Experimental and intelligent techniques to estimate bearing capacity of cohesive soft soils reinforced with soil-cement columns. (August 2018)

• Record Type: Journal Article
• Authors: Bunawan, Aliff Ridzuan; Momeni, Ehsan; Armaghani, Danial Jahed; Nissa binti Mat Said, Khairun; Rashid, Ahmad Safuan A.
• Is Part Of: Measurement. Volume 124(2018)
• Page Start: 529
• Page End: 538
• Publication Date: 2018-08
• Subjects: Deep soil mixing -- Cohesive soil -- ANN -- ANFIS -- Bearing capacity; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2018.04.057
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Deep Mixing (DM) method is a soil improvement technique, in which the soil is mixed in situ with stabiliser agents (cement and/or lime) in the form of slurry or powder. The group column is one of DM type that has been extensively used to treat subsoil. Limited attempts have been made to investigate the failure behaviour of a group of floating soil–cement columns. This paper presents a series of 1-G physical tests to identify the bearing capacity of floating soil–cement columns formed via DM method. Pre-formed soil cement columns of 24 mm diameter with different lengths (i.e. 50 mm and 100 mm) were installed beneath a rigid foundation of 80 mm width in a soft clay bed using replacing method. The study also compares the effect of different improvement area ratios i.e. 17%, 26% and 35% on the bearing capacity. Overall, the laboratory results suggest the improvement ratio in the range of 29–79% when the bearing capacity of the treated soil is compared with that of untreated soil. Additionally, using the results of the laboratory tests as well as a dataset comprising other related works (overall 55 sets of data), the paper proposes a predictive model of bearing capacity using adaptive neuro-fuzzy inference system (ANFIS) for cohesive soft soils reinforced with soil–cement columns. The prediction performance of the proposed ANFIS model (coefficients of determination equal to 0.989 and 0.960 for training and testing data, respectively) reveals the feasibility of the model.