Multi-scale metrology for automated non-destructive testing systems

The use of lightweight composite structures in the aerospace industry is now commonplace. Unlike conventional materials, these parts can be moulded into complex aerodynamic shapes, which are diffcult to inspect rapidly using conventional Non-Destructive Testing (NDT) techniques. Industrial robots provide a means of automating the inspection process due to their high dexterity and improved path planning methods.This thesis concerns using industrial robots as a method for assessing the quality of components with complex geometries. The focus of the investigations in this thesis is on improving the overall system performance through the use of concepts from the field of metrology, specifically calibration and traceability. The use of computer vision is investigated as a way to increase automation levels by identifying a component's type and approximate position through comparison with CAD models.The challenges identified through this research include developing novel calibration techniques for optimising sensor integration, verifying system performance using laser trackers, and improving automation levels through optical sensing. The developed calibration techniques are evaluated experimentally using standard reference samples. A 70% increase in absolute accuracy was achieved in comparison to manual calibration techniques. Inspections were improved as verified by a 30% improvement in ultrasonic signal response.A new approach to automatically identify and estimate the pose of a component was developed specifically for automated NDT applications. The method uses 2D and 3D camera measurements along with CAD models to extract and match shape information. It was found that optical large volume measurements could provide suffciently high accuracy measurements to allow ultrasonic alignment methods to work, establishing a multi-scale metrology approach to increasing automation levels. A classification framework based on shape outlines extracted from images was shown to provide over 88% accuracy on a limited number of samples.

Name dc:type.qualificationname

engd

Level dc:type.qualificationlevel

doctoral-pg

Grantor dc:publisher.institution

University of Strathclyde

Year dc:date.issued

2020

Author dc:creator

• Riise, Jonathan

Advisors dc:contributor.advisor

• Pierce, Gareth
• Gachagan, Tony

DOI

10.48730/s3j7-rg08

Identifier

T15598

Author Identifier

201474839

OAI identifier oai:identifier

oai:strathclyde:tq57nr01v