Discrete element analysis of recycled concrete aggregate responses during repeated load triaxial testing

Highlights

• Using of DEM in simulating recycled concrete aggregate under repeated load triaxial (RLT) test.

• A new approach developed to simulate stress-controlled conditions in the DEM simulation.

• The velocity of the loading-unloading stages of the RLT test determined for stable DEM simulation.

• Micro-scale behaviour of the RCA material compared with the macro-scale response.

Abstract

More effective and sustainable use of natural resources besides the mitigation of environmental impacts induced by their extraction could be achieved by reusing construction and demolition (C&D) materials. C&D materials are increasingly being used in pavement bases/subbases as a reliable alternative to virgin quarry materials. The behaviour of flexible pavements under dynamic loads conveyed from traffic load can be evaluated in the laboratory by the repeated load triaxial test (RLT). This research study aims to evaluate the macro- and micro-mechanical interaction and underlying mechanisms of the behaviour of unbound recycled concrete aggregate (RCA) materials under RLT test using discrete element method (DEM) and to compare the simulated results with experimental results. In this research a DEM numerical software was used to develop a novel stress-controlled test method for modelling the RLT test, using a flexible membrane boundary with spherical particles linked through linear contact model. During the simulation, the axial recoverable deformations were monitored to determine the resilient modulus (M_R) at each of the 16 loading-unloading stress level sequences. Furthermore, the micro-mechanics of the RCA samples were investigated using anisotropy tensor of fabric, contact orientation, contact force network and coordination number (CN). Laboratory RLT tests were conducted to verify the robustness of the DEM model. In general, both quantitative and qualitative agreements were attained between the physical tests and the numerical DEM simulations. The results of the micro-scale analyses revealed that the response of the RCA sample during the cyclic test was almost elastic. In other words, the almost-elastic behaviour was not accompanied by a significant change in the material fabric, rather was accompanied by a variation in the magnitude of the contact forces without variation in the contact force network and CN.

Introduction

Over the past decade, extraction of virgin materials from quarries have been reduced significantly due to the scarcity of natural materials, environmental impacts and economic issues. Construction and Demolition (C&D) materials have been extensively studied in recent years as a valuable alternative low-carbon material in the road construction industry [1]. The use of C&D materials in pavement bases/subbases is gaining increasing traction as a sustainable road construction material. In Australia, concrete waste comprises the largest proportion of C&D waste, with approximately 8.7 million tonnes per annum [2]. Recycled concrete aggregate (RCA) used for road base/subbase applications is typically 20 mm and finer. Several studies have investigated the feasibility of using RCA in road bases/subbases [3], [4], [5], [6].

Flexible pavement undertakes moving loads during its life span and unacceptable pavement deflection can affect the behaviour of the flexible pavement [7]. There are several testing methods to study the behaviour of the unbound pavement material [8], [9]. The resilient behaviour of pavement base/subbase materials can be characterised using three methods: (1) performing Repeated Load Triaxial (RLT) laboratory tests; (2) back-calculation using non-destructive field tests such as falling weight deflectometer test; and (3) using correlations with mechanical properties of soils [10].

The RLT test is the predominant testing method to characterise the pavement materials under repeated loading-unloading regime [11], [12]. Hveem in the 1950s [13] initially introduced the resilient properties for unbound granular materials and concluded that the behaviour of these materials are elastic considering that the deformations are recoverable. Seed et al. [14] presented a more robust concept of the resilient modulus (M_R) while studying the fatigue failure behaviour of the pavement materials and M_R was defined as the deviatoric stress over recoverable strain to deal with the accurate response of the pavement materials considering the nonlinearity in their behaviour.

Recently, several studies have been done to analyse the resilient behaviour of unbound pavement materials [15], [16], [17]. It is reported that several factors such as stress level, fine content, grain size, type of the particles, moisture content affect the resilient behaviour of these materials [18], [19], [20]. However, road pavement performance is not fully observed due to the complexity of modelling granular aggregates [21]. Research in recent years has focused on numerical modelling as a consequence of enhancement in computer processors. Discrete element modelling (DEM) is gaining increasing popularity in geotechnical and pavement engineering for its capacity to model granular materials, which are composed of distinct particles that demonstrate the complex macro- micro-mechanical behaviour during loading. Cundall and Stark (1979) [22] initially introduced DEM as a numerical modelling technique for two-dimensional disks.

The use of the DEM approach in studying the mechanical behaviour of granular aggregates was summarised by Ng (2004) [23]. It is worth mentioning that most of the research works in this field are focused on simulating triaxial test under cyclic loading. Lin et al. (2017) [24] studied the effect of the particle size distribution on the behaviour of the base layer under cyclic loading. Huang et al. (2017) [25] investigated the influence of several factors on the cyclic loading characteristics of granular materials. O’Sullivan and Cui (2009) studied the micro-mechanics of granular materials during load reversals [26]. Their DEM model consisted of several loading and unloading steps to give a pre-defined stress ratio. Asadzadeh and Soroush (2017) utilised DEM simulation to model cyclic simple shear test [27]. The cyclic load in their model applied by moving the plates at the constant value throughout the test. DEM has been used to simulate soil samples under cyclic load with almost constant amplitudes, however, based on AASHTO T307-99 testing standard [28] for modelling dynamic load of a moving vehicle at 100 km/h on the pavement, the loading of 0.1 s and unloading of 0.9 s are needed in each cycle. Meanwhile, the applied strain rates on boundary plates in previous simulations using DEM were constant between the loading and unloading stages. These values for RLT test are not constant between the stages and this gap will be addressed in this paper.

The application of DEM approach in pavement base/subbase materials was evaluated by Ngo et al. (2016) [29] by modelling reinforced railway ballast. The authors utilised DEM to simulate geogrid-reinforced ballast under monotonic and cyclic loading and found out that the stress-displacement responses obtained from DEM were in good agreement with the behaviour measured in the laboratory. Bian et al. (2016) [30] used DEM simulation to study the impact of gradation on void ratio and load-carrying performances of ballast material. Repeated train loading test was conducted in the laboratory to provide an insight into optimised ballast gradation. Meanwhile, DEM simulation had been utilized to study the effect of the characteristic gradations on volumetric properties of the assembly. Recently, numerical simulation has been used to study the behaviour of the RCA materials [31], [32], [33], [34]. Tan et al. (2019) investigated the failure processes of RCA and natural aggregate concrete using DEM and verified the results with experimental tests [32].

The primary objective of this paper is to propose a novel method to model RLT test using DEM and study the possibility of simulating cyclic responses of RCA. The proposed DEM model was validated with laboratory RLT tests on RCA to ensure a robust model is developed that can accurately replicate the physical and dynamic conditions. The size, number, behaviour, and the properties of the RCA particles are important as well as the geometry, boundary conditions, and dynamic loading-unloading conditions for the RLT test which should be accurately simulated. In order to model the stress-controlled conditions of the RLT test, a new algorithm was developed and checked for the validity based on the experimental results. Once a good agreement achieved between the macro-mechanical behaviour of the numerical and experimental tests, DEM simulation would be used to study the micro-mechanical behaviour of the sample under the RLT test. These three-dimensional particle-scale responses will provide unique results not only pertaining to the characteristic of the RCA sample under the non-destructive RLT test but also to establish the response of the granular material under dynamic loading-unloading sequences.