Ultrasonic Inspection for Complex Geometry

相控阵超声波检测是NDE中广泛使用的次表面检测方法，并且由于其能够从一个位置执行多个检测，因此能够比标准超声波检测更大程度地确定缺陷的位置和尺寸[1]。使用全矩阵捕获（FMC）获取数据，并使用全聚焦方法（TFM）[2]生成超声响应图像，从而提供检测区域中缺陷的指示。除了从阵列到缺陷的直接超声波之外，还将存在从检测材料中的后壁反射的波，其在时间上稍后到达缺陷。这似乎与FMC数据中的直接信号不同。这些反射波可以使用TFM [3]成像，从而能够从多个角度检查组件，以提高检测缺陷的速度，并改善表征。通常情况下，TFM的应用程序检查直接位于阵列下方或邻近阵列的组件区域[4-7] -换句话说，在阵列的视线（LoS）内。在无损检测中，通常的做法是优先检查最有可能出现缺陷的区域。在飞机发动机这样的复杂部件中，这些区域很可能不是LoS。这导致在检查之前需要拆卸，增加了检查过程的成本。该研究项目旨在开发和验证一种方法，该方法能够通过从组件的几何形状反射超声波来检查非视线（NLoS）区域。最初的目标是使用图1所示的简单测试用例来评估相控阵列在这些区域中的评估效果。后续目标是针对日益复杂的情况进行检测，包括抽象零件几何形状、各向异性材料或几何或材料属性不精确的样品。最终，一种能够检查真实的工业相关样品的方法，使得其在工业环境中的部署是合理的。到目前为止，工作重点是开发基于射线的模型，以产生FMC数据，包括来自组件几何结构中任何单个壁而不仅仅是后壁的超声波反射，以及探头对这些NLoS TFM图像中的侧钻孔（SDH）的预期灵敏度。32元件5 MHz探头对直径为0.4mm的SDH的灵敏度如图2所示，其中多模TFM从侧壁反射聚焦。这些结果表明，存在对NLoS区域中的理想缺陷具有不可忽略的灵敏度的视图。虽然这是用于确定理想条件下缺陷的预期信号的有用工具，但它没有考虑预期来自真实的系统的随机和相干噪声。因此，通过比较来自几何结构内的一系列位置中的缺陷的预期信号响应，针对有限元模拟来验证这些结果。该项目的下一步是调查这种灵敏度对探头位置相对于几何形状的依赖程度，以确保可以可靠地检测到缺陷，并根据实验结果进行验证。在此之后，模型的复杂性将增加，因为多边形各向同性固体的假设被放宽到包括任意的、各向异性的几何形状，其尺寸可能不确切知道。最终，需要根据真实的工业相关样本验证开发的工具。

Phased array ultrasonic testing is a widely used method of sub-surface inspection in NDE, and is able to locate and size defects to a greater degree of confidence and precision than standard ultrasonic testing due to its ability to perform multiple inspections from one location [1]. Data is acquired using Full Matrix Capture (FMC), and an image of the ultrasonic response can be produced using the Total Focussing Method (TFM) [2], providing an indication of flaws in the inspection region. In addition to the direct ultrasonic wave from array to defect, there will be a wave which reflects from a back wall in the inspection material, arriving at the defect later in time. This appears distinct from the direct signal in the FMC data. These reflected waves can be imaged using TFM [3], enabling inspection of the component from multiple angles to increase the rate at which defects are detected, and improve characterisation. Typically, applications of the TFM inspect regions of the component which are either directly below or adjacent to the array [4-7] - in other words, within the array's line-of-sight (LoS). It is common practice in NDE to prioritise inspection of areas most likely to develop defects. In a complex part such as an aircraft engine, it is likely that these areas will not be LoS. This leads to the requirement for disassembly prior to inspection, increasing cost of the inspection process. This research project aims to develop and validate a method which is able to inspect non-line-of-sight (NLoS) areas by reflecting an ultrasonic wave from the geometry of a component. The initial goal is to evaluate how well a phased array can assess these areas using the simple test case shown in figure 1. Subsequent goals are to target the inspection of increasingly complex cases, including abstract part geometry, anisotropic materials or samples whose geometric or material properties are not precisely known. Ultimately, a method capable of inspecting real, industrially relevant samples such that its deployment in an industrial setting will be justified. Work so far has focussed on developing a ray-based model to produce FMC data including reflections of the ultrasonic wave from any individual wall in the component geometry rather than just the back wall, as well as the expected sensitivity of the probe to side-drilled holes (SDH) in these NLoS TFM images. The sensitivity of a 32-element 5MHz probe to a SDH with diameter 0.4mm is shown in figure 2, where multi-mode TFMs are focussed from side wall reflections. These results indicate that there are views which have non-negligible sensitivity to an ideal defect in NLoS areas. While this is a useful tool for determining the expected signal from a defect in ideal conditions, it does not account for random and coherent noise expected from a real system. These results were therefore validated against finite-element simulations by comparing the expected signal response from defects in a range of positions within the geometry. Next steps in the project are investigate how dependent this sensitivity is to the probe location with respect to the geometry, to ensure that defects can be reliably detected, as well as performing validation against experimental results. Following this, the complexity of the models will be increased, as the assumptions of a polygonal isotropic solid are relaxed to include an arbitrary, anisotropic geometry whose dimensions may not be exactly known. Ultimately, it will be a requirement to validate the tools developed against real, industrially relevant samples.