 In previous lecture, we discussed condition monitoring of rotating machinery. Basically we try to see whether various kind of faults which reflect in vibration signal, how we can able to detect the various kind of faults in rotating machinery by just looking into the signal, either it is in time domain or frequency domain or whether it is orbit plot or maybe waterfall or cascade plot. So, these are the various forms of the plots which we require for analyzing the vibration signal, so that we can able to identify or locate or basically we can able to detect the fault in various kind of machine elements. Now today we will be extending the condition monitoring in a more broader perspective, especially condition based monitoring will be introducing this particular subject in more broader way and we will try to see what are the various tools are available to perform this and what are the research directions which are there in this particular subject. And then we will see few case studies at the end of this particular lecture, so in this initially is the overview of the condition, the overview of the presentation in which we will be introducing the condition based monitoring, even we will give some brief review especially of the requirement of the this condition monitoring. Few case studies we will try to see at the end, so coming to the condition monitoring of an industry, obviously any industry we need to capture various kind of signals, there is vibration signal or temperature, pressure, noise, various kind of signal we need to capture and then those signal we need to store in computer and try to analyze those signal and try to correlate various kind of faults which may appear in this kind of machinery. And then we can have some kind of data base based on that we will be having some kind of expertise in the form of may be heuristic expertise in which we can able to correlate those symptom of the fault and the whatever fault reflects what kind of symptoms. So, to introduce this particular subject, the equipment condition monitoring is essentially preventive failures that may cause human injury or equipment damage. So, obviously it can also enhance productivity if we are maintaining the machine properly, product quality will be good, life of the machinery will be more. This is the breakdown maintenance, so earlier the maintenance technique was basically the breakdown maintenance also called the unplanned maintenance or run to failure maintenance which take place only at breakdowns. So, when the machine what filled then you are calling for the maintenance, so that is call it a breakdown maintenance. Later maintenance technique is the time based preventive maintenance also called planned maintenance which sets a period interval periodic interval to perform preventive maintenance regardless of the health status of the physical asset. So, whether the machine is fine or not we schedule the maintenance and then we stop the machine and do the maintenance that is called preventive maintenance, but with this we unnecessary we stop the machine if machine is healthy. So, loss of economy will take place because of this. Therefore, more efficient maintenance approaches such as condition based maintenance are being implemented to handle the situation. A CMB or condition based maintenance if properly established and effectively implemented can significantly reduce maintenance cost by reducing the number of unnecessary scheduled preventive maintenance operations. So, we can able to see that through some monitoring technique we can able to see the condition of the machine and if there is unexpected signal coming in into the this kind of gadgets then we can able to plan the maintenance. So, now we will be briefly doing the machinery diagnostics and prognostic implementing condition based maintenance. In this diagnostics referred to the basically finding what kind of fault is that what is the severity of the fault and prognostic is whether we can able to a priorly we can able to have some indication that the faults are impending out is coming into the machine if we can able to detect that that will be a real challenge. So, abstract definition of the condition based monitoring is a maintenance program that recommends maintenance decisions based on the information collected through condition monitoring. So, based on information only we plan the maintenance not as a scheduled maintenance one. So, in this there are three basic component in the condition based maintenance program one is the data acquisition. So, we collect the data from various sensors and then we store that in a computer second is the data processing. So, once we have collected a huge amount of data we need to put them in a proper form and we need to plot them in bar chart or other kind of formats in which the analysis of the data can be done easily. So, that is the third one. So, once we have done the data processing then the most crucial part is the maintenance decision making. So, once you have the data then the experts will analyze the data and then make the decision for the maintenance. So, condition based monitoring so we will go in more detail of those three component of the condition based monitoring. So, one is the information collection. So, data collected in CMB program can be categorized into two main types one is the so called event data and condition monitoring data. So, there are two type of data. So, event data is that include the information on what happened as in the installation breakdown overhaul etcetera and what the causes were. So, what happened and what the causes were. So, that is some kind of event and or what was done minor repair preventive maintenance oil change etcetera to the targeted physical asset. So, this is some kind of event which will form the basis of the data decision that is event data apart from this we will be having condition monitoring data. These are measurement related to the health condition state of the physical assets like vibration, temperature, pressure, noise. These are the data which we collect that is that fall under the condition monitoring data. Then to collect this data we have various kind of sensors and in this we can able to see that this is one of the displacement measuring proximity probes. There are variety of accelerometer which. So, basically these are piezoelectric based accelerometer which gives vibration signal or the signal voltage signal corresponding to the proportional to the acceleration of the body on which they have been mounted. So, there are various kind of mounting like these are you can able to see this is a base. The holes are there. So, this we need to fix on to the body or sometimes we can able to glue it. So, various kind of mounting we can have in this kind of sensors. This is a hand held vibration meter in which we can able to get either the level of vibration or so simple rms value or that kind of informations we can able to get about the vibration with this vibration meter. Apart from this the acoustic measurement is very important. So, these are the typical various shape, various size of the sensors, the microphones and depending on upon the application their choice will be decided and this is the sound meter. So, this can measure the various parameter of the sound like power of the sound. This is another kind of instrument which is called impact hammer or model hammer. So, sometimes we need to give some kind of small force to the machinery and corresponding to this what is the change in the vibration we want to monitor and this is the purpose of this particular model hammer. The advantage of this is what particular impulse we are giving to the machinery we can able to that also we can able to measure it because there is a force measurement sensor is attached here and the tip of the hammer we can have variety of tips like rubber tip or aluminum tip or steel tip and depending upon what kind of frequency range we want to impart this tips can be chosen. This is another kind of instrument which is called vibration exciter or shakers. So, in this variety of shape, various size of the shaker we can have depending upon what kind of machinery we want to excite. So, with this we can able to excite the machinery externally and we can able to choose either the frequency of excitation or amplitude of that we can able to choose or even what kind of excitation force we want to give like sinusoidal or two frequency signal or multiple frequency signal or random signal. So, depending upon the application we can able to excite and then we can able to measure corresponding change in the vibration that basically we will in can be used to find the condition of the machine. This is for displaying the whatever the vibration signal captured through this these answers we can able to display the signal on this kind of screen that is oscilloscope and generally that is it displays signal in time domain, but frequency domain signal also can be seen in this and apart from this there is a another advance analysis that is spectrum analysis in which exhaustive analysis in the frequency domain of those time signal can be done and various form of the signal in the frequency domain can be plotted apart from the FFT simple FFT or if required we can able to transfer the data into the computer through a data equation system and here we can have some kind of virtual instruments. So, depending upon the need what kind of processing we want to do we can able to make our own instrument virtually and this will display various form of the signal which we want to monitor. So, this is very handy that whatever the signals we are capturing we can able to directly process it and display on the screen directly online signal we can able to see using this kind of virtual instruments and sometimes to pin point what kind of fault there is giving what kind of signatures and what kind of frequency this kind of laboratory instrument we can able to develop or fabricate in which various rotating components like this is a rotor with two discs there is a belt pulley arrangement gear box. So, it is a property mechanism so various kind of machine components we can able to design and fabricate and if you want to see the signature analysis of various kind of faults we can introduce in this and we can able to capture them in isolation of other kind of faults. So, that we can study a particular fault in more deep like even various kind of faults in the motor as there is a mechanical fault of the motor or electrical fault of the motor that we can able to study may be we can take out the current information from this kind of setup and even the current can indicate what kind of faults are appearing in the motor. This kind of a testing can be built to simulate the fault experimentally in their laboratory. So, these are the various kind of faults generally as we have discussed in the previous lecture also various kind of faults we can have and every fault gives a unique frequency in the vibration spectrum and with the help of that unique frequency we can able to identify which kind of fault has appeared in a particular machine. Now, coming back to the data processing that was the data capturing the second stage is the data processing in the condition based maintenance. So, this is the information handling stage of the condition based maintenance. So, we can have waveform data analysis. So, in the waveform we can have time domain analysis so various techniques are available I will be explaining this in the subsequent slide. Frequency domain analysis FFT, ARMA model these are auto regressive moving average model these are all time domain or frequency domain models. Time frequency analysis that is wavelength transform or short time frequency transform this kind of techniques are available to analyze the waveform data such as in vibrations. Another is the value type data like single value variable like temperature or pressure in which we get at one particular time may be one value. So, that kind of data also we can collect and there are various methods which are available to analyze this. So, various methods which are described there are listed here. So, describing each of them is beyond the scope of this particular lecture but some of them you will be knowing like Fourier fast Fourier transform short time Fourier transform wavelength transform artificial neural network neural Markov method support vector machine. So, these are various techniques by which we can able to process the data and we can have some kind of expert data base with us. Then data analysis combining event data and the conditional condition monitoring data. So, event and the condition monitoring data both we can able to combine because that is very important information. So, various artificial intelligent data processing techniques are there these are the techniques which are described in the subsequent slide. So, these can be used for processing of the data to take out the information from the enormous amount of data which comes from the machinery. So, some data are like frequency domain data free vibration analysis we can able to perform we can able to pinpoint what are the natural frequency of the system. So, predominant peak we can able to collect and that will be our data base or as we have seen in the previous lecture for gear. If we have chip tooth of gear then we can have this kind of impulses. So, those information can be collected these are the some of the data type, but we can have more of such type. Then the third stage which is most crucial one maintenance decision support that is the decision making process. So, diagnostic and prognostics are the two important aspect of condition base monitoring program. So, these are the two main component in this third stage which is most crucial stage. So, diagnostic deal with the fault detection, isolation and identification when it occurs. So, basically detection colliding that and finding the severity of that and prognostic deals with the fault prediction before it occurs. So, whether we can have technique before the fault appears into the system or at the beginning at very beginning of the fault appearance whether we can able to diagnose that or prognostic we can able to find that there is a fault appearing that will be the most challenging task. So, diagnostic is a posteriori event analysis and prognostic is a a priori event. So, before it happens we need to predict that then we can able to so together with these two we will be having a complete condition based maintenance program. So, diagnostics we have different statistical approaches like hidden Markov process, artificial intelligent approaches, artificial neural network, genetic algorithm, support vector machine, other applications like model based because various kind of fault like unbalance, misalignment we can have mathematical model of that and once we are measuring the vibration from the actual system we can able to curve it the mathematical model with the vibration signal and try to find out whether the fault is there or if it is there how much is the severity of the fault or the quantification of the fault we can able to do with the model based approach which is actually real research which is taking place in the present day on the rotor dynamics field. So, this is a very abstract way how we can able to basically diagnose the fault. So, in this particular case we have one we have collected various data from the machine and let us say there are n number of such data and those data in mathematics we can able to represent as a n dimensional vector. So, if we have in a vector n number of data or those data may be pressure, temperature, vibration, rms value various data we can have and we can put them in as a vector in n dimensional space it will occupy one particular vector direction then. So, in this particular case what we will be doing we will be introducing one fault in a machine at a time and we will be finding what is the change of these vectors are taking place these vectors are changing obviously their orientation in the n dimensional space will change. So, the first one is may be corresponding to one particular kind of fault this one is for one kind of fault this is for second fault. Similarly, we can have third fault state vector fourth fault in the machine. So, we need to introduce this kind of fault as shown earlier some kind of fault simulator in which we can introduce the fault one at a time and we can able to collect the data that will be the data base corresponding to various kind of faults. So, these are the five type of faults if a particular machine can have. So, we can able to first generate this kind of data base and once we have this five data base for five different fault then if let us say some unknown fault appears into the system. Then if we have measured all the data and if we plotted in this particular database space let us say this red one is for unknown fault we measured it and we can able to see that it is close to the fault three. So, that means it is representing that that whatever the fault is there in the system which is unknown to the user is actually from data base is predicting that there is a fault three in the particular machine this is this will happen when there is a single fault. So, the matching of these two vectors will be quite close, but if a vector comes like this. So, obviously in this particular case is not aligning with any of the fault, but they are close to four and five. So, we can able to attach some kind of a probability along with the because the fault four and five both are there is a possibility, but we can able to attach the probability of the four is this much and maybe 60 percent and 40 percent for the five that kind of based on the vector we can able to stimulate the probability of the fault if there are multiple faults appearing in the that particular machine. So, this is a very abstract definition how we can able to identify the fault using this enormous amount of data. So, this kind of a technique can be used to predict the fault in a machine. Now, coming to the prognostics there are several things which we can able to predict like remaining useful life of the machine. So, how much time is left before a failure occurs or one or more fault appears given the current machine condition and the past operation profile. So, once we have the condition based data based on that and previous history of the machine whether we can able to predict the remaining useful life of the machine this very pertinent prognostic prediction will be there for the machine. Apart from this prognostics incorporating maintenance policies optimize the maintenance policies according to the certain criteria such as risk, cost, reliability and availability. So, this if we can able to integrate the prognostic along with the maintenance policy that that will be very useful. Then third is the condition monitoring intervals periodic continuous. So, depending upon type of machine and critically of the machine we can able to decide whether a particular machine require the periodic monitoring or continuous monitoring. This is a typical onsite model of a condition monitoring system in which there is a power plants at different locations. We are extracting the data from this various kind of data we are storing at one place and there is expert system based on the data base of the history of the plant and this data base and the whatever the online data which is coming to the system to net can go to various web servers. From there experts can view this and they can see the condition of the machine and if there is some unusual case they can able to pinpoint if there is some fault is already appeared in some machine component or some sub plant or it is going to happen. So, that kind of interface we can able to build for such large systems apart from this this is some other requirement of the condition based monitoring which we may look into that one is the cost the low cost. So, system should be implemented by virtual instruments. So, it should we should have low cost of the CNB program portable it should be useful again I am repeating portable it should be used with notebook PC handheld computer mobile phones wireless communication etcetera easy to use detection of fault via instant exact wavelength analysis active noise cancellation and 3D trend plot etcetera. These are some other requirements automatic report generation and alarm, condition monitoring using automatic logging function, web and mobile based remote sensing and monitoring user friendly training video for self learning. So, these are the some of the requirement of CNB should have now after discussing very general topic on the condition based monitoring again we will take up some case studies. So, to start with we will take case study of gear in which various kind of faults have been introduced and based on the vibration signature the type of fault and the detection of the fault we will be doing it. So, in this particular case we have a double stage gear. So, you can able to see this is the input gear and this is the line diagram of that gear. So, input shaft there is a pinion 24 teeth is there and then this is gear 60 teeth then this is second stage in which 36 and 46 are the number of teeth this is the output shaft. Now, two stage transmission there are two machine frequencies because one between these two gear and another between these two gear and these are named as F 1, F M 1 and F M 2 and you can able to calculate this. So, first stage between these two that is. So, the input shaft speed is this one F I and if you multiply by the number of teeth of the pinion will get the F 1, F 1 will get the F M 1. So, this is one of the tooth missing frequency we could have taken this number of teeth and the speed of the shaft this one we could have got the same value another pair is this one. So, in this you can able to see that the speed of this particular gear. So, we are taking the speed of this gear and the number of teeth of this. So, this is the reduction with in the first stage and if we multiply with the number of teeth of this we will get the tooth missing frequency for the second pair. So, tooth missing frequency is a for pair not for a particular gear. So, this is for first pair and the second pair. So, we have this tooth missing frequency we should expect in the vibrational spectra apart from the rotational frequency of the input shaft. These are the various gears and especially the fault. So, you can able to see in the first one a small chip tooth. So, a small tooth has been chip off this is a large chip tooth has been removed and another one in which the whole teeth has been removed. So, there are three level of fault in which all of them are having fault related with the tooth of the gear. So, this is a typical signal this is the gear box acceleration spectra of baseline data first is without load and second one is the with load. So, as pointed out earlier whatever the dynamics is going on at the gear will reflect at the bearing location or that bearing location is because bearing is mounted on gear box. So, we can able to take the signal from there and that will reflect the dynamics of the gear. So, this is a typical signal. So, you can able to see the f. So, first gear miss frequency there is a peak for second one there is a large peak here also we have the value is around 4 to 5. That means, this is the first one 4 to 1 is this one red one and this f m 2 is this one that is 8 53 around. So, this is for f m 2 and this is for f 1 f m 1 apart from that there will be some other peaks corresponding to the structure of the gear box and other things. But, we must see we try to see the peaks basically this is a cursor which is showing the red line, but this peak is small you can able to see the peaks are small, but this is red dots are the cursor. So, in this 2 stage gear transmission process 2 missing frequency are there 2 stages are there. So, when the input shaft speed is little bit less that is 60 hertz for this has been measured f the first and second miss frequencies are roughly 1 4 2 5 and 8 5 4 5 4. So, you can able to see that these are matching with the actual experimental data. So, this is one and this is one. So, the theoretical calculated and the experimentally measured of frequencies are matching then this is the with test with a small chip tooth. So, with this obviously we expect the amplitude of the vibrations will increase. So, in this you can able to see the 24 tooth pinion gear with a smaller chip chip was used in this test the f m 1 is barely seen. So, this is not visible here. However, the second harmonics of the 2 missing frequencies are detected. So, basically you can able to we are not seeing the first harmonic, but second harmonic of the this particular miss frequencies are appearing in the spectra. So, the 1 7 2 5 that is this one and the second harmonic of this is 284624 you can able to see here. So, some higher multiples of the harmonic of the miss frequencies are visible, but f m 2 is more predominant, but f m 1 is not that much predominant, but second harmonics are picking up. So, then the next is this you can able to see the side band presented about the second missing frequency. So, we have this is the second miss frequency this is the side band. So, you can able to see this is coming from the modulation of the rotational speed of the input shaft. So, the difference between this frequency and the shaft frequency is giving as a side band similarly, multiple of that. So, f m 2 minus 2 f i this we already seen in the previous lecture that we can have one carrier frequency and modulated frequency. Then we can have this kind of side band, but in one side only that side bands are more predominant for this particular case. The shaft speed is 60 hertz in this particular case. Now, this particular signal is with large chip tooth the previous signal was for small chip tooth. So, in this you can able to see that the 24 tooth gear with a larger chip was used in this test. It can be seen that the f m 2 component and its second harmonics are dominant in the spectra. That means, this is f m 2, but second harmonic this is first harmonic of f m 2. So, they are predominant f 1 f m 1 is not that much predominant here also. And strong side bands are presented about both components. So, side bands are also there will see the zoom view of this also. Again the overall vibration level of the loaded case is higher even the vibration signal level. If we see this will be higher as compared to the previous one. So, this is with the missing tooth the most severe fault in which the tooth itself is removed completely. So, in this you can able to see let us. So, this is the most severe fault of all test then put gear has a missing tooth. In the previous figure it can be seen that the amplitude of the side band 7 9 6 and 7 3 6 are even higher than the missing frequency. So, 7 9 6 so in this f m 2 is more predominant, but there are side bands they are quite predominant here also. Here side band is more predominant can able to see 8 5 3 these are side bands f m 1 is present here. Here the second harmonic of the f m 1 is more predominant. So, this kind of modulation can take place in such even in the loaded condition the side bands mentioned above have higher amplitude then all other rotational related component except f m 2 component itself. So, side band is one particular phenomena which we could able to observe in such kind of fault. This is the basically zoom view of the previous slide. So, in this you can able to see that this is f m 2 there is a modulated frequency that is side band f m 2 minus f i corresponding to this spin speed of the shaft and this is f m 2 minus 2 twice of f i. So, these are the side bands which are predominant more than the actual this tooth missing frequency because of the severity of the fault. This is another zoom view of the same figure. So, in this you can able to see f or additionally we have 2 f m 1 that is in the higher range we have shown here and there is side band. So, side band is more predominant than the actual tooth missing frequency of the first tooth missing frequency. So, here it has been modulated even this is the higher harmonics of the tooth missing frequency second harmonic of the tooth missing frequency and side band of that here also the higher side bands twice of spin speed of the shaft. Now, we will take up some more case studies especially of the fans which are there in the industry. So, in this particular case we are describing about the industrial fans. So, each time a blade passes a point in space or an obstruction and impulse force fluctuation is experienced by the fluid of the solid body at the point. If a fan with n blades is running at f s r p m then the number of impulses experience per second will be n there is a number of blade into the speed of the fan. So, this is called the blade passing frequency as we had the tooth missing frequency similar to that blade passing frequency is there b p f. This frequency is inherent in pumps fans and compressor normally does not present a problem. So, normally it will not create any problem, but if there is some fault then we can have the difficulty in this particular frequency. However, large amplitude of b p f component and its harmonics can be generated in pumps or fan rotor system. If gap between rotating fan and the stationary diffuses is not equal and all the not equal all way round. So, in the whole circumference that gap is not equal then we can have this kind of large amplitude at blade passing frequency. These different gaps will cause air flow rate of the pump or fan to vary which makes the static and dynamic pressures of the blade changes as well. As it responds to the fluctuation of these pressure loads the larger amplitude of the b p f component will be generated. So, for a pure tone signal that let us say this is a signal sin omega t if its amplitude is modulated by another waveform like this. Then using trigonometric equation we can able to see that there will be not only the add and subtract of these two frequency modulation will take place. So, here as we have explained earlier also omega c is the carrier frequency and omega m is the modulation frequency. So, we can able to see these side bands in such a system. So, it is it can be clearly seen that the beside the carrier frequency component two side bands with frequency of the summation of two frequency and difference of this frequency are also present in the modulated spectrum. For fan vibration signal the b p f and its harmonics are the carrier frequency and the rotating speed of the is the modulation frequency. So, in the b p f and its harmonics are higher then the rotating frequency the amplitude modulation caused by 1 x vibration resulting side bands of b p f component as shown in the next figure. So, we can able to see that this is the b p f and there are side bands corresponding to 1 x r p m of the rotor speed. This is the second b p f higher harmonics of that and we have side bands of that and these are due to the unbalance 1 x and 2 x corresponding to the rotor speed. This is a typical fan. So, before and after installation the observation in the 12 blade axial fan. So, there are 12 blade axial fan and this we put this is taken from the spectra quest lecture notes. So, this is the obstruction which was put here. So, that we can get some kind of disturbance and this is a typical signal in this. So, we can able to see there is a 1 x 1 blade passing frequency is this one and there are side bands to that. This is component of the 16 x of the rotational speed. This is twice of the blade passing frequency peaks. So, this kind of spectra we can able to expect from such blades. In the present lecture I gave a brief introduction to the condition based monitoring which itself is a big subject and research area. So, in a single lecture is very difficult to cover all the aspect of the condition based monitoring. But I try to give various basic component of that particular condition based monitoring in which obviously the first component is the collection of the data. Second is the analysis of the data. Third is decision making based on the whatever the data we collected and analyzed. So, that was the overall overview of the condition based monitoring we presented in the present lecture. Apart from that few more case studies on gears and fans we presented and we try to see that how this condition based monitoring or especially looking at to the signal we can able to detect the fault in the rotating machinery. With this we basically conclude this particular subject of rotor dynamics lecture. So, generally this particular subject is taught as elective in our institute and this particular lecture generally we cover in 40 to 42 lectures. But with the power point presentation we could able to put more information and more illustrations in these lectures and hopefully it will be useful for the students as well as practicing engineers in India and all over the world. Now, I would like to acknowledge some of the people who were behind this preparation of this lectures. So, basically we can able to see all the amtech students and PhD students who took rotor dynamics course under me from 1999 up to 2012. They contributed in this particular subject. Their discussion with them were very helpful and because of the motivation from them only equitable to develop this particular lecture notes. Apart from this video recording and editing staff at CET they were very helpful in recording and editing of this particular video lectures. I am thankful to all of them apart from my family obviously because without their patience I cannot able to concentrate on developing this particular course. So, I acknowledge them also.