Transition from Multiple Macro-Cracking to Multiple Micro-Cracking in Cementitious Composites

  • Published on
    05-Jul-2016

  • View
    214

  • Download
    1

Embed Size (px)

Transcript

<ul><li><p>TSINGHUA SCIENCE AND TECHNOLOGYISSN 1007-0214 15/25 pp669-673 Volume 13, Number 5, October 2008</p><p>Transition from Multiple Macro-Cracking to Multiple Micro-Cracking in Cementitious Composites*</p><p>ZHANG Jun ( )1,2,** , LENG Bing ( )1</p><p>1. Department of Civil Engineering, Tsinghua University, Beijing 100084, China; 2. Key Laboratory of Structural Engineering and Vibration of Ministry of Education, </p><p>Tsinghua University, Beijing 100084, China </p><p>Abstract: This paper presents an experimental study of the possibility of transition from multiple </p><p>macro-cracking to multiple micro-cracking in cementitious composites. Conventional polyvinyl alcohol fiber </p><p>reinforced cementitious composites normally exhibit macroscopic strain-hardening and multiple cracking af-</p><p>ter the first cracks appear. However, the individual crack width at the saturated stage is normally 60 to </p><p>80 Pm. In the current study, the effect of fine aggregate size on the cracking performance, especially the in-</p><p>dividual crack width in the strain-hardening stage was studied by bending tests. The results show that the </p><p>individual crack widths can be reduced from 60-80 Pm to 10-30 Pm by modifying the particle size of the fine </p><p>aggregates used in the composites. </p><p>Key words: fiber reinforced cementitious composites; crack width; polyvinyl alcohol fibers </p><p>Introduction </p><p>Concrete is a typical brittle material with ten-sile-softening that behaves as a decaying load with immediate localizing of the deformation at the location of the first cracking after the peak load[1]. Therefore, in normal concrete structures, when the stress exceeds the tensile strength of the concrete, a single crack forms and the crack width quickly reaches a macro visible size to dissipate large deformations from both me-chanical and environmental loads, such as shrinkage and temperature induced loads. Cracking in concrete structures reduces the load-carrying capacity, and cracks allow water and other chemical agents, such as deicing salt, to get through the cover layer to the rein-forcements, finally leading to structural failure due to </p><p>the reduced durability of the concrete[2,3]. Many meth-ods have been proposed to improve the durability of concrete structures in the past[4,5], however few solu-tions have focused on how to overcome the brittle and tensile softening nature of the concrete, which leads to the formation of single large cracks. To effectively solve this serious problem in normal concrete struc-tures, a fundamental solution is needed to reduce the brittle nature of concrete, especially to reduce the crack width in concrete during its service stage. </p><p>In recent years, a new class of high performance fi-ber reinforced cementitious composites, called engi-neered cementitious composites (ECC) have been de-veloped, which are defined by an ultimate strength higher than their first cracking strength with the forma-tion of multiple cracks during inelastic deformation[6].After first cracking, the tensile load-carrying capacity continues to increase, resulting in strain-hardening ac-companied by multiple cracking. The individual crack widths first increase steadily to some level, after which the crack width stabilizes and tends to remain constant. </p><p> Received: 2007-01-17 </p><p>* Supported by the National Natural Science Foundation of China (No. 50178043) </p><p>** To whom correspondence should be addressed. E-mail: junz@tsinghua.edu.cn; Tel: 86-10-62797422 </p></li><li><p> Tsinghua Science and Technology, October 2008, 13(5): 669-673670</p><p>Further increases of the strain capacity are obtained by the formation of additional cracks until the material is saturated. After that, a single crack localizes and the load slowly drops with increased deformation. The spacing between multiple cracks in a typical ECC is 5 to 10 mm and the crack opening at the saturated stage is around 80 Pm[7]. With this crack width, the material durability is significantly improved compared to con-ventional concrete material with similar deformation capacity. However, if the crack width can be further reduced, the material permeability will be reduced. For cracked concrete, experimental results show that for crack widths smaller than 50 Pm as measured under load (crack width of about 30 Pm after unloading), the crack width has little effect on the concrete permeabil-ity. The permeability coefficient (k) is almost the same as the concrete without cracks (k=1u109 cm/s). By contrast, when the crack width is larger than 50 Pm, the concrete permeability increases rapidly[8]. There-fore, micro and macro cracks in concrete can be de-fined according to their influence on the concrete per-meability. If the crack width is less than 30 Pm, the crack can be classified as a micro crack. Otherwise, the crack is a macro crack. Therefore, a key target of on-going research is to develop ECC materials with even narrower saturated crack widths than currently achieved. This paper reports on a class of ECC materi-als with multiple micro-cracking having much finer crack openings and shorter crack spacing than in con-ventional ECC materials. </p><p>1 Experimental Program </p><p>The ECC material behavior was demonstrated using a four-point bending test. The tests were conducted with the deformation controlled by the displacement of the actuator at a constant rate of 0.10 mm/min. The speci-mens was 300 mmh75 mmh15 mm. The strain at the tensile face of the specimen was measured by a dis-placement transducer fixed to the bottom of the speci-men with a 50-mm gauge length. The load configura-tion is shown in Fig. 1. With a bending load, the tensile strain at the tensile face of the specimen can be calcu-lated as 'L/L0, where 'L is the deformation measured by the displacement transducer attached to the bottom of the specimen and L0 is the initial gauge length (L0=50 mm in the tests). Note that the measured tensile strain using this method is not exactly the same as that </p><p>measured by a uniaxial tensile test due to the effect of beam curvature. However in these tests, the effect of curvature on the measured tensile strain is limited be-cause the formation of multiple cracks is the key factor governing the magnitude of the strain and the portion of the strain resulting from beam curvature is relatively small. Therefore, the bending test is sufficient for evaluating the multiple cracking characteristics of the composite, including crack spacing and the crack width of individual cracks. All tests were carried out on a Toni Normal 2000 testing machine equipped for close-loop testing. The raw data consisted of time, load, piston position, and strain transducer displacement. Thus, the flexural stress and strain relation, as well as the multiple cracking characteristics, such as crack pattern, crack spacing, and average crack width, which can be calculated from the total displacement divided by the number of cracks within the gauge length, can be obtained. </p><p>Fig. 1 Test set-up for four-point bending test (mm) </p><p>The present investigation used ordinary Portland cement and two types of silica sand with different par-ticle sizes to form the matrix. One silica sand, the fine sand, had a single grade with an average particle size of 0.458 mm. The other sand, the coarse sand, had a continuous grade with the particle size distribution shown in Fig. 2. They are called coarse sand and fine sand respectively in the following statement. Polyvinyl alcohol fibers (PVA) supplied by the Kuraray Company in Japan were used as reinforcement with the fiber properties listed in Table 1. The mixture proportions used in this study are listed in Table 2. The principal difference between the two mixes is the sand particle size. The molds used to cast the specimens were made of plexiglass. The mixing procedure and the specimen preparation consist of the following steps: (1) Matrix preparation: The matrix was prepared in a mortar mixer. First, the cement and silica sand were mixed together for 2 min at low speed. Then water with a superplasti-cizer and viscous agent (methyl cellulose) was </p></li><li><p>ZHANG Jun ( ) et alTransition from Multiple Macro-Cracking to Multiple Micro-Cracking 671</p><p>gradually added and mixing was continued for 2 min which resulted in a uniform fluid matrix. Within this period, the bottom of the mixing bowl was scraped manually to ensure that no solid materials stuck to the bottom. After scraping, the matrix was mixed at a higher speed for 1 min before adding the fibers. (2) Addition of fibers: The fibers were gradually spread into the mixer by hand as the matrix was mixed at slow speed. The fibers must be added slowly to ensure proper distribution with no fibers bundled together. (3) Casting and curing: The composite material was care-fully cast into the mold in two layers. First, about half the material was placed in the mold. Then the mix was vibrated for 1-2 min to ensure that the material was well compacted. Then, the second half of the mold was filled by the composite in the same manner. After smoothing the surface, the specimens were covered with a polyethylene sheet to prevent moisture loss and stored for 24 h at room temperature prior to demolding. Then all specimens were removed from their molds and put into water at (20r2)oC for curing until the bending tests were performed. </p><p>Fig. 2 Coarse sand powder size distribution </p><p>Table 1 PVA fiber properties </p><p>Type Ef /GPa Vf /MPa df /mm Lf /mm PVA(REC) 42.8 1620 0.039 12 </p><p>Notes: Ef , fiber modulus; Vf , fiber tensile strength; df , fiber diameter; Lf , fiber length </p><p>Table 2 Mixture proportions by weight </p><p>Component Mixture I Mixture II Cement 1.00 1.00 Sand 0.33a 0.33b</p><p>Water 0.295 0.295 Superplasticizer 0.016 0.016 Methyl cellulose 0.000 635 0.000 635 PVA fibers 1.6% by volume 1.6% by volume</p><p>Notes: a, coarse sand; b, fine sand </p><p>2 Experimental Results and Analysis </p><p>The typical flexural test results for the flexural stress versus strain on the tensile side for the two mixes are shown in Fig. 3. The location of the first cracking in the material is illustrated in Fig. 4, which presents an expanded view of the initial portion of the curves. The specimen curing period was 7 days. The typical crack patterns formed at the tensile faces for both mixes at the end of the bending test is shown in Figs. 5a and 5b. The corresponded first crack strength, flexural strength, and ultimate tensile strain, as well as the average crack width and crack spacing at the ultimate flexural stress are listed in Table 3. </p><p>Fig. 3 Flexural stress with strain of mixture I and mixture II </p><p>Fig. 4 Expanded view on the initial portion of the flexural stress-strain curves </p><p>These experimental results show that uniformly dis-tributed cracks are generated along the tensile face of the specimens of the two mixes due to the significant strain-hardening behaviour of the materials under ten-sile stress. The strain-hardening leads the ultimate ten-sile strain capacity increasing about 100 times com-pared to that of pure cement matrix. The results also show that the multiple cracking characteristics of the </p></li><li><p> Tsinghua Science and Technology, October 2008, 13(5): 669-673672</p><p>Fig. 5 Crack patterns for mixture I and mixture II </p><p>Table 3 Bending performance and cracking characteristics of the two mixtures </p><p>Cracking strength (MPa) </p><p>Flexuralstrength (MPa) </p><p>Corresponding tensile strain (%) </p><p>Average crack width at ulti-mate flexural stress (Pm)</p><p>Average crack spacing (mm) </p><p>Mixture I Mixture II </p><p>3.5-4.02.0-2.5</p><p>8.84; 9.32 5.37; 5.53 </p><p>1.50; 1.75 1.75; 2.00 </p><p>9412</p><p>5-60.6-0.7</p><p>two mixes differ greatly. In general, much smaller crack widths and shorter crack spacings were obtained from the specimens of mixture II than those from mix-ture I. The average crack width at the flexural strength is about 12 Pm for the mixture II specimens. The crack widths of individual cracks before reaching the flexural strength are successfully controlled to be micro cracks according to the definition of micro and macro cracks based on the concrete permeability. By contrast, the corresponding average crack width with the conven-tional ECC (mixture I) is about 94 Pm, which is a macro crack. The transition from multiple macro- cracking to multiple micro-cracking is successfully realized by modifying the silica sand particle size used in the matrix. This transformation is significant in view of the material durability which is connected with the structure service life and maintance cost. If cracks are controlled within the micro level, i.e. the crack width is less than 30 Pm, the cracks will not influence the con-crete permeability since the durability of concrete structures is mostly governed by the concrete perme-ability. Secondly, in the multiple cracking stage, the flexural stress-strain curve is smoother for specimens with fine sand particles than with coarse sand. As new cracks occur, the stress decreases in the coarse sand specimens. When the coarse sand is replaced by fine sand, this stress decrease is eliminated and the stress increases until the saturated stage of multiple cracking. Thirdly, the 1.5% tensile strain capacity for both mixes </p><p>is more than 100 times the strain capacity of plain concrete, especially for the mixture II specimens, which have high strain limits without loss of concrete permeability. Some of the tests were stopped before reaching the saturation point due to limitation of the testing equipment (the maximum vertical displacement of the transducer was 10 mm). Therefore, the real strain capacity may be higher than the values measured with the present test set-up. Fourthly, the cracking and flexural strengths of the two mixtures differ. The aver-age cracking and flexural strengths of mixture I were 3.75 MPa and 9.08 MPa, while these of mixture II were 2.25 MPa and 5.45 MPa. Thus, the cracking and flexural strengths of the composite with the fine sand in the matrix were reduced. Despite this, the smaller crack widths and the lower permeability indicate that the fine sand mixture should be better. </p><p>Clearly, the sand particle size significantly influ-ences the multiple cracking characteristics of the composites, including individual crack widths and the crack spacing. Without many detailed experimental studies, the mechanism for these effects can simply be explained as follows. The cracking strength of the composites is obviously reduced by the fine particle sand as shown in Fig. 4 and Table 3. Therefore, the fine sand in the matrix tends to reduce the matrix frac-ture toughness, since the material fracture toughness is proportional to the load at which a crack starts to grow. In addition, the fine sand may also improve fiber </p></li><li><p>ZHANG Jun ( ) et alTransition from Multiple Macro-Cracking to Multiple Micro-Cracking 673</p><p>bridging because the matrix is more uniform which may lead to a better fiber distribution. These two as-pects, reducing the matrix fracture toughness and im-proving fiber bridging, both enhance the multiple cracking performance of the composite[9,10] and lead to significant differences in the multiple c...</p></li></ul>