Behaviour of Aluminium EN AW 6082 T6 Columns Exposed to Transient Heating—Experimental and Numerical Analysis

Abstract

The paper presents an experimental and numerical analysis of EN AW 6082 T6 aluminium alloy columns exposed to high-temperature creep in transient conditions. Transient tests with columns subjected to a constant heating rate for a persistent external load in the form of the horizontal and transversal forces were carried out. A total of ten columns were examined with varying ratios of horizontal and transversal loads. The test results were compared to numerical results obtained from ANSYS 16.2. The coefficients for an ANSYS built-in Modified Time Hardening creep model were calibrated from the previously conducted tests on coupons and used as a base for the numerical analysis of the column. The study results reveal that creep reduces column load-bearing capacity, starting at temperatures above 150 °C. Furthermore, the level of reduction in the aluminium column capacity, which manifests itself as a runaway failure of the column between the creep and creep-free model, deviates with a difference exceeding 160% in vertical displacement upon failure, while the creep model correlates very well with the results obtained from the tested specimens in terms of failure time and the displacement ratio.

1. Introduction

1.1. Research Motivation

Over the last few decades, aluminium provided great potential for implementation in everyday civil engineering practice. Significant quantities of aluminium worldwide, together with this metal’s advantageous mechanical properties compared with those of other metals, support its use as a standalone structural material. The main drawback of aluminium is fire sensitivity since its melting point is approximately 660 °C and reduction in its load-bearing capacity starts at temperatures over 100 °C [1]. Moreover, aluminium shows high thermal conductivity which results in faster heat transfer and, therefore, more rapid strength reduction compared with structural steel. Generally, metals exposed to fire have noticeable temperature-related strain components, consisting of three types: thermal strain, stress-related strain, and creep strain. Creep strain is time-dependent, and its development depends on the material characteristics and microstructure of the material altogether, applied physical loads, and temperature. Most metals are susceptible to significant creep strain starting at temperatures around 25–30% of their melting point, depending on the applied loads. From an engineering perspective, columns represent critical load-bearing structural elements and if the column fails, there is a high possibility that the structure will collapse [2]. Currently, there is a lack of research on the effect of creep on aluminium columns exposed to high temperatures. This is partly related to the relative complexity of the experiments and the general use of material test data based on transient tests.

Although the first creep development laws were proposed in the nineteenth century after Vicat [3], in 1834, discovered this phenomenon in tensile-loaded steel wire, more serious research on steel and aluminium alloys dates back to the second half of the 20th century. Currently, the design codes for aluminium do not cover the effect of transient induced creep in an explicit manner since its background was scarcely investigated. The creep effect on columns is included in structural fire design with a reduction factor of 1.2 according to EC 1999-1-2 Section 4.2.2.4 [4]. Current strict regulations for fire protection requirements in everyday use are based on accidents caused by fires that occasionally led to structure collapse, so there is a need to explore this possible structural failure mode in more detail. In order to start investigating this topic, experimental data obtained from tests conducted on coupons of aluminium alloy 6082 T6 [5] were used to enable the use of creep models offered by software ANSYS [6]. A verified creep model was subsequently used in the numerical study of the axially loaded aluminium columns, followed up by experimental research, and presented in this study.

1.2. Comparison between Stationary and Transient Creep Tests

Different loading regimes generally influence successive creep development. Tests carried out for creep modelling, in terms of member capacity where the load is increased while the temperature is held constant, are defined as stationary tests or steady-state tests. Transient tests (Figure 1) are used to show the member behaviour in a more realistic manner due to the impact of fire where the temperature load, according to the shape of nominal fire curve ISO 834, increases over time whilst the load remains constant.

1.3. Previous Research

The creep behaviour of aluminium is, currently, a scarcely investigated topic within the structural-fire research community. Significant research breakthroughs on this topic date back to the middle of the twentieth century. Harmathy [7] proposed a new approach to determining the creep development of metals under transient conditions. He applied the Dorn [8] creep model and adapted it to meet the practical requirements and to calculate displacements of protected steel elements exposed to transient high temperatures. Generally, Harmathy expressed time-dependent strain as (1):

$$\frac{1}{Z}\frac{d}{d\theta }=cot{h}^2\left(\frac{{\epsilon }_c}{{\epsilon }_I}\right)$$

(1)

where ε_c is time-dependent creep strain and θ is the temperature defined by Dorn (2)

$$\theta ={{\displaystyle \int }}_0^t{e}^{-∆H/RT}dt$$

(2)

where ∆H represents the creep activation energy (J/mol), R is the gas constant (J/mol°K), and T is the temperature (°K). Z and ε_I are the dimensionless creep parameters dependent only on the applied stress.

Previously published research on the influence of creep on aluminium alloys included studies conducted by Langhelle et al. [9] and Maljaars [10,11,12,13,14,15]. Langhelle et al. conducted research on the behaviour of aluminium AW 6082 columns of an alloy of T4 and T6 temper exposed to fire. Buckling tests of 31 RHS columns with a slenderness ratio of 41 were carried out in stationary (temperature up to 300 °C) and transient (heating rate range 5–12°C/min and the stress level range 75–110 MPa) heating conditions. The main research topic of this study was the influence of high-temperature creep effects on column load capacity and its comparison to design codes. Langhelle et al. reported that the creep strain for temperatures over 170 °C was significant, but insufficiently researched and covered by the current standards for aluminium. Maljaars in his extensive studies covered the numerical and experimental behaviour of 5xxx and 6xxx series aluminium alloys exposed to stationary and transient heating, with particular emphasis on creep strain. He applied the aforementioned Dorn and Harmathy creep models and carried out transient tests with a varying heating rate within the range of 1.6–11 °C/min and a constant load (stress range 20–150 MPa), where the critical temperature interval was within the range 170–380 °C and sufficiently valid to simulate insulated aluminium members exposed to a real fire. The main conclusion of his studies is that the Eurocode 9 approach for creep effects in structural fire design has certain limitations since it is based on conservative steady-state tensile tests instead of transient tests. He also pointed out that the creep model developed by Dorn and Harmathy is not suited for aluminium alloys in series 6xxx for creep strain representation. Kandare et al. [16,17] applied an analytical creep model developed by Maljaars for the plates simulated as columns of aluminium alloys 5083 H116 and 6082 T6 exposed to compression load and transient heating. The main conclusion of this study was a fairly good prediction of the failure of aluminium plates due to buckling at high temperatures with the visible representation of primary and secondary creep phases, but with a noticeable deficiency for complex structures and elements such as I profiles.

Further research on the topic of axially loaded aluminium columns exposed to high temperatures was put forward by Jiang et al. [18] in an extensive study where the authors examined 108 specimens (48 circular tubes and 60 rectangular tube cross-sections). Columns of different slenderness failed due to buckling for different levels of temperature load and the main conclusion was that the EC9 calculation methodology has certain limitations for temperatures above 300 °C so he proposed some additional formulas to estimate the flexural buckling behaviour of aluminium alloy columns for a temperature span of 100–400 °C. Wang et al. [19] carried out a study with a series of 6060 columns with different lengths and irregular-shaped cross-section thickness and stated that the current design codes for columns exposed to fire predict load capacity too conservatively. Based on his conclusions, he proposed the modification of the reduction in the Young’s modulus of aluminium alloys for temperatures above 250 °C. With the increase in interest for use of aluminium alloys as a load-bearing material in everyday structures, Zheng and Zhang [20] simulated the response of fire-protected and unprotected I cross-section aluminium alloy beams and compared the results to the Eurocode 9 standard and proposed additional formulas for calculating the critical temperature of 5xxx and 6xxx series aluminium alloys.

Due to very scarce research data on the behaviour of the I section columns exposed to high-temperature creep, Torić et al. [5,21,22,23,24] carried out a research study. Within the mentioned research, a material study on coupons of EN AW 6082 T6 was conducted as the main prerequisite to determine the creep behaviour of specified aluminium alloy columns exposed to transient heating presented in this paper.

2. Creep Model Calibration of Aluminium Alloy EN AW 6082 T6

2.1. ANSYS Creep Model—Theoretical Background

ANSYS software considers creep as a rate-dependent material nonlinearity where the material continues to deform under constant load over time. The software is capable of modelling creep following two different approaches, strain hardening and time hardening. Strain hardening is based on the stress relaxation law presented in the ANSYS Structural Analysis Guide [25] where the reaction force diminishes over time due to occurring creep-related displacements; this is more applicable to stationary tests since the creep strain rate is calculated as a function of the previous accumulated creep and stress. Time hardening represents creep strain due to constant applied stress and is more suited to high-temperature creep representation in transient conditions (Figure 2).

The software can model the first two stages of creep (primary and secondary), while the tertiary phase is omitted from the calculation since it implies impending rupture, which is not considered relevant in the case of large I-sections. ANSYS determines creep strain by means of numerical methods recommended for general use for problems involving large creep strains and large deformations caused by high temperature. It uses a Euler forward algorithm where the modified total strain in every substep (time step) n is computed as (3):

$$\left\{{\epsilon }_n^{\prime }\right\}=\left\{{\epsilon }_n\right\}-\left\{{\epsilon }_n^{pl}\right\}-\left\{{\epsilon }_n^{th}\right\}-\left\{{\epsilon }_{n-1}^{cr}\right\}$$

(3)

where ε stands for total strain vector, ${\epsilon }^{pl}$ plastic strain vector, ${\epsilon }^{th}$ thermal strain vector in the ongoing time step, and ${\epsilon }^{cr}$ the creep strain vector from the previous time step, calculated as a scalar quantity through the equivalent creep strain increment $∆{\epsilon }^{cr}$ in each master node m, which is calculated as (4):

$$∆{\epsilon }_m^{cr}={\epsilon }_{et}\left(1-\frac{1}{e^A}\right)$$

(4)

where the equivalent modified total strain ${\epsilon }_{et}$ is computed as a function of all three normal strains (${\epsilon }_x^{\prime }, {\epsilon }_y^{\prime }, {\epsilon }_z^{\prime }$) and shear (${\gamma }_{xy}^{\prime }, {\gamma }_{yz}^{\prime }, {\gamma }_{zx}^{\prime }$) components (5):

$${\epsilon }_{et}=\frac{1}{\sqrt{2}}{\left[{\left({\epsilon }_x^{\prime }-{\epsilon }_y^{\prime }\right)}^2+{\left({\epsilon }_y^{\prime }-{\epsilon }_z^{\prime }\right)}^2+{\left({\epsilon }_z^{\prime }-{\epsilon }_x^{\prime }\right)}^2+\frac{3}{2}{\left({\gamma }_{xy}^{\prime }\right)}^2+\frac{3}{2}{\left({\gamma }_{yz}^{\prime }\right)}^2+\frac{3}{2}{\left({\gamma }_{zx}^{\prime }\right)}^2\right]}^{\frac{1}{2}}$$

(5)

and e = 2.7182 (base natural logarithm).

A = $\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}$ (the creep strain ratio in the range from 1 to 10 for the current integration point).

Since creep is a path-dependent phenomenon, it is of great importance to adequately capture the response of the model; in ANSYS, this is represented by the creep strain ratio, which is dependent on the equivalent creep strain increment $∆{\epsilon }^{cr}$ and the modified equivalent strain ${\epsilon }_{et}$. In accordance with the law described above, the creep strain increment converts to a full strain tensor with six (three normal strain and three shear) components (6)–(11):

$$∆{\epsilon }_x^{cr}=\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}\frac{\left(2{\epsilon }_x^{\prime }-{\epsilon }_y^{\prime }-{\epsilon }_z^{\prime }\right)}{2\left(1+\nu \right)}$$

(6)

$$∆{\epsilon }_y^{cr}=\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}\frac{\left(2{\epsilon }_y^{\prime }-{\epsilon }_y^{\prime }-{\epsilon }_x^{\prime }\right)}{2\left(1+\nu \right)}$$

(7)

$$∆{\epsilon }_z^{cr}=-∆{\epsilon }_x^{cr}-∆{\epsilon }_y^{cr}$$

(8)

$$∆{\epsilon }_{xy}^{cr}=\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}\frac{3{\gamma }_{xy}^{\prime }}{2\left(1+\nu \right)}$$

(9)

$$∆{\epsilon }_{yz}^{cr}=\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}\frac{3{\gamma }_{yz}^{\prime }}{2\left(1+\nu \right)}$$

(10)

$$∆{\epsilon }_{xz}^{cr}=\frac{∆{\epsilon }^{cr}}{{\epsilon }_{et}}\frac{3{\gamma }_{xz}^{\prime }}{2\left(1+\nu \right)}$$

(11)

where the total creep strain is calculated separately for each component as the combination of the total creep strain from the previous time step and creep strain from the current increment, with an example for z component as follows (12):

$${({\epsilon }_z^{cr})}_n={({\epsilon }_z^{cr})}_{n-1}+∆{\epsilon }_z^{cr}$$

(12)

The time-step size in the integration procedure shown is presented for stress and strain in referred time interval t_n−1, while the creep strain rate is calculated in time t_n for temperature in the same t_n step. A detailed numerical procedure can be found within the ANSYS material [26] and theory [27] reference.

2.2. Available Predefined ANSYS Creep Models

ANSYS includes a total of 13 different creep models with temperature-dependent constants which can be entered directly by means of the creep model itself or by the material model. Most of the available creep models use a derivative of creep strain, which makes it difficult to obtain quality results due to the large number of noisy experimental data. Based on numerical simulations, the Modified Time Hardening model was chosen as it was the most applicable in terms of time consumption and the complexity of the calculations. Also, it is more applicable for transient tests since the creep strain rate is calculated as a function of stress and time when the stress level remains approximately constant during fire exposure. The creep equation for the corresponding model can be defined as (13):

$${\epsilon }_{cr}=C_1{\sigma }^{C_2}{t}^{C_3+1}{e}^{-C_4/T}/\left(C_3+1\right)$$

(13)

where ${\epsilon }_{cr}$ represents the creep strain, $\sigma$ is current stress (MPa), $T$ current temperature (°K), and t is the time (min), while $C_1,\dots ,C_4$ are the dimensionless coefficients for the adopted creep model. Presented coefficients for the Modified Time Hardening model have property values based on the initial value of strain resistances dependent on applied stress, Q is the activation energy (kJ/mol), R is the universal gas constant (kJ/mol°K), and an exponential factor based on time. The temperature component $C_4$ is included through the material model and the coefficient $C_2$ is fixed to a value of one in order to simplify the conducted robust calculations since it presents the stress exponent.

2.3. Transient Creep Model Calibration Using Material Data

In order to obtain the required coefficients of the presented MTH creep model, the test results for aluminium coupons of EN AW 6082 T6 alloy put forward by Torić et al. [5] were used as representative. The tests were performed by following the ASTM standard for high-temperature tests [28] and the ASTM standard for normal temperature tests [29]. The ANSYS mechanical built-in nonlinear model for rate-dependent creep-curve fitting was used for every temperature level in various stress ratios to acquire needed coefficients for the creep model. The results of the curve-fitting method for the referent four-hour interval are shown in

Table 1. Modified Time Hardening model coefficients.

Temperature (°C)	Applied Stress (MPa)	f0.2,θ (%)	MTH Creep Model Coefficients
			C1	C2	C3	C4
150	170.30	0.7	6.06 × 10−13	1.00	−0.2490	0
200	38.10	0.2	5.62 × 10−11	1.00	−0.8910	0
	57.19	0.3	8.20 × 10−11	1.00	−0.9250	0
	95.20	0.5	9.60 × 10−11	1.00	−0.8800	0
250	16.30	0.15	1.60 × 10−14	1.00	0.1380	0
	32.25	0.3	9.18 × 10−11	1.00	−0.7900	0
	53.76	0.5	5.00 × 10−11	1.00	−0.6160	0
300	8.73	0.15	3.75 × 10−13	1.00	0.0361	0
	17.46	0.3	2.95 × 10−10	1.00	−0.4670	0
	29.10	0.5	8.26 × 10−10	1.00	−0.3210	0

and Figure 3.

3. Test Study

3.1. Test Setup

Experimental analysis for this study was carried out at the specialised laboratory unit—Structures lab—at the Faculty of Civil Engineering, Architecture and Geodesy, University of Split. The setup consisted of the test frame, two separate hydraulic systems, an induction heating machine, an insulated steel tube serving as a furnace, a detached cooling system, and the measuring equipment (Figure 4).

The test frame was made of multiple welded and bolted UPN 280 steel sections to ensure sufficient support rigidity. The test setup was designed to test column specimens with an approximate length of 2590 mm in order to match the standard floor height for typical buildings. Boundary conditions for the columns were provided by high-quality steel pins lubricated with mechanical grease and used with thin lathed steel plates. This connection setup with a total diameter of 60 mm was used as a link between two welded steel bearings to ensure uninterrupted rotation and to minimise bolt friction during the experiments. The total axial distance between the two pins was approximately 2910 mm. The right bearing on the frame was fixed, while the left one was movable and used to apply the horizontal load on the column with a hydraulic ram with a capacity of approximately 1000 kN. A smaller hydraulic jack with a capacity of up to 60 kN with the load cell was used to counteract geometrical imperfections around the weaker axis of the column. Pressure gauges were used for constant monitoring of the pressure in the system and the load cell was used as validation.

An efficient heating regime was achieved with a high-frequency induction machine. Since aluminium is not ferromagnetic, heating was achieved by means of a radiative heat flux emanating from the surrounding steel tube of 406 mm serving as a furnace. A thick layer of insulation of 50 mm ceramic wool was placed on the tube and secured by wrapping with high-quality belts resistant to high temperatures which provided thermal resistance up to 800 °C. The same ceramic wool was used at both ends of the tube to prevent possible cooling and temperature leakage during the experiments in order to achieve the desired heating rate which was controlled by an induction machine with a power of up to 35 kW. This heating arrangement was chosen as being a more favourable test setup for temperature induction compared with electric heaters since the circular shape of the furnace provides more uniform heating of the cross-section of the column. The arrangement is also safer because the machine does not heat the specimen directly, which reduces the possibility of damage to the equipment, and it is aimed directly at the insulated steel furnace. This heating methodology is also energy efficient, and the heating rate can be corrected easily by increasing or decreasing the machine power.

The steel tube had an opening for a smaller hydraulic jack, used to apply the transversal force during the tests, since the oil in the hydraulic system must be maintained at a low temperature. It was necessary to include a steel extension for a transverse hydraulic jack with a cooling system because of the high metal temperature conductivity and the heating of the steel tube, which reaches 600 °C in these types of tests.

Measuring equipment consisted of displacement meters and high-accuracy (up to 0.001 mm) linear variable differential transformers (LVDT) which were connected to the data-acquisition card from National Instruments, and where the results were recorded by MatLab software [30] and later stored on a PC for further processing. The sensor for horizontal displacement was mounted on top of the horizontal hydraulic cylinder on the movable side of the frame (left), while the vertical LVDT was attached with 3D-printed accessories on a small hydraulic jack, used to inflict the transversal load (on the weaker axis).

3.2. Column Specification

The aluminium columns tested within this study were made of EN AW 6082 T6 alloy, which is normally used in construction due to its favourable mechanical properties. Alloys in the 6xxx series consist of a high percentage of silicon and magnesium, which increases weldability and plastic deformation, and they have good anti-corrosion properties and high strength. The proof strength for the presented aluminium alloy is generally higher than 260 MPa [4], which corresponds to the yield strength of standard S275 structural steel. Since the test frame was designed for a specimen length approximately equal to the average storey height, the tested I section aluminium columns were 2590 mm in length, with a total profile height of 220 mm (Figure 5).

A flange width of 170 mm and a thickness of 14 mm with a web thickness of 8 mm for specified length defined a column slenderness ratio of about 70.

The column temperature development was measured by 13 independent thermocouples in 5 different cross-sections on the flanges and on the web in order to obtain an accurate display of the generated temperature field over the column (Figure 5). The number of thermocouples was governed by the space limitation of the National Instruments data-acquisition card used to record the temperatures on the column in 0.5-s intervals. Temperature increase was defined by the power output of the induction machine. The optimal heating rate of the machine was within the range of 1.5–4.0 °C/min and the power output used for the experiments was approximately 17 kW.

3.3. Test Results

Ten aluminium columns of alloy EN AW 6082 T6 were tested by exposure to various load and heating arrangements. Stress levels based on column failure and obtained from the stationary tests [31] were compared to preliminary transient tests [32] and used to determine the input forces and temperature ratios applied within this study. Various horizontal forces in the range of 200–400 kN were applied to the column by two different transversal loads, equal to 10–15 kN, at the midspan of the column. A horizontal load with an increment of 50 kN was defined to induce stress levels below the 0.2 f_0.2,θ stress at ambient temperature; the tests conducted on columns could, therefore, fit the 0.2–0.5 f_0.2,θ interval conducted on coupons at elevated temperatures and subsequently used to obtain the MTH-creep coefficients presented in Section 2.2. The f_0.2,θ interval represents the proof strength which, for aluminium alloys, is the value of stress at 0.2% strain, obtained from the capacity tests for a given temperature θ. Detailed experimental data for each test are presented in

Table 2. Columns test parameters.

No.	Average Load (kN)		Average Load (bar)		Average Heating Rate (°C/min)	Specimen Name
	Horizontal	Transversal	Horizontal	Transversal
1	200	10	33	33	2.6	T1—H200 V10
2	250	10	41	33	3.2	T2—H250 V10
3	300	10	49	33	2.5	T3—H300 V10
4	350	10	56	33	2.9	T4—H350 V10
5	400	10	64	33	2.9	T5—H400 V10
6	200	15	33	50	2.9	T6—H200 V15
7	250	15	41	50	3.0	T7—H250 V15
8	300	15	49	50	2.8	T1—H300 V15
9	300	10	49	33	2.8	T1—H300 V10
10	350	15	56	50	2.8	T1—H350 V15

.

The application of the horizontal force was controlled by a built-in potentiometer, and it was applied on the column with approximately 120 kN/min ratio up to 350 kN; the interval from 350 to 400 kN had a rate of 20 kN/min due to the limitations of the expansive hydraulic vessels and hydraulic system (Figure 6a). The transverse force was inflicted at a rate of about 26 kN/min (Figure 6b). The accuracy of the applied forces during the tests was ±5 kN for the horizontal load and about ±0.5 kN for the transverse load. The average heating rate for the presented tests was up to 3.2 °C/min at midspan, as presented in Table 2 and in Figure 6e, and it was achieved at approximately 60 min from the start of the test. Although the temperature peak was at the centre of the column, it decreased towards the bearings, as shown in Figure 6c,d, but not excessively due to the high thermal conductivity of aluminium. This initial one-hour interval represents the stage needed for the temperature increase in the material crystal structure when significant strain due to thermal expansion and creep do not occur. The boundary conditions of the furnace with constrained vertical and horizontal displacement and the limitations of the hydraulic system were sufficient to conduct valid transient creep tests with the presented loads.

4. Creep Model Verification

4.1. ANSYS Column Model

The numerical model developed in ANSYS consists of seven solid bodies (aluminium column and six steel accessories (pins, bearings) visible in Figure 7a) to simulate the column behaviour and the behaviour of two surface bodies under the application of external mechanical load. The column was divided into five different sections according to the mentioned positioning of thermocouples in order to accurately simulate the temperature field (Figure 7b). SOLID187 was the element type used for calculations, with a coarse 25 mm mesh for the column and 50 mm mesh for the steel joints and bearings (Figure 7c). Predefined joints with a 0.2 coefficient for friction between the steel sections were used as contact elements.

4.2. Comparison between the Model and Test Data

Since aluminium alloys are susceptible to large plastic strains, it is necessary to define the failure criteria for the column. When plastic strain increases, the load-bearing capacity of a column diminishes but not necessarily to the point of rupture, as is the case in coupon tests. In a column, failure is evident from the appearance of the vertical asymptote which manifests in vertical displacements, and it utilises the plastic capacity of the column. According to this, the failure criteria for the columns in experimental tests were determined based on the rate of deflection [33] when, during fire exposure, the column exceeds the limit of (14)

$$\frac{L^2}{9000d} \mathrm{mm}/\min$$

(14)

where L is the total length of the column (mm), and d is the effective depth (mm) of the loaded, weaker axis of the column. The exact point of failure based on the aforementioned condition, which is 4.4 mm/min (14) for the given geometry, was obtained from the derivation of vertical displacements, and is marked on the experimental charts presented in Figure 8. Vertical displacements for the numerical model represent the exceeded capacity due to failure or due to the time limit of the experiments.

5. Discussion

5.1. Creep Effect and Its Development Time Span

The need to elaborate on the topic presented in this study arose from the assumption that creep strain in aluminium exposed to high temperature can have a crucial impact on the load-bearing capacity of the columns, and thus the entire structure. Various external loads with an approximate constant heating rate defined by machine power are presented in this study with the aim of defining when and in what way the failure of the aluminium columns occurs due to creep.

The hypothesis that significant creep development for aluminium alloy EN AW 6082 T6 occurs at a temperature above 150 °C proved to be correct since no noticeable vertical deflections due to creep were observed for the first 60 min of the test—the timespan required to achieve a constant heating rate. The creep in aluminium is dependent on the external mechanical and temperature load defined by the applied heating rate. It is possible that lower heating rates (below 1 °C/min) could more accurately capture the creep strain, although that approach is questionable due to difficulties in execution and the probability of the occurrence of that scenario in a real fire.

5.2. The Influence of the External Load on the Column Failure

The external load for the presented tests was defined based on the stress levels of previously tested coupons in order to better cover the dependence on monotonically increasing stress levels caused by the transient heating conducted on the aluminium columns. From the graphs presented in Figure 8. It can be observed that the increase in horizontal load for the same amount of transversal load leads to the accelerated failure of the specimen.

It is evident from the test results presented in

Table 3. Comparison of vertical displacements and time of failure based on transversal load.

Specimen Name	Displacements (mm)	Displacement Ratio (%)	Time (min)	Time Ratio (%)
T1—H200 V10	59.2	139.4	126	99.2
T6—H200 V15	42.4		125
T2—H250 V10	41.8	104.3	120	96.7
T7—H250 V15	40.1		116
T9—H300 V10	39.3	97.4	122	89.3
T8—H300 V15	40.4		109
T4—H350 V10	38.1	93.7	121	86.0
T10—H350 V15	40.7		104

and Figure 8 that the difference in vertical displacements upon failure is reduced with the increase in load. At lower stress ratios, it takes more time to satisfy the 4.4 mm/min failure criterion, due to the slower establishment of creep. An increase in creep strain, which leads to faster column failure, manifests for higher loads and is perceptible in a descending time ratio difference upon failure.

5.3. Comparison between the Creep-Free and MTH-Creep Models

Figure 8 shows the difference in midspan displacements of the column between numerical models without taking creep into account from the modified time hardening models. In a creep-free simulation, the failure occurs only due to material properties which cannot be applied to metals, such as aluminium, because of its nonlinear, inelastic, and rate-dependent behaviour. According to the presented graphs, the Eurocode 9 factor of 1.2 for the creep effect on fire-exposed aluminium columns needs further consideration since the creep-free numerical model deviates somewhat from the experimental data obtained from transient tests, and because it is based on capacity tests which do not fully represent a realistic fire scenario.

It can be seen from

Table 4. Vertical displacements—comparison of creep and creep-free model.

Specimen Name	Creep Displacements (mm)	Creep-Free Displacements (mm)	Displacement Ratio (%)	Temperature at Midspan (°C)	Failure Time (min)
T1—H200 V10	58.4	11.6 *	504.9	293	126
T6—H200 V15	42.6	21.1 *	202.4	287	126
T2—H250 V10	41.2	13.3 *	308.9	286	119
T7—H250 V15	42.2	16.5 *	255.7	280	116
T9—H300 V10	38.1	12.4 *	308.2	268	119
T8—H300 V15	37.4	18.0 *	208.2	245	110
T4—H350 V10	38.1	23.2	164.3	268	117
T10—H350 V15	36.7	20.0 *	183.5	232	105
T5—H400 V10	34.7	20.1	171.7	263	117

*—failure did not occur for the creep-free model

and Figure 8 that the test time span was not sufficient for the column to fail when using the creep-free model for nearly all tests. It is noticeable that the difference in vertical displacements ratio decreases with the increase in load. Although in Tests 4 and 5 the column fails in both models, the difference in vertical displacements upon failure is more than 160%.

5.4. Comparison of the MTH Creep Model with Test Results

The comparison of the numerical model and experimental data is shown in Figure 8 with a noticeable resemblance to the average failure time. Creep indicates evident plastic strains on aluminium columns at midspan with the vertical deflection of 34 mm visible in Figure 9a. This is substantiated by the numerical results obtained from the ANSYS MTH creep model displayed in Figure 9b for the column in Test 5 since it sustained the largest deflection at midspan. The average vertical deflection which manifests through the plastic strain for all columns was 17 mm.

The presented creep model matches very well with the test results in terms of vertical displacements since the failure times are within ±10% perimeter which validates the MTH creep model for further consideration in terms of high-temperature creep on aluminium alloys (Figure 10).

6. Conclusions

The influence of the transient heating regime on the failure of EN AW 6082 T6 aluminium alloy columns due to creep has been presented in this paper. The heating rate at midspan of up to 3.2 °C/min resulted in a temperature range upon failure from 230 °C to 300 °C, based on the applied load. Some conclusions and guidelines for further research can be obtained from this study:

• The applied creep model can be used to predict the failure of I section aluminium columns made of alloy 6082 T6 exposed to high temperature with sufficient accuracy;
• The noticeable strain due to creep starts to occur in temperatures over 150 °C in a transient heating regime;
• The significant difference in failure between creep and creep-free model confirms the assumption of a substantial impact of creep; this must be taken into account in the design of aluminium columns.

The presented experimental and numerical analysis favours a further and more detailed elaboration of this topic since the creep strain impact in transient heating is significant in the overall capacity of the aluminium columns; this is especially so for alloy EN AW 6082 T6, which is the most suited to application in structural engineering practice. A more detailed analysis on similar 6xxx aluminium alloys for variable heating rates should be considered for further research, based on the acquired data and observations from this study.