Predicting porosity in composite high-pressure hydrogen vessels using augmented fuzzy cognitive AI and manufacturing process parameters

利用增强型模糊认知人工智能和制造工艺参数预测复合材料高压氢气容器的孔隙率

阅读:2

Abstract

High-pressure hydrogen storage vessels type IV (HPV), made from carbon fiber-reinforced composites, are essential for hydrogen-powered applications due to their strength and light weight. However, they remain vulnerable to internal defects such as porosities, which can compromise structural integrity. While certification tolerates some porosity, excessive levels may lead to failure. This study aims to predict both the number of porosity and the porosity rate category (low vs. medium/high) using XTRACTIS (XTS), a general reasoning artificial intelligence system, capable of automatically inducing robust and intelligible predictive models. For porosity rate classification, XTS and Boosted Tree were unable to find a robust model, due to a lack of information and erroneous values of the variable to be predicted. To model the log10 of the number of porosities, XTS automatically selected 15 out of the available 58 manufacturing variables to build 56 fuzzy IF…THEN rules, achieving an RMSE of 7.94% and a correlation of 0.824 on an external test dataset. The interpretable rules reveal relationships between process parameters and porosity characteristics, thus facilitating the understanding of defect formation mechanisms. The analysis revealed that uniform and stable fiber tension improves compaction and reduces voids, while heterogeneous or insufficient tension increases porosity. Optimized mandrel speed and winding angle enhance fiber alignment and resin distribution. Precise control of winding layer volume and doctor blade angle prevents dry zones and resin-rich areas, reducing porosity risk. Furthermore, the results emphasize the importance of monitoring and tuning multiple interacting parameters simultaneously to ensure composite quality. These findings highlight the value of augmented fuzzy cognitive AI in uncovering interpretable, domain-relevant interactions, supporting quality control and process optimization in advanced composite manufacturing.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。