Proceedings of the SEM Annual Conference June 1-4, 2009 Albuquerque New Mexico USA 2009 Society for

Experimental Mechanics Inc.

Experimental and numerical analysis of full-locked coil ropes fire behaviour

V.Fontanaria, B.D. Monellia, F. Degasperi b

a

Universit di Trento, Dipartimento di Ingegneria dei materiali e delle tecnologie industriali, via Mesiano 77 38050 Trento, e-mail: vigilio.fontanari@unitn.it b Laboratorio tecnologico impianti a fune (LaTIF) Via Provina 24 38040 Trento

ABSTRACT In the present paper the fire behaviour of full-locked coil ropes under axial loading conditions is addressed. Since there are no standards for these structural elements, ISO 834 standard curve was considered as fire model for its severity. The analysis of full-locked coil ropes fire behaviour was carried out both from the experimental and numerical point of view, thus obtaining a complete description of rope responses. Two distinct full-locked coil ropes were tested and heat fluxes, temperature fields in the rope cross-section and fire endurance times were determined. An experimental campaign, aimed to establish how wires mechanical properties change with temperature, was also carried out. Subsequently, a parametric Finite Element model (FE) able to predict the rope response, undergoing to the experimental thermal and mechanical loads, was implemented, thus obtaining the time evolution of damage and the failure mechanisms for the rope. The numerical model was then validated by comparing the predicted fire endurance times with those obtained from the experimental tests and a very good agreement was found. A method for determining the fire endurance times of full-locked coil ropes by using the numerical model, once the material properties and loading conditions are established, was thus defined. INTRODUCTION The study of the mechanical behaviour of steel ropes during a fire transient has received an increasing attention after two serious accidents occurred in the Alps in the 2000/01 winter season (Kaprun funicular in Austria, 11/2000 and Zugspitze ropeway in Germany 02/2001). Oplatka [1] carried out an investigation on the recent most serious fire accidents involving ropeways. He found 35 cases, 10 of them lead to the ropeways collapse; in two cases the time to collapse was even shorter than 15 min. The need to deepen the knowledge on ropes fire behaviour is emphasized also by the ever increasing diffusion of ropes in civil buildings, bridges, cable car. The exposure to high temperatures determines an abrupt worsening of the wires mechanical behaviour, thus influencing the rope response up to the final collapse. The relatively high carbon content in the steel and the manufacturing processes of wires lead to fine pearlitic microstructures with an high level of work hardening, which are responsible both for the high strength and the relatively low ductility of the material [2-4]. The wire exposition to temperatures over 300C activates the dislocation movements responsible for the internal annealing, thus producing an irreversible loosening of the work-hardening effect. A further increase of the temperature up to 600 C determines the nucleation and growth of a new crystalline structure (recristallization) [5-8]. Longer expositions to high temperatures can finally produce viscous flow (creep). Exposure times to the high temperatures during fire is relatively short, so that the creep contributions on the ropes damage can be neglected. In the scientific and technical literature there is a lack of information on the rope behaviour during fire: the high costs of the experiments and the difficulties to set up the experimental devices still represent a strong obstacle towards the acquisition of a data base of experimental results on different classes of ropes. It is therefore necessary to refer to papers and standards specifically conceived for civil buildings and structures. Different time-temperature histories are suggested by the standards to simulate the fire resistance of a structure: one of the most used is defined in the ISO 834 standard, which aims to simulate the time-temperature evolution in closed environments surrounding the structure during a fire accident [9-11]. The structure has to be loaded by the in service typical conditions up to the final collapse or to the loosening of its functionality (i.e. excessive

deformation incompatible with structural integrity). The standard temperature curves represent an overestimate of the fire heaviness: only the flash over phase is considered, whereas the initiation and the expansion phases of the combustion, during which the temperature increase more slowly, are neglected. In international standards the fire resistance of a structural part (the component or the entire structure) is defined as the attitude to maintain for a specified time interval, when subjected to an appropriate thermal history, the load bearing function R, the integrity E and the thermal insulation I. With reference to these classification parameters, three classes of fire resistance can be identified: REI label: the component must maintain, for a specific time interval, the three aforementioned characteristics; RE label: the component must preserve load bearing function and integrity; R label: the element must maintain its load bearing attitude; The time interval guaranteed before collapse has to be expressed in min. so that REI (RE, R), 15, 30, 45, 60, 90, 120, 180 classes can be usually found. It is very difficult to transfer these concepts in the study of ropes behaviour. The ropes response and the thermal history of each single wire are hardly valuable during the fire transient. In scientific literature a great effort has been spent to understand the mechanical behaviour of wire ropes under different loading condition, whereas no theoretical, numerical or experimental contribution can be found dealing with the mechanical behaviour of ropes during heavy thermal transients. In particular, several numerical and analytical models have been proposed with the aim of predicting loads sharing among wires and wires interaction [2,12]. As far as strands ropes are concerned, the mechanical behaviour is strictly related to the mutual contribution of wires to carry the applied load. In the case of locked ropes the mechanical behaviour is dictated by the contribution of the entire wire layer more than by the behaviour of each single wire. As a consequence the ortho-tropic model suggested in [13] allows for a better reproduction of experimental results. It has to be noted however that even though analytical models can predict the load distribution among wires, they are unfortunately unable to correctly describe the rope behaviour when subjected to heavy in service loading because of the complex phenomena such as contact, friction, plasticity and large displacements (full-slip regime vs. no-slip regime), which noticeably influence the mechanical response. In order to face the analysis of these phenomena some numerical approaches based on the Finite Element (FE) Method have been successfully adopted [13-16]. In the present paper a numerical-experimental approach is developed for studying the mechanical behaviour of full locked ropes undergoing to the high thermal loads produced during the fire accidents simulation. The principal aim is to define a design tool for predicting the ropes behaviour under these very heavy conditions. To this purpose a parametric FE model was set up, which allowed for the simulation of the ropes thermomechanical response during the fire transient following the ISO 834 standard. The model was set up and calibrated on the basis of experimental data measured on the rope and on single wires: i.e. tensile tests at room temperature, thermal histories of different wire layers during the ISO 834 thermal transient. Moreover, in order to correctly simulate the mechanisms of load redistribution among wires during the thermal transient, an extensive campaign was carried out on single wires to build up a data base of wires curves at different temperatures. The FE modelling of the thermo-mechanical response of the rope allowed to predict the load redistribution, the collapse mechanisms and the time to collapse at high temperature. On the basis of these results an experimental procedure was set up and two destructive tests have been carried out on full locked ropes. 1. THE FULL LOCKED ROPE The present study is focused on a full locked rope having external diameter of 60 mm (fig.1). The rope is characterised by a central straight wire, three layers of circular wires with alternate right-left lay and by three external layers of Z shaped wires also with right-left lay. This ropes configuration is generally used for cable tensostructures. The wires present higher cross sectional areas as compared to wires of cableways ropes, characterised by lower flexural stiffness. The effective cross section of the studied rope is nearly 2485 mm2, with a filling coefficient (ratio between effective and nominal cross section area) of nearly 88%. The chemical composition of steel wires is reported in table 1. Elem. % C 0.825 Mn 0.725 S 0.003 P 0.013 Si 0.215 Sn 0.02 Cu 0.02 Al 0.029 Mo 0.003 Ni 0.016 Cr 0.12 V 0.066

Table 1 Chemical composition of the steel wires The tensile properties of the circular and the z shaped wires at room temperature were measured by carrying out tensile tests on wires segments extracted from the rope. Tests performed on a 100 kN servohydraulic machine. An extensometer with a gage length of 12.5 mm was adopted. The representative true stress true strain curves are plotted in fig. 2.

N wires Core straight wire First layer Second layer Third layer Fourth layer Fifth layer 1 6 12 18 23 29

/ H [mm] 4.26 4.09 4.09 4.09 H 5.20 H 5.20

Cross section [mm2] 14.24 13.14 13.14 13.14 22.5 23.2

H 5.20 Sixth layer 35 23.2 (=circular wires diameter, H=thickness of Z shaped wires) Figure 1: geometry of the locked rope 60 [1 + 6 + 12 + 18 + ( 23 + 29 + 35 ) Z;

2000 1800 1600 true [Mpa] 1400 1200 1000 800 600 400 200 0 0 0.01 0.02 true [/]
Figure 2 : examples true stress and strain curves obtained for different wires
2000 1750 1500
circular wires shaped wires

0.03

0.04

[MPa]

1250 1000 750 500 250 0 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Stabilization cycles

[/]

Figure 3: engineering stress-strain curve of the rope at room temperature

The tensile test of the rope was performed on a servohydraulic 10MN testing machine. An extensometer with a gage length of 1 m was applied on the rope. Several loading cycles in a load range between 5% and 50% of the estimated fracture load were applied in order to reach a stabilization of wires contacts. After this preliminary phase the rope presented an initial linearly elastic behaviour which allowed for determining an apparent Young modulus of nearly 175 GPa. The extensometer was removed at the onset of non-linear behaviour, the load was then increased up to the rope failure and the engineering strain was estimated on the basis of the gripping heads position control. In figure 3 the stabilizing cycles and the obtained curve, as evaluated using the effective cross section, are plotted. 2. THE FINITE ELEMENT MODEL The FE model was developed with Ansys rev. 10 Code. A parametric solid model was built in order enable the analysis extension to any type of locked ropes. The model is built up by a central straight wires, three inner spiral strands of circular wires with right-left alternate lay, three external layers of shaped wires even with right-left alternate lay. The axial length of the model was chosen as a compromise between the computational heaviness and the need to avoid any boundary effect. A picture of two details (core strand and external layers) of the solid model is presented in figure 4. The mesh was manufactured by 8-nodes structural elements, whereas the contact was simulated by surface to surface gap elements. The elements size was determined by a convergence analysis. In order to limit the computational effort, an iterative procedure was set up to identify the contact surfaces and to minimize the number of contact pairs. A picture of the final mesh and an example of the surface of possible contact between two wires, as identified by the iterative procedure is shown in figure 5.

a)

b) Figure 4: the solid model (a), detail of the core strand with alternate (b)

a)

b)

Figure 5: a) FE mesh developed for the analysis, b) identification of the contact surfaces Axial loading was implemented in the FE model by imposing an axial displacements on a cross sectional end, whereas the circumferential displacements was constrained in order to prevent the rotation of the strands along

their longitudinal axis. As observed in service, an axial twisting moment arises. The value of this moment is however moderate due to the alternate lay up of the wire layers. A comparison between numerical and experimental results was carried out to state the correctness of the model in the simulation of the tensile response of the rope. The numerical and experimental results can be rigorously compared only in the first part of the stress-strain curve, up to the extensometer removal. Experimental and numerical curves are plotted in fig. 6. The FE analysis reasonably reproduces the ropes behaviour. An estimated values of 172 GPa can be obtained for the apparent elastic modulus slightly lower (2.0 %) with respect to 175 GPa obtained in the experimental tests. The FE model seems to correctly predict also the onset of non linear behaviour of the rope. 2000
1750 1500 [MPa] 1250 1000 750 500 250 0 0 0.01 0.02 0.03 [/] 0.04 0.05 0.06
Experimental FEM

Figure 6: experimental and numerical results of the tensile on the rope at room temperature The FE results allow for evaluating the load redistribution among wire strands either in the elastic and in the plastic part of the engineering curve. This analysis is very powerful for checking the correct design of the rope and also for defining the damage scheme in the different wire layers. The damage scheme allows for predicting the onset of overloading in the different wire strands and enables a dynamic control of the internal actions redistribution (fig.7). The curves can be grouped into three sets: the three lowest curves refer to inner strands of circular wires, the three intermediate curves refer to the shaped wires layers and the upper curve is that of the core straight wire, which undergoes only to axial loading and shows an initial stiffer behaviour. The curve of shaped wires are nearly overlapped, thus showing a correct design of the external layers of the rope. On the contrary internal strands contribute to sustain a lower fraction of the load especially at service loading, where a difference of nearly 200 MPa between upper and lower curve can be measured. The difference becomes smaller in the elasto-plastic region due to the load redistribution among wires
1800 1500 [MPa] 1200 900 600 300 0 0
Strain at nominal service load 200 MPa

core 1st layer 2nd layer 3rd layer 4th layer 5th layer 6th layer

0.005

0.01 [/]

0.015

0.02

Figure 7: stress strain curves of different wire layers as predicted by the FE model.

3. THE ISO 834 CURVE: THE THERMAL TRANSIENT ON THE ROPE An electric kiln mounted in the testing machine has been specifically designed for reproducing the ISO 834 curve on a ropes segment of 1.5 m length. The short time interval allows for disregarding the effect of creep so that the rope collapse can be principally ascribed to the worsening of wires mechanical behaviour with temperature. It is therefore of paramount importance to correctly know the thermal transient of each wire of the rope, since this can have a strong influence on the load redistribution among wires. A thermocouple in the ropes core, four thermocouples on the ropes skin with angular spacing of 90, and finally four thermocouples in the kiln chamber to measure the environments temperature were adopted. The measured temperature profiles are plotted in fig. 8 a). The feedback and air thermocouples positioned in the centre of the kiln chamber point out that the kiln reach the ISO 834 temperature profile after nearly 3 minutes and then it can follow the curve quite satisfactorily. The thermal transients measured by the four thermocouples on the ropes skin are overlapped thus showing that the temperature field is axi-symmetric and a radial heat transfer can be assumed with reasonable approximation. The temperature ramp measured by the core thermocouple starts with some delay (nearly 3 minutes) with respect to the other curves. The temperature gap between skin and core becomes consistent and this can have very important effects on the wires behaviour during the test. The FEM was calibrated for reproducing the timetemperatures curves measured in different positions of the cross section (figure 8b). FE calibration was carried out iteratively by setting some parameters: the global heat transfer coefficient between air and rope skin, the contact heat resistance between shaped wires layers and finally the contact heat transfer resistance for internal strands. By means of the experimental measurements and introducing the material thermal properties found in the literature [5,6,17,18], an optimal reproduction of the thermal transient of each wires layer can be obtained.
900 800 700 600 T [C]

700 600 500

feedback air skin 1 skin 2 skin 3

400 300 200 100 0 0 400 800 1200

time [s]

T [C]

500

400 300 200 100 0 skin experimental core experimental FE skin FE core 0 400 time (s) 800 1200

skin 4 core ISO 834

1600 2000

a)

b) Figure 8: a) temperature ramps measured in the kiln chamber (air, feedback), on the ropes skin and core, b) results of the FE model calibration.

4. MECHANICAL PROPERTIES OF THE WIRES MATERIAL AT DIFFERENT TEMPERATURES The mechanical behaviour of wires depends on the local temperature. It is therefore necessary to collect a data base of experimental stress-strain curves from tests carried out at different temperatures after different stabilization time intervals. Even though the steel grade adopted for wire ropes is quite common, systematic information on its mechanical properties as a function of temperature are not available. Therefore a series of tests at different temperatures has been carried out on wires segments extracted from the rope. Tests were performed on a 100 kN servo hydraulic testing machine, mounting a heat furnace designed for testing materials at temperatures up to 1000 C. The temperature and stabilization time have been chosen in order to comply with the typical time-temperature histories that can be experienced by the wire during the fire simulation. Preliminary tests showed that shaped wires behave like circular wires when tested at T>300 C. The experimental campaign was then carried out on wires with circular section. The tests were performed at temperature ranging from 100C

to 600 C. An appropriate procedure was set up for introducing the wire in the furnace after chamber temperature stabilization, in order keep very short (t < 5 min) the time necessary for heating the wire to the testing temperature. At the end of the transient, the system was stabilised before to start the test. Two stabilization times were considered: the shorter was typically 3 min., whereas the longer was nearly 8 min. In fig. 9 the curves are plotted as a function of the testing temperature and stabilization time, whereas in table 2 the measured values of Young modules, Yield strength, ultimate tensile strength are summarized.
2400

2000

100 C 20'

200 C 20'

1600 [MPa]
400 C 3'

300 C 3' 300 C 8'

vrue

1200

800
500 C 3'

400 C 8'

400

600 C 3' 600 C 8'

500 C 8'

0 0 0.02 0.04 0.06

true

0.08

0.1

0.12

[/]

Fig. 9: curves at different temperatures and for different stabilizing time intervals T [C] 20 100 200 300 400 500 600 y,stab 8-10[MPa] R,stab 3[MPa] R,stab 8-10[MPa] E [GPa] y,stab 3 [MPa] 186 1560 1560 1885 1885 169 1630 1630 2078 2078 157 1490 1490 1864 1864 140 1280 1320 1503 1517 121 1055 890 1196 1039 100 510 505 575 563 57 285 214 306 238 Table 2: results of the tests carried out at different temperatures of circular wires

Negligible differences were observed at 100 and 200 C between curves obtained at the two stabilization time intervals: at these temperatures the stabilization times are not long enough to influence the material response. At higher temperatures the dependence on time is clearly distinguishable. The maximal difference can be observed for tests carried out at about 400 C. At lower temperatures the kinetics for microstructural modification is slow and longer exposition time should have been applied. On the contrary at higher temperatures the kinetics becomes very fast and the annealing phenomena take place in very short time intervals.

5. THE ROPES TENSILE RESPONSE DURING FIRE SIMULATION 5.1 The thermo-structural FE model. The final objective of the simulation is to predict the load bearing class R xx of the rope during a fire accident. The FE model was adopted and the wires tensile behaviours at different temperatures have been introduced in the mechanical FE model. Since material behaviour depends upon both temperature and exposure time to temperature, two different analyses have been carried out, by considering the results of tests obtained after the two stabilization times in the furnace. The time to rope collapse that can be obtained can represent the lower and the upper limit of the fire resistance of the rope. Structural boundary and loading condition resemble those described for the tensile test at room temperature: the service nominal stress (nearly 30% of the ultimate tensile strength of the rope) was firstly applied at room temperature (T=20C). After preloading, the thermal transient is started, keeping constant the axial loads. Each wire layer is forced to follow the temperature ramp measured during the experimental simulation of the ISO 834 curve. (fig. 10).

Time (s)
Figure 10: time-temperature curves applied to the different wire layers. In figure 11 the ropes nominal strain evolution with time is plotted for the two simulations: the lower limit is obtained by considering the materials properties previously measured on the wires after the longer stabilization time (8min.), whereas the upper limit is obtained by introducing in the FE model the materials properties corresponding to a shorter stabilization interval. The rope behaviour predicted by the two analyses is similar in the first part. A noticeable difference can be detected when the rope approaches the final collapse. Two phase can be clearly detected: a linear strain vs. time evolution, which can be essentially explained by considering the thermal expansion of the material, followed by an ever increasing strain rate up to the final collapse. In order to set up the experimental test it is necessary to know how the final collapse takes place. This analysis can be carried out by considering the mechanisms of load redistribution predicted by the FE analysis. With reference to the stress-strain curves plotted in figure 9, the external wire layers, which are subjected to the highest temperatures (fig. 10) can undergo to an extensive plastic deformation at a nearly constant, or even decreasing stress. As a consequence the load is progressively transferred to the internal wire strands. All wires undergo to the same axial elongation: the single wire will fail when it reaches its rupture elongation, that depends on the temperature and therefore by its position in the rope. As a consequence it is reasonable to assume that failure will start at wires belonging to the internal layers. Load will be instantaneously transferred to the external layer thus determining the final collapse. The rope breakage will not be catastrophic, in fact a consistent part of the energy will be dissipated by plastic deformation of the external wire layers, which behave following the curves determined at nearly 600C (fig. 9). The characteristic whip stroke experienced in the ropes tensile tests at room temperature can be strongly smoothed or even suppressed because external shaped wires can act as a protective shield, thus mitigating the danger of the final breakage. This conclusion allows to design with reasonable safety the experimental test on the rope.

Figure 11: upper and lower limit of the nominal strain vs. time curves during fire simulation 5.2 The experimental behaviour of the rope The experimental tests on the rope are performed on the 10 MN testing machine, using the heating system previously described. Since the kiln is unable to exactly reproduce the ISO 834 curve, it is necessary to preheat it before to introduce the rope and to start the test. Following the standard specifications the rope was preloaded up to the nominal load (30% of the service load) for, at least, 15 min, before to start the test, in order to obtain a mechanical stabilization. To this purpose the positioning mechanism described in fig. 12 was adopted. Two tests on two ropes segments have been performed. The obtained strain vs. time to collapse curves were nearly overlapped with a time delay at the final collapse shorter than 30 s.: the first test was terminated after nearly 18 30, the second presented a little bit longer life (nearly 19 min). In figure 13 the comparison between numerical and experimental results is shown. Machine frame

a)

b)
kiln

c)

rope

Figure 12: a) kiln preheating and rope preloading b) the kiln is lifted up ad opened; c) the kiln is closed and the test started The numerical simulation allowed to predict a time interval between a lower limit of 17 min and an upper limit of 19 min. The experimental curves are indistinguishable from the numerical curves in the first part, then an anticipated transition to a non linear behaviour and a less steep gradient in the final part can be observed.

0.045 0.04 0.035 0.03

lower limit

upper limit

experimental test

strain [/]

0.025 0.02 0.015 0.01

17' 19'

0.005 0 0 200 400 600 800

1000

18 ' 30''

1200

Time [s]

Figure 13: comparison among strain vs. time curves predicted by FEM and obtained by experimental tests CONCLUSIONS This work presents an experimental and numerical procedure for studying the fire resistance of full locked ropes. The FE analysis correctly reproduces the thermomechanical processes in the rope undergoing heavy thermal loading. The analysis, performed on a specific class of ropes, can be easily extended to all locked ropes and adapted after some adjustments for studying also half-locked and wire strand ropes. The FE analysis was able to correctly model the mechanical behaviour of the rope at room temperature and to predict the load redistribution during the test among different wire layers. This information is very powerful for carrying out a design optimization of the rope. The model can be used indeed to predict very important parameters, such as the stabilized elastic modulus, the apparent yield strength and also the collapse conditions. REFERENCES 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) 15) 16) 17) 18) G. Oplatka, Brand von Seilbahnen, ISR International Seilbahn Rundschau, 1, 8-11, 2001 K. Feyrer,: Drahtseile Bemessung, Betrieb, Sicherheit, Ed. Springer Verlag, 2000 J. G. Wistreich, The fundamentals of wire drawing, Metall. Rev., 1958, Vol. 3, pp. 97-142. V. Fontanari, M. Benedetti , U. Bulf , Numerical analysis of the rolling process of shaped wires for locked steel ropes. Journal of materials processing technology, 170, 97-107, 2005 G. E. Dieter, Mechanical metallurgy, Mc Graw-Hill Book Company, 1988. R. E. Reed-Hill, R. Abbaschian, Physical Metallurgy Principles, PWS Publishing Company, 1994. C.R. Chaplin, Failure mechanisms in wire ropes, Engineering failure analysis, 2, 1, 45-57, 1995 A. Ray, S. K. Dhua, K. B. Mishra, S. Jha, Microstructural Manifestations of Fractured Z-Profile Steel Wires on the Outer Layer of a Failed Locked Coil Wire Rope, Practical Failure Analysis, 3, 4, 51-55, 2003 A. Michael Hasofer: Risk Analysis in Building Fire Safety Engineering Butterworth-Heinemann; 1st Ed., 2007 Dougal Drysdale: An Introduction to Fire Dynamics J. Wiley; 2nd edition 1998 D. Rasbash, G. Ramachandran, B. Kandola, J. Watts, M. Law: Evaluation of Fire Safety Wiley 2004 G. A. Costello, Theory of Wire Rope, New York: Ed. Springer Verlag, 1990. M. Raoof, I. Kraincanic, Prediction of coupled axial/torsional stiffness coefficients of locked coil ropes, Computers & structures, 69, 305-319, 1998 W. G. Jiang, J. L. Henshall, J. M. Walton, A concise finite element model for three-layered straight wire rope, Int. J. of Mech. Science, 42, 63-86, 2000 K.H. Wehking, S. Ziegler, Berechnung eines einfachen Seils mit FEM, Draht, 5, 32-36, 2003 M. Giglio, A. Manes, Life prediction of a wire rope subjected to axial and bending loads, Engineering Failure analysis, 12, 549-568, 2005 ASM, ASM Handbook vol. 1, ASM Metals Park OHIO, 1998 P.L. Dowling, J.E. Harding, R. Bjorhovde: Constructional steel design an international guide, Elsevier applied science, 1992