1364631818

8/13/2019 1364631818

1/10

Indian Journal of Advanced Electronics EngineeringVolume.1 Number.1 January-June 2013, pp.1-10

@ Academic Research Journals, (India)

1

Neuro-Fuzzy Based Performance Evaluation of Dc

Shunt Motor

T. D. Sudhakar,Associate Professor, St. Josephs College of Engineering, Chennai.Email:- [email protected]

Abstract: This paper describes an alternative methodology intended for the mathematical

calculation designed to evaluate the performance for DC Shunt motor through neuro-fuzzy logic

controller (NFLC). Here by using experimental results for various load conditions of a motor

using fuzzy logic controller is developed. This fuzzy logic controller provides imitativeness,

simplicity, easy implementation and minimal knowledge of system dynamics. This type of fuzzy

logic controller has met the growing interest in many motor control applications due to its non

linearitys handling features and independence.

1. INTRODUCTION

The performance evaluation of a DC shunt motor is very crucial especially in application where

precision and protection are very important. This type of motors is normally used in commercial practiceand is typically recommended where starting conditions are not usually severe. Performance of the shunt

wound motors may be regulated in two ways: first, by inserting resistance in series with the armature, thus

decreasing speed: and second, by inserting resistance in the field circuit, the speed will vary with each

change in load: in the latter, the speed is practically constant for any setting of the controller. This latter isthe most generally used for adjustable speed service, as in the case of machine tools. In general the

performance of the machine depends on the load conditions, so if the performance is known then thenecessary correction can be performed automatically. As a result in this work, a neuro fuzzy logic based

meter is implemented to evaluate the performance, say various output characteristics (like efficiency, speed

& output torque) of the DC shunt motor.

It is seen that shunt motor has a definite no load speed. The speed for any load within theoperating range of the motor can be readily obtained by varying the field current by means of field rheostat.

The efficiency & the torque are calculated using some manual calculations. These manual calculations will

also lead some errors. To avoid these errors, a meter is developed using the neuro-fuzzy Logic Controller(NFLC) which shall indicate the values for a particular load condition through a seven segment display.

Similarly there are some mathematical calculations involved to calculate other system parameters liketorque, efficiency & etc. By using NFLC we can avoid this calculation.

The Adaptive Neuro-Fuzzy Inference System (ANFIS), developed in the early 90s by Jang,

combines the concepts of fuzzy logic and neural networks to form a hybrid intelligent system that enhancesthe ability to automatically learn and adapt. Hybrid systems have been used by researchers for modelling

and predictions in various engineering systems. The basic idea behind these neuro-adaptive learning

techniques is to provide a method for the fuzzy modelling procedure to learn information about a data set,

in order to automatically compute the membership function parameters that best allow the associated FIS totrack the given input/output data. The membership function parameters are tuned using a combination of

least squares estimation and back-propagation algorithm for membership function parameter estimation.

These parameters associated with the membership functions will change through the learning

process similar to that of a neural network. Their adjustment is facilitated by a gradient vector, which

provides a measure of how well the FIS is modelling the input/output data for a given set of parameters.Once the gradient vector is obtained, any of several optimization routines could be applied in order to

adjust the parameters so as to reduce error between the actual and desired outputs. This allows the fuzzysystem to learn from the data it is modelling. The approach has the advantage over the pure fuzzy paradigm

that the need for the human operator to tune the system by adjusting the bounds of the membership

functions is removed.

The NFLC system combines the input and output relationship and based on this relationship onlythe output values are calculated. Simulation studies are done using the ANFIS editor of the matlab, which

show the importance of the inclusion of the FLC.

8/13/2019 1364631818

2/10



2

2. SYSTEM DESCRIPTION

The performance parameters of DC Shunt motor and how it is calculated are shown below inTable 1. All motor parameters are obtained by standard system identification as given in Table 2.

Table 1 : Dc Motor Performance Parameters

Description Formulae Units

Torque T = F * 9.81 * R N-m

Spring Balance Reading F = F1 - F2 Kg

Output Power ( ) 602 TNW = WattsInput Power P = VL*IL Watts

Line Current IL= Ia + If Amp

Efficiency = (Output power / Input power) *

100

%

Table 2 : Dc Motor Parameters

Parameters Description Value

V Rated Voltage 240 VN Rated Speed 1750 RPM

F Rated Frequency 50 Hz

W Rated Output 5 HP

A mat lab model of the D.C. shunt motor is used in the load test and other further analysis. Thevoltage for driving the motor is 240V which is applied to the circuit as shown in fig. 2. The resistance is

varied numerically in the motor model until the rated speed is reached. In DC shunt motor on a no-load

condition the speed is measured to be 1750 RPM. The load variation is done by connecting a CONSTANTblock to the mechanical load terminals of the motor model and the values are varied for which current and

speed is noted. It is shown that as current increases, speed decreases, voltage is constant and other

parameters values also get varied. The performance characteristics of the motor are shown in fig. 1. The

maximum efficiency calculated during the performance calculation of the DC shunt motor was 75%.

Fig.1. Performance Characteristics of DC Shunt Motor Experimental Setup

Fig 2. Circuit Diagram for load test on DC Shunt Motor

(%)

N(rpm)

T(n-m)

IL(A)

N Vs O/P

T Vs O/PvsO/P

ILVs O/P

O/P(W)

8/13/2019 1364631818

3/10



3

3. SOFTWARE ASPECTS

Development of the meter which indicates the various parameters of the motor is used in this

paper to illustrate the NFLC design. The main objective is to examine the performance of the shunt motor.

This is carried out through a fuzzy logic evaluation block that receives input from the command signal.NFL is a problem-solving and self tuning control system methodology that lends itself to

implementation in systems ranging from simple, small, embedded micro-controllers to large, networked,multi-channel PC or workstation-based data acquisition and control systems. It can be implemented in

hardware, software, or a combination of both. NFL provides a simple way to arrive at a definite conclusionbased upon vague, ambiguous, imprecise, noisy, or missing input information and is also self learning.

NFL's approach to control problems mimics how a person would make decisions, only much faster.

NFL requires some many numerical parameters in order to operate such as what is considered

significant error and significant rate-of-change-of-error, but exact values of these numbers are usually not

critical unless very responsive performance is required in which case empirical tuning would determinethem. The first step in implementing NFL is decide exactly what is to be controlled and how NeuroFuzzy

logic defines rules by using the trained data that determine the behavior of the system and gives its outputwith fuzzy using word descriptions instead of mathematical equations, these include fuzzification, de-

fuzzification and rule base.

Fuzzification is the process which determines the degree of membership functions of the input

values to define the fuzzy sets (linguistic variables) while Neural Networks are information processing

system that learns and trains itself to adapt.

Basic Adaptive Neuro-Fuzzy Principle

A typical architecture of an ANFIS is shown in Figure 2, in which a circle indicates a fixed node,

whereas a square indicates an adaptive node. For simplicity, we consider two inputs x, y and one output z.Among many FIS models, the Sugeno fuzzy model is the most widely applied one for its high

interpretability and computational efficiency, and built-in optimal and adaptive techniques. For a first order

Sugeno fuzzy model, a common rule set with two fuzzy ifthen rules can be expressed as:

Rule 1: ifx isA1 andy isB1 , then z1= p1x + q1y + r1

Rule 2: ifx isA2 andy isB2 , then z2= p2x + q2y + r2

WhereAi and Bi are the fuzzy sets in the antecedent, and pi, qi and ri are the design parametersthat are determined during the training process. As in Figure 2, the ANFIS consists of five layers [8]:

Figure 2. Corresponding ANFIS Architecture

Layer 1: Every node i in the first layer employ a node function given by:Oi1= Ai(x), i = 1, 2

Oi1= Bi=2(y), i = 3, 4

Where Ai and Bi can adopt any fuzzy membership function (MF).

Layer 2: Every node in this layer calculates the firing strength of a rule via multiplication:Oi

2 = Wi = Ai(x). Bi (y), i=1,2

8/13/2019 1364631818

4/10



4

Layer 3: The i-th node in this layer calculates the ratio of the i-th rules firing strength to the sum of ail

rules firing strengths:

Oi3= Wi= Wi/(W1+W2), i=1,2

Where Wiis referred to as the normalized firing strengths.

Layer 4: In this layer, every node i has the following function:

Where Wi is the output of layer 3, and { pi, qi, ri} is the parameter set. The parameters in this

layer are referred to as the consequent parameters.

Layer 5: The single node in this layer computes the overall output as the summation of all

incoming signals, which is expressed as:

The outputz in Fig. 3 can be rewritten as

This is how ANFIS implements the logic.

4. CONTROLLER DESIGN

The operation of a NFLC is based on validity of the data used to train the neural network, heuristicknowledge and linguistic description to perform a task. The effects from inaccurate parameters and modelsare reduced because a NFLC does not require a system model. However building a NFLC from the ground-

up may not provide good results or sometime even a worst result than a conventional controller if there is

not enough knowledge of the system.

Procedure 1: Defining Inputs and Outputs

To apply the heuristic knowledge in the FLC, inputs and outputs are defined below:

Input 1: CurrentInput 2: Speed

Output 1: Efficiency

Output 2: Torque

Procedure 2: Loading, Plotting and Clearing of Data

The data collected from the load test of the D.C Shunt motor is loaded into the ANFIS editor in the

form of a .dat file. Then the fuzzy membership functions are generated using the GENERATE FIScommand by selecting Grid Patterning in the window, this automatically generates the FIS structure with

the rules based on the No. of input membership functions ,type of member function and type of

membership function for the output given by us. We have used the Sugeno Fuzzy model which is supported

in the ANFIS editor. The Sugeno model has only one output type and therefore two controllers one for

torque and one for efficiency are designed based on previous data. The no. of member functions and type of

8/13/2019 1364631818

5/10



5

member function is the basic constraint for reducing the average training error. The fuzzy constraints usedfor the Neuro-Fuzzy model are given below illustrated in Fig 3.

Table 3 : Fuzzy Linguistic Terms

Terms Definition

In1mf1 Current1In1mf2 Current2

In1mf3 Current3

In1mf4 Current4

In1mf5 Current5

In1mf6 Current6

In1mf7 Current7

In1mf8 Current8

In1mf9 Current9

In1mf10 Current10

In2mf1 Speed1

In2mf2 Speed2

In2mf3 Speed3In2mf4 Speed4

In2mf5 Speed5

In2mf6 Speed6

In2mf7 Speed7

In2mf8 Speed8

In2mf9 Speed9

In2mf10 Speed10

Ou1mf1 Efficiency1

Ou1mf2 Efficiency2

Ou1mf3 Efficiency3

Ou1mf4 Efficiency4

Ou1mf5 Efficiency5

Ou1mf6 Efficiency6

Ou1mf7 Efficiency7

Ou1mf8 Efficiency8

Ou1mf9 Efficiency9

Ou1mf10 Efficiency10

Terms Definition

In1mf1 Current1

In1mf2 Current2

In1mf3 Current3

In1mf4 Current4

In2mf2 Speed1

In2mf3 Speed2

In2mf4 Speed3

In2mf5 Speed4

Ou1mf1 Torque1

Ou1mf2 Torque2

Ou1mf3 Torque3

Ou1mf4 Torque4

Ou1mf5 Torque5

Ou1mf6 Torque6

Ou1mf7 Torque7

Ou1mf8 Torque8

Ou1mf9 Torque9

Ou1mf10 Torque10

Ou1mf11 Torque11

Ou1mf12 Torque12

Ou1mf13 Torque13

Ou1mf14 Torque14

Ou1mf15 Torque15

Ou1mf16 Torque16

8/13/2019 1364631818

6/10



6

3(a)Input1 for Eff : Current

3(b)Input2 for efficiency: Speed

3(c) Output 1: Efficiency

3(d) Input1 for Torque: Current

8/13/2019 1364631818

7/10



7

3(e)Input2 for Torque:Speed

3(f) Output 2: Torque

Fig 3. Membership Functions

Procedure 3: After the above defined fis is generated according to our desired constraints then the

Optimum training method, the error tolerance and the no. of epochs for training are to be given. We haveused optimum training method as HYBRID, with an error tolerance of .,and the no. of epochs of,

for EFFICIENCY and optimum training method as HYBRID, with an error tolerance of .,and the no.

of epochs of, for TORQUE. After the above constraints were mentioned the network is trained till the

error tolerance is achieved.

3(g)structure for efficiency

8/13/2019 1364631818

8/10



8

3(h) structure for torque

Procedure 4: After the training is done and the required error tolerance is achieved the trained datais tested for the Average Training Error. The Average Training Error is obtained as 0.0030451 for the

TORQUE Controller and 0.016772 for the EFFICIENCY controller. Instead of using mathematical

formulae, a FLC so designed uses fuzzy rules to make a decision and generate the control effort. The rules

are in form of IF-THEN statements. The matters in defining rules are how many rules should be used andhow to determine the relation in IF-THEN statements. The solutions are based on the experience of a

designer or the previous knowledge of the system. The performance parameters can be improved bychanging the membership functions and rules as given in Table 4.

3(i) Tested input and output for efficiency

3(j) Tested input and output for torque

8/13/2019 1364631818

9/10



9

This logic is implemented in the evaluation of the DC Shunt motorIn this kind of motor, the field winding and the combinations is connected across the supply. Load

test on motors are performed to know about the efficiency, torquecharacteristics, which enable us to select

an appropriate motor for an application. In DC shunt motor as phase is also a constant, then torque isdirectly proportional to the armature current. As the load on the motor increases, the drop Ia.Ra though

increases is negligible as Ra is very small and the speed is nearly constant. If it is started on load it drawsheavy armature current which in turn will damage the machine itself. Hence they are always started on no-

load. The basic principle of this is that the back emf is directly proportional to speed.In shunt motor the field winding is connected in parallel with the armature. The field winding is

placed in parallel with the armature, it is called a shunt winding and the motor is called a shunt motor..

Notice that the field terminals are marked Fl and F2, and the armature terminals are marked Al andA2.

The shunt field coil is made of fine wire, it cannot produce the large current for starting like the

series field. This means that the shunt motor has very low starting torque, which requires that the shaft load

be rather small. When voltage is applied to the motor, the high resistance of the shunt coil keeps the overall

current flow low. The armature for the shunt motor will draw current to produce a magnetic field strongenough to cause the armature shaft and load to start turning.

The shunt motor's speed can be varied in two different ways. These include varying the amount of

current supplied to the shunt field and controlling the amount of current supplied to the armature.Controlling the current to the shunt field allows the rpm to be changed 10-20% when the motor is at full

rpm. The voltage across armature and the field winding is same equal to the supply voltage V.As long as

the supply voltage is kept constant the flux produced is constant. The various output parameters are

calculated manually and implemented through the fuzzy logic with the rule base algorithm and the requiredresult is obtained via the rule viewer.

5. EXPERIMENTAL RESULTS

In this section, an experiment is setup to demonstrate the performance of the NFLC. The controller

is tested on different current, efficiency and velocity performance both on load and no-load conditions. Theline voltage, frequency and armature current are constantly monitored. In DC Shunt motor Current, speed

are taken as the input and torque, efficiency are taken as the output. The functions used in this example are(In1, In2, In3etc.) the various values obtained in the load test were then loaded in the anfis editor and

proceeded for the controller construction. The membership functions used in the performance evaluation of

a DC shunt motor is gbellmf for the input and a linear output for the EFFICIENCY controller,

triangularmf for the input and a Constant output for the TORQUE controller. To maintain a set rpm athigh speed when the load changes can be done by controlling the shunt field and armature voltage. Since

the armature begins to produce back EMF as soon as it starts to rotate, it will use the back EMF to maintainits rpm at high speed. This means that if the motor is operated on less voltage than is shown on its data

plate rating, it will run at less than rpm.The Efficiency calculated by performing the experiment was 75%

approximately. When developed through the Fuzzy Logic, Efficiency resulted to only 73.85% in bothtriangular and gaussmf membership functions.

6. CONCLUSION

This paper has demonstrated the implementation of a NFLC for the performance evaluation of a

DC shunt motor.The NFLC is easy to implement and requires less cost.The critical point is if there is not

enough knowledge applied in the design,the result could be drastically bad.

8/13/2019 1364631818

10/10



10

7. REFERENCES

1. Fuzzy Logic Toolbox: Users Guide2. Fuzzy Logic Motor Control with MSP430*14*:Application Report3. Neuro-Fuzzy DC Motor SpeedControl

Documents

1364631818