Three-Dimensional Residual Neural Architecture Search for Ultrasonic Defect Detection.

This study presents a deep-learning (DL) methodology using 3-D convolutional neural networks (CNNs) to detect defects in carbon fiber-reinforced polymer (CFRP) composites through volumetric ultrasonic testing (UT) data. Acquiring large amounts of ultrasonic training data experimentally is expensive and time-consuming. To address this issue, a synthetic data generation method was extended to incorporate volumetric data. By preserving the complete volumetric data, complex preprocessing is reduced, and the model can utilize spatial and temporal information that is lost during imaging. This enables the model to utilize important features that might be overlooked otherwise. The performance of three architectures was compared. The first architecture is prevalent in the literature for the classification of volumetric datasets. The second demonstrated a hand-designed approach to architecture design, with modifications to the first architecture to address the challenges of this specific task. A key modification was the use of cuboidal kernels to account for the large aspect ratios seen in ultrasonic data. The third architecture was discovered through neural architecture search (NAS) from a modified 3-D residual neural network (ResNet) search space. In addition, domain-specific augmentation methods were incorporated during training, resulting in significant improvements in model performance, with a mean accuracy improvement of 22.4% on the discovered architecture. The discovered architecture demonstrated the best performance with a mean accuracy increase of 7.9% over the second-best model. It was able to consistently detect all defects while maintaining a model size smaller than most 2-D ResNets. Each model had an inference time of less than 0.5 s, making them efficient for the interpretation of large amounts of data.

Clinical Monitoring of Tooth Wear Progression in Patients over a Period of One Year Using CAD/CAM.

PURPOSE: The aim of this study was to clinically monitor the progression of tooth wear over a period of 1 year in a cohort of referred tooth wear patients through the use of a computer-aided design/computer-assisted manufacture (CAD/CAM) scanner and a standardized scanning/assessment methodology. MATERIALS AND METHODS: Polyether impressions were made of 11 participants (130 teeth) at baseline and at 1 year. Impressions were poured in type IV dental stone and the anterior teeth were 3D scanned. A surface-matching software was used to compare 1-year and baseline scans and identify any dimensional differences. RESULTS: Parafunctional habits were reported by all patients. All participants exhibited tooth wear ≥ 140 µm in depth and extending to ≥ 280 µm in at least one tooth. Maxillary central incisors were the most commonly and severely affected teeth. CONCLUSION: The ability of the developed CAD/CAM scanning methodology in clinical monitoring of tooth wear was demonstrated. Further research is needed to assess its practicality in large-scale epidemiologic tooth wear studies.

A Proposed Methodology to Assess the Accuracy of 3D Scanners and Casts and Monitor Tooth Wear Progression in Patients.

PURPOSE: The aim of this study was to detail and assess the capability of a novel methodology to 3D-quantify tooth wear progression in a patient over a period of 12 months. MATERIALS AND METHODS: A calibrated stainless steel model was used to identify the accuracy of the scanning system by assessing the accuracy and precision of the contact scanner and the dimensional accuracy and stability of casts fabricated from three different types of impression materials. Thereafter, the overall accuracy of the 3D scanning system (scanner and casts) was ascertained. Clinically, polyether impressions were made of the patient's dentition at the initial examination and at the 12-month review, then poured in type IV dental stone to assess the tooth wear. The anterior teeth on the resultant casts were scanned, and images were analyzed using 3D matching software to detect dimensional variations between the patient's impressions. RESULTS: The accuracy of the 3D scanning system was established to be 33 µm. 3D clinical analysis demonstrated localized wear on the incisal and palatal surfaces of the patient's maxillary central incisors. The identified wear extended to a depth of 500 µm with a distribution of 4% to 7% of affected tooth surfaces. CONCLUSION: The newly developed 3D scanning methodology was found to be capable of assessing and accounting for the various factors affecting tooth wear scanning. Initial clinical evaluation of the methodology demonstrates successful monitoring of tooth wear progression. However, further clinical assessment is needed.

Simulation of ultrasonic lamb wave generation, propagation and detection for a reconfigurable air coupled scanner.

A computer simulator, to facilitate the design and assessment of a reconfigurable, air-coupled ultrasonic scanner is described and evaluated. The specific scanning system comprises a team of remote sensing agents, in the form of miniature robotic platforms that can reposition non-contact Lamb wave transducers over a plate type of structure, for the purpose of non-destructive evaluation (NDE). The overall objective is to implement reconfigurable array scanning, where transmission and reception are facilitated by different sensing agents which can be organised in a variety of pulse-echo and pitch-catch configurations, with guided waves used to generate data in the form of 2-D and 3-D images. The ability to reconfigure the scanner adaptively requires an understanding of the ultrasonic wave generation, its propagation and interaction with potential defects and boundaries. Transducer behaviour has been simulated using a linear systems approximation, with wave propagation in the structure modelled using the local interaction simulation approach (LISA). Integration of the linear systems and LISA approaches are validated for use in Lamb wave scanning by comparison with both analytic techniques and more computationally intensive commercial finite element/difference codes. Starting with fundamental dispersion data, the paper goes on to describe the simulation of wave propagation and the subsequent interaction with artificial defects and plate boundaries, before presenting a theoretical image obtained from a team of sensing agents based on the current generation of sensors and instrumentation.

Evaluation of neural network robust reliability using information-gap theory.

A novel technique for the evaluation of neural network robustness against uncertainty using a nonprobabilistic approach is presented. Conventional optimization techniques were employed to train multilayer perceptron (MLP) networks, which were then probed with an uncertainty analysis using an information-gap model to quantify the network response to uncertainty in the input data. It is demonstrated that the best performing network on data with low uncertainty is not in general the optimal network on data with a higher degree of input uncertainty. Using the concepts of information-gap theory, this paper develops a theoretical framework for information-gap uncertainty applied to neural networks, and explores the practical application of the procedure to three sample cases. The first consists of a simple two-dimensional (2-D) classification network operating on a known Gaussian distribution, the second a nine-lass vibration classification problem from an aircraft wing, and the third a two-class example from a database of breast cancer incidence.