The geotechnical ground investigation involves several testing techniques. Within such techniques, the Piezocone Penetration Test with seismic wave measurements (SCPTu) is one of the most popular. This test is widely used for in situ soil characterisation given its continuous measurements and repeatability. The interpretation of such test results has a strong theoretical background, namely allowing the identification the soil profile based on correlation of mechanical properties. In turn, the application of multivariate statistical methods allows identifying the soil stratigraphy by data association. This paper proposes a multivariate statistical approach, based on cluster analysis, for defining stratigraphic interfaces from SCPTu measurements. Using an extensive database from the Lower Tagus Valley region (near Lisbon, Portugal), this novel approach combines the four SCPTu parameters and recognises the association between the measurements. A comparison between statistical results against the profiles of soil behaviour index validates the efficiency of the approach, indicating a good agreement between data associations with the soil behaviour index profiles. The findings showed that the statistical procedure proposed in this study leads to reliable results for identifying soils with similar soil behaviour type.

岩土地面调查涉及多种测试技术。在这种技术中，带有地震波测量的压电渗透测试（SCPTu）是最受欢迎的技术之一。鉴于其连续测量和可重复性，该测试被广泛用于原位土壤表征。这种测试结果的解释具有很强的理论背景，即允许根据机械特性的相关性来识别土壤剖面。反过来，多元统计方法的应用允许通过数据关联来识别土壤地层。本文提出了一种基于聚类分析的多元统计方法，用于通过SCPTu测量定义地层界面。使用来自塔霍斯河谷下游地区（葡萄牙里斯本附近）的广泛数据库，这种新颖的方法结合了四个SCPTu参数并识别出测量值之间的关联。将统计结果与土壤行为指数分布图进行比较可以验证该方法的有效性，表明数据关联与土壤行为指数分布图之间具有良好的一致性。研究结果表明，本研究提出的统计程序可为识别具有相似土壤行为类型的土壤提供可靠的结果。