Compressive Behaviour of Sustainable Concrete-Filled Steel Tubes Using Waste Glass and Rubber Glove Fibres

Abstract

To reduce the carbon footprint of the concrete industry and promote a circular economy, this study explores the reuse of waste materials such as glass powder (GP) and nitrile rubber (NR) fibres in concrete. However, the inclusion of these waste materials results in lower compressive strength compared to conventional concrete, limiting their application to non-structural elements. To overcome this limitation, this study adopts the concept of confined concrete by developing concrete-filled steel tube (CFST) stub columns. In total, twelve concrete mix variations were developed, with and without steel tube confinement. GP was utilised at replacement levels of 10–30% by weight of cement, while NR fibres were introduced at 0.5% and 1% by volume of concrete. The findings demonstrate that the incorporation of GP and NR fibres leads to a reduction in compressive strength, with a compounded effect observed when both materials are combined. Steel confinement within CFST columns effectively mitigated the strength reductions, restoring up to 17% of the lost capacity and significantly improving ductility and energy absorption capacity. All CFST columns exhibited consistent local outward buckling failure mode, irrespective of the concrete mix variations. A comparison with predictions from existing design codes and empirical models revealed discrepancies, underscoring the need for refined design approaches for CFST columns incorporating sustainable concrete infill. This study contributes valuable insights into the development of eco-friendly, high-performance structural systems, highlighting the potential of CFST technology in facilitating the adoption of waste materials in the construction sector.

1. Introduction

The concrete industry, with an estimated consumption of 30 billion tonnes per year, is a significant economic sector that has a substantial environmental footprint in terms of energy consumption, CO₂, and other greenhouse gas emissions [1]. Therefore, promoting and implementing sustainable practices in the concrete industry is crucial for reducing the carbon footprint, conserving resources, and improving its energy efficiency. In recent years, different researchers have taken different approaches to make concrete more sustainable, reusing recycled waste materials (e.g., glass, rubber, etc.) in concrete either as a substitute for its constituents or as additive fibres is one such attempt to achieve this goal [2,3]. The use of recycled waste materials can improve the mechanical and durability performance of concrete, making it suitable for both above ground and underground structural applications [4,5].

Glass is one of the oldest and most used man-made materials around the world. In 2018, around 130 million tonnes of glass were produced, out of which only 21% was recycled [6,7]. From a theoretical point of view, glass is a 100% recyclable material [8]. Waste glass can be used as culets in the production of new glasses. However, different colours, chemical composition, and impurities of waste glasses makes it difficult to be reused in the glass production industry [9]. These waste glasses normally end up in landfills. Although glass is an inert and non-biodegradable material, landfills have finite spaces. Moreover, glass can take more than a million years to decompose [10]. During that time, it can break into small pieces and puncture the protective lining of landfills, allowing toxic liquid to seep through the soil and groundwater layer. Hence, alternative uses, such as incorporating waste glass into concrete, have been explored. Research into waste glass as concrete aggregates can be traced back to the early 1960s [11,12]. These early researchers noted a decline in concrete strength and durability properties upon the inclusion of glass aggregates, which can be attributed to the increased potential of alkali–silica reaction (ASR). Upon further investigation in later years, it was discovered that ASR is dependent on the particle size of glass. As the particle size of glass decreases, the probability of ASR expansion reduces [13,14]. The existing literature shows that, when the particle size of glass is less than 300 µm, ASR expansion is negligible, and it starts exhibiting pozzolanic behaviour [15]. Pereira et al. [16] showed that the increase in pozzolanic reaction is higher when the particle size is between 45 µm and 75 µm. Shao et al. [14] observed that the particle size of waste glass significantly influences the performance of concrete, with glass ground to a fineness below 38 µm exhibiting notable pozzolanic activity and contributing positively to strength development at later ages. This led to the use of finely ground glass powder as a replacement for cement [17]. Over the last century, different researchers have used different sizes of GP at different percentages of cement replacement in concrete and mortar. Idir et al. [18] studied the compressive strength development of GP mortars with respect to age, fineness, and substitution level. The results showed that increasing the GP content from 10% to 40% generally reduced compressive strength. However, the pozzolanic activity of GP became evident after 28 days, partially compensating for early-age strength losses. Taha and Nounu [19] replaced up to 20% of cement with finely ground GP (45 µm) and reported a 16% reduction in concrete strength due to the incorporation of glass powder as a partial cement replacement. Elaqra and Rustom [20] reported that while 10% and 20% GP replacements showed lower strength than the control at 28 days, they surpassed it after 90 days. Jiang et al. [21] conducted a comprehensive review of the existing literature, where they reported that most of the existing research used GP particle sizes less than 75 µm and the incorporation of waste GP as a partial cement replacement generally results in a reduction in mechanical properties, particularly compressive strength.

In the modern construction industry, the demand for high-strength concrete (HSC) is steadily increasing due to its superior mechanical properties. However, the brittle nature of HSC poses another significant challenge, as increased compressive strength is often accompanied by reduced ductility and limited post-peak deformation capacity. To address this limitation, the incorporation of high strain capacity materials has proven to be an effective strategy. Traditionally, fibres such as steel and polypropylene have been used to enhance the ductility and energy absorption capacity of concrete [22]. However, the growing environmental concern and the drive towards sustainable construction practices have led to research into more sustainable alternatives, including the use of recycled waste materials such as crumb rubbers from waste tyres or fibres from nitrile rubber (NR) gloves. Rubber absorbs a large amount of plastic energy in concrete, therefore improving the overall energy absorption capacity and ductility in concrete. However, higher rubber content adversely affects the compressive strength and elastic modulus of concrete [23,24,25]. While the use of crumb rubber as an aggregate replacement in concrete is well-established, the incorporation of NR glove fibres is a relatively novel approach [3]. NR gloves are predominantly used by the healthcare industry. The current market for NR gloves is USD 16.25 billion, which is expected to increase at a CAGR (compound annual growth rate) of 11.7% over the next 5 years [26]. These are synthetic gloves which mainly consist of acrylonitrile and butadiene [27]. The nitrile butadiene rubber (NR) contains considerable amounts of sulphur (0.73%) and nitrogen (0.16%), which makes it unfeasible for chemical recycling processes such as pyrolysis and steam gasification [27,28]. Instead of disposing of these waste gloves in landfills or incinerating them, both of which are harmful to the environment, they can be reused in the concrete industry. These recycled NR glove fibres not only promote environmental sustainability by reducing landfill waste but can also improve the fatigue resistance and energy absorption capacity of concrete [29]. So far, only three studies have been conducted on concrete with shredded rubber gloves. Kilmartin-Lynch et al. [3] were the first to investigate the application of NR fibres in structural concrete, reporting that incorporating up to 0.2% NR fibres significantly enhanced compressive strength. Microstructural analysis further revealed a strong bond between the NR fibres and the cement matrix. Ran et al. [29] investigated the fatigue resistance of concrete upon the inclusion of NR fibres at different concentrations (0.5% to 2%). Although the compressive strength of the concrete was reduced upon the inclusion of NR fibres, a higher percentage of NR fibres led to a reduced number of cracks and higher deformability upon repeated loading.

The incorporation of NR fibres into GP concrete offers twofold benefits. Firstly, it further reduces the carbon footprint by utilising additional recycled materials, contributing to the development of more sustainable and environmentally friendly concrete. Secondly, the combined use of GP and NR fibres will enhance the concrete’s energy absorption capacity and ductility. However, this combination may cause a more pronounced reduction in compressive strength. To offset the strength reduction associated with this inclusion, external confinement can be an effective solution. Over the last decades, steel confinement has proven to be an effective method for restoring the mechanical properties of sustainable concrete that experiences strength loss [30,31]. In concrete-filled steel tubular (CFST) columns, the outer steel tube does not only serve as formwork for the concrete core, but it can also provide confinement to the concrete by restricting the outward displacement of concrete, thereby improving overall axial load capacity. In addition, the concrete core reduces the steel tube’s sensitivity to local buckling [32].

Based on the existing literature, numerous studies have been carried out to evaluate the effects of GP as cement replacement in mortar and concrete at the material level. However, research on the incorporation of NR fibres in concrete is highly limited, with only two known studies addressing this topic. To date, no study has investigated the combined use of GP and NR fibres in concrete. Furthermore, the application of such GP-NR based concrete in concrete-filled steel tube (CFST) stub columns remains entirely unexplored. The novelty of this research lies in the fusion of these two waste materials in concrete along with confinement provided by the steel tubes. Investigating the mechanical performance of concrete incorporating both GP and NR fibres, particularly within CFST systems, can provide valuable insights into the structural applications of using these recycled waste materials.

With the aforementioned research gap in consideration, this study aims to investigate the effects of incorporating recycled GP as a partial cement replacement and NR gloves as additive fibres in concrete, both independently and in combination, to reduce carbon footprint and enhance the sustainability of concrete industry. This study also explores the potential of steel confinement as a method to mitigate strength loss and improve the structural performance of sustainable concrete. A total of twelve variations of sustainable concrete mixes were experimentally tested, both with and without steel confinement. The key parameters examined were the different proportions of GP and NR fibres, assessed individually and in combination, to evaluate their impact on the performance of confined and unconfined concrete. The findings of this study could contribute to the development of eco-friendly, high-performance construction materials, paving the way for more sustainable infrastructure solutions.

This study is organised into seven sections. Section 2 investigates the properties of the raw waste materials that were used in the concrete mix. Section 3 outlines the concrete mix design developed for the study. Section 4 describes the preparation of the CFST specimens and the compression testing setup. Section 5 presents the experimental results and their discussion. Section 6 compares the experimental findings of this study with existing theoretical prediction models proposed by various design standards and researchers. Finally, Section 7 summarises the conclusions drawn from the research.

2. Materials

2.1. Cementitious Materials

Two types of cementitious materials, general purpose cement and waste glass powder, were used in this study and their details are discussed in the following sub-sections.

2.1.1. General Purpose Cement

For this study, general purpose cement conforming to AS 3972 [33] was used. The particle size distribution curve of cement was investigated using the “Malvern Mastersizer 3000” particle analyser. The cumulative volume particle size distribution of cement is reported in Figure 1, along with d₉₀, d₅₀, and d₁₀, which are 46.5 µm, 17.7 µm, and 3.7 µm, respectively. Figure 2 shows SEM images of cement, which were obtained using the “Phenom XL G2” scanning electron microscope. Cement particles were irregular in shape and a wide range of particle size was visible, which corroborates with the particle analyser data.

2.1.2. Glass Powder (GP)

Existing studies have extensively explored the use of GP as a partial substitute for cement, examining a wide range of particle sizes. Some researchers have focused on coarse GP with particle sizes ranging from 100 to 600 µm [34,35,36,37], while others have investigated finer particles between 25 and 75 µm [20,38,39,40,41,42,43,44]. Additionally, several studies have examined ultra-fine GP with particle sizes below 25 µm [13,45,46,47,48,49,50]. This aforementioned research has shown that when the particle size of GP is less than 150 µm, the silica in the glass reacts with calcium hydroxide to form calcium silicate hydrate (CSH), exhibiting pozzolanic behaviour. This pozzolanic activity improves as the particle size decreases; for instance, glass particles with a size of 25 µm demonstrate better performance than those at 150 µm. However, grinding glass into ultra-fine sizes (<25 µm) is an energy-intensive and time-consuming process. Moreover, according to ASTM C1866 [51], ground soda–lime glass can be classified as a pozzolan if 95% of its particles are smaller than 45 µm. Aligning with these standards and the findings from the existing literature, fine waste glass powder (<45 µm) as a partial replacement for cement was used. Apart from particle size, the performance of GP concrete is influenced by several factors, including the type of glass, its chemical composition, pozzolanic reactivity, etc. Therefore, there is no universal consensus among researchers regarding the optimum percentage of cement replacement with GP. A review of the existing literature reveals that while most studies have investigated replacement levels up to 30% [20,38,39], some have extended this to as high as 60% [2,52,53]. However, it is widely reported that the 28-day compressive strength of concrete generally decreases with increasing GP content, primarily due to the dilution of cementitious materials and the relatively slower pozzolanic activity of GP compared to conventional cement. In the present study, three GP replacement levels, 10%, 20%, and 30%, were selected. These values were chosen to explore the balance between enhancing sustainability and maintaining adequate mechanical performance. The selected range also aligns with the most studied and practically feasible replacement levels reported in the literature. Waste glass in powder form was not commercially available from suppliers in the Greater Brisbane area. Therefore, to address this challenge, recycled crushed glass sand was sourced from a Brisbane City Council project. This material consisted of uniformly graded particles (<5 mm) that were cubical in shape, free of sharp or elongated edges, and mixed in colour. To produce the desired fine glass powder, the crushed glass sand was processed using a Fritsch Ball Mill (Pulverisette 5). The coarse glass sand was ground at 250 rpm for one hour, resulting in the required fine powder. Figure 3 shows the particle size distribution curve of glass before and after ball milling along with the d₉₀, d₅₀, and d₁₀ of GP, which are 37.4 µm, 11.4 µm, and 3.2 µm, respectively. The particle size distribution of glass sand was determined from sieve analysis following ASTM C136 [54] and for GP, the “Malvern Mastersizer 3000” particle analyser was used to determine the particle size distribution. Figure 4 shows SEM images of GP at 1500× and 5800× magnification, which was obtained using “Phenom XL G2” electron microscope. The glass powder was angular in shape with a wide range of particle size, which corroborates with the particle analyser outputs.

2.2. Fine and Coarse Aggregates

The coarse aggregates used in this study were sourced from the local quarries. “Blue metal” aggregates were collected from a local quarry. As for fine aggregates, coarse river sand was used. Sieve analysis for both coarse and fine aggregate was carried out as per ASTM C136 [54], which is shown in Figure 5. Coarse aggregates had a nominal maximum size of 10 mm, and the Fineness Modulus (FM) of the fine aggregates was 2.8. The physical property tests that were carried out for coarse and fine aggregates are shown in

Table 1. Physical properties of coarse and fine aggregates.

Physical Properties	Coarse Aggregate	Fine Aggregate
OD Bulk Density (kg/m3)	1544	2600
SSD Bulk Density (kg/m3)	1572	2620
Specific Gravity (OD)	2.68	-
Specific Gravity (SSD)	2.71	-
Apparent Specific Gravity	2.75	-
Voids	42%	-
Absorption	0.95%	0.7%

.

2.3. Rubber Glove Fibres

Due to the health and safety regulations, contaminated rubber gloves could not be used for this study. Moreover, the sterilisation of the contaminated rubber gloves does not fall under the scope of this study. Therefore, fresh NR gloves procured from local stores were used. A similar approach was adopted by Ran et al. [29]. The rubber gloves were first cut into long strands using scissors and then further processed into small rectangular fibres measuring approximately 5 mm × 15 mm. The fibre dimensions were chosen to align with the sizes used in the studies conducted by Ran et al. [29] and Kilmartin-Lynch et al. [3]. The selection of fibre dosage is a critical aspect in the development of fibre-reinforced concrete. Excessive fibre content can adversely affect workability and hinder uniform dispersion of fibre in concrete, whereas insufficient fibre content may result in similar performance as conventional concrete [55]. Research on NR fibre-reinforced concrete is currently limited, with only two studies available. The dosage parameters for the present study were determined based on a broader review of the literature on other fibre types (e.g., basalt, polypropylene, etc.). Existing studies suggest that, depending on the fibre type and stiffness, up to 1–2% fibre by volume can be incorporated into concrete without significantly compromising its mechanical or durability properties [55]. Kilmartin et al. [3] employed a very low NR fibre dosage ranging from 0.1% to 0.3%, while Ran et al. [29] explored moderate to high dosages between 0.5% and 2%. Kilmartin et al. [3] showed that NR fibres have very low tensile strength (2.73 MPa) compared the conventional fibre materials such as polypropylene or basalt fibres. Ran et al. [29] highlighted that using a higher dosage may further reduce the compressive strength of concrete, even though it can significantly enhance deformability and elasticity. Therefore, in this study, two fibre contents, 0.5% and 1% (by volume of concrete), were selected, considering a balance between maintaining sufficient strength and improving ductility and energy absorption capacity.

2.4. Admixture

In this research, Sika Viscocrete 1250, a high-range, water-reducing admixture conforming to the requirements of AS 1478.1 [56], was used to achieve the desired workability. This third-generation superplasticizer, composed of an aqueous solution of modified co-polymers, has a specific gravity of 1.07 g/cm.

3. Concrete Mix Proportions

Concrete mixing was carried out using a barrel mixer in QUT laboratories. A total of 12 concrete batches was prepared for this study, which is shown in

Table 2. Proportions of concrete mixtures.

Mix ID	Description	Water (kg/m3)	Cementitious Materials		Aggregates		Rubber Gloves Fibre (%)
			Cement (kg/m3)	Glass Powder (kg/m3)	Coarse Aggregate (kg/m3)	Fine Aggregate (kg/m3)
CS	Control	216	554	-	849	688	-
GP10S	GP	216	499	55	849	688	-
GP20S		216	443	111	849	688	-
GP30S		216	388	166	849	688	-
NR05S	NR fibres	216	554	-	849	688	0.5
NR10S		216	554	-	849	688	1
GP10-NR05S	GP + NR	216	499	55	849	688	0.5
GP10-NR10S		216	499	55	849	688	1
GP20-NR05S		216	443	111	849	688	0.5
GP20-NR10S		216	443	111	849	688	1
GP30-NR05S		216	388	166	849	688	0.5
GP30-NR10S		216	388	166	849	688	1

. The target compressive strength for the control concrete was set at 60 MPa, and the mix design was developed following the guidelines provided in ACI 211 [57]. One of the key parameters influencing the strength of concrete is the water-to-binder (w/b) ratio. To determine an appropriate w/b ratio for achieving the target strength, a brief literature review was initially conducted to examine mix designs with similar strength levels. Based on insights from the literature and a series of trial mixes, the w/b ratio was optimised and finalised at 0.39. This ratio was consistently maintained across all concrete batches. The total cementitious material content in each batch was set at 554 kg/m³. To achieve the desired workability without compromising strength, a uniform dosage of superplasticiser (Sika Viscocrete 1250) was used at 450 mL per 100 kg of binder. This dosage was also determined through preliminary trials, considering the manufacturer’s recommendations and the performance of the fresh concrete. The mix proportions for the remaining constituents (fine and coarse aggregates and water) were calculated using the ACI mix design methodology. Once the control concrete mix, denoted as CS, comprising only conventional raw materials, was finalised and verified for both strength and workability, it was used as the reference mix for all subsequent experimental batches. Three batches, namely GP10S, GP20S, and GP30S, were prepared by partially replacing cement with glass powder at substitution levels of 10%, 20%, and 30% by weight, respectively. To investigate the effect of rubber fibre inclusion, two additional batches, NR05S and NR10S, were produced by incorporating NR fibres at 0.5% and 1% by volume of concrete, respectively. The remaining six batches, referred to collectively as the GP-N series, were prepared to investigate the combined effect of GP as a cement replacement and NR as additive fibres. These mixes explored varying proportions of both materials to assess potential synergies or compounded effects on concrete performance.

4. CFST Specimen Preparation and Testing

This study used Grade C350L0 cold-formed circular hollow steel tubes conforming to AS/NZS 1163 [58]. These 90 mm nominal bore steel tubes had an outer diameter of 101.6 mm and a wall thickness of 3.2 mm. The tubes were used to develop stub column, which is defined as short members, having a length-to-diameter (L_e/d_o) ratio less than 4 [59]. Therefore, in this study, a L_e/d_o ratio of 2 was adopted. The steel tubes, which were originally supplied in 6500 mm lengths, were cut to a final length of 204 mm using a band saw (see Figure 6a). The rough-cut surfaces of the tubes were then smoothened with an angle grinder, and a die grinder was used to remove rust and debris from the interior surface of the tubes. Additionally, 5 mm thick, 125 mm square acrylic sheets were used as base plates prior to casting the concrete into steel tubes. As shown in Figure 6b, the acrylic sheets were attached at the bottom of the steel tubes using polyurethane-based adhesive Sikaflex 11 FC⁺. The adhesives were cured at room temperature for 24 h prior to concrete casting. After mixing the concrete in a barrel mixer, it was transferred to a tray and thoroughly mixed manually with a trowel. Once a uniform concrete mix was achieved, it was poured into the steel tubes and cylinder moulds using a scoop. The specimens were filled in two equal layers, with each layer being compacted 25 times using a tamping rod. After compaction, the top surface of each specimen was levelled and finished with a flat plasterer’s trowel to achieve a smooth finish. A rubber mallet was then gently used to tap a few times on the sides of the specimen for further release of any remaining air bubbles. These combined compaction techniques were employed to prevent defects such as honeycombs and voids, thereby ensuring a homogenous concrete mix. After 24 h of concrete casting, the acrylic sheets from the bottom of the steel tubes were removed and the CFST specimens were wrapped in plastic and left for curing at room temperature for 28 days (refer to Figure 6c). For each batch of concrete, along with the CFST stub columns, three concrete cylinders of 100 mm diameter and 200 mm height were cast in accordance with ASTM C192 [60]. In order to keep the curing regimen similar to the CFST specimens, the concrete cylinders were also wrapped in plastic and left for cutting at room temperature for 28 days.

During the end of the curing period, close observation revealed longitudinal gaps between the top concrete surface and the steel tubes (shown in Figure 6d). To address this issue, around 48 h before testing, epoxy-based two-component thixotropic rapid hardening mortar, Sikadur-31 CFN, was applied to fill the gaps. The CFST specimens were then placed in a shaded area to allow the mortar to cure effectively. Additionally, during the compression tests, to ensure the load was evenly distributed, a 6 mm thick marine plywood sheet was positioned between the top of the specimen and the loading plate.

The axial compression tests of the CFST specimens were carried out at QUT Banyo Pilot Plant Precinct using a 400-tonne hydraulic compression testing machine equipped with a universal data acquisition system. The short length of the stub columns, which was less than the machine’s safe manoeuvring distance, necessitated the use of additional steel plates. These plates ensured that the specimens could deform to the desired extent without exceeding the machine’s safety limit. Therefore, two 60 mm thick steel plates and one 20 mm thick steel plate were positioned at the bottom, and a 20 mm thick steel plate was placed at the top to bring the specimens closer to the load cell (refer to Figure 7). The load was applied from the top through Durapac hydraulic load cell at a low rate (less than 1 mm/min). The hydraulic pressure gauge of the load cell was calibrated before the tests and a 1.44% error margin was observed, which was within the acceptable limit. The axial shortening of the specimens was primarily measured using the linear variable displacement actuator (LVDA) integrated into the testing machine. The LVDA was calibrated before testing and had an error margin of 7.9%, which was within the acceptable limit. Nevertheless, to ensure accuracy, the axial shortening measurements from LVDA were verified using an additional laser displacement sensor (LDS). A Panasonic HG-C1200 LDS, capable of measuring displacements up to ±200 mm, was employed for this purpose. Two Omron ZX1-LD100A61 laser sensors, each with a measurement range of ±35 mm, were placed 180 degrees apart to measure the transverse displacement at the mid-height of the specimen. All laser sensors were calibrated prior to testing. The calibration results indicated error margins of 0.07% for the axial sensor and 0.27% and 1.27% for the two horizontal sensors, respectively. Moreover, to obtain the localised axial and hoop strain data, two strain gauges were attached at the mid-height on opposite sides of the specimen. BF350-3AA strain gauges were used for this study, which had a resistance of 350ohm and a gauge length of 3 mm. The outer surface of the steel tube was already coated with black paint to protect it against corrosion. Therefore, before attaching the strain gauges, the black paint was sanded off using an orbital sanding machine. At first, 120-grit sandpaper was used to remove the black paint and expose the bare metal surface. After that, 240-grit sandpaper was used to eliminate surface irregularities and achieve a smooth finish. The whole process was carried out with due care to avoid excessive removal of the underlying metal. Once the bare metal surface was exposed and smoothed, it was thoroughly cleaned with acetone. The strain gauges were then attached to the exposed steel surface using cyanoacrylate adhesive. To protect the installed strain gauges from moisture, dust, or oils, a thin coating of nitrocellulose-based lacquer was applied as a protective sealant. Then the specimens were placed concentrically on the lower loading plate and the upper loading plate was brought into close contact with the specimen. The load cell along with the strain gauges, LVDA and LDS, were all connected to the universal data acquisition system to obtain the load, strain, and displacement data throughout the test.

5. Test Results and Discussions

5.1. Steel Tubes

A hollow tube with a 101.6 mm outer diameter and 3.2 mm thickness was tested under compression. The length of the tube was kept as 204 mm. The test setup used for the hollow tube was kept the same as the test setup for the confined specimens, which is shown in Figure 7. The tubes had a yield strength of 350 MPa. The maximum squash load of the specimen was recorded at 370 KN. Figure 8 shows the load–displacement curve of the hollow steel tubes. The failure mode of the steel tubes was determined as local outward buckling (elephant foot buckling).

5.2. Concrete Cylinders

Compressive strength tests on the concrete cylinders were conducted in accordance with ASTM C39 [61] using the 200-tonne Instron 400RD compression testing machine. In order to distribute the load evenly onto the top concrete surface, a rubber cap was placed between the concrete cylinder and the upper loading plate. The compressive load was transferred at a low loading rate (1 mm/min) through the hydraulic load cell. Displacement and load values throughout the tests were obtained through the inbuilt Instron software. However, the strain values of the concrete cylinders were obtained using an extensometer (see Figure 9).

Figure 10 illustrates the compressive strength of various concrete mixes, represented in MPa, with different combinations of glass powder and rubber glove fibre percentages. The control concrete (CS) has the highest compressive strength at 62 MPa. In the GP series, adding 10%, 20%, and 30% GP results in compressive strengths of 53 MPa, 51 MPa, and 50 MPa, with corresponding reductions of 15%, 18%, and 20%. While the strength decreases with more glass powder, the rate of reduction seems to slow down at higher concentrations. In the NR series, specimens with 0.5% and 1.0% fibre content had 53 MPa and 49 MPa strength, with reductions of 14% and 22%, respectively. The NR fibre dosage used in this study is comparable to that in the work of Ran et al. [29] and the observed reduction in concrete strength also aligns with their findings, which reported 13% and 22% reductions for 0.5% and 1% fibre content, respectively. However, due to the limited availability of data on the use of GP as a cement replacement in concrete, the findings of this study were validated against existing works on both GP-based paste, mortar, and concrete. Studies by Zeybek et al. [34], Shayan & Xu [13], Li et al. [35], and Idir et al. [18] reported that replacing up to 10% of cement with GP resulted in a 5–10% reduction in compressive strength. Additionally, Taha & Nounu [19], as well as Shayan & Xu [13], observed a 16–25% reduction in strength when 20% of the cement was replaced with GP. Similarly, Zeybek et al. [34], Shayan & Xu [13], and Kalakada et al. [15] reported a 16–44% reduction in compressive strength at a 30% replacement level. The aforementioned data indicate that the experimental findings of GP series concrete are in good agreement with existing research. The combined GP and NR (GP-NR) series shows the most significant reduction in compressive strength, with higher concentrations of both additives producing a compounded effect. For example, a mix with 30% GP and 1.0% NR (GP30-NR10S) shows the largest reduction of 32%, with a compressive strength of only 42 MPa. This suggests that the combined use of GP and NR increases micro-voids and reduces bonding, thereby further compromising strength properties.

5.3. Failure Mode of CFST Stub Columns

Figure 11 shows the failure pattern (surface damage) of the confined specimens. The specimens demonstrated a local outward buckling of the tube (elephant foot buckling), which was most prominent at the ends of the specimens. For further damage investigation of the CFST stub columns, the outer steel tubes were cut with an angle grinder to see the damage of the concrete core. As shown in Figure 12, the outward buckling of the steel tube led to the crushing of the concrete core adjacent to the buckled areas of the steel tube, where the concrete was effectively “sandwiched” and subjected to high localised stresses. Additionally, as the infill concrete began to crush internally, it exerted lateral pressure on the surrounding steel tube. This pressure caused localised bulging, particularly noticeable around the mid-height of the columns, where the steel tube started to deform outward.

Overall, the specimens exhibited a ductile failure mode, which can be attributed to the small diameter-to-thickness (D/t) ratio, i.e., thick-walled outer tube. The inclusion of waste materials in the concrete had no significant effect on the failure mode of the test specimens. The failure characteristics remained consistent across all specimens, regardless of the waste material content in different concrete mixes.

5.4. Axial Load-Shortening Response of CFST Specimens

The load vs. displacement curves of the CFST columns with different types of concrete infill are shown in Figure 13. The load value was extracted from the load cell of the machine and the displacement values presented here were derived by axial displacement from the actuator and verified against laser displacement sensor. In all specimens, a slight nonlinearity was observed at the initial loading phase, which was due to the timber plate that was used to distribute the load evenly into the top concrete and steel surfaces. After that, during the linear-elastic phase, deformation increased at a low rate with increasing load. This continued till around 70–80% of the peak load for all specimens. Afterwards, during the plastic phase, displacement rapidly increased with increasing load. After reaching the peak load, the specimens behaved differently depending on the concrete infill. The curve characteristics can be classified in three types (see Figure 13a) namely, Type 1, after reaching the peak load, there is a descending branch, which ultimately reaches a stable residual strength plateau. Out of 12 specimens, 5 specimens (CS, GP10S, GP20S, NR05S, NR10S) followed this trend (see Figure 13b). There was a reduction of 10–15% from peak load then it reached a stable residual strength plateau. In Type 2, there was also a descending branch after the peak load. However, in these specimens, the residual strength reduces by only 3–7% of the peak load. Five specimens (GP30S, GP10N05S, GP10N10S, GP20N05S, and GP30N05S) followed this trend (see Figure 13c). In Type 3, after the peak load, a strain-hardening branch was observed, where the residual strength surpassed the peak load. Only two specimens, which have high glass content and maximum rubber content (GP20N10S and GP30N10S), followed this trend (see Figure 13d).

5.5. Load–Strain Relationships

The axial load vs. strain curves for the confined specimens are shown in Figure 14 and Figure 15. The load values were extracted from the load cell of the compression testing machine and the strain values were derived from strain gauges. Two strain gauges were attached at the mid-height of each specimen. One was attached for the axial strain and the second one was attached at the opposite face (180 degrees) for transverse strain. The positive value of the strain refers to the lateral strain and the negative value refers to the axial strain of the specimens.

Due to the excessive deformations of the steel tube in the post-peak stage of the curves, some of the strain gauges broke in the middle of the tests and stopped giving any outputs. As a result, some branches of the load–strain curve were terminated. However, the data before that point was found reliable and verified against the displacement laser data. The load–strain graphs were found to be bilinear with a high initial slope. The second part of the curve had a relatively flat plateau, where strain increases rapidly with a slight increase in load, which corroborates with the load–displacement curves of the specimen and the existing literature. Using glass powder and rubber glove fibres in concrete infill had no apparent effect on the shape of the load strain graphs of the confined specimens.

5.6. Peak Axial Load and Residual Strength

As illustrated in Figure 13, the specimens achieved their first peak load at an axial strain of approximately 5%. This initial peak load, which occurred within this strain range, is designated as the peak load in this study. Due to the confinement effect of steel tubes, after the ascending and descending branches of the load–deformation curves, it reaches a relatively stable residual strength. The values corresponding to this residual strength and the peak load are summarised in

Table 3. Details of the prepared confined specimens.

Specimen ID	GP as Cement Replacement	NR as Additive Fibres	Height (mm)	Diameter (mm)	Tube Thickness (mm)	Peak Load (KN)	Residual Strength (KN)
CS	-	-	204	101.6	3.2	893	757
GP10S	10%	-	204	101.6	3.2	817	742
GP20S	20%	-	204	101.6	3.2	803	720
GP30S	30%	-	204	101.6	3.2	775	727
NR05S	-	0.50%	204	101.6	3.2	862	773
NR10S	-	1%	204	101.6	3.2	848	748
GP10-NR05S	10%	0.50%	204	101.6	3.2	806	750
GP10-NR10S	10%	1%	204	101.6	3.2	798	740
GP20-NR05S	20%	0.50%	204	101.6	3.2	752	727
GP20-NR10S	20%	1%	204	101.6	3.2	740	745
GP30-NR05S	30%	0.50%	204	101.6	3.2	739	716
GP30-NR10S	30%	1%	204	101.6	3.2	708	727

.

The peak load of the control specimen (CS) was 892 KN. In the GP series, where cement of infill concrete was replaced with GP at concentrations of 10%, 20%, and 30%, a clear trend of reduced peak load capacity is observed. The specimen with 10% glass powder (GP10S) exhibited a peak load of 817 KN, corresponding to a 9% reduction from the control. As the GP content increased to 20% in GP20S, the peak load further decreased to 803 KN, resulting in a 10% reduction. At the highest concentration, GP30S, with 30% glass powder, the peak load dropped to 775 KN, a 13% reduction. The reduction in peak load of the CFST stub columns observed in the GP series can be attributed to the decreased strength of the infill concrete resulting from the replacement of cement with GP, which is due to the delayed pozzolanic reactivity of GP and the dilution effect [21]. In contrast, the NR series, which consisted of shredded NR gloves as additive fibres in infill concrete, showed a relatively smaller reduction in peak load. With 0.5% NR fibres (NR05S), the peak load reached 862 KN, representing only a 3% reduction from the control. When the fibre content increased to 1% in N10S, the peak load decreased to 848 KN, corresponding to a 5% reduction. The peak load reduction observed in the NR series CFST stub columns can also be attributed to the decreased strength of NR fibre-reinforced concrete, resulting from the inclusion of the fibres. The low tensile strength of NR fibres, combined with their inherently low stiffness, limits their ability to generate a “bridging effect”, which is typically responsible for strength enhancement of concrete upon incorporating relatively stiffer fibres such as steel. Additionally, NR fibres are softer (elastically deformable) than other constituents of concrete, which further contributes to the reduction in strength.

The combined GP-NR series, which contained both GP and NR fibres in various concentrations, showed a compounded effect on peak load reduction. When 10% GP was combined with 0.5% NR fibres (GP10-NR05S), the peak load was 806 KN, marking a 10% reduction from the control mix. As the fibre content increased to 1% (GP10-NR10S), the peak load slightly decreased further to 798 KN, an 11% reduction. Higher concentrations in both additives produced even more noticeable reductions. For instance, GP20-NR05S, which contained 20% GP and 0.5% NR fibres, showed a peak load of 752 KN, or a 16% reduction, while GP20-NR10S had a peak load of 740 KN, corresponding to a 17% reduction. The greatest reduction was observed in GP30-N10S, where the peak load dropped to 708 KN, representing a 21% reduction.

The data on residual strength reduction from peak load illustrates how glass GP and NR fibres influenced post-peak performance. The control specimen (CS) had a residual strength of 757 KN, indicating a 15% reduction from its peak load. In the GP series, as GP content increased, post-peak strength loss generally decreased. For instance, GP10S (10% GP) had a 9% reduction, GP20S (20% GP) showed a 10% reduction, and GP30S (30% GP) achieved the lowest reduction at 6%, suggesting that higher GP content improved residual strength retention. In the NR series, the addition of NR fibres at low concentrations also contributes to a moderate reduction in post-peak strength loss. Specifically, NR05S, which contains 0.5% fibres, showed a 10% reduction, while NR10S, with 1% fibres, exhibited a reduction of 12%. When examining the combined GP-NR series, certain mixtures demonstrated a notable enhancement in post-peak strength retention. For instance, GP20-NR10S showed a slight increase of 1% in residual strength, while GP30-NR10S showed a more pronounced increase of 3%. This suggests that the incorporation of a higher concentration of NR fibres in conjunction with GP in infill concrete can effectively bolster the post-peak residual strength behaviour of the CFST stub columns, which indicates that the specimens exhibit more ductile behaviour under the tested conditions, as evidenced by their ability to retain load-bearing capacity and undergo further deformation beyond the peak load.

5.7. Strength Index and Confinement Factor

To quantify the strength enhancement due to the confinement of concrete, two parameters, namely the strength index (SI) and confining factor of the CFST specimens, were determined. The SI of the confined stub column [62] is defined in Equation (1).

$$SI={\displaystyle \frac{N_{max}}{A_c{f}_c^{\prime }+A_s{f}_y}}$$

(1)

where $N_{max}$ is denoted as the maximum compressive load recorded during the compressive test. The sum of sectional capacities is defined as ${A_{c}f}_c^{\prime }+A_s{f}_y$, where $A_c$ is the area of concrete, $f_c^{\prime }$ is the compressive strength of concrete cylinders, $A_s$ denotes the cross-sectional area, and $f_y$ refers to the yield strength of the steel tube. Figure 16 shows the strength index values, which range from 1.09 to 1.23. It was observed that for all confined specimens, SI values were greater than 1, indicating that the confinement remained effective even with the inclusion of GP and NR fibres in infill concrete.

To evaluate the confinement contribution of the steel tubes on the compressive strength behaviour of CFST, a confinement factor $\theta$ is introduced [63], which is as shown in Equation (2).

$$\theta ={\displaystyle \frac{A_s{f}_y}{A_c{f}_c^{\prime }}}$$

(2)

The values of the confining factor of the specimens are shown in Figure 16b, which ranges from 0.79 to 1.15. It was observed that the confinement factor of the CFST specimens increases with increasing glass and glove fibre content infill concrete, demonstrating that compared to control concrete, glass and glove fibres, used both independently and in combination, showed a more significant confinement action of the steel tube.

5.8. Ductility Index and Fracture Energy

Ductility refers to a material’s ability to undergo significant deformation beyond its elastic limit while retaining a reasonable load-carrying capacity. To evaluate the ductility of confined specimens, an energy-based ductility index (µ) was calculated using the equation $\mu =E_c/E_y$, which was proposed by Duarte et al. [64]. In this equation, $E_c$ and $E_y$ represent the energy at peak load and yield load, respectively. The energy is defined as the area under the axial load-shortening curve, where $E=\int F du$. Fracture energy, on the other hand, refers to the energy absorbed to break a specimen, which is represented as the total area under the load displacement curve [65]. The calculated fracture energy and ductility indexes for the confined specimens are presented in Figure 17.

The control specimen (CS) exhibits a DI of 1.03 and a fracture energy of 3793 N-m. In the GP series, increasing the GP content improves ductility but reduces fracture energy. Whilst the GP series showed a 17–83% rise in DI, fracture energy declined by 14–28%. In the NR series, adding NR fibres improves ductility but reduces fracture energy. The NR05S specimen achieves a DI of 1.90 and a fracture energy of 4013 N-m, while increasing the fibre content to 1% (NR10S) slightly raises the DI to 1.94 but decreases the fracture energy to 3507 N-m.

The combination of GP and NR fibres (GP-NR series) produces the most significant improvements, demonstrating a synergistic effect. Combining GP and NR fibres significantly enhances both ductility and energy absorption capacity. GP10-NR05S and GP10-NR10S achieve DIs of 1.71 and 2.04 with fracture energies of 3647 N-m and 3747 N-m, respectively. Higher glass powder content further amplifies this synergy, with GP30-NR05S and GP30-NR10S achieving the highest DIs of 3.71 and 4.66, and fracture energies of 5019 N-m and 5320 N-m, respectively.

6. Comparison Between Experimental Tests and Theoretical-Predicted Strengths

Several well-known national standards and empirical models currently address the design of carbon steel CFST columns. However, these codes and empirical models primarily focus on evaluating the performance of conventional CFST columns. It remains uncertain whether the calculation methods used for conventional CFST columns are applicable to CFST columns incorporating GP and NR concrete. Therefore, in this section, the axial capacity of CFST columns is evaluated using commonly used design standards and empirical models proposed by various researchers. For this study, a total of six models was selected and the models are presented in Table 4. In order to keep the results uniform, all partial safety factors were taken as unity.

The comparison between the predicted and experimentally derived ultimate strength of the CFST columns are shown in Figure 18 to understand their variations. In order to compare the predictions with the experimental results consistently, a model error (ME) was introduced, which was defined as the ratio of the experimental and prediction of ultimate strength [66]. The statistical characteristics of ME values were evaluated including mean, minimum, maximum, COV, and confidence level at two different percentiles, which is shown in Table 5.

Table 4. Existing models for prediction of ultimate strength of CFST columns.

Reference	Model
Eurocode 4 [67] and AS5100 [68]	$P_{EC4,AS5100}={\eta }_a{A}_a{f}_{yd}+A_c{f}_{cd}\left(1+{\eta }_c{\displaystyle \frac{t}{d}}{\displaystyle \frac{f_y}{f_{ck}}}\right)+A_s{f}_{sd}$
AISC 360 [69]	$P_{ AISC}=F_y{A}_S+C_2{f}_c^{\prime }\left(A_c+A_{se}{\displaystyle \frac{E_s}{E_c}}\right)$
Giakoumelis & Lam [70]	$P_{ GL}=1.3f_c^{\prime }{A}_c+f_y{A}_s$
Sakino et al. [71]	$P_{ SK}={\gamma }_U{f}_c^{\prime }{A}_c+\left(1+\eta \right)f_y{A}_s$
Lu & Zhao (modified AIJ model) [72]	$P_{ LZ}=f_c^{\prime }{A}_c+\left(1+k_1\right)f_y{A}_s$
Han et al. [73]	$P_{ H}=\left(1.14+1.02\xi \right)f_{ck}{A}_{SC}$

Table 4. Existing models for prediction of ultimate strength of CFST columns.

Reference	Model
Eurocode 4 [67] and AS5100 [68]	$P_{EC4,AS5100}={\eta }_a{A}_a{f}_{yd}+A_c{f}_{cd}\left(1+{\eta }_c{\displaystyle \frac{t}{d}}{\displaystyle \frac{f_y}{f_{ck}}}\right)+A_s{f}_{sd}$
AISC 360 [69]	$P_{ AISC}=F_y{A}_S+C_2{f}_c^{\prime }\left(A_c+A_{se}{\displaystyle \frac{E_s}{E_c}}\right)$
Giakoumelis & Lam [70]	$P_{ GL}=1.3f_c^{\prime }{A}_c+f_y{A}_s$
Sakino et al. [71]	$P_{ SK}={\gamma }_U{f}_c^{\prime }{A}_c+\left(1+\eta \right)f_y{A}_s$
Lu & Zhao (modified AIJ model) [72]	$P_{ LZ}=f_c^{\prime }{A}_c+\left(1+k_1\right)f_y{A}_s$
Han et al. [73]	$P_{ H}=\left(1.14+1.02\xi \right)f_{ck}{A}_{SC}$

Table 5. Statistical parameters through model analysis.

Reference	Mean ME	COV of ME (%)	Minimum ME	Maximum ME	95th Percentile		99th Percentile
					Lower Bound	Upper Bound	Lower Bound	Upper Bound
Eurocode 4 [67] and AS5100 [68]	0.91	4.12	0.86	0.98	0.88	0.93	0.87	0.94
AISC 360 [69]	1.18	3.56	1.14	1.28	1.16	1.21	1.15	1.22
Giakoumelis & Lam [70]	1.00	3.33	0.98	1.09	0.98	1.03	0.97	1.03
Sakino et al. [71]	1.00	3.58	0.96	1.08	0.97	1.02	0.97	1.03
Lu & Zhao (modified AIJ model) [72]	0.94	3.90	0.90	1.02	0.92	0.97	0.91	0.98
Han et al. [73]	1.13	3.92	1.08	1.23	1.11	1.16	1.09	1.17

Table 5. Statistical parameters through model analysis.

Reference	Mean ME	COV of ME (%)	Minimum ME	Maximum ME	95th Percentile		99th Percentile
					Lower Bound	Upper Bound	Lower Bound	Upper Bound
Eurocode 4 [67] and AS5100 [68]	0.91	4.12	0.86	0.98	0.88	0.93	0.87	0.94
AISC 360 [69]	1.18	3.56	1.14	1.28	1.16	1.21	1.15	1.22
Giakoumelis & Lam [70]	1.00	3.33	0.98	1.09	0.98	1.03	0.97	1.03
Sakino et al. [71]	1.00	3.58	0.96	1.08	0.97	1.02	0.97	1.03
Lu & Zhao (modified AIJ model) [72]	0.94	3.90	0.90	1.02	0.92	0.97	0.91	0.98
Han et al. [73]	1.13	3.92	1.08	1.23	1.11	1.16	1.09	1.17

Among the design codes evaluated, EC4 [67] and AS 5100 [68] yielded a mean ME of 0.91, while AISC 360 [69] exhibited a mean ME of 1.18. This indicates that EC4 [67] and AS 5100 [68] tend to overestimate the ultimate axial strength of concrete-filled steel tube (CFST) columns, whereas AISC 360 [69] underestimates it. The minimum and maximum ME values observed for EC4/AS 5100 [67,68] and AISC 360 [69] were 0.86 and 1.28, respectively, which indicates that EC4 [67] and AS 5100 [68] overpredicted the axial strength by up to 14%, while AISC 360 [69] underpredicted it by as much as 28%, when compared to the experimental results. This discrepancy is primarily attributed to differences in how these design codes account for the confinement effect of the steel tube on the concrete core. Whereas EC4 [67] and AS5100 [68] take into account of the strength gain of concrete core due to confining effect of the steel tube, AISC [69] limits the compressive strength of the core concrete to $0.95 f_c^{\prime }$, which results in a more conservative estimate. Given the limitations and variability in predicted strengths among these codes, several researchers have proposed modified or new calculation models to better capture the composite action and confinement effects in CFST columns.

Among the empirical equations, the model proposed by Han et al. [73] yielded a mean ME of 1.13, with minimum and maximum ME values of 1.08 and 1.23, respectively. These indicate that the model was consistently conservative, underestimating the axial capacity of CFST columns by approximately 8% to 23% compared to the experimental results. In contrast, the empirical models developed by Sakino et al. [71], Giakoumelis & Lam [70], and Lu & Zhao’s modified AIJ model [72] exhibited mean model errors of 1.00, 1.00, and 0.94, respectively, demonstrating a good correlation with the experimental results of this study. The COV of ME for all six empirical models ranged between 3 and 4%, indicating low variability and suggesting a high level of consistency among the models.

7. Conclusions

This study investigated the compressive behaviour of CFST stub columns incorporating waste glass powder (GP) as a partial cement replacement and nitrile rubber gloves (NR) as additive fibres in concrete infill. The key findings and conclusions are as follows:

• The inclusion of GP reduced concrete compressive strength by 9–21%, with greater reductions at higher GP content. NR fibres caused a relatively smaller reduction in strength of 3–5%. However, the combined use of GP and NR fibres had a compounded effect in concrete, with the highest strength reduction of 32% observed in GP30-N10S. Upon confinement, the loss of strength became partially recovered, resulting in 5–17% of strength recovery.
• All CFST stub columns exhibited local outward buckling (elephant foot buckling) and crushing of concrete core near buckled areas. The presence of GP and NR fibres did not alter the failure characteristics of CFST specimens.
• Three distinct behaviours were observed in axial load responses, a stable residual strength plateau, minimal strength reduction, and strain-hardening beyond peak load, depending on the GP and NR content. As the GP and NR percentages increased, the residual strength of the confined specimens progressively improved.
• While increasing GP content improved ductility, it led to a reduction in fracture energy of the confined specimens. In contrast, increasing NR fibre content enhanced both ductility and fracture energy. A synergistic effect was observed when GP and NR fibres were used together, resulting in a significant improvement in ductility and energy absorption capacity. As the GP and NR content increased, both energy absorption capacity and ductility improved substantially.
• A comparison of the experimental results with theoretical predictions of the ultimate axial capacity of CFST specimens revealed that Eurocode 4 [67] and AS 5100 [68] overestimated the axial capacity, while AISC 360 [69] underestimated it. In contrast, empirical models proposed by Sakino et al. [71], Giakoumelis & Lam [70], and Lu & Zhao’s modified AIJ model [72] demonstrated greater accuracy, closely aligning with the experimental findings.

In summary, this study demonstrates that incorporating waste glass powder and nitrile rubber fibres in CFST stub columns can enhance sustainability while maintaining structural integrity. The findings emphasise the effectiveness of steel confinement in mitigating strength loss and improving ductility. Additionally, the results highlight the need for refinements in existing design standards to accurately predict the behaviour of sustainable CFST structures.

Future research could investigate the long-term durability and environmental performance of GP-NR concrete. Studies on full-scale structural elements and behaviour under different loading scenario (eccentric, flexure, etc.) and confinement (CFRP, GFRP, etc.) mechanisms are needed to implement practical applications. Additionally, exploring combinations with other recycled materials could advance sustainable construction solutions.