Abstract
Out-of-Autoclave manufacturing is an emerging lower-cost alternative to produce composite components, however, higher porosity (>2%) is often observed due to the 3–7x reduction of the consolidation pressure compared to autoclave manufacturing. For this reason, fully characterising the composite microstructure as it undergoes an Out-of-Autoclave process is pivotal to understand the routes to porosity and the effects of processing parameters on the porosity evolution. In this PhD project, developments in three areas (data acquisition, data processing and experimental setup customisation) enabled the in-situ characterization of the microstructural changes occurring during a typical Out-of-Autoclave process using lab-based X-ray computed tomography (XCT).The maximisation of the number of microstructural states captured during a lab-based in-situ XCT experiment is achieved by a substantial decrease of the scan time, which affects the quality of the output. A study was conducted to quantify the extent of this effect, showing that reducing the scan time to a value between 2 and 8 minutes, at a maximum resolution of 25 µm, allowed an accurate evaluation of the porosity content.
Deep Learning was implemented to the segmentation of composite X-Ray micrographs. This novel technique outperformed thresholding in the phase characterization of uncured Out-of-Autoclave prepreg laminates, as well as in the segmentation of porosity in low quality micrographs obtained in shorter scan times. The effect of the Convolutional Neural Network architecture and the annotator in the quality of the Deep Learning segmentation were investigated. U-Net provided the highest segmentation score, whereas the Deep Learning models trained with composite experts and non-experts provided an equivalent segmentation of the composite phases.
Finally, a custom XCT rig was developed to study the effect of the mould shape and laminate configuration in the microstructure Out-of-Autoclave processing, allowing the real-time visualisation of the void movement and thickness change.
| Date of Award | 24 Jan 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | James Kratz (Supervisor) |