| Popis: |
The focus of this thesis is evaluating the feasibility of experimental modal analysis for use in an industrial quality assurance setting. Manufacturers are always looking for ways to improve their quality assurance processes to ensure higher customer satisfaction and a better bottom line. The challenge they face is that many testing methods are either time consuming or result in the destruction of the test specimen, neither of which are desirable outcomes.Finite element equations show that measured physical properties, such as mass and thickness govern the resonant frequencies of a plate. Because of this, vibrational frequencies should be able to indicate any abnormalities. A case study using an acoustic method has shown that visibly chipped or cracked metal plates vibrate at different resonant frequencies than their non-defective counterparts and that the frequency signatures differ enough that the defective plates can be reliably sorted. It is hypothesized that a hammer test can be used to identify non-visibly defective plates as in need of further testing for quality assurance. Because of the distributions of the measured physical properties, the frequency of each mode of vibration should be distributed normally across a sample of plates. In this application of the modal testing method, it is suggested that departures from normality can indicate that additional tests may be needed to assure quality of the plate.In this study, a sample of 31 steel plates was gathered and each plate was subjected to a uniformly performed hammer test, and the basic assumptions of linearity, time invariance, reciprocity, and observability were all checked. Using a frequency domain modal parameter estimation procedure and the autonomous modal parameter estimation feature in X-Modal III, the frequencies and shapes of nine modes of vibration were extracted for each plate. Frequencies from two of the plates were noticeably lower and were confirmed as departures from normality using a Shapiro-Wilk test. Modal vector complexity plots showed that the modes of one of these plates had a lower mean phase correlation than the rest of the plates. Lastly, an equation for the resonant frequencies of a plate was used to show that the departures from normality also did not fall within the measurement uncertainty range.Because of the departure from normality, lower mean phase correlations, and frequencies outside the uncertainty range, it was suggested that the two plates should be set aside for further testing in a manufacturing environment, demonstrating the feasibility of experimental modal analysis in a quality assurance setting. |
