Peter Huthwaite

Data scarcity is a known limitation associated with many ultrasonic applications, particularly those that, by their nature, require large amounts of data to be set up and perform as intended. Examples include the in-silico qualification of nondestructive evaluation methods and training machine learning algorithms for engineering or medical applications. Experimental data is typically difficult to obtain in sufficient quantities, and, while synthetic data generation is possible, the data is often either not sufficiently realistic or the process is computationally expensive. The latter is a major limitation of finite element (FE) modeling, which otherwise generates highly accurate data. In our previous work, we presented a linear interpolation technique to significantly reduce the computational cost associated with FE modeling by using results from a small number of full FE models to populate an entire parameter space via interpolation with negligible additional computational cost. While linear interpolation was demonstrated to be a helpful tool, it is beneficial to develop stronger interpolation, capable of addressing wider and more complex needs. In this work, we extend our previous method to use quadratic polynomial interpolation. After demonstrating that this interpolation order is sufficient, our method is validated using test cases of increasing complexity; validation is achieved by comparing the interpolation-assisted data with their corresponding true FE-generated counterparts. A method for interpolating complicated signals containing multiple features is also presented, implemented, and its performance is assessed through the test cases. Our results clearly show an improved agreement between the estimated and true results compared with linear interpolation. We also demonstrate robust performance in complex cases, and a very substantial reduction in computational cost in all instances.

Ultrasound tomography struggles with high-contrast and/or extended-range problems, such as site-specific in vivo bone imaging, which is crucial for assessing fracture risk. Two main obstacles arise. (a) An external region prohibits access to the region of interest (ROI), and (b) high contrast exists between the two regions. These challenges hinder existing algorithms—including bent-ray tomography (BRT). Virtual initialised ray tomography (VIRT) is proposed, which addresses these challenges through (a) placement of virtual transducers on the ROI, facilitating (b) rapid initialisation before BRT inversion. In silico validation against BRT with and without prior information shows superior resolution and robustness while improving computational speed. Validation against BRT directly on the ROI shows that VIRT is approaching the resolution limit. Additionally, VIRT performs well under 20 dB signal-to-noise ratio. The ability to solve high-contrast extended-range tomography problems without prior knowledge about the interior of the ROI has many implications. VIRT has the potential to unlock site-specific in vivo bone imaging for assessing fracture risk. It may also replace BRT to improve resolution, robustness, and computational speed—especially where the ROI does not cover the entire imaging array.

Guided wave tomography plays a significantly important role in quantitatively mapping thickness variations in plate-like structures and pipelines in the petrochemical industry for accurate measurement of corrosion, as guided waves enable rapid screening of large areas without needing direct access. Several inverse algorithms have been implemented in guided wave tomography; however, almost all of them use a primary assumption: the three-dimensional (3D) guided wave thickness reconstruction is simplified as a two-dimensional (2D) acoustic wave velocity inversion, which is then mapped to thickness variation using the dispersive nature of guided waves. Although this assumption simplifies the inverse procedure, it makes it impossible to use non-dispersive modes in reconstructions. Geometrical full waveform inversion (GFWI) is promising to overcome this limitation since it can reconstruct corrosion profiles through geometry optimization using a data-fitting procedure, where velocity-to-thickness mapping is not needed. In this work, GFWI-based guided wave tomography is developed in plate-like structures, and is applied to reconstruct thickness maps in a series of corrosion defects using the non-dispersive SH0 mode, demonstrating high performance and achieving an improved reconstruction resolution.

Tomographic reconstruction is used extensively in medicine, non-destructive testing and geology. In an ideal situation where measurements are taken at all angles around an object, known as full view configuration, a full reconstruction of the object can be produced. One of the major issues faced in tomographic imaging is when measurements cannot be taken freely around the object under inspection. This may be caused by the size and geometry of the object or difficulty accessing from particular directions. The resulting limited view transducer configuration leads to a large deterioration in image quality, thus it is very beneficial to employ a compensation algorithm. At present, the most effective compensation algorithms require a large amount of computing power or a bespoke case-by case approach, often with numerous arbitrary constants which must be tuned for a specific application. This work proposes a machine learning based approach to perform the limited view compensation. The model is based around an autoencoder architecture. It is trained on an artificial dataset, taking advantage of the ability to generate arbitrary limited view images given a full view input. The approach is evaluated on ten laser-scanned corrosion maps and the results compared to positivity regularisation – a limited view compensation algorithm similar in the speed of execution and generalisation potential. The algorithms are compared for root mean squared error (RMSE) across the image, and maximum absolute error (MAE). Furthermore, they are visually compared for subjective quality. Compared to the conventional algorithm, the ML-based approach improves on the MAE in eight out of the ten cases. The conventional approach performs better on mean RMSE, which is explained by the ML outputting inaccurate background level, which is not a critical ability. Most importantly, the visual inspection of outputs shows the ML approach reconstructs the images better, especially in the case of irregular corrosion patches. Compared to limited view images, the ML method improves both the RMSE and MAE by 41%.

Ultrasound tomography generally struggles to reconstruct high-contrast and/or extended-range problems. A prime example is site-specific in-vivo bone imaging, crucial for accurately assessing the risk of life-threatening fractures, which are preventable given accurate diagnosis and treatment. In this type of problem, two main obstacles arise: (a) an external region prohibits access to the region of interest (ROI), and (b) high contrast exists between the two regions. These challenges impede existing algorithms -- including bent-ray tomography (BRT), known for its robustness, speed, and reasonable short-range resolution. We propose Virtual Extended-Range Tomography (VERT), which tackles these challenges through (a) placement of virtual transducers directly on the ROI, facilitating (b) rapid initialisation before BRT inversion. In-silico validation against BRT with and without a-priori information shows superior resolution and robustness -- while maintaining or even improving speed. These improvements are drastic where the external region is much larger than the ROI. Additional validation against the practically impossible -- BRT directly on the ROI -- demonstrates that VERT is approaching the resolution limit. The capability to solve high-contrast extended-range tomography problems without prior knowledge about the ROI's interior has many implications. VERT has the potential to unlock site-specific in-vivo bone imaging for assessing fracture risk, potentially saving millions of lives globally. In other applications, VERT may replace classical BRT to yield improvements in resolution, robustness and speed -- especially where the ROI does not cover the entire imaging array. For even higher resolution, VERT offers a reliable starting background to complement algorithms with less robustness and high computational costs.

Across non-destructive testing (NDT) and structural health monitoring (SHM), accurate knowledge of the systems' reliability for detecting defects, such as Probability of Detection (POD) analysis is essential to enabling widespread adoption. Traditionally this relies on access to extensive experimental data to cover all critical areas of the parametric space, which becomes expensive, and heavily undermines the benefit such systems bring. In response to these challenges, reliability estimation based on numerical simulation emerges as a practical solution, offering enhanced efficiency and cost-effectiveness. Nevertheless, precise reliability estimation demands that the simulated data faithfully represents the real-world performance. In this context, a numerical framework tailored to generate realistic signals for reliability estimation purposes is presented here, focusing on the application of guided wave SHM for pipe monitoring. It specifically incorporates key characteristics of real signals: random noise and coherent noise caused by the imbalance in transducer performance within guided wave monitoring systems. The effectiveness of our proposed methodology is demonstrated through a comprehensive comparative analysis between simulation-generated signals and experimental signals both individually and statistically. Furthermore, to assess the reliability of a guided wave system in terms of the inspection range for pipe monitoring, a series of POD analyses using simulation-generated data were conducted. The comparison of POD curves derived from ideal and realistic simulation data underscores the necessity of considering coherent noise for accurate POD curve calculations. Moreover, the POD analysis based on realistic simulation-generated data provides a quantitative estimation of the inspection range with more details compared to the current industry practice.