Detecting defects to improve the sustainability of composite manufacture

Posted on
A picture of composite materials

Adapted from this post on the Engineering Our Future Blog. Written by Louisa Cockbill.

Gary Atkinson, Associate Professor in Engineering at UWE Bristol and part of Lyndon Smith’s team in the Centre for Machine Vision, is developing the “Polarisation Vision for Composite Part Inspection” – a system to automatically detect defects in raw and processed composite materials and so reduce waste and huge expenses for composite industries.

The project, funded by Digital Engineering Technology Innovation (DETI), has potential to impact on industrial sustainability, where testing and subsequent discarding of materials is a big issue.

The Weakest Link

Composite material strength lies in the uniform pattern of either unidirectional or interwoven carbon fibres lying undisturbed in resin. But any composite component is only as strong as its weakest link, and weaknesses caused by fibre misalignment or gaps is a common problem in composite manufacture.

The dark, shiny nature of composite materials makes defects tricky to spot using visualisation techniques based on traditional camera technology. The current industry standard is to perform time consuming and eye straining manual inspections to ensure the safety of components, particularly those destined for applications in the aerospace industry.

Gary is utilising a different type of visualisation technique to identify the problem areas by using a specialised camera. The camera detects when light becomes linearly polarised – with its electric field confined to a single plane – on reflection from a surface. Instead of blocking glare from reflected light like polarised sunglasses, the camera detects the reflected polarised light and, with specially developed software, produces a “photo” of composite fibres – with each orientation of fibre shown in a different colour.

Polarised Vision setup imaging a composite sheet

The images produced by the software highlight tiny, localised defects, but individual fibre orientation can also be computed to give an overall report on a material’s quality. 

“There’s no way you’d consistently detect localised defects like that using standard manual inspection or traditional methods from machine vision,” said Gary, explaining how his instrumentation could provide rapid and detailed defect detection for composite manufacture.

In recent months, Gary and his team have painstakingly explored what type of defects his “Polarisation Vision” can detect. They’ve taken a huge range of lab-based scenarios, as well as composites in realistic factory conditions, and developed the detection algorithms to automatically find component flaws using the polarised images.

Now they’re ready to start real-world trials, collaborating with Airbus to feed more and more data into the machine learning algorithm, training it further in defect auto-detection.

View the full blog post here.

Back to top