118

CHAPTER 7

DESIGN AND SIMULATION OF AN ANFIS CONTROLLER

BASED DRIVE SYSTEM

This chapter presents the modeling and simulation of an adaptive

neuro-fuzzy inference strategy (ANFIS) to control one of the most

important parameters of the induction machine, viz., speed. IM’s are

non-linear machines having a complex and time-varying dynamics.

Some of the states are inaccessible during the operational stages and

also many of the states are not available for measurements; hence, it

is essential to design state observers.

The control of IMs is thus a challenging problem to be taken up in

the industries, especially in the control of speed in paper mills, etc.

Various advanced control techniques have been devised by various

researchers across the world, with some of them being based on

hybrid techniques. Some of them have already been explained in the

previous chapters.

These fuzzy-based controllers develop a control signal that yields

on the firing of the rule base, which is written on the previous

experiences, which is random in nature. Thus, the outcome of the

controller is also random and optimal results may not be obtained.

Selection of the proper membership functions and in turn selecting

the right rule base depending on the situation can be achieved by the

use of an ANFIS controller, which becomes an integrated method of

 119

approach for control purposes and yields excellent results, which is

the highlight of this chapter. In the designed ANFIS scheme, neural

network techniques are used to select a proper rule base, which is

achieved using the back-propagation algorithm. This integrated

approach improves the designed controller’s performance in many

ways in terms of cost-effectiveness and reliability.

The simulation results presented at the end of this chapter prove

that if the designed control is more effective, it has faster response

times or settling times. The sudden fluctuation or change in speed

and its effect on the various parameters of the dynamic system are

also considered further in this chapter. The designed controller not

only takes care of the sudden perturbations in speed, but also brings

back the parameters to the reference or the set value in a few

milliseconds, thus exhibiting the robustness in behavior. In other

words, the designed controller is robust to parametric variations. A

reasonable accuracy in the speed-control characteristics of the IM

could be observed using this robust control scheme.

7.1 INTRODUCTION

Artificial intelligence, ANN, Fuzzy Logic, hybrid networks, etc. have

been recognized as main tools to improve the performance of power

electronics-based drives in the industrial sectors. Currently, the

combination of this intelligent control with adaptiveness appears as

the most promising research area in the practical implementation and

 120

control of electrical drives [77]. A review of the research carried out by

various researchers [52]-[65] regarding the ANFIS control of AC

machines was discussed briefly in Chapter 2. The responses (speed)

had taken a long time to reach the set value. In the research work

presented in this chapter, an attempt is made to reduce the settling

time of the responses (speed) and make the response very fast by

designing an efficient controller using a hybrid type of ANFIS-based

control strategy taking into account some of the parameters

mentioned in the above papers.

An intelligently developed back-propagation algorithm could be

used for NN training for the proper selection of the rule base. Here, we

have formulated this complex control strategy for the speed control of

IM, which yielded excellent result compared to the other methods

mentioned in the literature survey.

7.2 ANFIS CONCEPT

ANN has strong learning capabilities at the numerical level. Fuzzy

logic has a good capability of interpretability and can also integrate

expert's knowledge. The hybridization of both paradigms yields the

capabilities of learning, good interpretation and incorporating prior

knowledge. ANN can be used to learn the membership values for fuzzy

systems, to construct IF-THEN rules, or to construct decision logic.

The true scheme of the two paradigms is a hybrid neural/fuzzy

system, which captures the merits of both the systems [89].

 121

This concept is made use of in developing the ANFIS controller in

this chapter. A neuro-fuzzy system has a neural-network architecture

constructed from fuzzy reasoning. Structured knowledge is codified as

fuzzy rules, while the adapting and learning capabilities of neural

networks are retained. Expert knowledge can increase learning speed

and estimation accuracy.

Fuzzy logic is one of the most successful applications in the control

engineering field, which can be used to control various parameters of

real-time systems. This logic combined with neural networks yields

very significant results. This merged technique of the learning power

of the NNs with the knowledge representation of FL has created a new

hybrid technique, called neuro-fuzzy networks [62]. This technique

gives a fairly good estimate of the speed and is robust to parameter

variation [58].

7.2.1 The ANFIS Model

The architecture of the ANFIS model is a graphical representation

of the TS-FLC model. The general ANFIS control structure for the

control of any plant is presented in this section [63], [57]. The

functions of the various layers are given in the form of an algorithm as

described below. The structure contains the same components as .fis,

except for the NN block. The network structure is composed of a set of

units arranged into 5 interconnected network layers, viz., l1 to l5 as

shown in Fig. 7.5.

 122

Layer 1: This layer consists of input variables (membership

functions), viz., input 1 and input 2. Here, triangular or bell-shaped

MF can be used. This layer supplies the input values to the next layer,

where i=1 to n. In other words, layer 1 is the input layer with n nodes.

Layer 2: This layer (membership layer) checks the weights of each MF.

It receives the input values xi from the 1st layer and acts as MFs to

represent the fuzzy sets of the respective input variables.

Furthermore, it computes the membership values that specify the

degree to which the input value xi belongs to the fuzzy set, which acts

as the inputs to the next layer. Note that layer 2 has nK nodes, each

outputting the MF value of the ith antecedent of the jth rule given by

yij(2)=µAji x(i), where A is a matrix defining a partition of the space xi

and µAji(x(i)) is typically selected as a generalized bell MF defined by

the equation: µAji(x(i))= µ(xi,cji,aji,bji). Note that the parameters cji,aji,bji

in the above equation are referred to as the premise parameters.

Layer 3: This layer performs the pre-condition matching of the fuzzy

rules, i.e., they compute the activation level of each rule, the number

of layers being equal to the number of fuzzy rules. Each node of these

layers calculates the weights that are normalized. Layer 3 has K fuzzy

neurons with the product t-norm as the aggregation operator. Each

node corresponds to a rule and the output of the jth neuron

determines the degree of the fulfillment of the jth rule given by

y (j3)  in1 Aij ( xi ) for j=1,………, K.

 123

Layer 4: This layer provides the output values y, resulting from the

inference of rules. Connections between the layers l3 and l4 are

weighted by the fuzzy singletons that represent another set of

parameters for the neuro-fuzzy network. Each neuron in layer 4

performs the normalization, and the outputs are called normalized

firing strengths, which are given by

y (3)
 for j=1,…..,K. (7.1)
(4) j
y j K

y
k 1
(3)
k

Layer 5: This layer is called the output layer, which sums up all the

inputs coming from layer 4 and transforms the fuzzy classification

results into a crisp (binary). The output of each node in the layer can

be defined by y (j5)  y (j4) f j ( x) for j=1,……,K, where f j (X ) is given for the

jth node in layer 5. The outputs of layer 5 are summed up and the final

output of the network can be re-written as y (j5)   j 1 y (j4) . The ANFIS

K

structure is tuned automatically by least-square-estimation and the

back-propagation algorithm [89]. The algorithm shown above is used

in the next section to develop the ANFIS controller to control the

various parameters of the IM.

Learning of the ANFIS Model: In the ANFIS model, functions used at

all the nodes are differentiable; thus the BP algorithm can be used to

train the network. Each MF μAij is specified by a pre-defined shape

and its corresponding shape parameters. The shape parameters are

 124

adjusted by a BP learning algorithm using a sample set of size N (xt ,

yt). For non-linear modeling, the effectiveness of the model is

dependent on the MFs used. The TSK fuzzy rules are employed in the

ANFIS model as follows:

Ri : IF x is Ai, THEN

   A x f ( x)i i i
y=f i(x)= yi  i 1
k
(7.2)
A x
i 1
i i

where, Ai xi   nj1 Ai j x(i, j )   nj1 Aij x(i, j ). Accordingly, the error

measure at time t is defined by

1 
y  Et  ( y t  yt )2 (7.3)
2

After the rule base is specified, the ANFIS adjusts only the MFs of

the antecedents and the consequent parameters. The BP algorithm

can be used to train both the premise and consequent parameters. A

more efficient procedure is to learn the premise parameters by the BP,

but to learn the linear consequent parameters ai,j by the RLS

(Recursive Least Square) method [89]. The learning rate η can be

adaptively adjusted by using a heuristic approach.

7.3 ANFIS CONTROLLER DESIGN

In the modeling and feedback control of any dynamical system, a

controller is a must for the plant as it takes care of all the

 125

disturbances and brings back the system to its original state in a

couple of second.

Fig. 7.1: Block diagram of the ANFIS control scheme

To start the design of the controller using the ANFIS scheme, first,

a mathematical model of the induction motor plant along with the

controller mathematical model is required, which can be further used

for simulation purposes.

The mathematical model of the plant is given by Eq. (3.56) in

Chapter 3 and it is a fourth-order mathematical model of size (4  4),

which is used here in the Simulink model. The basic structure of the

ANFIS coordination controller developed in this study to control the

speed of the IM. The block diagram of the developed controller is

shown in Fig. 7.1. Inputs to the ANFIS controller, i.e., the error and

the change in error, are modeled using Eqs. (7.4) and (7.5) as follows:

e(k) = ωref - ωr (7.4)

∆ e(k) = e(k) - e(k - 1) (7.5)

 126

where ωref is the reference speed, ωr is the actual rotor speed, e(k) is

the error and ∆e(k) is the change in error.

The fuzzification unit converts the crisp data into linguistic

variables, which is given as inputs to the rule-based block. The set of

49 rules is written on the basis of previous knowledge or experiences.

The rule-based block is connected to the neural network block. Back-

propagation algorithm is used for NN training in order to select the

proper set of rule base. For developing the control signal, training is a

very important step in the selection of the proper rule base. Once the

proper rules are selected and fired, the control signal required to

obtain the optimal outputs is generated.

The inputs are fuzzified using the fuzzy sets and are given as input

to the ANFIS controller. The rule base for the selection of proper rules

using the back-propagation algorithm is shown in Table 7.1.

Table 7.1: Rule base for controlling the speed of IM using ANFIS

e
nb nm ns ze ps pm pb
∆e 
nb nb nb nb nb nm ns ze
nm nb nb nm nm ns ze ps
ns nb nm ns ns ze ps pm
ze nb nm ns ze ps pm pb
ps nm ns ze ps ps pm pb
pm ns ze ps pm pm pb pb
pb ze ps pm pb pb pb pb
 127

The same rule base described in Table 5.1 in Chapter 5 (Section

5.2), which was used for Mamdani-based FLC and TS-based FLC for

decision-making purposes, is also used in this chapter for the

decision-making purposes to design the ANFIS controller. The same

49 rules have been used here for control purposes and are not shown

here for the sake of convenience.

The control decisions are made based on the fuzzified variables in

Table 7.1. The inference involves a set of rules for determining the

output decisions. As there are 2 input variables and 7 fuzzified

variables, the controller has a set of 49 rules for the ANFIS controller.

Out of these 49 rules, the proper rules are selected by training the

neural network with the help of back-propagation algorithm and these

selected rules are fired. Furthermore, it has to be converted into

numerical output. The output y is given as follows:

R R

  i a1i x1      i aqi xq
y i 1
R
i 1
(7.6)

i 1
i

This controlled output of the IM, i.e., y (here, it is the speed of the

IM), is the weighted average of the proper rule-based outputs, which

are selected by the back-propagation algorithm.

 128

7.4 DEVELOPMENT OF THE SIMULINK MODEL

Fig. 7.2: The Simulink model for the ANFIS controller.

 129

ANFIS controller based system was developed using the various

toolboxes in Matlab /simulink. The various waveforms are observed

on the corresponding scopes after running the simulations. The

specifications of the SCIM used for simulation purposes were

described in Chapter 3 (table 3.1).

7.5 SIMULATION RESULTS AND DISCUSSIONS

The Simulink model of ANFIS controller is as in Fig.7.2. In order to

start the simulations, the 49 fuzzy rule set has to be invoked first from

the command window in the Matlab. Initially, the fuzzy file where the

rules are written with the incorporation of the T-S control strategy is

opened in the Matlab command window, after which the fuzzy editor

(FIS) dialogue box opens. The .fis file (sugenosevenrules2.fis) is

imported using the command window from the source and then

opened in the fuzzy editor dialog box using the file open command.

Once the file is opened, the TS-fuzzy rules file gets activated.

Furthermore, the data is exported to the workspace and the

simulations are run for a specific period of time (say 3 second). The

fuzzy membership function editor is then obtained using the view

membership command from the menu bar.

The written TS-fuzzy rules also can be viewed from the rule view

command. The rule viewer for the 2 inputs and 1 output can be

observed pictorially. Now, after performing all the preliminary

operations, the simulations are run for a period of 3 second in Matlab

 130

 100  60 
7 with a set speed of 100 rad / second, i.e.,   =955 rpm and
 2 

with a 2 N-m load torque. Once, the simulation is run, various

parameters such as speed, flux, torque, currents, slip, etc. get stored

in the workspace. After running the developed Simulink model, we get

the error, change in error and an intermediate parameter.

Fig. 7.3: ANFIS editor window

Fig. 7.4: ANFIS editor: Loading the data from the work space

These 3 parameters are stored in a variable in the command window.

The ‘anfis’ editor opened in the command window is shown in Fig 7.3.
 131

These variables, which are in the form of data in the workspace, are

loaded into the ‘anfis’ editor (Fig. 7.4).

Once the data file is loaded, the ‘anfis’ has to be trained by

selecting a back-propagation algorithm with a suitable number of

epochs. In our work, we have used the back-propagation algorithm

with a suitable number of epochs being used for training the rules.

This is done by selecting these 2 items in the ‘train window’ of the

‘anfis’ editor and training the neural network for proper selection of

the rule base. The trained data are further exported to the workspace

using the file-export command.

Fig. 7.5: ANFIS model with 2 inputs and 1 output showing the

ANN architecture

The developed ANFIS model structure with 2 input neurons and 1

output neuron along with 4 hidden layers (input membership

function, rule base, membership function and aggregated output) is

shown in Fig. 7.5. The training of the ANN by using the fuzzy rule
 132

base for the selection of the proper and optimal rule is taken care of

by the designed ANFIS controller. Note that 7 by 7 rules are used in

the hidden layers. Both neuron 1 and neuron 2 are connected to 7

fuzzy rules. The hidden layers contains 49-49 neurons to deal with

the problem (for selection of the proper rule base, because the rule

base is written randomly in fuzzy, the neural network selects the right

optimal rule base to fire).

The 2 input neurons, viz., the error and change in error, are given

as input to the 1st hidden layer of the ANN as shown in Fig. 7.5. This

1st hidden layer deals with various input membership functions. In

the 2nd and 3rd hidden layers, the set of 49 fuzzy rules is properly

identified by training and the sets of optimal rules are selected. These

sets of optimum rules are available at the 4th hidden layer. Out of the

49 rules, the optimal rules are fired here and the de-fuzzified output is

obtained as the output neuron. The de-fuzzified output is further used

to generate the firing pulse to be applied to the inverter bridge, which

is further used to control the speed of the IM drive.

After the simulation is run, the performance characteristics are

observed on the respective scopes. The response curves of flux, load,

torque, terminal voltage, speed, rotor angle, stator currents, slip, id, iq,

rotor currents ( and d-q ) v/s time, slip vs. speed for a set speed

of 100 rad/second (955 rpm) and with a 2 N-m load torque are

observed and are shown in Figs. 7.6 – 7.17, respectively.

 133

The simulation results in Fig. 7.6 show that the stator current does

not exhibit any overshoots or undershoots. The response curves takes

lesser time to settle and reach the desired value compared to the

results presented in [84], [76], and [90].

Fig. 7.6: Plot of speed vs. time

The simulation results showed that by using the neuro-fuzzy

(ANFIS) control, for the set speed of 100 rad/second and for the 49

rules, the speed reaches its desired set value at 0.3 second. This

shows the effectiveness of the designed neuro-fuzzy controller, which

tries to speed up the performance of the drive, thus showing faster

dynamism.

The motor speed increases like a linear curve up to the set speed of

955 rpm (100 rad/second) in 0.3 second as shown in Fig. 7.6.

Furthermore, it can also be observed that using the ANFIS control, the

system stabilizes in a less time compared to the other methods.

 134

Fig. 7.7: Plot of flux vs. time

From the variation of flux with time as shown in Fig. 7.7, it can be

observed that when the motor speed increases (during the transient

period); more stator current is required to develop the requisite flux in

the air gap. Hence, the flux also starts increasing during the transient

period (0 to 0.3 second) exponentially. Once, the motor attains the set

rated speed, the flux required to develop the torque almost remains

constant after >0.3 second. Once the flux in the air gap remains

constant, the variation of the load torque and speed will not disturb

the flux curve. Hence, the IM will operate at a constant flux. The

is with time is shown in Fig. 7.8.

 135

It can be clearly observed from this figure, that at lower speeds, the

slip is more, and hence the flux required to develop the suitable

torque is also more. Moreover, the torque required to reach the set

speed is also more. Hence, the magnitude of the stator currents will

also be more during the transient periods (starting periods) of the

induction motor. When the speed reaches the set value from zero, the

speed at 0.3 second, it requires a nominal stator current to drive the

IM system.

Torque characteristics for a reference speed of 100 rad/second

(955 rpm) are shown in Fig. 7.9. From this figure, we can conclude

that when the motor operates at lower speeds, the slip is more. Hence,

the machine requires more torque to attain the set speed. Once the

machine reaches the set speed of 955 rpm, the average torque of the

machine becomes nearly zero after 0.3 second, which is justified from

the torque simulation result shown in Fig. 7.9. The terminal voltage of

the IM is shown in Figs. 7.10 (a) and (b), respectively.

Fig. 7.9: Plot of torque vs. time

 136

(a) Plot of voltage vs. time (normal)

(b) Plot of voltage vs. time (normal and zoomed) between t=0.984s

to 1.033s

Fig. 7.10: Plot of voltage vs. time (normal and zoomed)

Fig. 7.11 shows the variation of slip vs. time characteristics for a

speed of 100 rad/second (955 rpm). From this simulation result, we

infer that the IM attains the set reference speed of 955 rpm in 0.3

second using the ANFIS (neuro-fuzzy) controller. At that instant, the

 N  N  1800  955
slip being  s = =0.46 can be verified from the result
 N s  1800

shown. Note that the slip decreases from 1.0 to 0.46 linearly in a time

span of just 0.3 second.

 137

Fig. 7.11: Plot of slip vs. time

The slip-speed characteristics are shown in Fig. 7.12. When the

speed is varied from 0 to the rated speed, the slip decreases, i.e., the

slip is inversely proportional to the speed, which is a characteristic of

IM. When the speed is zero, the slip is 100% whereas the IM operates

close to the rated speed (180 rad/second), the slip is very low. The

plots of the direct axes id and i q and quadrature axes currents iq

versus time are shown in Figs. 7.13 and 7.14. From these figures, it

can be inferred that the machine reaches the set reference speed of

955 rpm at a time interval of 0.3 second.

Fig. 7.12: Plot of speed vs. slip

 138

Fig. 7.13: Plot of direct axis id v/s time

The variation of rotor currents ir-abc with time is shown in Fig. 7.15.

It can be inferred that at lower speeds, the slip is more, and hence the

flux required to develop the suitable torque is also more. Moreover,

the torque required to reach the set speed is also more. Hence, the

magnitude of the rotor currents will also be more during the transient

periods (starting periods) of the induction motor.

Fig. 7.14: Plot of rotor current iq v/s time

Whe

currents decrease exponentially up to 0.3 second; thereafter it

ir-abc are transformed to the

direct axes and quadrature axes currents using the d–q

 139

transformation techniques; the variation of the transformed currents

with time is shown in Fig. 7.16. Here, only two phases d and q axes of

the currents can be observed in the characteristic curve. In this case,

also, once the motor achieves the set speed at 0.3 second, it requires a

nominal current to drive the IM system.

Fig. 7.15: Plot of rotor current ir abc v/s time

Fig. 7.16: Plot of rotor current irdq v/s time

7.6 JUSTIFICATION OF ROBUSTNESS ISSUES IN SPEED

CONTROL

A significant contribution of the developed controller in the work

presented in this chapter is that, the designed controller can also be

 140

used for variable speed. When the system is in operation (when the

simulations are going on), due to sudden changes in set speed (say,

the set speed immediately changed from 100 rad/second

 100  60   140  60 
   955 rpm to140 rad/second    1337 rpm or anything
 2   2 

else and then suddenly decrease the speed back to normal), with the

incorporation of the designed controller in loop with the plant, the

system comes back to stability within a few millisecond (ms).

The simulation results due to the parametric variations of speed

from 100 to 140 rad/second and then back to normal are shown in

Figs. 7.17-7.25, respectively. It is clearly observed that the dynamic

performance of the system is quite improved, and also insensitive to

parametric variations with the incorporation of the ANFIS-based

control scheme. Furthermore, it can be also concluded that although

some motor parameters are non-linear, it appears linear in nature.

Fig. 7.17: Speed curve for variation in speed from 100–140 r/s

and back
 141

Note that when the speed is varied from 100 to 140 rad/second at

say t =0.75 second, the motor takes very less time (only 0.26 second)

to reach the new set speed point (140 rad/second) to become stable.

Again when the IM runs at 140 rad/second, the speed suddenly varies

from 140 to 100 rad/second at say t =1.6 second, and the motor again

takes very less time (only 0.45 second) to reach the new set speed

point (100 rad/second) to become stable as shown in Fig. 7.17. From

this, it can be observed that the speed of the IM is robust (insensitive)

to sudden changes in speed, which is because of the ANFIS controller.

This speed characteristic curve shows that even when there is a

sudden change in speed, the controller takes very less time to reach

stability. The stator current does not exhibit any overshoots or

undershoots. The motor speed increases similar to a linear curve up

to the set speed of 955 rpm in 0.3 second. The torque vs. time for

variation in speed from 100 to 140 rad/second and back to 100

rad/second is shown in Fig. 7.18. It can be seen that when the speed

of the IM increases from 0 to the set value (100 rad/second), the

torque required to reach the set speed is high.

After the motor reaches the set speed of 100 rad/second, the

average torque required to run the motor at the set speed of 100

rad/second will be zero (average of upper and lower half

characteristics of the torque during the rated speed). The motor

reaches the rated speed in just 0.3 second. Now, if the speed is

suddenly increased from 100 to 140 rad/second, the torque

 142

requirement also increases between the period from t =0.75 second to

1.01 second.

Fig. 7.18: Torque characteristic for variation in speed

After the motor reaches the new set speed of 140 rad/second, the

average torque required to run the motor at the new set speed of 140

rad/second will be zero between the period from t =1.01 second to 1.6

second. Now, if the speed is suddenly decreased from 140 to 100

rad/second, the torque requirement is less between the period from t

=2.1 second to 3.0 second. After the motor reaches the original set

speed of 100 rad/second, the average torque required to run the

motor at the original set speed of 100 rad/second will be zero from t

=2.1 second onwards. From Fig. 7.18, we can conclude that when the

motor operates at lower speeds, the slip is more. Hence, the machine

requires more torque to attain the set speed.

Once the machine reaches the set speed of 955 rpm, the average

torque of the machine becomes nearly zero, which is justified from the

is-abc with time

for the variation in speed from 100 to 140 rad/second and back to
 143

normal is shown in Fig. 7.19. There is a change in the stator current

variation during the change in speed from one value to another during

the time interval from t =0.75 to 1 second and from t =1.6 to 2.1

second.

Fig. 7.19: Stator current characteristics for variation in speed

Once the stable point is reached, the stator current becomes

normal. This figure clearly shows that at lower speeds, the slip is

more, and the flux required to develop the suitable torque is also

more.

The torque required to reach the set speed is also more. Hence, the

magnitude of the stator currents will also be more during the

transient periods (starting periods) of the induction motor. When the

exponentially as shown in Fig. 7.19. Once it attains the set speed at

0.3 second, it requires a nominal stator current to drive the IM

system.
 144

One observation that can be made in the flux characteristics

during the change in speed is that, during speed variation, the flux

varies slightly only during the time interval at which the speed

variation occurs. The flux characteristics are shown in Fig. 7.20.

From the variation of flux with time, it can be observed that when the

motor speed increases (during the transient period); more stator

current is required to develop the requisite flux in the air gap.

Fig. 7.20: Flux characteristics for variation in speed

Hence, the flux also starts increasing during the transient period (0

to 0.3 second) exponentially. Once the motor attains the set rated

speed, the flux required to develop the torque almost remains

constant at 0.9 wbs after ≥0.3 second. Once the saturation of the flux

takes place in the air gap, the variation of the load torque and speed

will not disturb the flux curve. Hence, the IM will operate at a

constant flux.

The plots of the direct axis current id and quadrature axis current

iq versus time are shown in Figs. 7.21 and 7.22, respectively. From
 145

these figures, it can be inferred that the machine reaches the set

reference speed of 955 rpm at a time interval of 0.3 second initially.

When there is a variation in speed (either increase or decrease), the

current also changes approximately during that transition period, i.e.,

t =0.75 to 1 second and from t =1.6 to 2.1 second. It can be inferred

that at lower speeds, the slip is more, and hence the flux required to

develop the suitable torque is also more.

Fig. 7.21: Direct axes current characteristics for variation in

speed

Moreover, the torque required to reach the set speed is also more.

Hence, the magnitude of the rotor currents will also be more during

the transient periods (starting periods) of the IM. When the speed

exponentially and finally reach a steady-state value.

ir-abc are transformed to the direct axes and

quadrature axes currents using the (d and q) transformation

techniques and the variation of the transformed currents with time is

 146

shown in Figs. 7.21 and 7.22 ir-

abc with time is shown in Fig. 7.23

Fig. 7.22: Quadrature axes current characteristics for variation in

speed

Fig. 7.23: 3 Φ rotor current characteristics for variation in speed

Furthermore, the combined curve of the 2-phase rotor currents in

the d-q plane is shown in Fig. 7.24. Here, only two phases (d and q

axes) of the currents can be observed in the characteristic curve. In

this case, also, once the motor achieves the set speed at 0.3 second, it

requires a nominal current to drive the IM system. When there is a

variation in speed (either increase or decrease), the current also

 147

changes approximately during that transition period, i.e., from t =0.75

to 1 second and from t =1.6 to 2.1 second.

Fig. 7.24: Rotor current: 2 phase characteristics for variation in

speed

Fig. 7.25: Plot of load torque v/s. time

The load torque is set at a constant value of 2 N-m throughout the

process of simulation at the time of change in speed, which can be

seen in Fig. 7.25. The performance of the developed method in this

chapter also demonstrates the effectiveness of the sudden variation of

speed from clockwise to anticlockwise rotation (-50 rad/second to +50

rad/second with specific amount of time) to obtain stability (which is

shown in Fig. 7.26). The simulation results demonstrate the good

 148

damping performance of the designed robust controller despite speed

fluctuations.

Fig. 7.26: Plot of variation of speed with time for clockwise and

anticlockwise rotation of IM

7.7 SUMMARY

A novel and systematic method of achieving robust speed control of

an IM system using SVPWM technique for voltage source inverter by

means of adaptive neuro-fuzzy inference system has been investigated

in this chapter. The SVPWM is used to control the firing angle of the

inverter. This, in turn, controls the speed of the IM. The Simulink

model was developed in Matlab 7. Due to the incorporation of the

ANFIS controller, it was observed that the motor reaches the set speed

very quickly in a shorter period of time, i.e., only 0.3 second to reach

the set speed (100 rad/second).

 149

One of the main advantages of the ANFIS scheme is that, it is

computationally efficient, increases the dynamic performance and

provides good stabilization when there is a sudden fluctuation in one

of the system parameters, say speed of the IM. This shows the

excellent response of the proposed control scheme as it has the

learning capability using the neural networks.

The ANFIS systems handling more complex parameters, which

demonstrates the effectiveness of the sudden variation of speed

(because of parametric variation) from the normal value and its effects

on the various parameters (such as slip, current, torque, etc.) to

obtain the stability.