Home HomeJoseph Mathew, Monash University (Australia)
Vibration condition monitoring is now a well accepted part of an effective plant-wide condition monitoring program. Several techniques have been developed in recent years to assist vibration engineers detect and diagnose faults in machinery (Mathew 1987). This paper is concerned with two of these newer techniques, viz, the application of chaos theory and neural networks to vibration monitoring. The article is presented in two parts; Part I is concerned with chaos theory and Part II describes an application of neural networks to rolling element bearing fault diagnosis.
sv970555.pdf (Scanned)
TOPG. Krishnappa, Integrated Manufacturing Technologies Institute (Canada)
Vibration analysis techniques are frequently employed to detect faults in mechanical components. The amplitude and phase modulation analysis techniques can be used to detect the faults in rotating machinery components, which include bearings and gears. This paper discusses the development of several parameters based on amplitude and phase modulations and their applications to detect faults in helicopter transmissions. The parameters based on amplitude modulations, absolute measure of the percentage modulation of the amplitude envelop, negative peak taunt to measure the number of times the envelop modulated and kurtosis of the envelop. The phase modulation parameters include, percentage modulation of phase, kurtosis of wrapped and unwrapped phase and kurtosis of the first derivative of phase modulations. The vibration data collected from the endurance test runs of OH-58A helicopter transmissions were examined to validate the effectiveness of these modulation parameters in detecting both incipient and advanced faults. Using a combination of these parameters, most of the faults were detected.
sv970384.pdf (Scanned) TOPBrian Rebbechi, Department of Defence, Australia (Australia)
Aircraft power take-off shafts operate at high speeds and in a difficult operating environment. This paper describes the results of an investigation into the dynamic behaviour of the F/A-18 Aircraft Mounted Accessory Drive gearbox drive system. The investigation was carried out to determine the cause of failure of the input bearing in this gearbox. Measurements of gearbox vibration and shaft dynamic orbits were carried out to determine the origin of the high dynamic loads, and to evaluate the system critical speeds. These measurements indicate that the system behaviour was determined by a large initial unbalance due to component clearances, and a dynamic magnification due to the proximity of running speed to the system critical speed. Possible system improvements and proposals for design changes are discussed, as are ways to provide an interim alleviation of the problem.
sv970523.pdf (Scanned) TOPIstvan Nagy, Hungarian Academy of Science (Hungary)
A recent study of the Predictive Maintenance Program for approximately 300, different types of rotating machinery in NPP Paks is reviewed. The program is based on an automated vibration analysis expert diagnostic system. The data measurement and evaluation methods, the fault identification accuracy, the expert system analysis steps and the rule-base development techniques are reviewed. The process of order normalization of vibration data, the accumulation of average data as a baseline for automated analysis, the cepstrum analysis for bearing wear detection are presented. The fundamental assumption used in expert system is that the health of the machine as a whole can be assessed by isolating the spectral data pertinent to each major component. Details of machine configuration such as minor component arrangement, types of bearings, gearbox details and coupling types are stored in knowledge base via a component coding scheme. The major component groups are defined and the proper identification of the major components of a given machine and the location of the test points are shown. The method of fault diagnostics using the expert system rule-base, the fault severity assessment and fault severity trending are demonstrated through concrete practical examples of machines (motor - coupling - gearbox - coupling - pump, and high - pressure turbine - low - pressure turbine - generator, etc.) of NPP Paks.
sv970506.pdf (Scanned) TOP
D. Ho, University of New South Wales (Australia)
R.B. Randall, University of New South Wales (Australia)
When vibration analysis is used for bearing fault diagnostics in gearboxes, the accelerometer measures both bearing vibration and extraneous noise which may be discrete frequency and/or random. The bearing vibration appears in the spectrum as broadband frequency components because of its random-like characteristics and these may be masked by the gearmesh frequencies and their sidebands which are more discrete than the bearing components. In such cases a signal processing technique called "Self Adaptive Noise Cancellation" (SANC) may be used to separate the discrete and random-like vibrations. However the performance of this technique is controlled by the time delay, order of the FIR filter and the convergence factor and the optimal values for these variables depend on the particular signal. This paper presents the effects of these three factors on the performance of SANC in separating simulated gear vibration signals (discrete frequency components) from simulated bearing vibration signals (broadband frequency components).
sv970495.pdf (Scanned) TOPKrystof Kryniski, Royal Institute of Technology, KTH (Sweden)
Over the past few years advances in electronics totally changed the concept of vibration monitoring. Currently available monitoring systems are capable of collecting and processing a large amount of data. It is quite obvious that the problem of today is not associated with the lack of information but with its extend. To avoid the information overload new methods need to be devised which are capable of processing the information in real time and extracting the relevant information. Presented approach promotes such a concept by introducing the design of a stand-alone monitoring device, which can be programmed to track a specific machine problem. The paper briefly reviews some elements of architecture of the TMS320C31 DSP processor, which are essential for understanding the methods used for machine fault detection and to appreciate their effectiveness. However, the major part of the work is focused on monitoring and processing methods devised for detecting cracks in the meshing zones of slow running gearboxes and to monitor their growth. Additionally,the concept presented is discussed with reference to monitoring a planetary gearbox.
sv970284.pdf (Scanned) TOP
A. Ghasempoor, Queen's University (Canada)
Thomas N. Moore, Queen's University (Canada)
J. Jeswiet, Queen's University (Canada)
On-line wear detection in turning operation is considered in this paper. A wear monitoring system based on hierarchical neural networks is suggested for this purpose. The changes in cutting force components are used for monitoring three wear components. The hierarchical neural network structure uses multilayered, feedforward, static and dynamic neural networks as specialized subsystem for each wear component to be monitored. Simulation studies are performed to investigate the overall suitability of the system.
sv970154.pdf (Scanned) TOP
Yujin Gao, University of New South Wales (Australia)
Robin Ford, University of New South Wales (Australia)
R.B. Randall, University of New South Wales (Australia)
Signature analysis deals with the extraction of information from measured signal patterns and this paper proposes a new procedure to enhance the mechanical signature related to a faulty element in a rotating machine. The enhancement operation is implemented in the time lag domain using envelope signals which are obtained using a multiple carrier amplitude demodulation technique. This procedure is studied theoretically. Also it is evaluated using simulated digital sequences and vibration measurements from a railway track and a paper machine.
sv970498.pdf (Scanned) TOP
Chong Won Lee, Korea Advanced Institute of Science and Technology (Korea)
Yun Sik Han, Korea Advanced Institute of Science and Technology (Korea)
Young Seob Lee, Korea Advanced Institute of Science and Technology (Korea)
In this paper, a new vibration signal processing technique is applied to a laboratory rotor system for characterization of its misalignment. This technique utilizes the directional spectra of the complex-valued vibration signals measured from two vibration transducers, placed perpendicular to each other. The directional power spectrum preserves the important directivity and shape information of whirling orbits, whereas the phase of the directional cross-spectrum indicates its inclination angle. Experimental results support that the directional power- and cross-spectra can be effectively used for diagnosis of the misalignment in test rig.
sv970407.pdf (Scanned) TOP
Fengfeng Xi, Integrated Manufacturing Technologies Institute (Canada)
Q. Sun, University of Calgary (Canada)
G. Krishnappa, Integrated Manufacturing Technologies Institute (Canada)
In this paper a new method is proposed to diagnose bearing defects based on pattern recognition of statistical parameters. This pattern recognition problem is described as transformations from the pattern space to the feature space and finally to the classification space. Based on trend analysis of various statistical parameters, four parameters Peak, RMS, CrestFactor and Kurtosis are selected to form a pattern space. A feature space is then formulated by a nonlinear transformation end its dimension is defined as two, for the purpose of displaying the results on the 2-D plane. Then, through a linear transformation a classification space is formed on which the data of different bearing defects can be clustered in different regions. Classification of bearing defects is done by the discriminan t function which is generated through a lemming process. The discriminant function can relate the different bearing defects to the different regions in the classification space. The proposed method is tested to diagnose the bearings with seeded defects and the results show that it is an effective method.
sv970356.pdf (Scanned) TOP
Bernard Garnier, METRAVIB RDS (France)
D. Vaucher de la Croix, METRAVIB RDS (France)
D Webster, Pacific Noise and Vibration (Australia)
Very quiet ships are now built not only for military purposes (minehunters, ASW frigates, submarines, but also for oceanography and fishery research. It is proven that fishing boats could catch more fish schools if they were quieter. However, the perfect vibro-acoustic control achieved by a careful design, the systematic introduction of rubber isolators, the use of large amounts of sound and vibration absorbing coatings and sound proofing materials is a costly investment to maintain. As organizing adequate sea trials for measuring the ship radiatcd noise is very demanding - and sometimes even impossible if the vessel is quiet and the sea noisy - the only realistic solution for controlling periodically, or even better permanently, the vessel signature is to reconstruct it from on-board measurements. Practical seaproven strategies for evaluating the sound radiation of a complete vessel from hull vibrations (accelerometers), machinery vibration (accelerometers), and machinery noise (microphones) will be presented. A sufficient emphasis will be given to the delicate balance between a machine per machine monitoring philosophy (already widely used for machinery health monitoring in industrial plants), and a global approach of the vessel as an integrated system. Another point to be presented with some details is the difficult task of data reduction - it is easy to see that processing at several kilohertz several dozen of sensors brings to handle megabits per second of raw data, when the final answer to provide relates to a binary question: is my instant signature OK? Specific system architectures can help to solve this difficulty, intrinsic to any dynamic monitoring system. Last but not least, the major issue of such a system is the acoustic calibration, i.e, the proper sensors localization strategy and associated database providing the solution for estimating the radiated signature from onboard measurements. An original holography technique is recommended and will be presented in a companion paper.
sv970291.pdf (Scanned) TOP
Ian Howard, Curtin University of Technology (Australia)
Adem Omerovic, Curtin University of Technology (Australia)
Laurie Morgan, Curtin University of Technology (Australia)
Journal bearings are used extensively on small and large turbogenerator units in the process and power generation industries. The ability to predict the condition of the rotor system using proximity probe measurements of the shaft motion in two orthogonal planes can lead to a reduction in maintenance costs and can also improve the operational safety of the high speed rotating machines. The output signals from the proximity probes are typically used to produce orbit plots representing machine shaft centre line motion which can be used as an indicator of general machine condition as well as for detecting specific machine malfunctions. This paper presents an introduction to the frequency analysis techniques which relate specifically to turbomachinery with journal bearings. It is shown how the complex discrete Fourier transform of the orthogonal orbit shaft motion produces the full asymmetrical spectrum, which further leads to the concepts of forwards and backwards shaft whirl. The full spectrum can be used on run up and run down tests to highlight changes in the dynamic amplitude and phase motion of the shaft and is a very useful diagnostic tool for turbomachinery. The techniques are illustrated by experimental examples.
sv970269.pdf (Scanned) TOP
Y Zhang, Changsha Communications Institute (China)
Z. P. Zhang, Changsha Communications Institute (China)
Y.C. Zhou, Wuhan Transportation University (China)
In this paper, the principle to diagnose the fault compressing rings with the block vibration signal is analyzed by considering of the excitation resources, the excited objects and the drift process of the rings: the equation to calculate the rings nature frequency in the cylinder is given; the simulation tests of the fault ring diagnosing are carried out by the two types of the diesel engines. The tests results coincide with the values obtained from theoretical calculation in trend, therefore the method to diagnose the fault compressing rings by the block vibration signal is got: In the domain of the low frequency of the block vibration signal there is a characteristic frequency which is closed to the single rings nature frequency in the cylinder, the PSD value at the characteristic frequency is grown, in the narrow frequency band including to the characteristic frequency the power value is grown obviously, then we can believe that there are fault compressing rings (usually it is the first ring).
sv970253.pdf (Scanned) TOP
Takeo Yoshioka, Mechanical Engineering Laboratory (Japan)
H Mano, Tokyo University of Agriculture and Technology (Japan)
This paper describes the principal of an acoustic emission source location method for a radial rolling bearing and proof of the location method. The method locates contact points between balls end an inner ring or an outer ring within a loading zone of a radial rolling bearing under a radial load at the moment of acoustic emission generation as possible acoustic emission sources. The location result is shownas a histogram which has several peaks and is symmetrical at the highest peak after locating repeatedly. In the location result the X-axis shows the position on a raceway of the inner race or outer race, the Y-axis is the number of acoustic emissions. The position of the highest peak agrees with the position of the actual acoustic emission. The principal of the location method was proven by rolling contact fatigue tests of radial roiling bearings under a radial load. The location system was comprised of acoustic emission measurement equipment and position detectors of the inner race and the balls. The AE sensor and the detected sensors were fixed on the housing of the test bearing. The locator in the system decided the source position of acoustic emissions under signals from the equipment and the detectors. The located positions indicated by the method agreed with the positions of spallings. It was conformed in the rolling contact fatigue tests that acoustic emission was caused by propagation of rolling contact fatigue crack and occurred before the generation of spalling.
sv970251.pdf (Scanned) TOPAndrzej Wilk, The Silesian University of Technology (Poland)
Tooth working surfaces wear in wheels and other transmission elements at work increases dynamic loads. Optimum symptoms of dynamic load changes were sought for diagnosing in the elaboration. Simulating examinations of gear wheels with spur or helical teeth were carried out. Non-linear dynamic models of a pair of gear wheels were chosen as the basis of examinations. Their changing stiffness meshing, interteeth backlash and teeth errors were taken under consideration. It has been established that there is a linear dependence between (Ku-1=Pd) additional dynamic load and the effective value of circumferential vibration accelerations of wheels. Vibration measurements, most frequently for selected points of housing, are carried out in industrial gear weels. The examination of a pair of gear wheels working at circulating power system of the test stand were carried out. Circuit vibrations of rotating gear wheels and those of selected points of the housing were measured and analyzed simultaneously at the test stand by piezoelectric transducer. Marginal correlation between non-filtrated effective values of vibration in transmission housing and circumferential vibrations of gear wheels has been observed. The correlation has increased significantly after selecting the housing vibration signal. This selection eliminated the dominating frequencies of free vibrations in the measured points.
sv970238.pdf (Scanned) TOPWen Yi Wang, Monash University (Australia)
In the diagnosis of rolling bearing defects, the envelope spectrum technique is regarded as an effective method. Ensemble averaging of envelope spectra can be used to further enhance the detectability of the technique. However, in cases where signal-to-noise-ratio (SNR) is poor, the inherent non-linearity of the enveloping process limits the effectiveness of this technique. In this paper, the synchronous autocorrelation technique is proposed to detect bearing defects under poor SNR conditions. The idea comes from the synchronous demodulation of amplitude modulation (AM) signals in communication systems. Using this technique, the original vibration signal is first bandpassed at a chosen high frequency resonance to produce the impulsive AM signal related to faulty bearings. The instantaneous angle information of the AM carrier is identified using Hilbert transform and the AM signal is then synchronously demodulated. The resulting signal may be resampled at a lower rate if required and then autocorrelated. Because this technique only involves linear operation, the signal identity should be maintained regardless of the noise level. Ensemble averaging can also be used to improve the autocorrelation estimate. Finally, the defect period will be identified in the autocorrelation function. The effectiveness of the synchronous autocorrelation technique is demonstrated in this paper using both numerical and experimental data. It is found that the synchronous autocorrelation technique provides good results for a SNR of -11dB.
sv970236.pdf (Scanned) TOPStanislaw Radkowski, Warsaw University of Technology (Poland)
The work undertakes an issue of estimating and predicting the level and structure of the vibroacoustic signal generated by a pair of toothed wheels and propagated by the structure of the gears housing. Taking into account the existence of a number of internal disturbances in a toothed gear, apart from the external ones, we have built a model of generation of a disturbed vibroacoustic signal which refers to the friction power. Assuming that the basic role in transmission of such diagnostic information is played by the phenomena of amplitude and phase modulation of the vibroacoustic signal, the author points to the significance of the signals non-linear components. For that reason in this work a model of signal in the form of a Volterra series is assumed, which enables examination of the influence that disturbances have upon non-linear components, and assessment the technical condition of the assembly based on the results from demodulation of choosen frequency bands. At the same time, utilizing the higher - order spectral analysis method, a procedure to determine the changes in the spectrum structure caused by a local failure was developed. Additionally, the author took into account the precision of manufacturing of toothed wheels and the influence that operating wear and tear of individual elements of a unit has upon variations of the general vibroacoustic activity. In the final part of the work is presented an algorithm of simulation of signal generation that accounts for the effect of apparent interference of teeths flexibility, for the related disturbances of contact (too early and too late contact) in the dynamic model, and for the changes of period of meshing and contact, as well as for the related phenomenon of amplitude and phase modulation in the kinematic model.
sv970197.pdf (Scanned) TOP
Seung P. Lee, LG Electronics Inc (Korea)
C.O. Ahn, LG Electronics Inc (Korea)
H.S. Rew, LG Electronics Inc (Korea)
S.C. Park, LG Electronics Inc (Korea)
Y.M. Park, LG Electronics Inc (Korea)
D.I. Rhee, LG Electronics Inc (Korea)
A sirocco fan, as a compartment of a window type room air conditioner, has been used in the indoor part to suck the warm room air and to blow the chilled air into the room. Because the air conditioner goes on and off according to the room termperature repeatedly, the sirocco fan also repeats its drive and repose action as long as the air conditioner operates. And since a sirocco fan usually has been made by engineering plastic, it is tend to deform and crack easily as comparing with the metals. Thus, there are possibilities of the fracture by the periodic loading, fatigue, and natural frequency. To understand the mechanism of failure in a fan, we chose a kind of sirocco fan and experimented. After 150 days continuous operation, we could see the cracks. To check the possibilities of the failure due to the periodic loading and resonance, both numerical and experimental approaches were carried out. In case of fatigue, numerical method based on F.E.M. is applied. Based on von Mises failure criterion, we can conclude that the sirocco fan is safe in case of the fatigue. And the other possibility is the resonance between a natural frequency and the aerodynamic force. Experiment is performed to see the natural frequency and numerical method based on the V.E.M. is applied. The numerical calculation is going on. In this paper, the details of calculation results will be included and discussed.
sv970123.pdf (Scanned) TOP
Wenhu Huang, Harbin Institute of Technology (China)
Xuexiao Wang, Harbin Institute of Technology (China)
Xinwei Jiang, Harbin Institute of Technology (China)
In this paper the failure records of spacecrafts over the world are reviewed to confirm the urgency of the diagnosis of faults. The special features of the dignostic technology for a spacecraft are discussed, and accordingly the problems of diagnostic technology for a spacecraft system are reviewed and prospected such as: the strtegy and architecture of the intelligent diagnostic system, the hierarchical structure of the knowledge model, the inexact reasoning and heuristic reasoning, the distributed diagnostic technology and so on. Finally, the prospects of the intelligent diagnosis technique for a spacecraft are reviewed.
sv970084.pdf (Scanned) sv970084.pdf (From Postscript) TOPJanusz Gardulski, The Silesian University of Technology (Poland)
Mathematical model of suspensions together with its geometrical dimensions hasbeen presented in this paper. It is a 6 mass discreet dynamic system with 36 freedom grades. Its concentrated masses are connected by elastic and attenuating elementsof non-linear characteristics. Apart from springs and silencers there are also wheel guiding elements ( rookers, Iateral control rods). The model is activated to vibrations by road unevenness. Input functions are of stochastic nature. Changes of many element parametres of the model were taken under consideration in simulating examinations (type of suspension, mass distribution, stiffness and attenuation characteristics changes, etc.). The results were presented in a graphic form together with comparative analysis.
sv970058.pdf (Scanned) TOP
Yu-Fang Wang, Australian National University (Australia)
Peter J. Kootsookos, MV Technology Ltd. (Ireland)
In this paper, a general model of the signal from faulty rolling element bearings under the condition of a heavy load is given. The envelope-autocorrelation of this proposed model in the case of very low shaft speed is given with a mathematical description. The simulated signals of rolling element bearings under the condition of a heavy load with inner race fault, outer race fault, and roller fault are generated using the model. In the power spectrum of signals, the characteristic frequency and its harmonics are submerged in the white noise, but they are obvious in the envelope-autocorrelation and envelope-autocorrelation power spectrum. It is demonstrated that the envelope-autocorrelation and its power spectrum are effective as to a fault detection technique.
sv970234.pdf (Scanned) TOP
Yu-Fang Wang, Australian National University (Australia)
Peter J. Kootsookos, MV Technology Ltd. (Ireland)
Faulty rolling element bearings under very low shaft speed and light load exhibit vibrations which possess periodic envelope-autocorrelations. The main frequencies of this envelope-autocorrelation are the fault characteristic frequency and its harmonics. In this paper, one of the improved Notch Filtering Techniques with a designed filter is used to estimate the fault characteristic frequency and its harmonics. The designed filter is used to remove the estimated frequency. With this technique, the fault characteristic frequency and its harmonics can be accurately estimated.
sv970295.pdf (Scanned) TOP