Current methods for inspecting subgrade compaction quality are time-consuming and destructive. This paper presents an in-situ measurement methodology for rapid detection of subgrade compaction quality using Ultrasonic Pulse Velocity (UPV) and Intelligent Compaction (IC). Field compaction tests, field UPV test, and laboratory tests are performed to construct the heterogeneous datasets. A set of intelligent models are established to estimate the subgrade compactness in terms of the in-time measured UPV. The training strategy is proposed by expanding dataset quantity and dataset dimension to enhance the model performance. The model demonstrates high accuracy in compactness estimation and strong generalization in verification. This contribution can be utilized in engineering applications to effectively detect the compaction quality during the construction.

中文翻译：

---

路基压实度原位快速测量训练策略及智能模型


目前检查路基压实质量的方法既耗时又具有破坏性。本文提出了一种使用超声波脉冲速度（UPV）和智能压实（IC）快速检测路基压实质量的现场测量方法。进行现场压实测试、现场 UPV 测试和实验室测试来构建异构数据集。建立了一套智能模型，根据实时测量的 UPV 来估计路基压实度。提出了通过扩展数据集数量和数据集维度来提高模型性能的训练策略。该模型在紧致度估计方面表现出较高的准确性，在验证方面具有较强的泛化能力。这一贡献可用于工程应用，以有效检测施工过程中的压实质量。