Compressed Sensing in Material Diagnostics via Ultrasound Imaging (CoSMaDU)

无损检测（NDT）是确保制造过程中没有关键缺陷和维护大型基础设施（如铁路轨道）的核心工具。随着这两个领域的需求不断增长，快速可靠的无损检测技术的重要性日益增加，其目的是在整个生命周期内检查组件。超声由于其简单和安全的适用性，是最成熟的无损检测技术之一。然而，到目前为止，超声检测主要是在定性的基础上进行的（即，样品是否有缺陷）。缺陷的定量评估实际上是有限的，因为超声数据的准确解释是困难的。由于超声波传播的复杂性，所获得的数据可能是高度模糊的，这就需要对被测部件或结构以及测量设置进行大量的先验知识，以便进行可靠的解释。为了实现这一点，适当的后处理变得必不可少。尽管如此，目前在商业超声无损检测产品中使用的后处理方法相当简陋，或者由于缺乏可靠性，只能与原始数据的检查一起使用。先进的信号处理技术没有得到应用的主要原因是缺乏适当的正演模型，这种模型既考虑到超声波传播的复杂性，又适合于信号处理应用。本项目旨在通过追求两个目标来提高技术水平：第一个目标是研究超声波传播的正演模型，目的是在色散、声阴影和散斑噪声等物理传播效应的真实建模与信号处理方法在模型复杂性方面的适用性之间找到最佳折衷。我们的研究将联合考虑测量设置和后处理阶段。基于此，该项目的第二个目标是设计有效捕获相关（用于后处理）信息的度量体系结构。为了实现这一目标，最先进的亚奈奎斯特采样技术，如压缩传感或有限创新率采样将应用于超声波无损检测。特别是，我们的建议旨在缩小新的理论结果（通常依赖于理想化的代数公式）和超声波传播的物理之间的差距，使这些结果适用于实际设置。

Non-destructive testing (NDT) is a core tool to ensure the absence of critical flaws in fabrication and enable maintenance of large infrastructure such as railroad tracks. With growing demands in both areas, the importance of fast and reliable NDT techniques has increased aiming at examining a component throughout its entire life.Ultrasound is, due to its simple and safe applicability, one of the most established NDT techniques. However, so far ultrasound testing is performed mostly on a qualitative basis (i.e., is a specimen is flawed or not). The quantitative assessment of defects is in fact limited, since the exact interpretation of ultrasound data is difficult. Due to the complex nature of ultrasonic wave propagation, the acquired data can be highly ambiguous which necessitates a large amount of a priori knowledge on the component or structure under test and the measurement setup to allow for a reliable interpretation.To facilitate this, appropriate post-processing becomes indispensable. Notwithstanding, the post-processing methods currently used in commercial ultrasound NDT products are quite rudimentary or are used only together with the inspection of the raw data in a complementary fashion for lack of reliability. The main reason advanced signal processing techniques have not been applied is the lack of appropriate forward models that account for the complexity of the ultrasonic wave propagation and yet are still suitable for signal processing applications. The proposed project aims at improving the state-of-the-art by pursuing two goals: The first objective is to investigate forward models for ultrasonic wave propagation with the goal to find the best compromise between the realistic modeling of physical propagation effects such as dispersion, acoustic shadowing and speckle noise on the one hand and the suitability for signal processing methods in terms of the model complexity on the other hand. Our investigations will consider the measurement setup and the post-processing stage jointly. Based on this, the second objective of this project is to design measurement architectures that capture relevant (for post-processing) information efficiently. To achieve this, the state-of-the-art in sub-Nyquist-Sampling techniques such as Compressed Sensing or Finite Rate of Innovation Sampling will be applied to ultrasonic NDT. In particular, our proposal aims at closing the gap between novel theoretical results (that usually rely on idealized algebraic formulations) and the physics of ultrasonic wave propagation to make these results applicable to real setups.