SHAFT MISALIGNMENT PREDICTION ON BASIS OF DISCRETE WAVELET TRANSFORM

International Journal of Mechanical Engineering and Technology (IJMET) Volume 9, Issue 7, July 2018, pp. 336 344, Article ID: IJMET_09_07_038 Available online at http://www.iaeme.com/ijmet/issues.asp?jtype=ijmet&vtype=9&itype=7 ISSN Print: 0976-6340 and ISSN Online: 0976-6359 IAEME Publication Scopus Indexed SHAFT MISALIGNMENT PREDICTION ON BASIS OF DISCRETE WAVELET TRANSFORM Amit Umbrajkaar Research scholar, Department of Mechanical Engineering, Sathyabama Institute of Science and Technology, Chennai, Tamilnadu, India A. Krishnamoorthy Professor, Department of Mechanical Engineering, Sathyabama Institute of Science and Technology, Chennai, Tamilnadu, India ABSTRACT In different kind of industries like automobile, ship building, manufacturing, textile industry, rotating machinery are observed in major stake. In due course of time it approaches to service condition due to unbalance, misalignment, insufficient lubrication, bearing fault etc. It is seen that misalignment caused in rotating machinery is major contributing part of annual maintenance of industry. During installation of machine though alignment in static condition is ensured, misalignment gets introduced in due course of time of service period. Several attempts are made for analysis of misaligned shaft in rotating machinery using Fast Fourier Transform (FFT). The FFT is effective in analysis of vibration signals with frequency domain. It has a constraint in reveling data from non-stationary signals. It is ineffective in multilevel signal decomposition of vibration signals. This constraint can be minimized by using Discrete Wavelet Transform (DWT).Selection of appropriate mother wavelet is mainly focused in DWT analysis. Different mother wavelets are compared for different output condition in the form of vibration signals. Exact selection of mother wavelet helps in predicting misalignment involved in any unknown system whose aligned condition is to be checked. It is observed that result obtained with this approach are very well in close proximity. Key words: Misalignment, Discrete Wavelet Transform (DWT), Overall Vibration Level(OVL), Misalignment Prediction. Cite this Article: Amit Umbrajkaar and A. Krishnamoorthy, Shaft Misalignment Prediction on Basis of Discrete Wavelet Transform, International Journal of Mechanical Engineering and Technology 9(7), 2018, pp. 336 344. http://www.iaeme.com/ijmet/issues.asp?jtype=ijmet&vtype=9&itype=7 1. INTRODUCTION In all cases of industries where use of rotating machinery is applicable, maintaining aligned state in operating condition is challenging task. Different causes which may introduce misalignment in machine are viz. inertial forces, meeting condition in assembled parts, http://www.iaeme.com/ijmet/index.asp 336 editor@iaeme.com

Shaft Misalignment Prediction on Basis of Discrete Wavelet Transform improper lubrication and unbalance present. All kind of machines are ensured well for aligned state in their static condition. In due course of their service time, the misalignment get introduce due to one of the reason as cited above. The kind of misalignment considered here are offset and angular type. Condition monitoring of any rotating machine involves monitoring output vibration signals. The obtained vibration signals are compared with healthy condition signals of machine and misalignment induced with its degree of severity can be predicted. This helps to avoid catastrophic failure which may be caused during run time failure. Several attempts are done for the analysis of misalignment in rotary machines. In majority of cases FFT is used for the analysis of output vibration signals. The FFT gives analysis on frequency based in frequency domain. Each type of fault has been observed with its unique frequency as shown in table1. When different type of faults are existed in common system, the frequency domain results may observed with many faults reporting at same frequency. The fault wise separation and identification of faults can not be obtained with frequency domain analysis. Hence, the unique features of fault identification and fault occurrence of DWT technique is recommended in this work. D.F. Ramirez et.al have presented vibration analysis of misaligned shaft accounting effect of type of coupling and shaft stiffness. Author have made an attempt to analyze shaft misalignment with the help of frequency spectrum and phase shift [1]. Diangni Huang have claimed that any presence of misalignment in the rotary machine gives excessive raise in torsional vibration response. These torsional vibrations are measured successfully by DK-II approach [2]. Oliver Tanks proposed a method of detecting shaft misalignment of wind turbine using thermal parameters with the help of infrared thermometer. Author claims this method as one of effective online control and measuring of data for misalignment. The results predicted are based on temperature variation of various parts. It is necessary to point out here that temperature variation of different parts can also be effect of other faults associated with system. This is not clearly mentioned how only misalignment can be correlated to observations [3].Anthony Simm et.al have made an attempt to visualize run time alignment detoriation in close limit upto 05mm using laser beam arrangement[4]. F Gu and A Ball et.al [5], have proposed a method of condition monitoring of aligned shaft on precise scale using MEMS based wireless accelerometer. The output signal accuracy has been compared with different sensors such as laser vibrometer, shaft encoder and an accelerometer. Different levels of misalignment are diagnosed and observed clear distinction in their measurement. The shaft misalignment can also be diagnosed on the basis of motor current signature analysis using machine fault simulator [6]. L.D.Mitchellet.al,have made an attempt to understand vibration response of coupled disc under misaligned condition. Author verified 2X and 4X as a dominant frequency to misaligned state. This has also been verified theoretically and experimentally[7]. Many times in vibration signal analysis, mixed mode results are seen i.e. more than one faults get reported at same frequency. In such cases, fault separation at such frequency can be done with analysis of non-stationary signals using continuous wavelet transform (CWT) and wavelet packet transform(wpt). This has been observed as an effective tool compared with FFT results [8]. Effects of flexible coupling on vibration pattern in misaligned condition has been studied and result obtained are verified through FEM analysis. An analysis of misaligned coupled shaft has been presented considering an effect of bending moment and shaft defection [9]. Jose M Bossio have studied effect of angular misalignment for induction motor. Author gave an analysis of power and current variation due to misalignment in dynamic condition [10]. In theoretical analysis of misalignment, modelling of shaft is done in consideration with Timoshenko beam. To bring all accountability in analysis 16 degrees of freedom is considered. Experimental results are verified with orbit http://www.iaeme.com/ijmet/index.asp 337 editor@iaeme.com

Amit Umbrajkaar and A. Krishnamoorthy analysis to determine presence of misalignment. The inner and outer loops observed are an indication of parallel and angular misalignment present in coupled rotors [11]. Different techniques fault diagnosis of electrical systems is proposed. These methods of fault diagnosis are based on pattern recognition, wavelet analysis and Artificial Intelligence [12-18]. It is found that in majority of rotary machinery cases misalignment analysis is done with FFT and FEM technique. The information in non-stationary signal cannot retrieved effectively in FFT analysis. Also multilevel decomposition of signal is not possible with FFT for misalignment prediction. It loses the necessary information available with time domain. Hence, Discrete Wavelet Transform (DWT) is used for analysis of output vibration signal. It involves selection and application of different mother wavelet for prediction of shaft misalignment. Table 1 Faults and related frequency (Courtesy ISO 10816 Mechanical vibration) Sr. No Mechanical Fault Related frequency 1 Primary Cause: Oil Whirl ½ X RPM Secondary cause: Bad drive belt, Sub-harmonic resonance, background vibration 2 Primary Cause: Unbalance 1X RPM Secondary cause : Eccentricity in gears and pulley,, Bent shaft, Electrical interference, Misalignment 3 Primary Cause : Misalignment( high axial vibration) 2X RPM Secondary cause: Mechanical Looseness, unbalanced reciprocating forces, bad belts, Mechanical Looseness 4 Primary cause : Crack,Combination of Misalignment Secondary cause: Mechanical looseness or excessive clearance 3X RPM 2. METHODOLOGY The outline of flow process considered for shaft misalignment analysis is as depicted in fig(1). The output signals are collected at second bearing with provision of signal measurement in all three directions viz. x, y and z. These output vibration signals are sensed at second bearing by an accelerometer. The output signals from all three directions are given as input to DWT. This is a suitable tool for analysis of non-stationary signals and multi-level decomposition of vibration signals. In DWT analysis, high pass and low pass filters are used which gives an output of signal in the form of Approximate coefficient (Ac) and Detailed coefficient (Dc) respectively. These are coefficients related with information in time domain and frequency domain respectively. In entire assessment of signal information Approximate coefficient (Ac) is referred to correlate with fault information. The expression for Ac and Dc are mentioned by equation (1) and (2). [ ] (1) [ ] (2) Where, P and q are filter coefficient, m is number of sample and i is shifting parameter. http://www.iaeme.com/ijmet/index.asp 338 editor@iaeme.com

Shaft Misalignment Prediction on Basis of Discrete Wavelet Transform Figure 1 Process outline for DWT analysis 3. EXPERIMENTAL SET UP An experimental set up shown in Fig. (2), consists of heavy base plate supported with the vibration isolator pad. The motor base plate is assembled with foundation plate. The lateral slots provided at common interface and base plate facilitates to introduce offset and angular misalignment externally. The displacement of motor plate in lateral direction is controlled with fine screw provided with additional firm locking between the plates. The accuracy of any lateral displacement is monitored through dial gauge. The rotor and motor shaft is coupled together by flexible coupling. In static condition, an alignment is tested by face and rim method. The face and rim dial is facilitated through separate fixture designed as shown in Fig. (3). The face arrangement suitably tells about angular misalignment and rim arrangement explains about offset condition of meeting shaft. These conditions are verified in static condition of machine. Initially, at the installation machine may have unbalance, runout or misalignment present in it. In order to refrain setup from all other faults, the output shaft has tested on multiplane dynamic balancing machine. Figure 2 Experimental set up The setup is then tested with face and rim method to ensure aligned condition. An accelerometer (model 352B70) is used to sense output vibration signals in all three directions. The vibration signals obtained through accelerometer is collected in Data Acquisition System (DAS). The DAS contains with four channel digital oscilloscope. A variable frequency drive (VFD). The required constant speed has been delivered through VFD to 3ɸ A.C. induction motor. The setup has been developed for speed range up to 3000 rpm. The facility in setup has been provided to facilitate to induce artificial misalignment. The range for misalignment is 0- http://www.iaeme.com/ijmet/index.asp 339 editor@iaeme.com

Amit Umbrajkaar and A. Krishnamoorthy 0.2 with incremental value of 0.02mm. The different speed considered for testing is 500, 740, 1000, 1440, 1800 and 2100 rpm. The combination of different speed and misalignment is prepared to collect data exhaustively for analysis of offset and angular misalignment. Figure 3 Fixture for Face and Rim method Figure 4 Vibration signal at healthy and 0.03mm offset at 1800RPM. Figure 5 Vibration signal at different conditions http://www.iaeme.com/ijmet/index.asp 340 editor@iaeme.com

Shaft Misalignment Prediction on Basis of Discrete Wavelet Transform http://www.iaeme.com/ijmet/index.asp 341 editor@iaeme.com

Amit Umbrajkaar and A. Krishnamoorthy 4. RESULT AND DISCUSSION The pattern of few sample vibration signal are presented as shown in Fig.(4). It is seen from theoretical analysis that misalignment present in system shows higher overall vibration level (OVL) at 1X, 2X operating frequency. It is also revealed that OVL in axial directions are dominating for angular misalignment and more in lateral directions for offset misalignment. The sample vibration signal obtained from experimental setup are shown in Fig.(5).The obtained signals confirms theoretical conclusion that OVL are seen at 1X, 2X operating frequency for misalignment problem. In order to evaluate misalignment cases more than 2000 samples are considered for different operating conditions. It is necessary to get information associated with non-stationary signal for accurate fault prediction. In order to retrieve information from non-stationary signals for fault detection, a second level decomposition is applied and 500 detailed coefficient (D c ) are obtained. The packet size of 500 is used for analysis. The Maximum, minimum, mode, median, mean are the featured values of detailed coefficient which may considered in selection of suitable mother wavelet. The Max value of detailed coefficients are considered for further selection of suitable mother wavelet. The accurate selection of mother wavelet is key part in prediction of misalignment using DWT. The extracted feature of D c for different mother wavelet for different misalignment condition is mentioned as shown in Fig (6-15). These results are obtained for various levels of decomposition. The Maximum feature obtained for different values of detailed coefficient are as shown in Table (2). Table 2. Mother wavelet comparison with Max feature values Sample signals with condition Mother Wavelet 500 1800 1800 RPM, 1800 RPM, RPM RPM 0.03 mm offset 0.09 mm offset BIOR1.1 0.118 0.112 0.300 0.214 BIOR1.3 0.160 0.200 0.384 0.244 COIF1 0.163 0.120 0.420 0.230 COIF2 0.151 0.121 0.370 0.213 COIF3 0.144 0.130 0.352 0.202 DMEY 0.110 0.138 0.298 0.161 HAAR 0.115 0.117 0.298 0.214 RBIO1.3 0.126 0.197 0.344 0.276 SYM2 0.149 0.212 0.343 0.365 DB2 0.149 0.212 0.343 0.365 http://www.iaeme.com/ijmet/index.asp 342 editor@iaeme.com

Shaft Misalignment Prediction on Basis of Discrete Wavelet Transform 5. CONCLUSIONS It is seen that to obtain detailed information of fault occurrence and nature of fault, an information associated with non-stationary part of vibration signal is equally important. The multi-level decomposition of non-stationary signal has been done with the help of DWT technique. Proper selection of mother wavelet helps in close prediction of misalignment. Different mother wavelet on the basis of Maximum extracted feature are compared. It seen that, after many trials DB2 and SYM2 are suitable mother wavelet for misalignment prediction for considered sample of analysis. REFERENCES [1] P.N. Saavedra, D.F. Ramirez, A vibration analysis of rotors for the identification of shaft misalignment, part I and II, Journal of Mechanical Engineering science, 218(9), 2004. [2] Diangui Huang, characteristics of torsional vibrations of a shaft system with parallel misalignment, Proceeding of Mechanical Engineering science, vol.219, part c, 2005. [3] Oliver Tanks, Qing Wang, The detection of wind turbine shaft misalignment using temperature monitoring, CIRP Journal of manufacturing science and technology, 2016. [4] Anthony simm, Qing Wang et.al, Laser based measurement for the monitoring of shaft misalignment, Measurement,87, 2016. [5] L.Arebi. et.al, A Comparative study of misalignment detection using novel wireless sensor with conventional wired sensor, IOP publishing, Huddersfield, UK, 2012. [6] Alok kumar. et.al, shaft misalignment detection using stator current monitoring, International journal of advanced computer research, vol.3, Issue 8, March 2013. [7] D.L. Dewell, L.D. Mitchell, detection of misaligned disk coupling using spectrum analysis, Journal of vibration, acoustic, stress and reliability in Design, ASME, vol.106,1984. [8] F.Al.Badour, M.Sunar et.al, Vibration analysis of rotating machinery using time frequency analysis and wavelet techniques, Mechanical systems and signal processing, vol.25, 2011. [9] B.S. Prabhu, A.S. Sekhar, Effects of coupling misalignment on vibrations of rotating machinery, Journal of sound and vibrations, 185(4), 1995. [10] Jose M. Bossio et.al, Angular misalignment in induction motor with flexible coupling, IEEE, 2009. [11] Tejas Patel, Ashish Darpe, Vibration response of misalignment rotors, Journal of sound and vibration, 325(2009). [12] R Ubale, R. B. Dhumale, S. D. Lokhande, Open switch fault diagnosis in three phase inverter using diagnostic variable method, International Journal of Research in Engineering and Technology, Vol. 2, iss. 12, pp 636-640, 2013. [13] Lalit Patil, R. B. Dhumale, S. D. Lokhande, Fault Detection and Diagnosis Technique gor Power Inverter using Discrete Wavelet Transform, International Journal of Electronics Circuits and systems, Vol. 3, iss. 2, pp 174-178, June 2014. [14] M. R Ubale, R. B. Dhumale, S. D. Lokhande, Method of Open Switch Fault Detection in Three Phase Inverter using Artificial Neural Network, International Journal of Research in Science and Advance Technology, Vol. 3, iss. 3, pp 78-82, June 2014. [15] R. B. Dhumale, S. D. Lokhande, N. D. Thombare, M. P. Ghatule, Fault Detection and Diagnosis of High Speed Switching Devices in Power Converters, International Journal of Research in Engineering and Technology, Vol. 4, iss. 2, pp 253-257, Feb 2014 http://www.iaeme.com/ijmet/index.asp 343 editor@iaeme.com

Amit Umbrajkaar and A. Krishnamoorthy [16] R. B. Dhumale, S. D. Lokhande, Diagnosis of multiple open switch faults in three phase voltage source inverter, Journal of Intelligent & Fuzzy Systems, vol. 30, no. 4, pp. 2055-2065, 2016. [17] R. B. Dhumale, S. D. Lokhande, Neural Network Fault Diagnosis of Voltage Source Inverter under Variable Load Conditions at Different Frequencies, Measurement, Available online 26 April 2016. [18] R. B. Dhumale, S. D. Lokhande, Condition Monitoring of Voltage Source Inverter, International Journal of Emerging Trends in Electrical and Electronics Vol. 11, Issue. 6, pp. 21-26 Ocobert-2015. [19] Priya Sahu and Dr. Paresh Rawat, VLSI Architecture For Discrete Wavelet Transform Using CSD Based Technique, International Journal of Electronics and Communication Engineering and Technology, 7(6), 2016, pp. 48 55. http://www.iaeme.com/ijmet/index.asp 344 editor@iaeme.com