VIBRATION ANALYSIS OF MACHINE FAULT SIGNATURE

Transcription

1 Abstract VIBRATION ANALYSIS OF MACHINE FAULT SIGNATURE 1 Amit Aherwar, 2 Md. Saifullah Khalid, 3 Hemant Kumar Nayak 1, 2, 3 Department of Mechanical Engineering, Anand Engineering College (SGI), Agra amit @yahoo.com Mob: Condition monitoring is becoming back bone of the industries with medium to high automation. These industries can t afford the downtime cost because of quantity of production loss per unit downtime. Vibration signature is the most versatile and highly responsive fault signature is normally used for identification of fault in rotating machines. In the present work we have used the Fast Fourier Transform of the time domain vibration signal to identify the characteristic fault frequency representation corresponding to the calculated fault frequency from the bearing with known fault. Signature pattern graphs so obtained are used as knowledge base for the future fault prediction by comparison of the fresh signature with the knowledge base. We have done the experiment with one healthy and three faulty bearings with fault at outer race, inner race, ball respectively. Result obtained were promising to identify outer race fault and ball fault but found difficult to discriminate between healthy and inner race fault bearing. Keywords: Condition Monitoring, Vibration Signatures, Fast Fourier Transform, Time Domain Signals, Frequency Domain Signals. 1. Introduction The prevention of potential damage to machinery is necessary for safe and reliable operation of process plants. Failure prevention can be achieved by sound specification, selection and design audit routines. When a failure occurs, then finding the absolute root cause of that failure is the main prerequisite to the prevention of future failure events. The extensive research has been carried out in the area of machine health condition monitoring (MHCM) and the work is still going on. During the twentieth century Machine Health Condition Monitoring has evolved from breakdown maintenance to the developing stages of diagnosis of machinery and its components. The future of MHCM is the design of smart machinery with the built-in diagnostic capabilities. The machine health condition monitoring is based on a systematic engineering approach of data acquisition, signal processing, fault pattern recognition and classification of fault s features. In many domestic and industrial applications, rolling element bearings are regarded as critical components with the strong influence on fundamental functionality of rotating machinery. For this reason,various measurement methods have been developed with the aim of detection and identification of faults in bearings,among these methods,vibration analysis has been established as the most common and reliable technique in order to detect the anomalous behavior. Most of the industrial rotating machinery work continuously for several hours. Therefore it is very important to find out the s of bearing that create different types of vibration characteristics. For this we have to calculate the characteristic vibration frequencies due to the bearings s, with the help of rotor speed and bearing geometry.for each type of characteristic frequency varies simultaneously. Here we are focusing on the outer race, inner race and ball 2. Literature Review M.Torbation, M.H.Kahaei and J.Poshtan[2004] presented a work in level dependence noise reduction technique the period of the impulsive vibration signals generated by healthy and faulty bearing.they also compared the results to the conventional FFT algorithm are shown the higher resolution of the applied technique in detecting bearing s. Jason R. Stack,Thomas G. Habetler and Ronald G. Harley[2004] has introduced in their research work the notion of categorizing bearing 487

2 Amit Aherwar, Md. Saifullah Khalid, Hemant Kumar Nayak fault as either single point as generated roughness. They shown fault classification according to the fault signature that are produced rather than by the physical location of the fault.they shown single point produce the four predictable characteristics fault frequency while faults characterized as generalized roughness produce unpredictable broad band changes in the machine vibration and stator current. James Robinson Emerson process management, Mitchell Illig IMI Sensors, Thomas Brown NSK corporation, Ray Limberg PCB used the accelometer to meter the peak g-level which rises sufficiently just before catastrophic failure can be used as an indication of failure. For this he has inducted fault in ball, inner race, outer race, under lubrication, over lubrication and studied peak g- level, corresponding to some speed for which alert time (peak g-level) and Alarm limit (peak g-level) are recommended value They concluded the peak g-level has proven through long term experiences with peak value to be both a reliable detector of the presence of a fault and a reliable indicator of the severity of the fault. Jason.R.Stack, Thomas G. Habetler, and Ronald G, Harley [2003] (7) has studied load and vibration level, variable speed and vibration level in comparison to base level.in there work they tested the bearing at different load level and speed having bearing life passed 90%.They concluded that when comparing vibration data to baseline values in variable speed measurement must be obtained at the same machine speed to provide a valid comparison. 3. Signature Analysis The word signature has been coined to designate signal patterns which characterize the state or condition of a system from which they are acquired. Signatures are extensively used as a diagnostic tool for mechanical system. In many cases, some kind of signal processing is undertaken on those signals in order to enhance or extract specific features of such vibration signatures. It is very important to consider type and the range of transducers used as pickup for capturing vibration signal. Signatures based diagnostic make an extensive use of signal processing techniques involving one or more methods to deal with the problem of improvement in the signal to noise ratio, say identification, data production and transformation etc. These techniques have been broadly classified in these areas namely: (a) Amplitude versus frequency; (b) Amplitude versus time; (c) Time waveform analysis; (d) Lissajous patterns (orbits); (e) Phase (relative motion) analysis; (f) Mode shape determination 3.1 Vibration amplitude verses frequency analysis The procedure of obtaining and displaying the amplitude of vibration for all frequency present is perhaps the most useful of all analysis technique. It is estimated that over 85% of the mechanical problems occurring on rotating machinery can be identified by displaying the vibration amplitude verses frequency data however, successful identification of the problem in a machine utilizing this or any other analysis technique requires that complete data to be obtained in a systematic way that will simplify interpretation. 3.2 Rotating amplitude verse frequency signature Some analyzer can be connected to a standard X-Y recorder to obtain graphical plots, or signature of machinery vibration amplitude verse frequency. On older analyzer it was generally necessary for the operator to manipulate the filter. Thus, as operator would manually adjust the filter over the frequency range, the recorder automatically plotted the amplitude verses frequency data Plotting vibration amplitude verses frequency signature has much advantage over manual analysis. First many sources of human error in absorbing and recording the data are eliminated, there is less chance of missing significant vibration frequency, and analysis time is greatly reduced by eliminating the need to fine tune to each frequency found. However as with tabular analysis for a complete analysis of vibration, plot must be made in horizontal, vertical and axial direction at each bearing of the machine. A recommended procedure is to record all three signatures for a particular bearing on the same sheet. This greatly simplifies data interpretation by concentrating all the vibration data for a particular bearing on a sheet. In almost all cases the amplitude resolution provided with data recorded in this way is quite adequate for making necessary 488

3 Vibration Analysis of Machine Fault signature comparisons. Another recommendation to simplify data interpretation is to use the same amplitude range for all vibration signatures wherever possible. It should be apparent that if the horizontal signature were plotted with the analyzer amplifier range selector on the 1.0 in/sec range, it would be possible to obtain three signatures which visually appeared quite similar. Therefore it is advisable to note and record the overall or filter-out amplitudes for each bearing before carrying out the analysis. Based on filter-out readings, one amplitude range should be selected, which is appropriate for all measurement points. With all signatures plotted on the same range, data evaluation and interpretation will be greatly simplified and far less confusing then if different amplitude ranges had been used. 3.3 Time Waveform Analysis The vibration signal is applied to the scope vertical input. The vertical axis on the CRT is scaled in amplitude. The horizontal axis of the scope is scaled in time, such as seconds or milliseconds. For example, the time waveform generated by the ive gear tooth would differ noticeably from that caused by unbalance. The unbalance would produce a sinusoidal waveform, whereas the ive tooth, being in contact for only a brief instant in the 1 x rpm cycle, would show a distinctive spike like appearance. In addition, due to the short duration of the tooth-excited vibration, it is doubtful that the filtered amplitude-versusfrequency analysis would reveal the true amplitude of the vibration. When a spike-like signal is filtered, the amplitude may be significantly attenuated. The oscilloscope allows the true peak amplitude to be observed. 3.4 Lissajous pattern analysis The resultant display on the CRT will be a plot of the total motion of the shaft within the bearing. Such plot or display are called lissajous pattern, and are also referred to a shaft orbit. In recent year, it has become common practice to install dual no contact (X-Y) pickups on the critical high-speed turbo machines. Made part on an electronic logic system, they provide redundant, fail-safe protection and avoid false shutdown if one of the pickup fail. 3.5 Phase analysis Another very useful analysis technique, which can help, detects and identifies machinery problems in the phase measurement and analysis. Phase measurements are normally expressed in degrees, where one complete cycle of vibration equals 360. Phase can be defined as that part of cycle (0-360) through which one part of a machine has traveled relative to another part of a fixed reference. However, from a practical standpoint, phase measurement simply provides a means of determining the relative motion of various part of machine. 3.6 Mode shape analysis Mode shape analysis is a technique of determining the vibratory or mode shape of a machine, a structure, piping or other components. To determine the mode shape or a machine or structure, vibration amplitude and phase reading must be taken at numerous measurement points and plotted on a chart to reveal the vibratory shape. This analysis technique is extremely useful for confirming resonance condition, identifying nodal points, and revealing sources of structural weakness. The time and effort to obtain mode shape information will often result in a clear indication of where structural modification are needed and may avoid costly and embarrassing trial and error approaches. 4. Data Acquisition and Analysis 4.1Experimental Setup Experimental data were collected from double row, self aligning bearings mounted on a shaft driven by a 3-phase electric motor (0-55KW, V) of a experimental setup. This Experimental setup has facility for easy removal of bearings to be tested.the bearings housing of the experimental setup has a facility for the placement of sensors on it. The motor connected with the shaft rated max 3310 rpm. For running at low load two disks are mounted on the both sides the one bearing housing, where experimental data is collected. 4.2 Instruments Used Vibscanner 489

4 Amit Aherwar, Md. Saifullah Khalid, Hemant Kumar Nayak It is a measuring instrument for the diagnosis and recording of machine conditions. The transducers required for this are already integrated so that the time consuming task of changing to different transducers and cables is no longer necessary. Vibscanner is also a data collector. In this experiment this used for speed and temperature measurement. We can use it to transfer the stored measurement results to our PC. Its comprehensive measurement and analysis functions and the convenient joystick for navigation make this handy instrument ideal for every day inspection routines. In this present work, instrument is used in recording Bearing temperature and RPM Dynamic Signal Analyzer: The model 2400 is a portable, battery operated, multi channel signal analyzer. This is controlled by the application packages which are downloaded in the analyzer. The Model 2400 Signal Analyzer includes the standard application package FFT Analysis which allows the Model 2400 to function as a digital oscilloscope and FFT spectrum analyzer. With the addition optional downloadable application package, this instrument can also perform continuous monitoring, transient data capture, balancing, structural analysis and routine data collection. This analyzer is capable to communicate with the computer system for the future analysis. geometry. The characteristic vibration frequency is the inverse of time between occurrences of bearing impulse. An outer race causes an impulse when a ball or roller passes the. And thus the formulas for all these frequencies are: Bearing Name Fault-1 Fault-2 Fault-3 Table 1: Bearing fault descriptions Bearing Defect Ball Defect Inner Race Outer Race Location of Number of s 2 balls in one row, one by one 3 s at each row. 3 mm distance, both rows with s 3 s at each row. 3mm distance, both rows with s Defect size 1*1mm, 0.1mm, deep 1*1mm, 0.2 mm, deep 1*1mm, 0.2mm, deep Table 2 Characteristic frequencies of different bearings used Bearing Bearing Frequency Multiples Name Defect at 2980 where fault to rpm,f (Hz) be identified Faulty-1 Ball Defect 261 3*f,4*f,5*f Faulty-2 Faulty-3 Inner Race Outer Race 385 3*f,4*f5*f 245 3*f,4*f5*f 4.3 Bearings Used The bearings used in the experiment are double row self-aligning by SKF have 13 balls in a row. The inner race diameter is 25mm, the outer race diameter is 52mm, and there are three types of faulty bearings used in our experiment and one healthy bearing for the comparison. The details of these faulty bearings are tabulated in table Analysis Accelerometer is used to acquire data from shaft. It is then feed to the dynamic data analyzer, which uses FFT transform to detect the characteristic frequency, and its harmonics present in vibrations acquired from test rig. The characteristic frequency for the bearing can be calculated from the bearing Fig 1: Line diagram of MFS 490

5 Vibration Analysis of Machine Fault signature The outer race frequency is given by: FOD bd n / 2 N / 60 1 cos The inner race frequency FID, the ball pass frequency of the inner race is given by: pd a) FFT of Filtered signal. b) Wavelet Transform. c) Artificial Neural Networks. These methods are found to show much better results and building an expert system makes it feasible for getting fast and accurate results. FID bd n / 2 N / 60 1 cos pd The ball frequency (or ball spin frequency) is given by FBD pd bd N / bd pd 2 cos 2 Where: = Contact angle; pd= pitch diameter; bd= ball diameter; n= number of balls; N = rotational speed in rpm. The characteristic frequencies of the different bearings, multiple of which the faults presence to identified are shown in the table (3). The characteristic frequency of the shaft is 50 Hz. While analyzing the signal, FFT in the Dynamic Signal Analyzer, Various peaks large and small were present in the desired multiples of the fault frequency. At the same time FFT of healthy bearing was also performed, it was analyzed that there was no peaks present in the signal corresponding to the characteristic frequencies of the fault Conclusion and Future Scope The signals are obtained from the three faulty bearings and one healthy bearing. During the analysis results obtained for the outer race fault bearing and ball fault bearing were found to be promising, but results for inner race fault bearing were found indeterminate in comparison with the healthy bearings. Thus a conclusion was drawn that much for more prominent results some other methods should be deployed rather than using Fast Fourier Transform. As by using Fast Fourier Transform the results obtained are incoherent so for better results characteristic frequencies related to faults can be obtained using following methods: Fig. 2: Experimental Setup Fig. 3: Waveform of Vibration for duration of sec of bearing Acknowledgement The authors would like to thanks Madhav Institute of Technology and Science, Gwalior, India for experimentation facility and their support throughout the work. A special thanks to Dr. Pratesh Jayaswal, Reader, Mechanical Engineering Department, MITS Gwalior, India, for the motivation throughout the work. References [1] Jason R.Stack,Thomas G.habelter and Ronald G.harley, Effects of Machine Speed on the Development and Detection of Rolling 491

6 Amit Aherwar, Md. Saifullah Khalid, Hemant Kumar Nayak Element Bearing Faults, IEEE power electronics letters, vol. 1, no. 1, march 2003 [2] Jason R.Stack,Thomas G.habelter and Ronald G.harley, Fault-Signature Modeling and Detection of Inner-Race Bearing Faults, 2003 IEEE. [3] Gary G.Yen, Health Monitoring of Vibration Signatures, IEEE Transactions on Control Systems Technology, Vol. 2, [4] Jason R.Stack,Thomas G.habelter and Ronald G.harley, Fault Classification and Fault Signature Production for Rolling Element Bearings in Electric Machines, IEEE transactions on industry applications, vol. 40, no. 3, may/june [5] Julian Blair and Dr. Amir Shirkhodaie, Diagnosis and Prognosis of Bearings Using Data Mining and Numerical Visualization Techniques, 2001 IEEE. [6] Cristina Cristalli and Jiri Vass, Automatic Identification of Mechanical Defects in Electric Motors Applied to Production Line Reality,CIMSA 2005 IEEE International Conference on Computational Intelligence for Measurement Systems and Applications Giardini Naxos, Italy, July [7] M.C.S. Young and McMahon, DSP (Digital Signal Processing) Application in the Condition Monitoring of Bearings [8] Loven Eren and Michael J.Devanex, Bearing Damage Detection Via Wavelet Packet decomposition of the Stator, IEEE transctions on instrumentation and measurement,vol.53,no.2,april [9] Kaptan Teotrakool, Michael and Leven Eren, Adjustable Speed Drive bearing Fault detection via Wavelet packet decomposition, IMTC-20006,April [10] Yu-Fang Wang and Peter J. Kootsookos, Frequency Estimation in the Fault Detection of Rolling Element Bearing. [12] Jason R. Stack, Thomas G. Habelter and Ronald G. Harley, Bearing Fault Detection via Autoregressive Stator Current Modeling, IEEE Transactions on Industry Applications, VOL. 40, NO. 3, MAY/JUNE [13] Heinz P. Bloch and Fred K, Geitner, Machinery Failure Analysis and Troubleshooting. About the Authors: Amit Aherwar was born in Gwalior (M.P.), India, in 1982, received M-Tech in Production Engineering from the Rajiv Gandhi Technical University, Bhopal, Madhya Pradesh, India, in He is currently assistant Prof. in Dept. of Mech. Eng. in Anand Engineering College (SGI) Agra (U.P.), India. His current research interests are condition monitoring, machine fault signature analysis, application of AI and fuzzy techniques in mechanical systems. He has published 04 international papers and more than 12 national publications in different journals and proceedings. Assistant Professor (Mechanical Engineering Department) in Anand Engineering College. Having academic and research experience of 7 years and industrial experience of 4 years. Expertise in Hybrid energy analysis and exhaust engine analysis. 492