Fuzzy Logic based fault detection in induction machines using Lab view

Transcription

1 226 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 Fuzzy ogic based fault detection in induction machines using ab view R.SaavanaKuma K.Vinoth Kuma D. K.K.Ray VIT Univesity,velloe,INDIA, Kaunya Univesity, Coimbatoe, INDIA, VIT Univesity,velloe,INDIA Summay The main objectives of this pape ae to pefom fault analysis on an induction moto using both expeiments and simulation, and to study failue identification techniques applied fo condition monitoing of the moto and finally to design an On-line condition monitoing system with fuzzy logic contolle using abview. The wok in this pape was divided into two phases. The phase one was concened with the modeling of thee-phase induction moto in thee phase efeence fame using Matlab/Simulink and designing an intelligent system fo condition monitoing of the motos. The phase two concened with the implementation of On-line condition monitoing system though abview. Key wods: Condition monitoing, induction moto, ab view, fuzzy logic.. Intoduction The thee phase Induction motos due to thei simple constuction, high eliability and low cost, have dominated in the field of electomechanical enegy convesion by having moe than 75% of motos in use. Though the pobability of beakdowns of Induction motos is vey low, the fault diagnosis has become almost indispensable. Paticulaly when they ae woking in sophisticated automated poduction lines. To decease the machine down time and impove stability on line diagnostic featues ae to be necessaily incopoated with the dives. In moden industy lots of machines depend on mutual opeation, and the cost of unexpected beakdowns is vey high. Thus condition monitoing techniques compising of fault diagnosis and pognosis ae of geat concen in industy and ae gaining inceasing attention. A thee-phase Induction motos may expeience seveal types of fault conditions which include Ove load, Gound fault, ine to ine fault, Unbalanced supply voltage, Ove voltage, Unde voltage, Single phasing, Tun to Tun fault. When one of the thee-phase to the moto is open, single phasing situation occus. This esult inceased positive and negative sequence cuents and hence excessive heating. Unbalance supply voltage esults in negative sequence voltage. It leads to incease in positive and negative sequence cuent components. Gound and line faults ae pevalent in motos than othe powe system devices. These faults ae detected by obseving the zeo sequence cuent. Similaly Tun to Tun shot and coil open faults can cause cuent unbalance. Fom the foegoing analysis it is clea that the appeaance of vaious faults is simply detemined by the stato cuent values. In geneal stato cuents and voltages ae pefeed because the sensos equied ae usually pesent in the dive consideed. The conventional monitoing systems have a numbe of limitations such as inflexible, high cost, hadwae limitations which ae heavily dependent upon specialized instuments. In ecent yeas the monitoing and fault detection of electical machines have moved fom taditional techniques to Atificial intelligence and vitual instumentation techniques. But the Vitual instumentation appoach helps us to get id of the above said dawbacks. In addition to this, additional scientific visualizations and advanced analysis can also be added in the fom of vitual instuments with minimal cost. Because of its ove all vesatility as an engineeing tool, the softwae package ab View is chosen in most of the engineeing poblems. In this pape Condition monitoing system fo Induction moto has been developed in both simulation model (Matlab/Simulink) as well as in eal time (abview). Fo the past fifteen yeas thee has been a substantial amount of eseach into the ceation of new Condition Monitoing techniques fo Induction Moto dives. New methods have been developed which ae now being used in industies and eseach is continuing with the development of new and altenative on-line diagnostic techniques. Howeve it is still the end uses who have to make the selection of most appopiate and effective monitoing system to suit thei paticula Induction Moto dive systems. The possible detection methods to identify the moto faults ae listed as follows.. Vibation Monitoing. 2. Moto Cuent Signatue Analysis (MCSA).. Electomagnetic field monitoing using seach coils. 4. Chemical analysis (ubicating oil debis, Cooling gas). Manuscipt eceived Septembe 5, 2009 Manuscipt evised Septembe 20, 2009

2 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe Tempeatue measuement. 6. Infaed measuement. 7. Acoustic noise measuement. 8. Radio fequency emission monitoing. 9. Patial Dischage measuement. The Fig. 2. pesents an oveview of poblems and possible On-line monitoing techniques. The focus has been on the use of On-line condition monitoing to detect degadation pocesses and failue mechanisms. It is necessay to teat each induction moto dive as a unique entity and the potential failue modes fundamental causes, mechanical load chaacteistics, and opeation conditions have all to be taken into consideation when a condition monitoing system is being selected. MCSA, VIBRATION AND AXIA FUX ANAYSIS VIBRATIO N ANAYSIS STATOR 2. ROTOR MECHANICA PROBEMS BEARING WEAR MISAIGNME NT Failue of insulation between winding and gound can cause lage gound cuent, which would esult in ievesible damage to the coe of the machine. If the fault is detected at an ealy stage, the machine can be put back into sevice by just ewinding the stato on the othe hand, eplacing the moto means inceased downtime. Typical potection fo stato insulation failue in an industial envionment includes the use of gound fault elays and negative sequence o phase cuent balance elays would cause nonsense tips with unbalanced line voltages. Gound fault elays would not be effective fo ealy fault detection. Senso less methods fo detection of tun-faults ae based on monitoing the line cuents and voltages fo an electical asymmety caused by tun-tun shot cicuits in the winding. Fo line connected machines, the negative sequence cuent is the fundamental fault signatue fo a winding fault. R Tun to Tun Shot STATOR Open coil BROKEN ROTOR BARS MCSA AND VIBRATION ANAYSIS. OW/MEDIUM VOTAGE 4.kv AND BEOW STATOR WINDING PROBEMS ine to ine Fault ine Gound to.. INSUATION BREAKDOWN SHORTED TURNS SHORTED COIS OPEN COI INE TO GROUND SHORT Fig.. Poblems, Failues and Possible on-line Monitoing Techniques fo Induction Moto Dives B Fig. 2. Thee phase Stato Winding Y

3 228 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 The Figue 2shows the thee phase stato winding with diffeent faults. Tun fault detection based on the axial leakage component flux and electically excited vibations have also been existing, howeve, the sequence components appoach povides a bette signal to noise atio, is easie to measue and does not equie any special instumentation besides cuent sensos. The sensitivity of sequence components based tun-fault detection schemes depends on the extent to which the effects of non-idealities ae compensated. The negative sequence cuent is affected by unbalance in powe supply and inheent asymmeties in the machine and in instumentation. Moto Cuent Signatue Analysis can detect the fault befoe a phase-phase o phase-eath failue. It is theefoe possible with a low voltage stato winding to have some lead time between shoted tuns developing and actual failue. Method Vibatio n MCSA Axial flux ubicat ing oil debis Cooling gas Patial Discha ge Table. Compaison of diffeent fault detection technologies Insulati on Stato winding Faults can be detected Roto Windin g Roto Eccenti city Beaing Damage No No Yes Yes Yes No Yes Yes Yes Yes No Yes Yes Yes No No No No No Yes Yes Yea Yes No No Yes No No No No The possible detection methods to identify the moto faults ae Vibation, Moto Cuent Signatue Analysis (MCSA), Axial flux, ubicating oil debis, Cooling gas, and Patial Dischage. A compaative analysis of diffeent monitoing techniques is listed in Table.. Fom this MCSA is the most appopiate technique which can be used fo both stato as well as oto fault detection. ine cuent can be monitoed non-invasively via clip-on cuent tansduces hence no inteuption to the machine opeation is equied, and the system can be made vesatile and potable. MCSA has shown to be capable of detecting, diffeentiating, and tacking the machine abnomalities, it has numbe of inheent stengths, the most notable ae. Povides nonintusive monitoing capability at a location emote fom the equipment. 2. Povides degadation and diagnostic infomation.. Offes high sensitivity to a vaiety of disodes affecting opeational eadiness 4. Can be pefomed apidly and as fequently as desied by elatively unskilled pesonnel using potable, inexpensive equipment. 5. Is equally applicable to high-poweed and factional hosepowe machines, ac and dc motos. MCSA can be implemented using time domain o fequency domain analyses and is also capable of analyzing systems duing tansients, such as duing initial o final opeation of the system. 2. Modeling and simulation of thee phase induction moto fo vaious fault conditions 2.. Intoduction A theoy is often a geneal statement of pinciple abstacted fom obsevation and a model is a epesentation of a theoy that can be used fo pediction and contol. To be useful, a model must be ealistic and yet simple to undestand and easy to manipulate. These ae conflicting equiements, fo ealistic models ae seldom simple and simple models ae seldom ealistic. Often, the scope of a model is defined by what is consideed elevant. Featues o behavios that ae petinent must be included in the model and those that ae can not be ignoed. Modeling hee efes to the pocess of analysis and synthesis to aive at a suitable mathematical desciption that encomposes the elevant dynamic chaacteistics of the component, pefeably in tems of paametes that can be easily detemined in pactice. In mathematical modeling, we ty to establish functional elationships between entities that ae impotant. A model supposedly imitates o epoduces cetain essential chaacteistics o conditions of the actual often on a diffeent scale. It can take on vaious foms: physical, as in scale-models and electical analogs of mechanical systems; mental, as in heuistic o intuitive knowledge; and symbolic, as in mathematical, linguistical, gaphical, and schematical epesentations. Simulation can be vey useful in many scientific studies that poceed as follows:. Obseving the physical system. 2. Fomulating a hypothesis o mathematical model to explain the obsevation.. Pedicting the behavio of the system fom solutions o popeties of the mathematical model. 4. Testing the validity of the hypothesis o mathematical model.

4 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe Evey model has paametes, which ae to be estimated. The model must lend itself to methods by which these paametes can be detemined expeimentally; othewise the model will be incomplete. The developed model must be veified and validated. Veification involves checking the consistency of the mathematics, the solution pocedue, and the undelying assumptions. Validation is the detemination of how adequately the model eflects petinent aspects of the actual system that ae epesented by it. When a discepancy in unacceptably lage, the model must be evised and the cycle epeated. Modeling and simulation have appopiate uses. They ae especially beneficial in situations whee the actual system does not exist o is too expensive, time-consuming, o hazadous to build, o when expeimenting with an actual system can be cause unacceptable disuptions. Changing the values of the paametes, o exploing a new concept of opeating stategy, can often be done moe quickly in a simulation than by conducting a seies of expeimental studies on an actual system Simulation can also be a vey useful taining aid; it is a technique by which students can lean moe and gain geate insight and bette undestanding about the system they ae studying. A fequent question about simulation is its validity. Do the simulation esults eflect those of the actual system fo the condition simulated? Even with valid component models, the use of them in a lage simulation must be done caefully with consistency; othewise, the esults could be meaningless. Finally, in intepeting the esults, we should not ovelook the simplifications and assumptions made in devising the model. In this chapte, the thee-phase induction moto model has been deived instead of two phase model (d-q epesentation), which is vey commonly used. This is because the two-phase model is diven unde balance opeation. This model constained the diven solution fo any poblem to balance opeation, which seveely limits the scope of many pactical poblems that may be studied. Also, it impais the accuacy of these solutions since absolute balanced conditions aely apply. Thee-phase epesentation of all system components has been chosen so as to allow fo unsymmetical opeation and to enhance the pactical significance of the deived model. Such a epesentation is able to epesent in accuate detail the many unsymmetical phenomena encounteed, which is not possible when using d-q modeling methods. So, in ode to gain an undestanding of poblems that does not comply diectly with balanced opeating conditions and fo which d-q methods of analysis ae inappopiate, theephase epesentation becomes essential. 2.2 Induction Machine Equations The figue. shows the two dimensional diagam of thee-phase induction moto with stato and oto windings. Fig.. Thee phase Induction moto The voltage equations fo a thee-phase induction machine can be expessed as: Stato equation is dλ A VA RAiA + dt dλb VB RBiB + dt dλc VC RCiC + () dt Roto equation is dλa Va Raia + dt dλb Vb Rbib + dt dλc Vc Rcic + (2) dt The flux linkages associated with the inteactions between stato and oto windings ae epesented by: Stato: λa AAiA + ABiB + ACiC + Aa cos( θ ) ia + Ab cos( θ + ) ib + Ac cos( θ ) ic

5 20 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 λ B + Bb C + i BA A λ + cos( θ ) i Cb i CA A b + i BB B + CB B + BCiC + Ba cos( θ ) i cos( θ + ) ic Bc + CC C cos( θ ) ib + i i Cc + Ca cos( θ ) i cos( θ + ) i Roto: λa aa cos( θ ) ia + ab cos( θ + ) ib + ac cos( θ ) ic + Aaia + Abib + i λb ba cos( θ + ) ia + ab cos( θ ) ib + ac cos( θ ) ic + baia + bbib + bcic λc ba cos( θ ) ia + cb cos( θ + ) ib + cos( θ ) i + i + i + i cc C ca a cb b c cc c The electomechanical toque equation is: a Ac c i a[ Aa + aa]sin( θ) + ib[ Ab + ba] i A sin( θ + ) + ic[ Ac + ca]sin( θ ) ia[ Ba + ab]sin( θ ) + ib[ Bb + bb] (5) T e + i 2 B sin( θ ) + [ + ]sin( + ) ic Bc cb θ ia[ Ca + ac]sin( θ + ) + ib[ Cb + bc] + ic sin( θ ) + [ + ]sin( ) ic Cc cc θ a () (4) The dynamic load equation is: T e dω T J + Dω (6) dt dω Te T (7) dt J ω ( Te T ) dt (8) J 2.. Induction Machine model in a thee-phase efeence fame When induction machines ae expessed in thee-phase axes, many of the inductances ae function of the oto displacement and theefoe functions of oto speed and time as shown in the following Stato Inductances: It is assumed that the ai gap of the induction machine is unifom and the stato and oto windings ae sinusoidally distibuted, all the stato self-inductances ae identical. + (9) AA BB CC ls The mutual inductance between any two stato windings is the same due to symmety 0.5 AB BC BA CB ms 0.5 ms CA AC 0.5 ms Roto Inductances: In the same manne to that given fo the stato, the oto self-inductances and mutual inductances ae: + aa ab bc ca Aa Ac Ab bb Bb Ba Bc ba cb ac cc l Cc Cb Ca ms ms ms m m m m ms cosθ cos( θ 20 ) cos( θ + 20 ) (0) () (2) The mutual inductance between a stato winding and any oto winding vaies sinusoidally with oto position.

6 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe Induction Moto Model implementation with Matlab/Simulink In this section, the implementation of the stationay efeence abc model of a thee-phase induction moto using simulink, using the equations listed in the pevious section has been given. Figue 4.shows an oveall diagam of the induction moto in the stationay thee-phase efeence fame. The details of the subsystems in the main blocks ae given in figue 5. Fig. 5. Simulink Model of Induction Moto Subsystem Fig. 4 Simulink Model of Condition Monitoing System The paametes inside the induction moto thee-phase model and the thee phase souce can be set by executing a m-file which stoes the all paametes used in the model. By unning the m-file all the values of the paametes can be accessed by the model fom the wokspace. Machine Paametes: The paametes of the machine used fo simulation ae listed below: Rated Voltage V 80v Fequency f50hz, Stato Resistance R stato 5.Ω Roto Resistance R oto 7.46Ω, The stato and oto self-inductances ae equal to stato oto leakage + mutual h The mutual inductance between any two stato and any tow oto windings is equal to ss,mutual,mutual -0.5 mutual H The mutual inductance between a stato winding and any oto winding is equal to s,mutual mutual 0.55H Numbe of Poles p 4, Inetial constant J 0.02kg.m 2 Fig. 6. Model of Symmetical components subsystem 2.5. Design of Condition monitoing of Induction Moto using Fuzzy ogic Contolle Condition monitoing of the induction moto is a pocess of continuous evaluation of the health of the moto thoughout its seviceable life. Condition monitoing and potection ae obviously closely elated functions.

7 22 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 Condition monitoing should be designed so as to peempt faults that can occu in the induction moto. It can be extended to povide pimay potection, but its eal function must always be to attempt and to ecognize the development of faults at an ealy stage. Such advanced waning is desiable since it allows maintenance peson geate feedom to schedule outages in the most convenient mannes. By condition monitoing we can educe unexpected failues and downtime, incease the time fo standads maintenance, and educe maintenance and opeational cost Oveview of Fuzzy ogic Fuzzy ogic has emeged as a pofitable tool fo the contolling of subway systems and complex industial pocesses, as well as fo household and entetainment electonics, diagnosis systems and othe expet systems. Fuzzy logic is basically a multivalued logic that allows intemediate values to be defined between conventional evaluations like yes/no, tue/false, etc. Notions like athe wam o petty cold can be fomulated mathematically and pocessed by computes. In this way an attempt is made to apply a moe human-like way of thinking in the pogamming of computes. Fuzzy contolles ae the most impotant application of fuzzy theoy. They wok athe diffeent than conventional contolles; expet knowledge is used instead of diffeential equations to descibe a system Constuction of Fuzzy and Membeship functions Fuzzy ules and membeship functions ae constucted by obseving the data set. Fo the measuements elated to the stato cuents, moe insight into the data ae needed, so membeship functions will be geneated fo Negative Medium(NM), Negative Small(NS), Zeo(Z), Positive Small(PS), Positive Medium(PM). Fo the measuement elated to the stato condition, it is only necessay to know if the moto condition is Good, o o Seiously. Optimized membeship functions fo this ae shown in figue 6 and figue 7. The stato elated faults that have been investigated in this study ae. Tun-to Tun shot, 2. Beak in stato winding. Unbalance in input Voltage 4. Open phase The steps in building Condition Monitoing system ae:. Detemine the fuzzy contol input. The cuent is the input to the fuzzy contolle. 2. Detemine the fuzzy contol output. The output is the condition of the moto.. Choose the wod desciptions fo the status of input and output. Input Status Wod Desciptions. Negative Medium(NM) 2. Negative Small(NS). Zeo(Z) 4. Positive Small(PS) 5. Positive Medium(PM) Output Status Wod Desciptions. Good 2.. Seiously The next step is to detemine the degee of membeship fo both input and output vaiables. This can be done by knowledge base only.(the membeship function values fo input and output vaiables and Rule base ae given in chapte 4). Implementation of On ine condition monitoing system. Intoduction The thee phase induction motos due to thei simple, obust design and constuction, have been widely used in industies. In spite of the fact that the pobability of beakdowns of induction moto is vey low, 2 to % pe annum, monitoing thei opeating condition is essential paticulaly when they ae employed in automated poduction lines. The induction motos ae subject to the occuence of incipient faults. If these incipient faults ae left undetected, the occuence of an incipient fault will eventually degeneate into a machine failue. In many industies lots of machines depend on mutual opeation, and the cost of unexpected beakdowns is vey high. In such cases the beakdowns of machines in opeation usually involve highe losses in poduction pocess than the cost of thei epais o even the initial costs of the machine themselves. In addition, the unexpected beakdowns may have safety implications and may even expose human being to dange. Both technical and economical consideations suggested the development of a new compute-based monitoing system to pedict the faults by means of acquiing the data and the analysis of that data. An efficient Condition Monitoing system can educe manpowe consumption as well as guaantee that the unning will neve halt accidentally. It should be capable of monitoing the unning machines with the existence of electical intefeence, pediction the need fo maintenance befoe seious detoiation o beakdown occus, identifying and locating the defects in detail, and even estimating the life of machines.

8 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe Oveview of Vitual Instumentation Vitual Instumentation is defined as combining hadwae and softwae with industy-standad compute technologies to ceate use defined instumentation solutions. abview is a gaphical pogamming language that uses icons instead of text to ceate applications. Using abview we can ceate test and measuement, data acquisition, instument contol, data logging, measuement analysis, and epot geneation applications. Thee Phase Cuent Values Signal Analysis Thee Phase Voltage Values FFT Hamonic Distotion Machine Paamet Powe Flow Symmetical Components Fig. 7. Schematic diagam of a Vitual Instument In geneal abview pogams ae called Vitual Instuments (VI) it is shown in figue 7. VI's contain two main components, the font panel, and the block diagam. The font panel is the use inteface of the VI. The block diagam contains the gaphical souce code. abview is integated fully fo communication with hadwae such as GPIB, RS-22, RS-485, and Plug-in data acquisition boads. It also has built in libaies fo using softwae standads such TCP/IP netwoking and Activex. It is a poweful gaphical development envionment fo signal acquisition, measuement analysis, and data pesentation, without the complexity of taditional development tools. Seiously Fuzzy Health Evalua tion Dam og Data Isolate Moto STOP Y o g Da og Data G N og Data Wan ing. Development of ab View based Monitoing system The development of compute technologies, tansduce technologies, signal pocessing techniques togethe with atificial-intelligence(ai) techniques has made it possible to implement Condition Monitoing (CM) moe efficiently on any electical equipment. It is expected to make CM systems moe eliable, moe intelligent and cheape, so that they can be widely employed even in small scale industies. CM is a technique seved fo Condition-Based Monitoing (CBM). Ealie to this time-based maintenance had been the mainly used maintenance stategy fo a long time. Fig. 8. Functionality of the VI system designed Time-based maintenance, to examine and epai the machines offline eithe accoding to a time schedule o unning hous, may pevent many failues. Howeve, it may also cause many unnecessay shutdowns and unexpected accident will still occu in the intevals. Manpowe and time and money wee wasted because the activity of maintenance with little infomation of the cuent condition of the machines. But the CBM will let opeatos know moe about the state of machines and indicate clealy when and what maintenance is needed.

9 24 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 In a abview pogam the font panel simulates the panel of a physical instument.the esults can be viewed on the compute sceen though the font panel. Actually the VI eceives instuctions fom a block diagam constucted in gaphical souce code. The block diagam is a pictoial solution to a pogamming poblem. The figue 8. shows the functionality of the VI system fo the Condition Monitoing of induction moto. Hee the input quantities ae the thee phase input voltage and the thee phase stato cuent signals. The acquied signals ae pocessed fo diagnostic pupose though Fast Fouie Tansfom (FFT), Total Hamonic Distotion (THD), and symmetical components analysis. The powe flow calculations ae done using the acquied signals and the paametes of the specific moto. A fuzzy logic contolle has been designed to evaluate the state of the moto based on the stato cuent magnitude. Based on the output of the fuzzy contolle potective actions ae caied out. The total Instumentation system incopoates. Data Acquisition. 2. Signal Pocessing. Fuzzy Contol.4 Data Acquisition A data acquisition system in the boad sense could mean any method used fo obtaining o acquiing any type of data, howeve, a data acquisition system (DAQ) usually denotes a goup of electonic devices that ae connected to pefom the measuement and quantization of analog electical signals fo digital pocessing. The DAQ is the inteface between analog and digital electical signals. Although most physical attibutes ae analog in natue, many ae not electical and must be conveted to the electical wold using sensos o tansduces. The figue 9. illustates a typical DAQ system with a minicompute. Key pats of the system include: Analog Multiplexe This function allows the selection of any one of seveal analog signals. Signal Conditione This function usually follows the multiplexe, but could be used on any o all input signal lines, to pepae the signals fo convesion. The conditioning equied might include one o moe of the following: inea o logaithmic amplification, filteing, peak detection, sample-and hold, o tack-and-hold. This pat of the DAQ system will incease the signal to noise atio level of the input signal. Analog-to-Digital (A/D Convete) The A/D convete tanslates the analog signal to a digital fomat. I/O Contolle The I/O contolle geneates the system timing and contols the memoy READ/WRITE functions. Minicompute An opeational compute system based on a micopocesso chip. It contains memoy and is contolled by softwae. Output Buffe The output buffe combines the data with the pope signal fomat fo output to the specific peipheal devices. The output buffe is also called a peipheal contolle. Peipheal Diffeent micocompute type output devices such as a line pinte, floppy-disk, o magnetic tape stoage unit. Fo a DAQ system to function, the input signal must be sampled. Sampling may be defined as measuing a continuous function at discete time intevals. These sampled signals epesent some analog paamete conveted to a seies of discete values. In a DAQ system the quantize comes next to the sample. Quantization is the pocess of appoximating the linea cuve by a seies of stai-step values called levels. Afte quantize Digital pocessing takes place, it is evaluating the linea function as a seies of discete values epesented by ones and zeos. The sampling theoem states that signals with a finite bandwidth of f-hetz can be descibed by sampling the signals at instants sepaated by T/2f seconds. This means the signal must be sampled at a ate at least twice the fequency of the highest fequency in the spectum being analyzed. National Instuments E-seies, 6024-E DAQ cad has been used in this system. The compute eceives aw data. Softwae takes the aw data and pesents it in a fom the use can undestand. Softwae manipulates the data so it can appea in a gaph o chat o in a file fo epot. The softwae also contols the DAQ system, telling the DAQ device when to acquie data, as well as fom which channels to acquie data. Cuent Voltage Speed Ana log Mul tiple Sign al Fig. 9. Typical Data Acquisition System When measuing analog signals it is necessay to conside the mode of input (efeenced input o diffeential input), Resolution, Range, sampling ate, Accuacy and Noise. In Geneal it is good to use diffeential inputs to avoid noise eos. The esolution coesponds to the numbe of bits that the ADC uses to epesent the input analog signal. The highe the esolution, highe will be the accuacy in epesentation of analog signal. The ange efes to the A / D C O N V E R T E R I/O Co nto l Mini Comp ute PERIPERA ab View

10 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe maximum and minimum voltage levels that ADC can quantize then sampling ate detemines how often an analog to digital convesion takes place. Signal conditioning is necessay to shield the signals fom noise and distotion. The actual cuent measued has to be stepped down and isolated to potect the DAQ boad. The most common type of signal conditioning is amplification, cuent signals ae amplified to impove the digital signal accuacy and to educe noise..5 Signal Pocessing The signals to be analyzed ae acquied though signal conditioning cicuits and DAQ cad. This is shown in Fig. 0. Afte acquisition the signal is bought to the actual magnitude by suitable multiplication facto. epesentation could give moe insight about the signal and the system. The algoithm used to tansfom samples of the data fom the time domain into the fequency domain is known as the Fast Fouie Tansfom (FFT). The FFT establishes the elationship between the samples of a signal in the time domain and thei epesentation in the fequency domain. The FFT is widely used in the fields of spectal analysis. If the signal is sampled at a sampling ate of f s Hz, then the time inteval between the samples is Δt, (e.g., see Eq. ) whee Δ t () f s The Fast Fouie Tansfom is given by X k N i 0 x i e k j 2πi N fo k0,,2, N- (4) Whee X[i] is the fequency domain epesentation. Fig 0. Schematic diagam of the actual setup The signal pocessing has been divided into two pats.. Spectum analysis 2. Symmetical component analysis.5.. Spectum analysis The samples of the signal obtained fom the DAQ cad constitute the time domain epesentation of the signal. This epesentation gives the amplitudes of the signal at the instants of time duing which it had been sampled. Howeve, in the fault diagnosis of induction moto, it is equied to know the fequency content of the cuent signal athe than the amplitudes of the individual samples. The epesentation of the signal in tems of its individual fequency components is known as the fequency domain epesentation of the signal. The fequency domain Δ f f N s N Δ t (5) The time domain 'x' and the fequency domain 'X' have a total of N samples. If the FFT is complex, it contains two pieces of infomation-the amplitude and phase. n f st f ( s) ± k p (6) The objective of pefoming FFT analysis on the cuent signal is to identify cuent components in the stato winding that ae function of shoted tuns. The following equation gives the components in the aigap flux wavefom that ae a function of shoted tuns. The diagnosis of shoted tuns via MCSA is based on detecting the fequency components given by equation 6 in that these otating flux waves can induce coesponding

11 26 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 cuent components in the stato winding. The components ae clealy a function of moto design, in a fou pole, thee phase, squiel cage induction moto, fom equation (6) thee will be a whole seies of components, 2 s With k, n, f st f, and, ( + s) With ( ) k, n 2, f st f sf f 2 2 s, and, 5 s f With 2 k, n 4, f st f s, and, f( 2s) With 7 5s 5s k, n 5, fst f, and, f With k, n 6, f st 4 f s, and, f ( 2 s) 4 With k, n 7, fst f s, and, f s (7) With k, n, f st f, and, ( s) and so on with combinations of k and n values. The figue 4..2 shows the fequency spectum of the induction moto unde diffeent fault conditions Symmetical Component Analysis The ocuence of shoted tuns and phase imbalanced in the stato winding of induction motos cause pedictable hamonics to appea in the spectum of line cuents and in the axial leakage flux. These changes howeve, can be small elative to the fundamental. The situation is futhe complicated because the induced hamonic components can exhibit beating at slip fequency o at a low intege multiple of slip fequency. This means that the detection of changes in the amplitude of such components can be uneliable, since it both fequency and time dependant. To ovecome these dawbacks, the symmetical component theoy, and the appoach is quite geneal, allowing fo any asymmetical aangement of stato winding faults. By analyzing the elation of the positive and negativesequence cuents to that of toque and slip, when the machine is subjected to vaious opeating conditions, the faults can be diagnosed. An unbalanced system of elated phasos can be esolved into n systems of balanced phasos called the symmetical components of the oiginal phasos. In symmetically wound machines thee is no eaction between the diffeent sequency quantities, if balanced voltages ae applied to the stato, only balanced cuents of the same sequence will flow. Due to the mutual winding coupling and othe effects, the impedance displayed by machines will be diffeent when exited by diffeent sequence souces. The pe phase impedance of the positive-sequence netwok will be diffeent fom those of negative and zeo-sequence netwoks. The expession fo the thee unbalanced phasos as a function of the balanced phaso components ae (8) The positive, negative, and zeo sequence vecto components of any phase always have the angula elationship with espect to one anothe. V V V a 0 a a 2 a a 2 V a 2 a Vb a Vc (9) (20) (2) If suppose C phase is lost and the moto is unning at a steady speed. Then Va and Vb ae the emaining voltages and they poduce equal and opposite line cuents Ia and Ib, which incease to aound.7 to 2.00 times that of the nomal. The back EMF geneated in the machine amatue will be appoximately balanced in all thee phases, since both the moto stato windings and oto cage have distibuted conductos. By ignoing stato esistance, the intenal voltage Vma and Vmb of phase A and B espectively will add vectoially with the espective stato leakage eactance voltage dops Vsa and Vsb to equal the applied voltages Va and Vb. Fo ungounded neutal, zeo-sequence cuent is neglected. The negativesequence voltages poduce cuents which ae limited by impedanced that closely appoximate those that apply when the moto is stated. This negative-sequence cuent esults in the poduction of counte toque. To compensate negative toque poduced, the positivesequence cuent inceases. The moto slip inceases to allow the additional positive toque to develop.

12 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe The simila appoach can be applied fo the tun faults too. The symmetical component wavefoms of the input voltage and stato cuents fo diffeent fault conditions ae shown in figue 8..6 Fuzzy Contol In the moto fault diagnosis pocess, time domain cuent signals ae captued fom sensos. The diagnostic expet then uses both time domain and fequency domain signals to study the moto condition and detemines what faults ae pesent. Howeve, expeienced enginees ae often equied to intepet measuement data that ae fequently inconclusive. A fuzzy logic appoach may help to diagnose induction moto faults. Fuzzy logic is eminiscent of human thinking pocess and natual language enabling decisions to be made based on vague infomation. Fuzzy logic is shown in figue. it allows items to be descibed as having a cetain membeship degee in a set. While conducting fault diagnosis, thee ae seveal situations in which an object is not obviously "Good" o "", but may fall into some inteio ange. Accoding to the fact that induction moto condition intepetation is a fuzzy concept. Hee the moto condition is descibed using linguistic vaiables. Fuzzy subsets and the coesponding membeship functions descibe stato cuent amplitudes. A knowledge base, compising ule base is built to suppot the fuzzy infeence. The induction moto condition is diagnosed using a compositional ule of fuzzy infeence. The obtained esults indicate that the poposed fuzzy logic appoach is capable of highly accuate diagnosis. humans expess knowledge. (like an electical machine efeed as "somewhat secue", "little oveloaded"). This linguistic input can be expessed diectly by a fuzzy system. The intenal stuctue of fuzzy contolle is shown in figue 2. The stato cuent signal contains potential fault infomation. Fuzzy systems ely on a set of ules. These ules, allow the input to be fuzzy, i.e. like the natual way that I a { μ ia (I aj ) Σ I a } I b { μ ib (I bj ) Σ I b } I c { μ ic (I cj ) Σ I c } CM { μ cm (cm j ) Σ CM } (22) Whee I aj, I bj, I cj and cm j ae elements of the discete univese of discouse I a, I b, I c and CM.the optimized ule base has been developed so as to cove all the healthy and the faulty conditions of the moto. Fig.. Complete stuctue of Fuzzy Contolle Fig. 2. Intenal stuctue of Fuzzy Contolle In this case the stato cuent Ia, Ib, and Ic ae consideed input vaiables to the fuzzy system. The stato condition, CM is chosen as output vaiable. All the system inputs and outputs ae defined using fuzzy set theoy. The input vaiables ae intepeted as linguistic vaiables, with Negative Medium(NM), Negative Small(NS), Zeo(Z), Positive Small(PS), Positive Medium(PM). Similaly the output vaiable stato condition (CM) is intepeted as linguistic vaiables, with Good,, and Seiously. Fuzzy ules and membeship functions ae constucted by obseving the data set. Thee ae if-then ules used and the membeship function fo input and ouput vaiables ae shown in figue and figue 4 espectively.

13 28 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 μ(cm) Fig.. Fuzzy membeship functions fo stato cuents Fig.4. Fuzzy membeship functions fo Condition of Moto CM CM.6.. Rule Base. If Ia is NS then CM is Seiously 2. If Ib is NS then CM is Seiously. If Ic is NS then CM is Seiously 4. If Ia is PM then CM is Seiously 5. If Ib is PM then CM is Seiously 6. If Ic is PM then CM is Seiously 7. If Ia is NM and Ib is NM and Ic is NM then CM is Good 8. If Ia is NM and Ib is NM and Ic is Zeo then CM is. 9. If Ia is NM and Ib is NM and Ic is PS then CM is 0. If Ia is NM and Ib is Zeo and Ic is NM then CM is. If Ia is NM and Ib is PS and Ic is NM then CM is 2. If Ia is Zeo and Ib is NM and Ic is NM then CM is. If Ia is PS and Ib is NM and Ic is NM then CM is 4. If Ia is NM and Ib is Zeo and Ic is Zeo then CM is 5. If Ia is NM and Ib is Zeo and Ic is PS then CM is Seiously 6. If Ia is NM and Ib is PS and Ic is Zeo then CM is Seiously 7. If Ia is NM and Ib is PS and Ic is PS then CM is 8. If Ia is Zeo and Ib is NM and Ic is Zeo then CM is 9. If Ia is Zeo and Ib is NM and Ic is PS then CM is Seiously 20. If Ia is PS and Ib is NM and Ic is Zeo then CM is Seiously 2. If Ia is PS and Ib is NM and Ic is PS then CM is 22. If Ia is Zeo and Ib is Zeo and Ic is NM then CM is 2. If Ia is PS and Ib is Zeo and Ic is NM then CM is Seiously 24. If Ia is Zeo and Ib is PS and Ic is NM then CM is Seiously 25. If Ia is Zeo and Ib is PS and Ic is Zeo then CM is 26. If Ia is Zeo and Ib is PS and Ic is PS then CM is 27. If Ia is Zeo and Ib is Zeo and Ic is PS then CM is 28. If Ia is PS and Ib is Zeo and Ic is Zeo then CM is 29. If Ia is Zeo and Ib is Zeo and Ic is Zeo then CM is Good 0. If Ia is PS and Ib is PS and Ic is Zeo then CM is Good. If Ia is Zeo and Ib is PS and Ic is PS then CM is Good. Defuzzification is defined as the convesion of fuzzy output to cisp output. Thee ae many types of defuzzification methods available. Hee we used Cente of Aea (COA) method fo defuzzification. Despite its complexity it is moe populaly used because, if the aeas of two o moe contibuting ules ovelap, the ovelapping aea is counted only once. Output of Fuzzy ogic Contolle. Good TO To 70. Seiously To 0 The output of the fuzzy contolle is used as the command signal fo the closed loop opeations. If the fuzzy contolle output is Good, then the pogam goes fo next set of data to be acquied. Meanwhile if the opeato wants, the data like thee phase cuent, thee phase

14 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe voltage, fequency of input voltage, powe facto, total hamonic distotion of both the cuent and voltage and the state of the moto can be stoed in a file. If any incipient faults ae slight voltage unbalance occus, then the output of the fuzzy contolle will go. Immediately the fault data and the cuent spectum ae stoed in a file fo analysis pupose with time as long as fault pesists. At the same time a waning indication will be given to the opeato, and a beep sound will be geneated at the cental pocessing unit of the compute to alet the opeato at the shop as well as the contol oom enginee. The font panel of the monitoing system will also display the possible cause fo the damaged state of the moto. The instantaneous cuent and voltage wavefoms, and the spectum of the cuent can be stoed as a HTM file to find the oot cause of the fault. Fig. 5. Thee phase Stato cuents and Pecentage Health of Induction moto (Nomal Opeation) Fo the sevee faults such as open phase, open coil, single line to gound shot and line to line to shot, the fuzzy contolle output will be seiously damaged. In this state the machine should not be allowed to opeate any futhe. Wheneve the fuzzy contolle output goes seiously damaged, the machine will be isolated fom the supply and the instantaneous fault data ae stoed, also the font panel of the monitoing system will display the possible cause seiously damaged state of the moto. Afte pefoming the above said opeations the pogam will be stopped. 5. Simulation and Results 5. Nomal Opeation Fo the values given in section 2.4 and using a simulation stop time of 2.0 seconds, the moto was simulated duing stating fom est with ated voltage applied and no mechanical load. Figue 5 and 6 shows the Stato cuent and health of induction moto, Stato input voltage, speed and toque, symmetical components of stato cuent, symmetical components of stato induced voltage, symmetical components of stato input voltage ae shown. Fom these esults it can be concluded that afte the tansient peiod is ove, the health of the moto is Good, and thee is no negative sequence component in both stato induced voltage and stato cuent. Fig. 6 Developed Toque and Speed of Induction moto (Nomal Opeation) 5.2 Tun-Tun shot in one phase winding Afte the simulation fo nomal opeation of the induction moto model, simulation fo the shot cicuit in the pat of the winding in R phase has been caied out. At this condition the value of the stato esistance at shot cicuit fault is equal to R stato, fault.ω, we can find the value of the inductance at the fault state by using the atio between the value of the esistance at both state(nomal and fault). Thus the value of the inductance is

15 240 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 R Stato, nomal R 5.. Stato, fault Stato, fault n Stato, nomal Stato, fault Stato, fault 0.5H Replacing the values of the stato esistance and stato self-inductance in phase R by these values the esults can be obtained. Figue 7 shows the Stato cuent and health of induction moto, speed and toque, symmetical components of stato cuent, symmetical components of stato induced voltage, ae shown. The simulation is stated up with nomal state paametes. Afte obtaining steady state at 0.5 second the tun fault has been ceated by changing the above said paametes. Fom these esults it can be concluded that duing nomal opeation(befoe fault), the health of the moto is Good, and thee is no negative sequence component in both stato induced voltage and stato cuent. As soon as the fault is ceated the stato cuent becomes unbalanced, and the health of the induction moto goes seiously damaged and finally settles to state, and we can notice that thee is pesence of negative sequence component in both stato induced voltage and stato cuent wavefoms duing fault conditions. stato esistance is vey lage and coesponding to this value we can calculate the value of the inductance by this equation: R Stato, nomal R Stato, fault n Stato, fault Stato, nomal Stato, fault Stato, fault 82.5H Replacing the values of the stato esistance and stato self-inductance in phase R by these values, the fault state esults can be obtained. Figue 8 and 9. shows the Stato cuent and health of induction moto, speed and toque, symmetical components of stato cuent, symmetical components of stato induced voltage, ae shown. The simulation is stated up with nomal state paametes. Afte obtaining steady state at 0.5 second the beak in winding fault has been ceated by changing the above said paametes. Fom these esults it can be concluded that duing nomal opeation(befoe fault), the health of the moto is Good, and thee is no negative sequence component in both stato induced voltage and stato cuent. As soon as the fault is ceated the stato cuent becomes fully unbalanced, and the health of the induction moto goes seiously damaged and finally settles to the same state, and the pesence of negative sequence component in both stato induced voltage and stato cuent wavefoms duing fault conditions can be noticed. Fig. 7. Symmetical component wavefoms of stato cuent (Tun-Tun shot) 5. Beak in stato winding Fo simulation of the beak fault in the stato winding at R phase, it is not possible to apply a beak in the phase by putting the value of the stato esistance and the stato inductance to infinity. It is assumed that the value of the Fig. 8 Stato cuent and Pecentage health of induction moto (Beak in winding)

16 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe Fig. 9. Symmetical component wavefom of stato cuent (Beak in winding) 5.4 Unbalance in input voltage The simulation of induction moto with voltage unbalance can be simulated by simply vaying the voltage magnitude in any one of the phase, no othe paametes need to be changed. As in the pevious case, the machine is stated up with nomal value, and at 0.5 second, the cuent takes its steady state value, now the fault has been ceated by changing the voltage of B phase. In this case a 6% of the ated voltage in C phase was educed to ceate unbalance. Figue 20 and 2 shows the Stato cuent and health of induction moto, speed and toque, symmetical components of stato cuent, symmetical components of stato induced voltage, and symmetical components of input voltages ae shown. The simulation is stated up with nomal state paametes. Afte obtaining steady state at 0.5 second the tun fault has been ceated by changing the magnitude of B phase voltage. Fom these esults it can be concluded that duing nomal opeation(befoe fault), the health of the moto is Good, and thee is no negative sequence component in both stato induced voltage and stato cuent. As soon as the fault is ceated the stato cuent becomes unbalanced, and the health of the induction moto goes seiously damaged and finally settles to state, and we can notice that thee is pesence of negative sequence component in both stato induced voltage and stato cuent wavefoms duing fault conditions. Fig. 20 Stato cuent and Pecentage health of induction moto (voltage unbalance) Fig. 2 Symmetical component wavefom of stato cuent (voltage unbalance) 5.5 Open Phase fault In this case afte nomal statup, at 0.5 second, R phase was open cicuited and the coesponding esults ae shown in figues 22.and figue 2. Shows the condition monitoing of moto using labview Fig. 22 Stato cuent and Pecentage health of induction moto (one phase open)

17 242 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe 2009 Fig. 2 Stato cuent and Pecentage health of induction moto (one phase open) using labview 4.Conclusion A abview based measuement and health evaluation system has been developed and implemented. This application allows fast failue state estimation. The moe detailed investigation to point out the difficult conditions of the machine unde diffeent stato fault conditions of induction moto can be pefomed. This is a highly vesatile technology fo condition monitoing and fault analysis of motos. It solves the shutdown Poblems and ensues safe woking envionment in continuous pocess industy. 5.Futue Impovement In futue this wok can be extended fo monitoing the oto elated faults. The developed condition monitoing system equies a pesonal compute fo data acquisition and diagnosis; this can be miniatuized by using an embedded system which does not equie any pesonal compute fo health evaluation. Refeences [] W. T. Thomson : "A Review of On-ine Condition Monitoing Techniques fo Thee-Phase Squiel-Cage Induction Motos -Past Pesent and Futue" Keynote addess at IEEE Symposium on Diagnostics fo Electical Machines, Powe Electonics and Dives, Gijon, Spain, pp -8 Sept [2] YE Zhongming and WU Bin, Ryeson "A Review on Induction Moto Online Fault Diagnosis" Poceedings of thid intenational confeence on Powe Electonics and Motion Contol, IPEMC [] Paul V. Goode, Mo-Yuen Chow "Using a Neual/Fuzzy System to Extact Heuistic Knowledge of Incipient Faults in Induction Motos: Pat II Methodology", IEEE Tansactions on Industial Electonics, Vol.42, no.2, pp: 8, Apil 995. [4] Randy R. Schoen, Bain K. in, Thomas G. Habetle, Jay H. Schlag and Sami Faag. "An Unsupevised, On-ine System fo Induction Moto Fault Detection Using Stato Cuent Monitoing" IEEE Tansactions On Industy Applications, Vol., No.6, pp: , Novembe/Decembe, 995. [5] M.E.H. Benbouzid, H. Nejjai, Univesity of Picadie, Fance "A Simple Fuzzy ogic Appoach fo Induction Motos Stato Condition Monitoing" IEEE Intenational Electic Machines and Dives Confeence, 200. IEMDC 200. Volume, Issue, 200 Page(s): [6] Sang-Bin ee, Rangaajan M. Tallam, and Thomas G. Habetle, Geogia "A Robust, On-ine Tun-Fault Detection Technique fo Induction Machines based on Monitoing the Sequence Component Impedance Matix" IEEE Tansactions On powe electonics, Vol. 8, No., Page(s): , May 200. [7] Subhasis Nandi, Hamid A. Toliyat, "Novel Fequency Domain Based Technique to Detect Incipient Stato Inte- Tun Faults in Induction Machines" poceedings of industy application confeence, Page(s): vol. 2000, [8] Fco. Segio Sellschopp S., Maco A. Ajona., Toeon Coahuila, "A Vitual Instument Fo Fequency Analysis of Electical Machines" poceedings of electical machines and dives confeence, On page(s): , 200 R.Saavanakuma eceived his B.E. Degee in Electical and Electonics Engineeing fom Thiyagaaja College of Engineeing, Maduai, Tamilnadu India in 996.He obtained M.E., in Powe Electonics and Dives fom College of Engineeing Guindy,Tamilnadu,India in 998.Since 999,He is with the Velloe Institute of Technology (VIT)Univesity Tamilnadu,India woking in the School of Electical Science. Pesently he is woking as a eseach schola fo PhD degee in the School of Electical Science at the VIT Univesity. His pesent eseach inteests ae Condition Monitoing of Industial Dives, Modeling and simulation of Electical system, Special machines, Application of Soft computing Technique.

18 IJCSNS Intenational Jounal of Compute Science and Netwok Secuity, VO.9 No.9, Septembe K.Vinoth Kuma is a membe of IEEE. He is eceived his B.E. Degee in Electical and Electonics Engineeing fom Anna Univesity, Chennai, Tamilnadu India in 2006.He obtained M.Tech., in Powe Electonics and Dives fom VIT Univesity, Velloe,Tamilnadu,India in Now he is with the Kaunya Univesity Coimbatoe, India woking in the School of Electical science. His pesent eseach inteests ae Condition Monitoing of Industial Dives, Modeling and simulation of Electical system, Special machines, Neual Netwoks and fuzzy logic. K.K.Ray eceived his B.E. Degee in Electical Engineeing fom Jabalpu Univesity, Calcutta India in 967.He obtained M.Tech fom Indian Institute of Technology Kanpu, Kanpu,India,in 977. Fom 967 to 2002, D Ray was with the Indian School of Mines Univesity, Dhanbad (Jhakhand),India woking in the Depatment of Engineeing and Mining. He eceived his PhD degee fom the Indian Institute of Technology Delhi, New Delhi,India in 99. Since 200 he is with the School of Electical sciences, VIT Univesity Tamilnadu, India. His pesent eseach inteests ae Powe Electonics and Dives, Instumentation and machines, Reactive Powe Compensation, Switchgea Potection.