3D convex surface reconstruction based on visual hull
In order to assure the continued airworthiness of aircraft and safe flight operations, engine maintenance, repair and overhaul (MRO) is essential. Engine blades are inspected on a frequent basis due to stringent requirements. These engine blades are visually inspected for surface and edge defects b...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis-Master by Coursework |
Language: | English |
Published: |
Nanyang Technological University
2024
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/181407 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-181407 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1814072024-12-07T16:53:41Z 3D convex surface reconstruction based on visual hull Girish, Balaraman Murukeshan Vadakke Matham School of Mechanical and Aerospace Engineering Advanced Remanufacturing and Technology Centre Cheng Fang MMurukeshan@ntu.edu.sg , chengf@artc.a-star.edu.sg Engineering In order to assure the continued airworthiness of aircraft and safe flight operations, engine maintenance, repair and overhaul (MRO) is essential. Engine blades are inspected on a frequent basis due to stringent requirements. These engine blades are visually inspected for surface and edge defects by trained aircraft maintenance personnel. However, many challenges persist during the inspection of blade profile, and many blades are scrapped as there is no established method to quantify the profile directly and the decision of whether the blade is to be scrapped or not is completely dependent on the experience of the maintenance inspector. Automated defect detection plays a pivotal role in the reduction of maintenance costs as well was material replacement costs and maintains performance of the aircraft while strictly adhering to safety standards. The current use of machine learning models for the automation of aircraft maintenance is challenging and time consuming due to millions of trainable parameters and lack of available data for existing models of aircraft as flight data is considered confidential thereby, demanding high initial setup costs and excellent proficiency in software knowledge. The absence of a commercially available product for blade inspection drives this research to mainly focus on implementing a compact blade profiling system which aids the traditional visual inspection method by using conventional image processing techniques thereby eliminating the use of machine learning models and their large quantities of their trainable parameters and datasets. The main objective of this study is to investigate into an optics-based approach for 3D convex surface reconstruction of HPC blades by using the visual hull algorithm. This involves the hardware setup, integration and error measurement of a 2D laser optical micrometre and a high accuracy motorized rotator which facilitated the image capturing process of the HPC blade as a single system without the need for a multiple camera setup. Various conventional image processing techniques were applied to the captured images as part of the visual hull algorithm such as grey-scale to binary conversion, voxel grid creation, projection of voxels for each individual image, voxel carving and finally mesh creation. A detailed study on how varying different parameters affected the final reconstructed 3D model was made based on the required computational power, time and the accuracy of the model. Increase in accuracy led to increased computation power and time while reduced accuracy required significantly lower power and processing times. The implemented system along with the visual hull algorithm produced 3D models with high accuracy while preserving the geometrical features of the HPC blade. The system and the algorithm were also used to perform defect detection for an alternative test object which yielded positive results. The use case of the aforementioned reconstruction software and hardware was also extended to other applications such as reverse engineering which was achieved via additive manufacturing processes. This research is expected to contribute to the domain of aircraft maintenance industry by aiding the traditional visual inspection process while integrating the perks of human expertise as well as automation. Master's degree 2024-12-02T02:36:54Z 2024-12-02T02:36:54Z 2024 Thesis-Master by Coursework Girish, B. (2024). 3D convex surface reconstruction based on visual hull. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/181407 https://hdl.handle.net/10356/181407 en application/pdf Nanyang Technological University |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering |
spellingShingle |
Engineering Girish, Balaraman 3D convex surface reconstruction based on visual hull |
description |
In order to assure the continued airworthiness of aircraft and safe flight operations, engine
maintenance, repair and overhaul (MRO) is essential. Engine blades are inspected on a frequent basis due to stringent requirements. These engine blades are visually inspected for surface and edge defects by trained aircraft maintenance personnel. However, many challenges
persist during the inspection of blade profile, and many blades are scrapped as there is no established method to quantify the profile directly and the decision of whether the blade is to be
scrapped or not is completely dependent on the experience of the maintenance inspector. Automated defect detection plays a pivotal role in the reduction of maintenance costs as well was
material replacement costs and maintains performance of the aircraft while strictly adhering to
safety standards. The current use of machine learning models for the automation of aircraft
maintenance is challenging and time consuming due to millions of trainable parameters and
lack of available data for existing models of aircraft as flight data is considered confidential
thereby, demanding high initial setup costs and excellent proficiency in software knowledge.
The absence of a commercially available product for blade inspection drives this research to
mainly focus on implementing a compact blade profiling system which aids the traditional visual inspection method by using conventional image processing techniques thereby eliminating
the use of machine learning models and their large quantities of their trainable parameters and
datasets. The main objective of this study is to investigate into an optics-based approach for 3D
convex surface reconstruction of HPC blades by using the visual hull algorithm. This involves
the hardware setup, integration and error measurement of a 2D laser optical micrometre and a high accuracy motorized rotator which facilitated the image capturing process of the HPC
blade as a single system without the need for a multiple camera setup. Various conventional
image processing techniques were applied to the captured images as part of the visual hull
algorithm such as grey-scale to binary conversion, voxel grid creation, projection of voxels
for each individual image, voxel carving and finally mesh creation. A detailed study on how
varying different parameters affected the final reconstructed 3D model was made based on the
required computational power, time and the accuracy of the model. Increase in accuracy led
to increased computation power and time while reduced accuracy required significantly lower
power and processing times. The implemented system along with the visual hull algorithm produced 3D models with high
accuracy while preserving the geometrical features of the HPC blade. The system and the algorithm were also used to perform defect detection for an alternative test object which yielded
positive results. The use case of the aforementioned reconstruction software and hardware was
also extended to other applications such as reverse engineering which was achieved via additive manufacturing processes. This research is expected to contribute to the domain of aircraft
maintenance industry by aiding the traditional visual inspection process while integrating the
perks of human expertise as well as automation. |
author2 |
Murukeshan Vadakke Matham |
author_facet |
Murukeshan Vadakke Matham Girish, Balaraman |
format |
Thesis-Master by Coursework |
author |
Girish, Balaraman |
author_sort |
Girish, Balaraman |
title |
3D convex surface reconstruction based on visual hull |
title_short |
3D convex surface reconstruction based on visual hull |
title_full |
3D convex surface reconstruction based on visual hull |
title_fullStr |
3D convex surface reconstruction based on visual hull |
title_full_unstemmed |
3D convex surface reconstruction based on visual hull |
title_sort |
3d convex surface reconstruction based on visual hull |
publisher |
Nanyang Technological University |
publishDate |
2024 |
url |
https://hdl.handle.net/10356/181407 |
_version_ |
1819112970919084032 |