The dynamic characteristics (resonance frequency, damping ratio and mode shape) of a mechanical system are usually estimated by fitting a model to the Frequency Response Function, a function derived from the input excitation (excitations) and output response (responses) of the system under investigation. The conventional modal analysis requires the measurement of the response (responses) as well as the excitation force (forces). The methods can be applied to a structure that can be easily excited into vibrations. The force is measured instantaneously by a load cell during its application to the structure. The response of the excitation is measured by an accelerometer (accelerometers) and from the input-output data, the dynamic characteristics of the system are measured. However, some mechanical structures like drag line machines and nuclear power plants are very difficult to excite and the forces acting on them during their normal operations are not possible to measure. However, the response of these structures can be measured quite comfortably. Several algorithms have been developed which estimates the dynamic characteristics of the system by analysing the response data only and have been applied to limited cases. This technique is known as the Output-Only meth od in the mechanical engineering discipline. In this study, some of these methods have been used to characterise a dragline boom for the first time.
This study has been carried out in three stages. In the first two stages both the conventional and Output-Only techniques have used to characterise an electrodynamic shaker and a model boom. Finally Output-Only methods have been used to characterise the dragline machine DRE23 of Peak Downs Coal Mine. The experimental work of this study has started with the calibration and understanding of the equipment as it is believed that the proper understanding of the equipment is the prerequisite for any experimental scheme. The algorithms used in this study are Fast Fourier Transform, the Complex Exponential (CE) Meth od, the Balanced Realization algorithm, Canonical Variate Analysis (CVA) and the Frequency Domain Decomposition (FDD) techniques. The latter three algorithms are classified as Output-Only algorithms. The CE meth od was chosen for three reasons:
1. It is one of the oldest methods primarily used for signal processing;
2. The CE method works very well under impulse excitation conditions; and
3. The CE method uses only the impulse response (not the excitation as used in FRFs calculation) and has similarity with the Output-Only methods.
The BR, CV A and FDD methods were chosen because of their recent interest among the researchers. The implementation was carried out in three stages. In stage one, Output-Only algorithms were applied to a 2-dof simple structure and the results obtained were compared with the traditional modal analysis ie. the CE method, FRFs plots, etc. The simple structure which was chosen is an LDS electrodynamic shaker, which has two resonance peaks at more than 10,000 Hz. apart, one at 35 Hz and the other at 10,351 Hz. Four types of excitations were used: namely, hand excited free decay (in this case the shaker table was pulled by hand and suddenly released), stepped sinusoidal, sinusoidal logarithmic chirp, random white noise excitation. The stepped sinusoidal excitations were not applied at resonance peak in the apprehension that they might damage the shaker but were applied just to check the whereabouts of the distinct modes. The plot of FRF (acceleration over Current) versus frequency indicates clearly that there are two modes: one between 27 and 30 Hz and the other after 8000 Hz. In this case, the shaker first mode was lowered because of the extra load placed on the shaker table.
Then the shaker was excited by hand and released suddenly to vibrate freely. The responses were measured and CV A, FDD and CE methods were applied to the responses. All three methods could detect the first mode of vibration. None of the methods gave any clue of the second mode; it is probable the hand excitation could not excite the structure at second resonance frequency, which is very high.
However, both the modes were identified with different excitations (noise and chirp) by the Output-Only methods. The conventional method could not pinpoint the second mode of the shaker. The FDD, BR and CVA methods identified the second mode.
After the implementation of the algorithms on the simple 2-dof structure we moved to the experimental boom which was built to simulate the dragline boom. Both the shaker and the hammer were used to excite the boom. Four types of excitations- namely, impulse (by hammer), chirp, random noise and simulated field excitations (the laterthree by the shaker) - were applied to the experimental boom. The test data obtained were analysed by both the conventional and Output-Only methods. The results obtained by both methods are quite comparable.
Finally the field tests data were analysed by the Output-Only methods. The conventional modal analysis of the dragline machine was not possible because the excitations of the boom due to normal working conditions could not be measured. The natural frequencies of the boom obtained by the output-only method were very close to those obtained by the FEM of the boom. However, mode pairing was not possible as the number of accelerometer readings required for the complete mode shape visualisation for this complex structure were far less than adequate. The MAC pairing of the CVA and FDD estimates has identified four global modes of the DRE23.
It has been revealed from the study that Output-Only methods estimate the dynamics of the structure quite accurately even in the cases where conventional methods fail.