Experimental study of a very high performance concrete slab subjected to fire on its underside and numerical modeling of the temperature field

Abstract

The present study deals with the numerical thermal odelling of two slab-specimens made of high performance concrete that underwent thermal spallin g tests. The specimens were equipped with thermocouples located at selected points and subjec ted on theirs undersides to a fire of the order of Eurocode 2 ISO 834 standard fire. The thermocouples responses were recorded and reported and some of them revealed the effects of presence of water i n zones near the specimens heating faces. Temperatures of the heated faces of the specimens w re measured and theirs evolutions with respect to time were used as inputs of the numerical modelling . The numerical modelling of the temperature field considering the selected points and using the stand ard concrete thermal characteristics was not satisfactory because the discrepancies between expe riment and modeling reached 41 to 63 °C at the highest temperatures. The study showed that by redu cing the thermal conductivity of the concrete, it i s possible to predict satisfactorily the specimen’s t hermal behavior during the spalling tests. Hence, t h numerical modelling of spalling tests with the use of thermo-mechanical approach becomes then possible. Key-words: Spalling tests, High performance concrete, spalling test bench, transient nonlinear thermal, thermal numerical modelling, thermo-mechanics approach, tem perature evolution prediction 1.Introduction The study of the thermal spalling of concretes is a complex subject that has been of interest to many researchers for a long time. It has an experimental dimension of observation of the phenomenon in order to understand the leading mechanisms and to e xplain it. Most of the studies conducted fall under this approach [1-3]. The various studies have shown that the causes are multiple but that in summary two main causes can explain these thermal instabili ties: the presence of a thermal gradient and the presence of water vapor in concrete migrating under th action of heat [4]. Studies have shown the existence of vapor pressure and thermal gradients [ 5]. Others focused on the influence of concrete constituents, on the differential expansion between c ment paste and aggregates, on the role of water in the process of thermal instability. Although the understanding of the phenomenon is quite advanced, further studies are still needed to reinforce or re fute certain assumptions currently used. It has a s econd dimension of modeling to help the partial understan ding of observed phenomena and whose ultimate goal would be to predict these thermal instabilitie s. There are far fewer studies in this area [6-7]. The phenomenon is complex because it induces phenomena of heat transfer and water transfer in a complex environment where chemical reactions contin ue to occur. If the aim is to treat the problem as a whole, we should adopt a thermo-hydro-chemo-mecha ni al approach and take into account the multi-scale dimension of the problem. The more comp lex the approach of the problem, the better the quality of the analysis, the longer the calculation s will be and the more likely are the risks of nonconvergence of the calculations. In this area, stud ies based on THM approaches have yielded very interesting results [Schreffler et al.]. In the present study, we have restricted the field of investigation to a thermomechanical approach which favors the effect of the thermal gradient as he main cause of the thermal spalling of concrete. One of the difficulties of this approach, as of any other approach, lies in the correct calculation of the temperature field in the structure, a prerequisite for the thermomechanical calculation of the stresse s and the damage to be realistic. To try to answer th is problem, a spalling test bench was designed and realized. The first tests were carried out on speci m ns equipped with thermocouples which measure the temperature at certain points during the test. The proposed study investigates the extent to which the thermal behavior of the specimens can be predic ted by numerical modeling of the test using the finite element method. 2.Materials and properties The studied material is a high performance concrete which formulation is based on normalized sand. It is constituted of 0-2 mm normalized sand of the Soc iéte Nouvelle du Littoral, cement CEM 1 52.5 produced by Vicat, silica fume Condensil SD95 produ ced by Condensil Company, superplasticizer Sika Viscocrete Tempo 10 produced by Sika and tap w ter. The W/C ratio is 0.25. The various constituents of the material were mixe d according to a specified protocol and cast. The molds are constituted of the assembly of Plexig las and steel walls and measure 250x250x100 (mm). The fresh concrete was put in the molds and v ibrated during 1 min 30s in two steps. In order to measure the temperature at selected points of the s pecimen, some thermocouples, placed along the specimen thickness, were embedded in them during th e casting. Generally three thermocouples were used and located at various distances to the face o f the specimen to be heated. In the first spalling test studied (Test_1) the three thermocouples were place d : first on the specimen heating face, second at 1 0 mm and third at 20 mm. Whereas for Test_2, the loca ti ns were : first on the specimen heating face, second at 5 mm and third at 10 mm. The thermocouple s locations are displayed in Fig.1. After the vibration process, the molds are covered with a pla stic film during 24 hours and then demolded. Hence, the cast specimens are squared slabs with 25 0 mm in length and 100 mm in thickness equipped with K type thermocouples. Fig.1. Sketches of the location of the measurement poi ts for a-specimen 1 and 2specimen 2. In order to widely study spalling phenomenon, an or iginal test bench was designed and fabricated. It consists of a 32 KW power burner associated with a cer mic structure forming then a sort of furnace. The burner was designed so that it heats uniformly the specimen face. The ceramic structure is constituted of vertical walls which one of them is equipped of a porthole where a camera can be placed to film the test, and an upper horizontal slab pier ced with a central hole and peripheral holes. The peripheral holes are aimed at evacuating the burnin g gas. The central squared hole measures 245x245 in mm and faces the burner; the specimen is placed t the top of the furnace and is then heated throug h this hole. The burner is powered with propane gas a nd air and controlled by a control unit. Hence the thermal cycle to apply to the specimen can be contr olled automatically with a program or manually. At the starting of the burner, a special regime is app lied during the first minute where the power is hig h enough in order to check the good working of the ap paratus and then it decreases down to a minimum value at which the spalling test can begin. The con trol unit is connected to two thermocouples placed in the furnace aimed at insuring the thermal regula tion: regulation and security thermocouples. Another independent type K thermocouple is also pla ced in the furnace beside the regulation thermocouple in order to measure the temperature in the furnace. During the spalling test, the three embedded thermo couples and the independent thermocouple placed in the furnace are connected to an independent data acquisition system that acquires the data all alon g the spalling tests. The test bench is under develop ment and other sensors are planned to be used later , especially acoustic and pressure sensors. The globa l view of the apparatus is displayed in Fig.2. a b Fig.2. Photo of aa global view of the spalling te st bench and ba detailed view of the specimen equipped with the three thermocouples. 3.Experimental study The test consisted in placing the instrumented test specimen on the furnace once the burner reached it minimum power after the starting regime and in begi nning the data acquisition. During the tests presented here, the burner power was increased manu ally in order to estimate the heating rate close to the ISO 834 standard curve. The evolutions of the v arious acquired temperatures with respect to time were visible in real time thanks to data acquisitio n software. Occurring events such as noise or explosion are collected during the test. Some photo s f water coming out of the lateral or upper faces of the specimen were taken and dated. The thermocou ples responses gave interesting information on what happened during the test. Two tests that diffe r from the embedded thermocouples locations and the heating rate are presented in this study. In th e two cases, thermal spalling occurred during the t ests. In Fig.3 are displayed the response of the thermoco uple located in the furnace (TH_Furn), the responses of those embedded in the specimen at vari ous d stances to the heated face (Th_S0, TH_S10, TH_20) and the fire standard ISO 834 curve of Euro code 2 (TH_ISO834). Globally, the responses of all the thermocouples increased with increasing tim e. The study concerned the ten first min of the spalling test (600 s). The evolution with respect t o ime of TH_Furn is higher at any time than the curve Th_ISO834 indicating then that the applied po wer cycle was more severe than the recommended standards. It also presents sudden increase (at tim e 75 and 310 s) corresponding to the sudden increas e of the power that was controlled manually. The evol uti n of the temperature of the thermocouple which tip was located in the burned face is quite d fferent in intensity from the temperature in the furnace but present the same discontinuities. The d iff rence of intensity is due to the influence of t he thermal inertia of the concrete surrounding the the rmocouple. The sudden decrease of the temperature observed at time 487s was the consequence of the oc currence of thermal spalling that was noisy. The TH_Furn temperature decreased from 857 to 737 °C wh ereas the one of TH_S0 decreased from 424 to 358 °C