ABSTRACT Pipe joints mostly form the weakest points in pipeline networks. In-field joints are prone to various flaws. Thus, the infrastructure industry requires an effective inspection technique. Our work focused on evaluating the performance of chord-type transducers for flaw detection in polyethylene (PE) pipe joints. Various artificially introduced flaws were fabricated and tested for statistical estimation of system performance. A-scans data was gathered to develop and assess the viability of a deep learning approach for automated flaw detection. Such an automated “smart” quality control method aligns with requirements of an nondestructive evaluation (NDE) 4.0 platform which can be utilized to achieve reliable and real-time inspection. In this we will introduce results of our current development, starting with approaches to generic data formats, communication protocols, signal processing, artificial intelligence-based (AI) information generation, and decision making. For each of the aspects, results and prototypical implementations will be provided. This includes a pilot development for modern human-machine-interaction using assistive technologies for manual NDE 4.0 inspection. This gives an outlook on further challenges and possible approaches for requirements in the context of secure data exchange, trusted and reliable AI processing, new standardization procedures, and validation of new “smart” NDE 4.0 ultrasonic inspection systems.

中文翻译：

---

使用由定制深度学习模型支持的弦式换能器，对中密度聚乙烯燃气管道的对接接头进行 NDE 4.0 兼容超声波检测

摘要 管接头大多是管网中最薄弱的环节。现场接头容易出现各种缺陷。因此，基础设施行业需要有效的检测技术。我们的工作重点是评估用于聚乙烯 (PE) 管接头缺陷检测的弦式传感器的性能。制造并测试了各种人为引入的缺陷，以对系统性能进行统计估计。收集 A 扫描数据以开发和评估用于自动缺陷检测的深度学习方法的可行性。这种自动化的“智能”质量控制方法符合无损评估 (NDE) 4.0 平台的要求，可用于实现可靠和实时的检测。在此，我们将介绍我们当前开发的结果，从通用数据格式、通信协议、信号处理、基于人工智能 (AI) 的信息生成和决策的方法开始。对于每个方面，将提供结果和原型实现。这包括使用辅助技术进行手动 NDE 4.0 检查的现代人机交互试点开发。这给出了在安全数据交换、可信和可靠的 AI 处理、新的标准化程序以及新的“智能”NDE 4.0 超声波检测系统验证的背景下的进一步挑战和可能的要求方法的前景。这包括使用辅助技术进行手动 NDE 4.0 检查的现代人机交互试点开发。这给出了在安全数据交换、可信和可靠的 AI 处理、新的标准化程序以及新的“智能”NDE 4.0 超声波检测系统验证的背景下的进一步挑战和可能的要求方法的前景。这包括使用辅助技术进行手动 NDE 4.0 检查的现代人机交互试点开发。这给出了在安全数据交换、可信和可靠的 AI 处理、新的标准化程序以及新的“智能”NDE 4.0 超声波检测系统验证的背景下的进一步挑战和可能的要求方法的前景。