In the field of civil engineering, nondestructive evaluation (NDE) can be understood as the process of a material or structural element assessment based on techniques that do not introduces damage to the assessed sample. NDE can be used to inspect the quality of materials or to find some damages, as cracks and voids. This way, ultrasonic data monitoring has become one of the most important tools for detecting local damages in civil engineering elements. Concerning to masonry application, this method consists in the measurement of the time that the ultrasonic wave takes to pass through a material and changes in the ultrasonic wave propagation can be used for identifying damages presence and anomaly. Despite statistical methods be widely used for processing large data sets, in the case of ultrasonic waves some issues need to be overcoming, as the analysis of the path traveled by the ultrasonic wave. Thus, the main aim of this work is to analyze the ultrasonic wave propagation behavior through a new approach of a statistical model. Using a Bayesian semiparametric method with Dirichlet process, a data-driven hierarchical regression model is applied to simulated and experimental data sets. For simulated data sets, the sensibility of the model in clustering profiles in an unknown number of groups and also detecting outliers is observed. The application in experimental data shows the potential use of the model for real case applications due to its efficiency in identifying different behaviors of the wave propagation in materials. Therefore, considering the advanced of automation and Internet of Things, statistical methods based on clustering are important and useful tools for being explored in the masonry conditions monitoring.

中文翻译：

---

聚类超声波传播时间：一种分层多项式半参数方法

在土木工程领域，无损评估 (NDE) 可以理解为基于不会对评估样本造成损害的技术进行材料或结构元素评估的过程。NDE 可用于检查材料质量或发现一些损坏，如裂缝和空隙。这样，超声波数据监测已成为检测土木工程元件局部损坏的最重要工具之一。关于砌体应用，该方法包括测量超声波穿过材料所需的时间，并且超声波传播的变化可用于识别损坏的存在和异常。尽管统计方法被广泛用于处理大数据集，但在超声波的情况下，一些问题需要克服，作为超声波行进路径的分析。因此，这项工作的主要目的是通过一种新的统计模型方法来分析超声波的传播行为。使用带狄利克雷过程的贝叶斯半参数方法，将数据驱动的层次回归模型应用于模拟和实验数据集。对于模拟数据集，观察模型在未知数量的组中聚类配置文件以及检测异常值的敏感性。实验数据中的应用显示了该模型在实际案例应用中的潜在用途，因为它可以有效地识别材料中波传播的不同行为。因此，考虑到自动化和物联网的先进性，