Woo, Wing Hon (2018) Development Of Moiré Fringe Recognition System Using Artificial Neural Network For 2-D Displacement Measurement. Masters thesis, Universiti Sains Malaysia.
|
PDF
Download (283kB) | Preview |
Abstract
Pelbagai kaedah telah dicadangkan untuk mendapatkan maklumat anjakan dalam analisis corak moiré. Kaedah-kaedah ini boleh dikategorikan kepada analisis manual oleh inspektor manusia, kaedah komputasi dan kaadah analisis berasaskan imej. Analisa manual terdedah kepada ralat manusia kerana ia bergantung kepada keputusan manusia dalam analisa corak moiré. Penggunaan kaedah pengiraan dalam analisa corak moiré adalah terhad kepada corak moiré yang dihasil daripada parutan berfrekuensi tinggi yang sinusoid. Dalam kaedah berasaskan analisis imej, Algoritma yang kompleks menyebabkan butir-butir halus dalam corak moiré terhilang dalam operasi pra-proses imej. Situasi ini menyebabkan ketidakpastian dalam analisa corak moiré. Untuk mengatasi kelemahan yang disebut di atas, kaedah rangkaian saraf buatan (ANN) dicadangkan untuk sistem pengenalan corak moiré dalam pengukuran anjakan 2-D. Sistem pengenalan corak moiré terdiri daripada dua ANN dengan dua tugas yang berbeza iaitu (i) penentuan pusat pinggiran moiré dan (ii) penentuan kesipian berdasarkan corak moiré. Kaedah ANN dibandingkan dengan kaedah analisa grafik (GAM), sejenis kaedah analisa berasaskan imej, dari segi ketepatan dan masa pengiraan untuk pengukuran anjakan 2-D pola moiré. The experiments prove that ANN approach has a higher accuracy to GAM with mean errors with 95% confidence of 0.068 ± 0.013 mm for eccentric magnitudes and 1.85 ± 0.465º. An improvement of 66.18% in the computation time is also reported in the comparison. A straightforward solution for the moire fringe recognition system of circular grating moire pattern is achieved using ANN approach. _______________________________________________________________________________________________________ Various methods have been proposed in the analysis of moiré pattern. These methods can be categorized into manual inspection by human inspector, computational methods and image analysis based methods. Manual interpretation of moiré patterns is prone to human errors as it is highly dependent on the decision of the human inspector. The computational methods are lack of flexibility as they are limited to high frequency gratings which are sinusoidal in the transmittance of grating. As for the image analysis based methods, complex algorithms can unintentionally remove the fine details in the moiré patterns and cause uncertainty in the analysis. To overcome the above mentioned drawbacks, an artificial neural network (ANN) approach is proposed for a moiré fringe recognition system in 2-D displacement measurement. The moiré fringe recognition system consists of two ANNs with two different tasks : (i) the determination of moiré fringe centers of the circular grating moiré patterns and (ii) the determination of eccentricity magnitudes and eccentricity directions of the circular grating moiré patterns. The ANN approach is compared to graphical analysis method (GAM), an image analysis based method, in terms of accuracy and computational time for 2-D displacement measurement of circular grating moiré patterns. The experiments prove that ANN approach has a higher accuracy to GAM with mean errors with 95% confidence of 0.068 ± 0.013 mm for eccentric magnitudes and 1.85 ± 0.465º. An improvement of 66.18% in the computation time is also reported in the comparison. A straightforward solution for the moire fringe recognition system of circular grating moire pattern is achieved using ANN approach.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | Full text is available at http://irplus.eng.usm.my:8080/ir_plus/institutionalPublicationPublicView.action?institutionalItemId=4702 |
| Subjects: | T Technology T Technology > TJ Mechanical engineering and machinery > TJ181-210 Mechanical movements |
| Divisions: | UNSPECIFIED |
| Depositing User: | Mr Mohd Jasnizam Mohd Salleh |
| Date Deposited: | 13 Jun 2019 09:11 |
| Last Modified: | 13 Jun 2019 09:11 |
| URI: | http://eprints.usm.my/id/eprint/44598 |
Actions (login required)
 |
View Item |
