and ambient temperature. This has important applications in motor thermal protection and aging control. The thermal model has been applied to three induction motors of different ratings and manufacture. Tests performed on these machines demonstrate that the thermal model is able to predict the important stator winding temperatures accurately during load variations that include periods of overload and stationary cooling. The success of these predictions leads to the conclusion that this form of lumped-parameter thermal model would be likely to give accurate results for most TEFC machines, provided they have a similar basic construction to that found in small-to-medium-rated TEFC induction motors. Larger machines, which have a more complex thermal construction, would require a more detailed thermal model with possibly a large number of components. For example, if the stator core has radial cooling ducts, its representation would be in the form of a number of lumped components connected in parallel across the ducts. Provided that sufficient detail is incorporated in the model, the lumped-parameter approach can be a very useful tool in the thermal modelling of the majority of electrical machines.
IEE PROCEEDINGS-B, Vol. 138, No. 5 , S E P T E M B E R 1991

Acknowledgments

The authors wish to thank GEC Electromotors and the UK Science and Engineering Research Council for their support of the project.
6
References

1 MELLOR, P.H., and TURNER, D.R.: Real time prediction of temperatures in an induction motor using a microprocessor, Electr. Mach. Power Syst., 1988,13, pp. 33S352 2 MELLOR, P.H.,TURNER, D.R., and ROBERTS, D.: Microprocessor based induction motor thermal protection. 2nd Intemational Conference on Electrical Machines, Design and Applications, IEE Con$ Publ. 254,1985, pp. 16-20 3 ROBERTS, D.: The application of an induction motor thermal model to motor protection and other functions. PhD thesis, University of Liverpool, 1986 4 PEREZ, I.J., and KASSAKIAN, J.K.: A stationary thermal model for smooth air-gap rotating electrical machines, Electr. Mach. Electromech., 1979,3, (U), 285-303 pp. 5 TAYLOR, GI.: Distribution of velocity and temperature between concentric cylinders, Proc. Roy. Soc., 1935,159, Pt. A, pp. 5 6 5 7 8 6 GAZLEY, C.: Heat transfer characteristics of rotating and axial flow between concentric cylinders, Trans. ASME, January 1958, pp. 79-89 7 LUKE, G.F.: T h e cooling of electric machines, Trans. AIEE, 1923, 45, pp. 1278-1288 8 MORI, Y., and NAKAYAMI, W.: Forced convective heat transfer in a straight pipe rotating about a parallel axis, lnt. J. Heat Mass Trans., 10, pp. 1179-1194 9 BRUNOT, I.W., and BUCKLAND, F.F.: Thermal contact resistance of laminated and machined joints, Trans. ASME, April 1950, pp. 253-257 10 SHLYKOV, Y.P., and GANIN, Y.A.: Thermal resistance of metallic contacts, Int. J. Heat Mass Tram$, 1964,7, pp. 921-929 11 Heat transfer and fluid flow data book(General Electric, 1969) 12 SAY, M.G.: Alternating current machines (Pitman, 1976)

7.2 Stator iron L = Stator length rl = Stator outer radius 1, = Tooth outer radius s = Lamination stacking factor k,, = Lamination axial conductivity k , = Lamination iron conductivity c, = Lamination iron specific heat p, = Lamination iron density

LIZ

Fig. 13

Stator iron

R3 =

L 6nk,,(r: - rf)

Appendix

(33)

7.1 Frame L = Stator length r , = Stator outer radius M e = Endcap mass M , = Frame mass c, = Endcap specific heat c, = Frame specific heat h, = Frame-core contact coefficient

(34)

(35)
7.3 Stator teeth L = Stator length 1 , = Tooth outer radius r3 = Tooth inner radius q5,, = Tooth pitch 6 = Equivalent tooth arc , n = Number of slots s = Lamination stacking factor k,, = Lamination axial conductivity k,, = Lamination iron conductivity c, = Lamination iron specific heat p, = Lamination iron density

; i
ambient
2
Fig. 12

P
(28)
(29) (30)
Fig. 14

P
A

Frame

R,

= measured

R, =nh, L r , C,
=

M,c,

+~ M , c ,

Stator teeth
215

IEE PROCEEDINGS-B, Vol. 138, NO. 5, SEPTEMBER 1991

(36) (37)

7.5 Airgap L = Stator length 1 , = Tooth inner radius rs = Rotor outer radius $ J = Tooth pitch ~ 4, = Equivalent tooth arc h,, = Rotating airgap film coefficient h,, = Stationary airgap film coefficient

r:

r:

(38)

(39)

17

(40) (41)
Fig. 16
Air gap

7.4 Stator winding L = Stator length r4 = Equivalent winding radius di = Insulation thickness A, = Copper cross section in slots n = Number of slots F = Radial conductivity factor k, = Copper conductivity ki = Slot liner conductivity k, = Varnish conductivity c, = Copper specific heat pc = Copper density

QQ
a
R13
-c4

di

Fig. 15

Stator winding

7.6 Endwinding R = Endwinding toriod radius r, = Equivalent stator winding radius r6 = Endwinding cross section radius I, = Slot winding overhang A , = Copper cross section in slots V, = Copper volume in entire winding n = Number of slots F = Radial conductivity factor a = Slot to endwinding volume ratio o = Hot-spot to mean temperature ratio k, = Copper conductivity k, = Varnish conductivity c, = Copper specific heat p, = Copper density

R12 = R 1 3=

2di nki Lr, n L 6k, A , n

+ 2nk,1LFn
~

i--J-4
I f 4

'\A,
~

4di R,, = + 1 nk, Lr, n nk, LFn


~

1 R l s =nk, LFn

M
Fig. 17
Endwinding

cy----..!?._
3-

216

1EE PROCEEDINGS-B, Vol. 138, N o . 5 , S E P T E M B E R 1991

R2, = 16z2RFk,
R21

cor; 8 z r t lo Fk, n
7.8 Rotor winding L = Rotor length r, = Rotor outer radius r7 = End-disc inner radius r8 = Equivalent rotor winding radius I, = End-disc width V , = Volume of end-disc and fan k, = Aluminium conductivity c, = Aluminium specific heat p a = Aluminium density

7.7 Endcap air A, = Contact area of endcap A, = Contact area of stator iron A , = Contact area of stator teeth A, = Contact area of endwinding A , = Contact area of rotor end-disc A, = Contact area of rotor iron A , = Contact area of rotor cooling hole h3, = Rotating endcap film coefficient h,, = Stationary endcap film coefficient h4, = Rotating rotor hole film coefficient h4, = Stationary rotor hole film coefficient

8 Fig. 19

b
Rotor winding

L
R29

6zk,(r:
-1

+
- 1;)

le zk,(r: - r:)

R30

4zkaL4r: - r i )

R3, = -

Fig. 18

Endcap ar i

7.9 Rotor iron L = Rotor length r8 = Equivalent rotor winding radius r9 = Shaft radius rl0 = Cooling hole radius s = Lamination stacking factor m = Number of cooling holes k,, = Lamination axial conductivity k,, = Lamination iron conductivity c, = Lamination iron specific heat p , = Lamination iron density
IEE PROCEEDINGS-B, Vol. 138, NO.5, S E P T E M B E R 1991 217

f'to

7.10 Shaft L = Rotor length r , = Shaft radius 1, = Bearing housing width 1, = Distance of bearing centre to rotor mean k, = Shaft steel conductivity c, = Shaft steel specific heat p, = Shaft steel density

&
Fig. 20

t
I

Lt?

Rotor iron

R33

L 6nkl,,(rf - rf

-1 R3.4 = 4nk,,Ls(r; - r f

mr:o)

Fig. 21

Shaft

1 1, R3, = 2nk,L 2nk,r;

+-

R3,=-+1 4nk, 1,

1, 2nk, rf

(72) (73)

c, =

c,p, nLs(rf - rf - mr:o)

Clo = P, c, nr&,

+ +Ib + &L)

218

I E E P R O C E E D I N G S - B , Vol. 138, N o . 5, S E P T E M B E R 1991

Você também pode gostar