Utilizing an Ultrasonic Inspection System Operating Inside an Autoclave and Machine Learning to Quantify Porosity within Composites During Cure

ABSTRACT

Composite materials are increasingly being utilized in aerospace applications for their high stiffness and strength to weight ratios and fatigue resistance. However, defects in the composite may arise during cure (e.g. porosity, delamination, fiber waviness), and current technology only allows for post-cure evaluation (e.g. microscopy, ultrasonic inspection). A high-temperature ultrasonic scanning system was developed for deployment in an autoclave, which can detect porosity in composites during the cure process. This study focused on the implementation of machine learning techniques to help generate a model that can quantify porosity, in addition to detection and localization that has previously been demonstrated. Two, six-hour-long experiments were conducted on curing of 762 mm × 305 mm (30 in. ×12 in.) composite panels with a [0/45/90/-45]_4s layup and varying regions of high and low pressure due to its tapered geometry in contact with a flat caul plate. The first experiment utilized a thick (12.7 mm) caul plate and the second utilized a thin (3.2 mm) caul plate. During experimentation, within the scan area (406 mm × 13 mm), data was recorded and stored for ultrasonic amplitude. Additional variables were measured or predicted including temperature, autoclave pressure, number of plies, slope of the composite panel surface with respect to the transducer, viscosity, and glass transition temperature. The pre-processed data was entered into the Regression Learner Application in MATLAB^Ⓡ,1 and a rational quadratic Gaussian process regression (GPR) was chosen for the machine learning algorithm. The model was then trained on a larger data set to make it more robust and capable of predictions using a function callout. The result was a machine learning algorithm that can reliably quantify porosity in a composite panel during cure based on measured amplitude response and generate images for intuitive visualization. This tool can be further trained with more experimentation and potentially employed for real-time porosity detection and quantification of composite components during cure in an autoclave. Practical use of this technology is the potential to dynamically control processing parameters (e.g. autoclave pressure) in real-time to reduce the level of porosity within the laminate to acceptable limits (e.g. 2% by volume).

KEYWORDS:

• Defect Detection
• Inspection During Cure
• Machine Learning (ML)
• Porosity
• Process Monitoring
• Ultrasonic Testing (UT)

Acknowledgments

The authors would like to acknowledge Hoa Luong and Sean Britton for their contributions to the experimental setup and data collection. Research reported in this publication was supported by funding provided by the Aeronautics Research Mission Directorate (ARMD) of the National Aeronautics and Space Administration (NASA).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Specific vendor and manufacturer names are explicitly mentioned only to accurately describe the hardware and software used in this study. The use of vendor and manufacturer names does not imply an endorsement by the U.S. Government nor does it imply that the specified equipment and software programs are the best available.

Additional information

Funding

This work was supported by the Langley Research Center.