An Innovative Approach for Condition Monitoring and Fault Diagnosis of Induction Motor by estimation of Rotor Time constant using

Extended Kalman Filter

Prof.P.Parthasaradhy, Suri Sathya Prashant, student IEEE member, S.Shyam Sunder, K.Satya Krishna Abstract:
75%of the electrical energy generated is utilized for running industrial and domestic motors. The manufacturers and consumers of these motors are now keen to include condition-monitoring equipment, which improves the safety and reliability of the equipment. Early detection of abnormalities in the motors will help to avoid expensive faults. In this paper, we consider a 400V, 50Hz squirrel cage Induction motor and estimate the Rotor resistance, inductance, stator current, rotor flux (along both, direct and quadrature axis) by means of extendable kalman filter, which takes into account measurement and modeling inaccuracies. Vas model of Induction motor deduced from the park transform is used for simulation and the Rotor inductance, resistances are considered as parameters as well as states. It has been observed that the parameter estimates (estimated values of stator currents, rotor flux, rotor rsistance, inductance) converges to the true values with a reasonable degree of accuracy. The method for analyzing internal faults based upon the estimated data is explained in the paper. Thus continuous monitoring of machine parameters by the above method can be very useful in detecting different internal faults occurring in Induction motor. Keywords: Fault diagnosis, Induction motor, Vas model, extended kalman Filter, (T =L /R ) is needed. However the electric parameters may
R R R

deviate from the original values due to changes in the working environment, temperature, speed, external load and noise. The inaccurate knowledge and parameter variation can deteriorate the achieved control and performance of the motor. Thus it becomes necessary to estimate the real time parameters. In this paper the value of T is estimated using
R

the EKF and the measurement of the stator currents and voltages. The ability of the extended kalman filter [3]-[7] to simultaneously estimate the states and the parameters of a dynamic system is well known. It allows for the real evaluation of the parameter uncertainties and errors due to the model simplification. A kalman filter estimates the states of a dynamic system with two different models namely the process model and the observation model. The dynamic model describes the behavior of the different dynamic states while the observation model establishes the relationship between the measurement and the state vector. Both models are associated with the statistical properties to describe the accuracy of the models. It uses the underlying process model to make an estimate of a system and corrects the estimate using available sensor measurements. Using this predictor corrector mechanism, it approximates an optimal estimate due to linearization of the process and measurement models. Section II deals with the induction motor model for the kalman filter algorithm. Section II shows the results of the estimation of states and parameters and finally section IV draws a conclusion and the future work envisaged.

I. Introduction:
Induction motors are widely used in industry since they are very economical, rugged and reliable, and are available in wide ranges of capacity. Unlike D.C motors, they have highly non-linear characteristic, requiring more complex control algorithms. When variable speed is required the D.C machines are the most suitable electromechanical device, as they have the advantage of precise speed control when utilized for the purpose of accurate driving. On the other hand induction motors are robust, simple, small in size, low in cost and almost maintenance free. The main obstacle in using induction motor drives is high cost of conversion equipment, the complexity of signal processing and poor precision in achieving effective speed and torque control. Speed control with a dynamic performance similar to that of a D.C motor can be achieved using field-oriented control(FOC), of the rotor flux, the knowledge of the machine parameter and in particular the rotor time constant

II. The Induction motor model for the Kalman filter

The model of the induction motor used is the Vas model deduced from the park transform [2]. This model considers
T

the stator voltages

T s s

U=[U

U ] as input and the stator

s T s s R r

currents Y=[I I ] as output. The stator currents and rotor flux are chosen as the state variables: X=[I I ] . The fourth order model in the mechanical frame is given by the following equations.

where X (k) ,U(k) and y(k) are the augmented state vector,
e

input vector and output vector. W (k) and V (k) are the
e e

system and measurement noises respectively. These noises are white with zero mean and characterized by The process covariance matrix Q(k) and measurement

covariance matrix R(k) are symmetric and semi-definite. The discrete kalman filter algorithm [9][10] is as follows. Prediction

Where is the state estimate and PeX is the estimation error

e

covariance and

R = stator resistance
S

R = rotor resistance
R

Updating .

L = stator inductance
S

L = rotor inductance
R

Sigma = motor disperson factor L = main inductance

H

K = R + ((1-sigma)L /T
R S S

K = sigma *L
L

, i - stator currents
s s

U ,U
s

- stator voltages SIMULATION AND RESULTS Simulation using Matlab has been carried out using the experimental data (stator currents and voltages) obtained for a 400 Volts, 50 Hz, 2.2 kw cage induction motor. The state
T

The discrete time versions are given by the following formulas.

variables are X=[I , I , , ] .The rotor resistance and

s s r r

where A , B , C represents the discretised versions of A, B

d d d

rotor inductance are taken as the parameters. The variations of the estimated parameters are accounted by tuning the parameter components of the covariance matrix Q .
e

and C.To estimate the parameters, the parameter vector is chosen as

Assuming there is no parameter coupling Q is taken as

e

diagonal. The first four components of W are related to the

e

Where N =1/LR To apply the EKF for simultaneous state

R

and parameter estimation, the state vector is augmented by the parameter vector. : Xe(k) = [X(k) (k)] which makes the system non-linear. The new dicretised state equations are

stator currents and rotor flux. The last two components characterize the parameters variations and permit to tune the dynamics of the parameters. The measured voltages are assumed to be uncorrelated and hence R is diagonal. Figures
k

1a and 1b demonstrates the convergence of estimated parameters. Figures 2a and 2b shows the estimated direct

axis and quadrature axis stator currents and figures 3a and 3b shows the estimated direct axis and quadrature axis rotor fluxTable 1 shows the estimated and true values of the parameters and the rotor time constant.

Table 1

Rotor resistance(R )
R

Rotor inductance(L )
R

Time constant

Estimated True

2.62 Ohms 2.72 Ohms

680 mH 718.81 mH

0.2595 sec 0.2642 sec

IV. CONCLUSION
In this paper we have estimated the stator currents, rotor flux, rotor resistance and rotor inductance of 400 Volts, 50 Hz, 2.2 KW squirrel cage induction motor using fourth order model of the induction motor and EKF. Simulation results show the ability of the EKF to track the rotor flux and the electrical parameters to a reasonable degree of accuracy. Different internal faults(e.g. partially broken rotor bars, partially broken end rings) can occur due to ageing and due to the environmental conditions under which the machine operate. These faults may cause gradual deterioration of the motor, which may lead to failure of the motor. The estimation procedure adapted in this paper can further be used for early detection of faults, by continuously monitoring the machine parameters. This will reduce the machine down time and enhance the reliability in critical application.

REFERENCES
1) C. DURIEU, L. LORON, E. SEDDA, I. ZEIN, Fault Detection of an Induction Motor by Setmembership 2) Filtering and Kalman Filtering, ECC99, September 1999, Karlsruhe. 3) P.Vas, Vector Control of AC Machines(1994), Oxford University Press, New York 4) M.A. Ouhrouche, (2003), EKF Based On-line Tuning of Rotor Time Constant in an Induction Motor Vector Control, International Journal of Power and Energy Systems, Vol.20(2), pp. 103-110 5) L. Salvatore, S.Stasi, and L.Tarchioni, (1997), A New EKF Based Algorithm for Flux Estimation in Induction Machines, IEEE Transactions on Industrial Electronics, Vol.40(5), pp. 496-504. 6) S.Wade, M.W. Dunnigan and B.W. Williams, (1997), Modeling and Simulation of Induction Machine Vector Control with Rotor resistance Identification, IEEE Transactions on Power Electronics, Vol.121(3), pp.495-506 7) L.C. Zai, C.L. Demacro, and T.A.Lipo, (1992), An Extended Kalman Filter Approach to Rotor Time Constant Measurement in PWM Induction Motor Drives , IEEE Transactions on Industry Applications, Vol.28(1), pp.96-104. 8) R.E.Kalman, R.S.Bucy, (1961), New Results in Linear Filtering and Prediction Theory, Journal of Basic Engineering, pp.95-108. 9) ATKINSON D.J., J.W. FINCH AND P.P. ACARNLEY, Estimation of rotor resistance in induction motors , IEE Proc. Electr. Power Appl., Vol 143, No 1, January 1996. 10) Shan-Shu Xiong, Zaho- Yug, (2003), Neural Filtering of Coloured Noise Based on Kalman Filter Structure, IEEE Transaction on Instrumentation and Measurement, Vol.52, pp. 742-747. 11) KIM Y.-R., S.-K.SUL AND M.-H. PARK, Speed sensorless vector control of induction motor using extended Kalman filter , IEEE Trans. Ind. Appl., Vol. 30, No 5, September/October 1994.

AUTHOR INFORMATION
[1] Prof. P. PARTHASARADHY, Is the Dean - Student Activities, Head of the Department EEE at Guru Nanak Engineering College, JNT University. He pursued his Masters degree in Industrial Drives & Control and is currently pursuing his Ph.D in Power Electronics he has with him more than two decades of teaching and industrial experience to his credit. He underwent special training programs at MAILLEFER, Switzerland.. He has guided several Hardware projects for B.Tech. and M.Tech. Students.

[2]S U R I S AT Y A P R A S H A N T:
Is pursuing his Bachelors degree in Electrical and Electronics Engineering. from Guru Nanak Engg. College. He passed the HAM Radio exam conducted by ministry of communications, Government of India. He is a Summer Research Fellow- Indian Academy of Sciences. He has 7 International and 8 National level papers to his credit. He is currently serving M-CRA 2007, Jordan as a TPC member, Reviewer. His areas of interests are medical electronics, power transformers, control systems. He is presently, working on two R&D projects in medical electronics, control systems.

[3]SISTLA SHYAM SUNDER

Is pursuing his Bachelors degree in Electrical and Electronics Engineering. from Guru Nanak Engg. College He has 5 International and 2 National level papers to his credit. His areas of interests are medical electronics, mechatronics, scada and network security. He is currently indulged in an R & D project on medical electronics.

[4] K.SATYA KRISHNA

Is currently pursuing his bachelors degree in electrical and electronics engineering. From Guru Nanak Engineering College. He has 2 International and 2 National Papers to his name. His areas of interests are Neural Networks and Fuzzy logic.