Defect Identification in Dielectric and Composite Materials Using Noise-Based Microwave Nondestructive Testing Techniques
Open Access
- Author:
- Navagato, Marc Dominic
- Graduate Program:
- Electrical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 07, 2022
- Committee Members:
- Michael Lanagan, Outside Unit & Field Member
Ram Narayanan, Chair & Dissertation Advisor
Julio Urbina, Major Field Member
David Jenkins, Major Field Member
Kultegin Aydin, Program Head/Chair - Keywords:
- Nondestructive Testing
Microwave Testing
Neural Networks
Dimensionality Reduction
Noise Radar
Noise Waveforms
Microwave Nondestructive Testing
Stacked Autoencoders
Anomaly Detection - Abstract:
- Nondestructive testing (NDT) is a class of material inspection where the external and internal states of a material can be determined without altering the integrity of the material. Many NDT methods are available, such as visual inspection, ultrasonic testing, eddy-current testing, radiographic testing, and microwave testing. In general, the use of a specific method over another depends on the properties of the material under inspection. Unlike some of the mentioned NDT methods, which require in-contact configurations, high-power systems, or the application of couplant gels, the (free-space) microwave testing technique offers a unique solution that allows for non-contact configurations, does not require couplant gels, and can be conducted using low-power systems. Microwave nondestructive testing (MNDT) is a well-suited NDT method for inspection of dielectric materials, as microwaves can penetrate the inner layers of dielectric materials. Composite materials tend to possess dielectric properties, making them ideal candidates for MDNT. Composite materials are widely used in aerospace, automotive, civil, and consumer industries, where they are subject to various hazards, such as impact damages, weathering effects, and salt-spray corrosion. Identifying such defects is critical for preventing extended physical damage, performance degradation, and high financial costs. MNDT can be used to identify these types of internal defects. This dissertation presents the development of a noise-based MDNT system used to identify flaws within the internal layers of dielectric and composite materials. The MNDT system operates in the X-band frequency range (8.2-12.4 GHz) and uses noise waveforms. Noise waveforms are useful for MNDT as they exhibit excellent autocorrelation properties, which aids in the defect identification problem. A description of the noise waveforms and the MDNT system are provided in this dissertation. The MNDT system is validated through various inspections of dielectric and composite materials with hidden anomalies, where the resultant microwave images reveal the presence of the hidden defects. This work also shows that supervised and unsupervised techniques, based on Neural Networks and dimensionality reduction techniques, can be used to enhance the contrast of hidden defects within composite materials. It is then shown how these techniques can remove the impacts caused by non-uniform standoff distances between the sample under test and the microwave sensor, which tend to mask the presence of minor defects.