 In the last class, we discussed about faults in electrical motors and transformers and how through current monitoring, we could find out the defects in these motors, in particular the rotor broken bar cases or the static eccentricity. Towards the end, I have discussed about how faults could be detected in the bearings of a motor. Now, from that you will get a clue as to a bearing is a mechanical system in a motor or mechanical component is a mechanical component in a motor. And essentially this bearing fault was nothing but the supply frequency plus minus a mechanical frequency. This gives us a clue that if I have any mechanical defect frequency in a system, where it is being driven by an electrical motor by looking at the motor current spectrum, I can see side bands around the supply frequency. I will repeat that again. If I have a system driven by an electrical motor and this system is a mechanical system, this mechanical system has defects. So, this is how the defect frequencies are given. So, if I now monitor the frequencies of the electrical motor, I will see these defect frequencies come as side bands around the supply frequency. We will see the genesis of this, how this happens in an electrical motor and you will see how powerful this technique is for the fact that it is non-intrusive. I need not go near the mechanical unit like I was telling in the last class remotely at my control room just by accessing the conductors which are carrying current of the electrical motor, I can be able to find out the faults in the electrical motor. So, in this lecture on motor current signature analysis, I will basically tell you about the theory of MCSA motor current signature analysis. And I will be proud to say that we at IID Khadakpur has been a pioneer in this since last 15 years, we have done research in this and many of our publications are now referred worldwide. And we have we are the first to find out faults in gears or gearboxes which were driven by electrical motors actually not by monitoring the vibration of the gearbox, but just by monitoring the current drawn by the electrical motor driving the gearbox we could find out the faults in the gear box. Imagine now I will just give you a wild imagination, imagine now that you have to measure or find out the faults in a nuclear power plant wherein you do not have any access, you obviously physically cannot go there to mount your transistors, but you have the access to the current carrying conductors to the electrical motors operating the different valves, gear boxes, pumps etcetera and that is how you can do them. Think of a submersible pump wherein the pump is underground, you obviously cannot put a virus or transducer going underground, but you can definitely access the current carrying conducted to the pump which to the motor which is driving the submersible pump and thus be able to monitor the effect of the motor. In fact, in this class I will be telling you about the theory of the motor current signature analysis and particularly how signal processing can be used to find out the faults in this systems and I will tell you about two case studies, one about the gear box and another one regarding the submersible pump and tell you the features of, future of the motor current signature analysis and I believe in the years to come motor current signature analysis about a decade ago motor current signature analysis was very much in a nascent research stage, but nowadays in fact there are industrial products which have come out wherein you can fit on a system and find out the fault only through current monitoring and I believe in the future using wireless techniques and motor current signature analysis you could be measuring remote off places very conveniently. So, you all know this by now in this class that we were somewhere here we know the 70 percent of the wireless monitoring where every analysis where the parent material composition is to be known, but this motor current signature analysis is an upcoming emerging field of CDM. Of course, the essential elements in this motor current signature analysis is based on the fact that the currents have to be measured. Current cannot be measured by an ammeter because the ammeter will only give an RMS value and at most with load or defect only the value the waveform the magnitude of the current is going to change, but I have no clue when somebody says my ammeter reads 10 amperes I do not know whether it is 10 amperes a neat sine wave or 10 amperes a neat modulated wave or 10 amperes with lot of peaks I do not know that. So, once I see the current waveform to begin with in an oscilloscope and then do its frequency analysis I get a better idea. So, everything depends on this current waveform current could be as needed as this or as modulated as this this is how a modulated current looks like actual measured modulated current. The statistical parameters of these two waveforms may be similar some of the statistical parameters could be same. So, we should not be misled that looking at a single number that they are the same current, but if you look at the waveform look at the frequency content they are different and this frequency content is actually the indicator of the fault in the system. In fact, people use Hall effect sensors like this wherein they put the conductor here and these are current transformers basically current transformers are used to measure very high currents and where in turn ratio is 1000 is to 1 I obviously cannot measure 1000 ampere current because the voltage induced will be large. So, we will have a transformer to reduce the magnitude and that is why when you go to the any plants they have C T coils just to reduce the turn ratio so that meter can be used to display and the same meter could be used to measure 100 ampere 1000 ampere of current same to measure the 1 ampere of current, but when you measure 1000 ampere of current you have to give a step down transformer and that is why this current transformers are used and particularly in the field this Rogoski coils are used, but we will be using Hall effect sensor because they have a wide frequency range and as you will see this green signal here the frequency content of the signal is different and then this is very very cheap all you have is a data equation unit which will carry the conductor and then or use one of these analyzers like we have in the laboratory then then you use it and then do the current analysis, but now let us go to the theory of this motor current signature analysis and we will see why this is important. I am talking about the current which is there in a motor so for example in a motor I have three phases T 1, T 2, T 3 or let me let me simply look into here so there will be two types of current in the motor one is the current which will be responsible for producing the magnetic field in this stator and other 90 degree to it is the current which is responsible for producing the torque and which is rotating the rotor of the motor. Now if there are three defects at frequencies f 1, f 2 and f 3 these are three mechanical defects which have taken and basically they have this three torques responsible for this defects produce three torques T 1, T 2, T 3. If I sum them up one is the static component of the magnetic the first equation here this magnetic component the DC component and the time varying component because f 1, f 2, f 3 and quadrature to it is the torque producing current IST which is given by this. Now if I just add them up they are all similar you can you could have taken also any one of them to demonstrate. Now if I see the current in any one phase if I add them up it will be nothing but the component of the these two together in their respective phase if I just add them up the magnetic current and the torque producing current. If I add them up together like this I will come up with an expression wherein I will get cosine f e minus f 1 f e plus f 1 f e minus f 2 f e plus f 2 f e minus f 3 f e plus f f e plus f 3 where f e is the supply frequency and f 1, f 2, f 3 are the defect frequencies. So, you see the current which is add in the r phase you know there are r y b 3 phases in any one phase the current drawn is no longer I naught psi 2 pi f t f e t but these components. So, this becomes I s r is I naught sin 2 pi f e t would have been my ideal case where this is the supply frequency but as soon as I have the defect frequencies f 1 f 2 f 3 these are my mechanical defect frequencies. I will have components of f e plus minus f 1 or f 2 or f 3 in the current spectrum and looking at this I will know whether if these come around the basically they are modulating this torque which is created because of a fault is modulating the current and because of modulations again I am seeing side bands. So, in effect if I have to explain you through a flow chart I have a mechanical unit which is coupled triple motor if this mechanical unit has defects f 1 f 2 f 3 anything they are basically going to produce a loading torque and this load torque is responsible for drawing the current because there is a load torque the flux is going to change. So, the rotor has been given with a load torque. So, I need to produce on contracting torque to this load so that the motor is going to run and this has to be at these frequencies because they have f e plus minus f 1 2 3. So, the current drawn will be having this kind of characteristics. So, side bands of components of torsional vibration across supply line frequency is the strongest indicator of the faults in the mechanical system because torsional vibration are created because of this faults. So, we have presented this in one of our paper. In fact, you can go to our website iitnoise.com to see more about our research findings and the you can see some of the details of the literatures of the reference papers regarding motor current signature analysis which we are doing since the 2000s. So, in fact, I will now describe about the experimental setup which we did in the laboratory to explain how motor current signature analysis is used to find out fault in an electrical motor. So, in this case what we have is a three phase 2 pole 7.5 kilowatt induction motor which is actually driving a 4 stage multi stage automotive gearbox transmission gearbox from a car and this gearbox is driving a generator and by giving a resistance load to the generator which is typical setup in early lab we are able to load the motor through by loading the generator. So, this experiments initially we did at a single speed because now we have provisions we can through a V F D drive also we can do this study at different speeds. But most important was that we artificially created faults in the gearbox in the automobile gearbox and then we know the characteristic frequencies of this faults basically around the gear machine frequencies and then we at different load conditions by loading the generator we loaded the gearbox we then analyze the current and then found out that the gear machine frequencies were actually showing a side bands of the current being drawn by that electrical motor. And this is the setup which we had we still have it in the laboratory this is an induction motor which is driving an automobile as you can see this is a car gear box we have a coupling and the DC generator and a loading unit from a resistance band and the control panel you will see three hall effect sensors because initially we are measuring the three phases and since this is a balance supply later on we just use the single hall effect sensor and this is the close of view you can see the three hall effect sensor to measure the three gear box and this is an automobile gear box wherein we could remove the gears and then put in defect in the gear and then run it and this is just to give you an idea of the experimental setup we have a three phase two pole induction motor driving a four stage gear box driving a generator a generator is being loaded and we have the current probes to measure the supply frequencies supply current and then of course to correlate or to find out the faults we also used vibration transducer and tri-axial accelerometer here and we recorded all of them in a tape recorder and then took it into a data acquisition system where we did the analysis and this is the internal of the gear box is a four stage gear box. So, this is if my speed was 49.1 hertz and this is the lay shaft and this is the output shaft and then the fourth gear has a 19 teeth third gear 24 teeth second gear 29 teeth. So, we can find out the machine you know we operated this at second gear and based on that we can calculate the output shaft speed for a supply frequency of 49.1 hertz that is close to about 3000 rpm and then we are having because of the reduction we are having the output shaft for the second gear I think it is a 30.1 is this hertz and output shaft is only 13 to 2 rpm. So, when we run in the second gear the gear missing frequency is 630 hertz and third gear it is 780 and the fourth gear it is 930 hertz gear machine is nothing, but this time just not to confuse. So, f m is nothing, but n by 60 times t for a linear spur gear helical gears. So, this are the views of the gear which we introduced defects in the second gears. In one case we remove one complete tooth and these are very hard. So, we use an wire cutting EDM and we remove two teeths by a disyncing EDM and then we again put them back in the gearbox. So, some of the data acquisition parameters we studied our focus was 0 to 2 kilohertz, sampling frequency was 4.096 kilohertz, number of data points was 8192 data points, time record was 2 seconds of data which we took and then we applied no load at you know this is the full load this was our 5.65 half load quarter load etcetera and then we measure the current spectra and then the lot of signal processing to find out the defect in the gears. So, some of the signal processing techniques which we are used we know the FFT analysis normal FFT analysis, but as you know all these signals become modulated. So, we have to do demodulation and in this class I will be mostly giving examples from the demodulation though there are other techniques of signal processing which is not there in the scope of this class regarding wavelet transform and multi resolution Fourier transform which could be used to find out the signal features or features of signals which are non stationary and some of the signals out of the gear box are non stationary, but we are not going to discuss about them in this class, but I will just show you the amplitude demodulated current and this is what is there and if you will see these are the publications from where this we have had this publications done and you will see just as a comparison FFT analysis and amplitude demodulation do zero defect means no defect in the gear box. If I do the vibration monitoring by putting the triaxial on the tail end of the gear box I am seeing the F1 is the input speed, F3 is the output speed and F2 is the second gear laser speed. These are pretty much available in the resonance spectrum as was expected, but if I look at the current signal you will see around the 50 hertz FE, I will see FE plus F3, FE plus F2, FE plus F1 the all the side bands come up. The side bands are because of the modulations obviously they are modulated and to find out these side bands I did a demodulation and this demodulations by this demodulations you will see F1, F2, F3 sticking up and with load or with defect severity these amplitudes are going to increase. So, nowadays there are commercial indicators based on these algorithms where people just use on a pump, on a motor, on a mechanical unit being driven by a gear box they know what F1, F2, F3 and they put lot of signal processing features to find out how much is the magnitude of F1 compared to the no defect case has the magnitude of F1 increase F2 increase and then they will set up triggers and alarms to tell you that there is a fault in the mechanical system. Now, as you see with defect D1 same frequencies you will see the severity of vibration increases severity of the amplitudes also in F1, F2, F3 are increasing. So, imagine in a current these are all current in a current you are able to get the mechanical frequencies of the system which is being driven by the motor. You can see the power of such a unit. Now, imagine you do not have to go inside your mind to close to the gear box inside underground to the pump you can be sitting in a control room measuring the current doing this kind of analysis and all you do is find out the amplitudes at F1, F2, F3 in the current spectrum. Similarly, again with the defect severity D2 is where when we have two teeth removed you can see the amplitudes again going up and these are because of side bands. So, key catch is the torque or the defect is going to modulate modulation means the current will be having side bands and side bands have to be detected by demodulation and the amplitude of side bands will increase with the defect severity. Of course, we also saw the high frequencies because we wanted to see around the gear measuring frequencies the earlier case was 0 to 100 hertz, but if you recall the second gear mass gear machine frequency around 630 hertz. So, you will see lot of side bands around the gear machine frequencies of course, they look very gibberish right now, but if you have the right kind of frequency estimation and then you have frequency detectors in your algorithm that I will only be looking for these frequencies and monitoring their amplitude you will do a good job in using motor current signature analysis, but because this came out of our research I am showing all the possible scenarios which are available. I could be monitoring the fourth gear or the third gear or the second gear depending on my frequency of interest and this is just to summarize what we have found out in the viruses and current signatures. You can see the monitor viruses, following frequencies can be tracked, input shaft frequency from vibration signatures, second gear mass frequency from vibration signatures, supply life frequency from current signatures. This is with vibration and with defect how the current will behave d 0, d 1, d 2 the current gives a good indicator and of course you know you can also do there are many algorithms to do, variations in amplitudes of side bands and harmonic of line frequency in the current signal with the amplitudes of side bands are going to increase because it is in second gear mass frequency f m 2 is 630 plus minus 50 hertz minus 50 plus 50 and so on and so on. So, you will see with defect severe to these amplitudes keep on increasing. So, this were out of the research as I was telling, but this only demonstrate that any mechanical unit can be used I mean its fault can be detected or estimated by monitoring the current being drawn by that motor which is driving that unit and of course we could do the transient loads also to do the because the loads are transient or the speeds are transient this will become non stationary signals I will not go in the details of the non stationary signals, but a certain algorithm can be developed and this actually gives you the motor current monitoring strategy which is being used in for MCSA. It can have a steady current signal and current transients we are not discussing about current transients transients occur when you are initially switching on the machine or switching off the machine, but a steady state current signature we can do FFT analysis. Sometimes it is difficult to trace the side bands in the normal FFT analysis because of these things become modulated, but once you do the demodulations we would have seen the shaft frequencies are detected, but of course in high frequency demodulation because of this gibberish nature of the high frequency signals it becomes difficult to trace the gear missing frequencies, but then there are other methods like the multi resolution Fourier transform. If you want to know more about these details it is there in our research publications which you can see you can very easily see the side bands of the second gear mass frequency and the defects can be very easily classified. This kind of an introduction to motor current signature analysis in fact in many of the early classes when I talked about fault prognosis which we did in the case of a machine in a lathe machine wherein we wanted to prognosis do prognosis on a cutting tool life fault. We did in fact an experiment on a lathe machine wherein we monitored the motor current of the lathe machine once it was cutting aluminum and steel and this is about 15 years from now we had seen that how this current in the lathe machine changes once it is machine steel once it machines aluminum and then we found out that this current definitely carries information regarding load on the lathe spindle and so a defect also gives a load and the load will reflect in the current. So, by monitoring and analyzing current we can find out faults in this case in the gear box in the previous case we found out faults in the lathe machine and we took it a little further in the sense we went ahead and started to use this technique to find out fault detection in submersible pump and this is currently also we are doing this kind of a research to find out faults in submersible pump wherein if you see here this is a submersible pump which is a drain type pump, but there are many submersible pumps and if you think of may be here oil well drilling imagine we spend you know crores of rupees to drill under the ocean bed and they are continuously being drilled of course if and there is one we drill and then we have of course because the shear pressure the oil gush out the pipe and then we tap it off and tap the crude line, but once you are drilling it drill has gone in a may be 50 meters, 100 meters below the ocean bed and if the motor is developed and some fault how do you monitor it? Large tunnel boring machines, submersible pumps, oil drilling, nuclear plants, wind mills basically think of scenarios where we are not able to position our virus sensor, position our virus sensor. So, MCSA is a relief in such a case we can use MCSA to find out faults. So, here in the laboratory in fact we are doing research in this areas into finding out more signal processing algorithms to find out these features better. So, we have an there are actually six radial veins in this pump and you will see my student has actually removed one of the veins here because we wanted to again artificially induce and see what are the effects of this on the pump function and again through doing motor current signature analysis analysis whether we can measure it. So, we again drove this pump through a V F D drive and then we have this hall effect sensor which is used to measure the current and then we have this analyzer system wherein you can you can see the current time history and then you can do see the current spectrum and you see this at a 40 hertz supply frequency the current spectrum for a normal impeller is given like this you will see many frequencies coming up and we can see at a particular gear mashing frequency is because you know you can again in pumps like we have in the case of gear gear boxes we have the gear mashing frequency in pumps we have known what is known as the vein pass frequency which is nothing but rotational speed the number of veins in the impeller. So, again if this vein pass frequency given by F V I will see F E plus minus F V or the other way also and this is what is there in the frequency spectrum and of course, there will be lot of harmonics in fact in this signal you will see lot of harmonics multiples are there harmonics of supply frequency. So, if you have the N M C S A you will if you have the appropriate signal frequency detecting or tracking algorithm in the software in the analysis software we can detect the signal faults in mechanical systems driven by electrical motors using. So, if we answer very very important just you may not get just mounting the peak at the supply frequency you may not get that. So, in this example at 40 hertz we again had defects of 1 vein removed 2 vein removed 3 vein removed and then we see the side band values at the these frequencies the lower side band and it is on the sorry left hand side band and the right hand side band and their amplitudes you will see it is observed that the when the severity of the defect increases height of the side band increases because you will see from minus 47.4 it has become minus 44.71 from minus 62 it has become minus 58.35 that means, around the supply frequency of. So, this is around 189. So, these values are going to increase and then another case was 267 this is the right side band this is the left side band. So, once the defect severity increases they will increase. So, if somebody is monitoring this ratio. So, you can develop many algorithms I monitor this ratios or I monitor this differences if this differences go beyond a certain value I know there is a fault. So, many algorithms can be developed as I was telling you just by having algorithms better algorithms to detect this you can find out the fault in this mechanical systems. Now, another technique how MCSA is powerful to find out faults in any mechanical system MCSA can be used to find out faults in any mechanical system. So, you can find out fault in any rotating it has to rotate rotating mechanical system imagine now I have an IC engine. Now, somebody ask me how do I use MCSA to find out fault in this IC engine all I can do is you know you have a crankshaft to this you use a taco you put a taco generator. So, because of this rotational taco generator is going to produce an EMF. So, if you this EMF will have the defect mechanical frequencies of this system will be there buried in the this. So, if you do a signal analysis on this EMF which is generated by the taco generator coupled to the mechanical system which is rotating I can find out the fault in the mechanical system. In fact, we in IIT Kharagpur have a patent on this system this is our patent actually this is our invention which we have done that by using a taco generator on any rotating system you just attach it and then we have used it successfully in mounting the health of gas turbines. We have done that for the case of gas turbines that we can find out fault in any mechanical system it is an IC engine gas turbine etcetera here I do not have a motor driving an IC engine. IC engine I just couple it to a taco generator because the taco generator is rotating and the rotational power power here is the crankshaft which is being driven which is the prime mover in this case and then I am seeing a EMF which is having a lot of ripples and all this characteristic frequencies are going to show up defect frequencies are going to show up in this spectrum. So, you can realize the power of the motor current signature analysis it is a very very powerful clean neat technology provided you do your measurements right and your signal processing right submersible pumps we just saw the example reactor coolant pumps we cannot go inside the reactors. So, we can monitor the coolant pumps particularly coolant pumps are very critical to nuclear reactors that is where they control the temperature of the nuclear reactor I mean you know what happened in Japan because there is a you know the cooling pumps did not function they got flooded the reactor you know we had an uncontrollable reaction. So, this is very very important rolling mills rolling mills there are a lot of motors which are driving the rolls. So, if one of the motor fails it is going to have a serious effect on the rolling production, reduction gear boxes in gas turbines power plants and now it is a good candidate for motor current signature analysis is this electrical vehicles which are which are moving around basically they are all high power battery driven cars wherein we have motors basically driving the wheels and if there is a defect in this electrical motor I am going to have a problem or this motors drive a gear box and gear box with a speed direction are driving the wheels. So, I can I can have some sort of an monitoring mechanism in that think of the traction drives and locomotive there also again the DC motors are used you know here we talked about AC motors even if you have DC motors same thing would happen only thing that there will be ripples after you remove after you remove the DC current you will be finding lot lot of ripples and this ripples do contain the frequencies of the defect frequencies of the mechanical system which are being driven by such units. So, in conclusion if I talk up to the previous class faults in electrical systems of course, nobody is asking me not to use virus and monitoring I can still do virus and monitoring I can do oil analysis particularly for the case of transformers I can do thermography something which you are going to study in the subsequent few lectures because there will be lot of heat generation in contact and switch gears and then we will we can use or MCSA to find out faults in electrical motors and also faults in mechanical systems and I believe if I use MCSA plus wireless techniques. In fact, we have done experiments on using MCSA and wireless and if you go to our website and then if you will see the experiments on the virtual lab on mechanical systems you will see virtual lab. In fact, you will see the virtual lab on mechanical systems and signal processing. You will find out there are two experiments one on using MCSA to find out faults in gear boxes another is to find out faults through MCSA and wireless techniques in industrial arm blowers. Thank you.