Citation: Xu S., Lai J., O’Kelly B.C. and Zhao B. (2023) Reverse extrusion test for fine-grained soil characterisation: internal flow pattern with ANN-enhanced particle tracking. In: Proceedings of the Fourth International Symposium on Machine Learning and Big Data in Geoscience, Cork, Ireland, 29th Aug – 1st Sept 2023. Extended Abstract #94. 3 pp. 1 Reverse Extrusion Test for Fine-grained Soil Characterisation: Internal Flow Pattern with ANN-Enhanced Particle Tracking Shuoshuo Xu 1 , Jinxing Lai 1 , Brendan C. O’Kelly 2 and Budi Zhao 3* 1 School of Highway, Chang’an University, Xian, 710064, China 2 Department of Civil, Structural and Environmental Engineering, Trinity College Dublin, Dublin, Ireland 3 School of Civil Engineering, University College Dublin, Dublin, Ireland * Budi Zhao (email: budi.zhao@ucd.ie) Keywords: Artificial neural network, extrusion test, flow pattern, particle tracking, shear strength 1 INTRODUCTION The reverse extrusion test involves one-dimensionally (1D) compressing a fine-grained soil sample contained in a cup container of cross-sectional area A. The force Fe applied by the loading platen causes extrusion of the soil to occur via a small and centrally located orifice on the platen. The applied force is continuously monitored during the test, with the pressure value causing extrusion (i.e.,  e = e  ⁄ ) proposed as a means of quantifying the undrained shear strength and consistency limits of fine-grained soils [1]. Previous experimental results showed that the pe value correlates well with undrained shear strength at various water content for a specific soil type [2]. However, their relationship varies between soils and also depends on the compression velocity. Gas bubbles entrained during sample preparation and the complex internal flow pattern and possible localized sample consolidation that occur during the extrusion test [2] hinder understanding of the mechanism of the extrusion test. Thus, we adopted X-ray computed tomography (CT) combined with a recently developed marker-based tracking algorithm [3,4] to quantify the internal flow pattern of saturated kaolinite during extrusion tests. 2 METHODOLOGY We developed a reverse extrusion apparatus manufactured from nylon plastic, which was transparent to X-ray. The internal dimensions of the sample container were 38 mm in diameter and 40 mm deep. The extrusion apparatus was fitted into an axial loading frame with detachable loading platen that included a central 6 mm diameter orifice. A stepping motor vertically displaced the loading platen, inducing a build-up in the platen axial force and causing extrusion to occur on reaching the pe value. A load cell and displacement transducer recorded the axial force and platen displacement throughout the test. We prepared the test sample by mixing kaolinite (IMERYS, Speswhite) with deionised water to produce a water content of 57% (i.e., lower than its liquid limit of 67%). Some sand particles (300–600 μm size range) were added to the kaolinite paste as markers for particle tracking. The kaolinite sample within the extrusion apparatus was compressed at a platen axial