GUO Xiaolu, WANG Sijia, ZHANG Hongmei

((1. Key Laboratory of Advanced Civil Engineering Materials Ministry of Education, Tongji University, Shanghai 201804, China; 2. School of Materials Science and Engineering, Tongji University, Shanghai 201804, China; 3. College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China)

Abstract: The 21 dog-bone specimens with different fiber contents and fiber distribution (random chopped fiber or directional continuous filament fiber bundles) were designed and tested under uniaxial tension using domestic PVA (polyvinyl alcohol) fiber. High fiber content exerted positive influences on cracking stress,peak stress and deformation capacity of specimens with random chopped fiber, compared with the decrease shown in cracking stress of specimens containing directional fiber bundles. There were multiple cracks in specimens containing directional fiber bundles, while only 1-2 typical cracks could be shown in chopped fiber specimens after being broken. Random chopped fiber connected more closely with matrix compared with that only part of fiber bundles could contact with matrix. Double-fold line model and parabolic model could be used simultaneously to fit well with the uniaxial tension constitutive relations of engineered cementitious composite (ECC). Although the performance of PVA produced in China can not reach to the same level of those from Japan, there exists certain practical value in engineering according to its contribution to deformability of structure.

Key words: fiber reinforced cement composites; fiber produced in China; fiber distribution; micromorphology; constitutive relation

1 Introduction

As the most widely used building material, concrete materials have made great contributions to social development. However, it should not be overlooked that the nature of the concrete material is brittle, the bending-resistance is low, and compressive strength is better than axial tensile strength[1], and the deformation capacity is insufficient. In addition, existing conventional concrete materials have obvious disadvantages in terms of resistance against impact, fatigue, leakage, and corrosion. Therefore, researchers have been looking for optimization of concrete materials.

In 1992, American scholar Liet al[2]obtained a fiber reinforced cement-based composite material with strain-hardening characteristics and strong deformation ability, based on the basic principles of micromechanics and fracture mechanics, and named it as “Engineered Cementitious Composite (ECC)”. Later, many scholars carried out further research on this kind of materials[3-6]. According to the typical uniaxial tensile stressstrain curve of ECC[7], the specimen exhibits obvious strain-hardening characteristics under the axial tensile load, and the peak tensile strain exceeds 5%, which is about 500 times higher than ordinary concrete, and the super high toughness is fully demonstrated. In addition,ECC also exhibits high performance in terms of fatigue resistance[8], penetration resistance[9,10], impact resistance[11], anti-flaking[12], corrosion resistance[13,14],etc.Since the matrix is usually composed of cement and mineral admixtures (such as fly ash, silica fume,etc.)and fine aggregates with finer particle size, so it also shows light weight. Fly ash used in the matrix of ECC not only plays a role in reducing hydration heat and improving the performance of the interface transition zone,but also contributes to saving resources and energy.

From the above, ECC has superior performance and broad development space. However, this type of material still needs further study in the following aspects:

(1) Effect of fiber distribution on material properties

For ECC, there is little research about the effect of fiber distribution on material properties. The mainstream research uses chopped fibers with three-dimensional random distribution in the matrix, ignoring the effect of fiber distribution on material properties. Leeet al[15]found that for fibers with random distribution,the effective coefficient can only reach 0.405 in either direction, which means that a large part of fiber-reinforced ability was not fully utilized. It is a relatively common orientation method to lay the fiber woven mesh in the matrix during the pouring process, but in this method, the effect of the warp fiber on the weft fiber is not clear, and warp fiber is equivalent to the internal lateral defect of the matrix[16], the effect of warp fiber on material properties cannot be ruled out.

(2) Localization of fiber materials

Chinese scholars often use imported PVA(polyvinyl alcohol) fiber produced by Kuraray Company of Japan as the reinforcing material for research. However, the unit price of Kura PVA is several hundred yuan/kg, and the relatively high price restricts the application of materials in practical engineering. The unit price of PVA produced in China is often 20-30 yuan/kg, which has a very large price advantage over Kura PVA. Therefore, the research on ECC with PVA produced in China is more meaningful for lower cost.

Thus, we designed 21 dog-bone specimens using PVA produced in China for the uniaxial tension test,based on two variables (fiber content and fiber distribution). Combined with scanning electron microscopy(SEM), the deformation and cracking of the specimens were observed and analyzed. The practical value of PVA produced in China was evaluated. The uniaxial tensile stress-strain curve of the specimens under different working conditions was drawn and the uniaxial tension constitutive relation of ECC was given by different fitting models as the preliminary preparation for finite element analysis and structural design.

2 Experimental

2.1 Raw materials

The matrix of ECC mainly comprised following parts: dry powder raw materials（PO 52.5 cement, ultra-fine sand, ASTM class F fly ash, nano-SiO2, superplasticizer(SP) with a water-reducing rate of more than 35%, and air entraining agent); water; PVA. Dry powder raw materials were purchased from Guyiqiang New Materials Company of Hangzhou, and the mix ratio was provided by the manufacturer. Chopped fibers with length of 6mm and continuous filament fiber bundles were used, supplied by Sinopec Chongqing Chuanwei Chemical Company, and the fineness of single filament was consistent with the chopped fibers. The relevant parameters are shown in Table 1. Each bag of dry mortar (25 kg) was mixed with a bag of admixture (superplasticizer and air entraining agent) dry powder (72.5 g)and 5 kg water; PVA was added according to different working conditions.

Table 1 Physical and mechanical characteristics of PVA fiber

Table 2 Parameters of 21 dog-bone specimens

The material preparation was carried out by using JJ-5 planetary cement mortar mixer. The preparation process is shown in Fig.1. It should be pointed out that the directional fiber should complete the fiber bundle pre-wearing work before preparing specimens, and then the stirred matrix slurry was poured into the dog-bone mold. After being demoulded, specimens were moved into curing room for 28-day standard curing, and then tested.

Fig.1 Specimens preparation process

2.2 Specimen overview

The dog-bone specimen had variable cross-section, and transition section was set to an arc shape. Its size was appropriately increased to reduce the effect of factors such as eccentricity on the test result. Dimensions of dog-bone specimen are shown in Fig.2. Fiber contents were 0.5%, 1.0%, and 1.5%, respectively for both random chopped fiber and directional continuous filament fiber bundles. Different end plate opening methods(7×7, 8×8) were set to explore the effect of different fiber distributions on results in the condition of 0.5% continuous fiber filament bundles(Fig.3).

Fig.2 Dimensions of dog-bone specimen/mm

Fig.3 End plate opening methods

21 dog-bone specimens were numbered as follows: R represented random chopped fiber; D represented directional continuous filament fiber bundles;C represented the fiber content (volume fraction); H represented end plate opening methods of continuous fiber filament bundles; S represented the parallel specimens under the same work condition, and there were 3 parallel specimens in the same working condition (S1,S2, S3). For example, “R-C0.5-S1” represented the first parallel specimen with 0.5% random chopped fiber, and“D-C0.5-H7-S1” represented the first parallel specimen with 0.5% directional continuous filament fiber bundles and end plate opening method was 7×7. Table 2 shows different work conditions of 21 dog-bone specimens.

2.3 Experimental methods

A four-column microcomputer-controlled electro-hydraulic servo-controlled universal testing machine (mode WAW-500J) was used in this experiment in order to obtain stable uniaxial tensile stress-strain curves[17]. The maximum of applied load was 500 kN and accuracy class was 1. The loading head and fixture of this experiment used the device designed by Zhouet al[18](Fig.4), and the installation was adhesive type.Fig.5 shows the bonded steel plate. The position of the screw plane on the back side of the steel plate was the same as that of the pre-embedded bolt hole of the connecting plate in Fig.4. The two ends of the specimen and the steel plate were bonded with structural adhesive, and could be used for installation after curing.

Fig.4 The loading head and fixture

Fig.5 (a) Connection plate; (b) Bonded steel plate

The loading type was displacement control loading at a speed of 0.25 mm/min. Deformation was measured by a displacement meter (mode YHD-30)mounted on both sides of the specimen. The maximum range was 30 mm, and the gauge length was 160 mm. A displacement meter was also mounted at the upper and lower connecting plates respectively for monitoring the deformation of the whole specimen as a reference.Scanning electron microscopy (SEM) was used to observe the internal microscopic morphology of the loaded specimens. The uniaxial tensile stress-strain curve of the specimen was fitted by the double-fold line model and parabolic model to obtain the constitutive relation expression.

3 Results and discussion

The point at which the bearing capacity was first suddenly dropped was selected as the initial cracking point of the specimen. If the bearing capacity sudden drop was not observed, the point where the slope of the curve changed significantly was selected as initial cracking point. The ultimate point was selected as the corresponding point when the stress dropped to 0.85 times the peak stress. The cracking mode of some specimens was single-point cracking, and the peak point coincided with the initial cracking point. In particular,D-C0.5-H8-S2 had a tendency to rise gently in the next step even after the single-point cracking, although the bearing capacity dropped suddenly. It had sufficient tensile deformation capacity, and the peaks of the gradual segments are also shown in Table 3. The ultimate point of D-C0.5-H8-S2 was determined by the peak values of the gradual segments.

Overflow of the end of the specimen caused stress concentration because of adhesive installation. Some ends of specimens had early failure during the loading process, and effective data did not be obtained. Specimens with partial valid feature point’s data are shown in Table 3 and corresponding situations are indicated.

Table 3 Characteristic points of different specimens in uniaxial tension test

3.1 Effect of fiber content on properties and crack characteristics of ECC

Fig.6 shows uniaxial tensile stress-strain curves of specimens under different working conditions. O in the curve number represents the stress-strain curve of the corresponding specimen was directly measured by the test. For example, “O-R-C0.5-S1” represents the uniaxial tensile stress-strain curve of R-C0.5-S1 obtained in the test. In Fig.7, N in the curve number represents the uniaxial tensile stress-strain curve of the corresponding specimen obtained by fitting, for example, “N-RC0.5-S1” represents the uniaxial tensile stress-strain curve of R-C0.5-S1 obtained by fitting.

Fig.6 Uniaxial tensile stress-strain curves of specimens with different fiber contents and distribution

Fig.7 Uniaxial tensile stress-strain fitting curves of specimens

3.1.1 Effect of fiber content on properties and crack characteristics of ECC with random chopped fiber

From Table 3 and Fig.6, fiber content has different effects on properties and crack characteristics of specimens with random chopped fiber and directional continuous filament fiber bundles. With the increase of fiber content, cracking stress, peak stress and deformation ability of chopped fiber specimens were improved.For directional fiber specimens, the cracking stress was not affected significantly by fiber content, but the peak stress improved with the fiber content increase, and the slope of the stress-strain curve increased in crack propagation.

As the fiber content increased from 0.5% to 1.5%,the average cracking stress of chopped fiber specimens reached 1.39, 1.85, and 2.44 MPa, respectively, and the peak stress average reached 1.44, 1.92, and 2.60 MPa,respectively, with a significant rise. Random chopped fiber across the cracks provides tensile forces to hinder crack development, but the lack of effective specimens is not sufficient for quantitative analysis. The average ultimate strain of chopped fiber specimens reached 0.1%, 0.31%, and 0.3%, respectively. The ratio of the ultimate strain of chopped fiber specimen with 1.0%fiber content to that with 0.5% fiber content was more than 3 times, indicating the deformation ability was improved significantly. The ultimate strain of chopped fiber specimen with 1.5% fiber content was close to that of the 1.0% fiber content, which may be due to the grooved treatment of the middle part of the specimen in order to reduce the effect of the stress concentration at the edge of the specimen ends. After the test, the specimens with 0.5% chopped fiber were divided into two halves, while the remaining specimens with higher fiber content maintained their integrity (shown in Fig.8).

Fig.8 Failure forms of random chopped fiber specimens with different fiber contents

3.1.2 Effect of fiber content on properties and crack characteristics of ECC with directional continuous filament fiber bundles

The average peak stress of these specimens with 0.5% fiber content was 1.52 MPa. When the fiber content was increased to 1.5%, the peak stress was more than 2.90 MPa and 2.12 MPa, respectively, showing a significant rise (Fig.6(b)). The specimens with the fiber content of 0.5% and 1.0% showed obvious bearing capacity drop at the initial cracking, while that curve of specimens with 1.5% fiber showed only the sudden change of the rising slope. The large fiber content allows the fiber to be quickly replenished after the matrix is cracked. In the stable development stage of the crack,the slope of curve was improved significantly with the increase of fiber content since load was mainly carried by the fiber at this stage.

3.2 Effect of fiber distribution on properties and crack characteristics of ECC

Due to different fiber distribution patterns, the cracking stress of specimen showed difference obviously. The cracking stress of specimens with random chopped fiber was obviously affected by fiber content,while that of directional specimens was more stable and significantly lower. It’s mainly due to the difference in the synergistic performance between the fiber and matrix in the early stage. The overall properties between random fiber and matrix are better, and have the ability to work together in the early stage of loading. For the directional specimens, firstly, only the outermost ring of fiber bundles distributed in each hole can be in direct contact with the matrix, and there is no effective connection between fiber bundles, resulting in insufficient stress transmission. Secondly, there is a time difference between the fiber perforation and specimens casting,and the fiber bundles appear slack, thus leading to insufficient synergistic performance in the early stage.

Different fiber distribution patterns also caused difference in deformability and failure process of specimens. Although the deformation ability of specimen with random chopped fibers was much larger than that of the ordinary concrete specimens, it was seriously insufficient compared to the directional specimens with the same fiber content. Only 1-2 obvious cracks could occur in the random chopped fiber specimens during loading, and the crack width could not be fully developed, while directional specimens had multiple cracks during the cracking process (shown in Fig.8 and Fig.9).Continuous filament fiber bundles distributed along the direction of the pulling force are effectively connected to both sides of the crack, and the crack width can be continuously increased. The fiber bundle exhibit good performance by stretching, sliding and fracture in the matrix to consume energy, and still have enough bearing capacity in the stress drop section.

Fig.9 Failure forms of directional continuous filament fiber bundles specimens with different fiber contents and end plate opening methods

There were three directional specimens of 0.5%fiber content (D-C0.5-H8-S1, D-C0.5-H8-S2, and D-C0.5-H8-S3), of which end plate opening method was 8×8. Failure process of D-C0.5-H8-S1 was similar with D-C0.5-H7-S1, D-C0.5-H7-S2, and D-C0.5-H7-S3, but the deformation ability decreased, while D-C0.5-H8-S2 performed completely different single-point cracking mode. The loading condition of D-C0.5-H8-S2 is similar to the description of the fiber woven mesh concrete with low mesh ratio in the uniaxial tension condition[16]. The maximum stress in the steady stage was 1.48 MPa, the strain was 2.85%, and the corresponding ultimate strain was 5.03%, which is shown a large increase.

Since the end plate opening method was 8×8, the design distance from the center of the outermost fiber to specimen surface was only 5 mm, and fiber bundles were easily exposed in the pouring process. The fiber failed to exert an effect in stress transfer in the steady stage of crack propagation. In this sense, the actual fiber content was less than 0.5%. From Fig. 9, the fiber exposure degree of D-C0.5-H8-S1 was relatively slight,and that of D-C0.5-H8-S2 was more severe, causing different failure forms. The cracking stress of D-C0.5-H8-S2 was obviously larger than other directional specimens. It may due to its directional continuous filament fiber bundles had been initially tightened and could cooperate with the matrix at the early stage of loading.The relationship of cracking stress and fiber bundle prestress needs to be further studied.

3.3 Microscopic morphology

The internal microscopic morphology of the loaded specimens was observed by scanning electron microscopy (SEM) (shown in Fig.10). Chopped fibers had tighter bond with the cement matrix. Obvious wear could be observed in the protruding portion and ends of the fibers were often broken. In the directional specimens, the distribution of internal fiber filaments showed certain dispersion, a part of the fiber filaments were wrapped by the cement matrix to form a whole,while the remaining only had slight contact between each other. The fiber filament surface was generally smooth and had slight wear, indicating that the bonding constraint between the fiber filament and the cement matrix or other fiber filaments was not sufficient, thereby affecting the synergistic performance.

Fig.10 Internal microscopic morphology of loaded specimens

3.4 Uniaxial tension constitutive relations

The ascending section of uniaxial tensile stressstrain curve is often fitted by double-fold line model.There may be a large deviation due to the initial eccentricity in the determination of initial cracking point, and the line between initial cracking point and peak point as the strain hardening curve may also generate deviation comparing with the actual curve. Liet al[19]linearly fitted all the data points in the strain hardening stage and obtained the expression, then defined the intersection point between the fitted straight line and the initial straight line as the “nominal initial crack point”, which became the turning point of the two straight lines.However, there is lack of research on the descending section of uniaxial tensile stress-strain curve.

Double-fold line and method proposed by Li[20]were used to fit the ascending section. Especially, the fitted straight line was required to pass through the peak point measured by test to reflect the peak deformation and bearing capacity of the specimen. According to the characteristics of the stress-strain curve, there were linear decreases occurring in the random chopped fiber specimens. However, those curves of most directional specimens showed a decline tendency whose rate was first slow and then steep, and the shape was similar to the parabola with the opening downward, since the fitting models of the descending section were different.

Nominal initial crack point was (εc0, σc0), peak point was (εm, σm), the expressions of the ascending section for specimens with random chopped fiber and specimens with directional continuous filament fiber bundles were described as equation (1) and equation (2):

Expression of the descending section of specimens with random chopped fiber is described as equation (3):

Expression of the descending section of specimens with directional continuous filament fiber bundles was described as equation (4):

wherek1,k2,k3, andaare parameters to be calculated.

The fitting curves are shown in Fig.7. From Fig.7,the fitting effect is expected, indicating the model has certain applicability, which can provide reference for subsequent finite element analysis and structural design.

3.5 Evaluation and future work

3.5.1 Evaluation of PVA produced in China as reinforcement

Specimens with random chopped fiber failed to adequately exhibit the expected multiple cracking and strain hardening effects. The realization of multiple cracking and strain hardening of specimens with random chopped fiber depends on the bridging stress of the fiber and the fracture toughness of the matrix. The stronger the bridging stress, the lower the toughness of the matrix, the more the effect can be played[20].

PVA produced in China used in this research,compared with PVA from Japan, the biggest difference is whether the surface is specially treated. PVA imported from Japan with special surface treatment is more easily dispersed in the matrix, and the cohesive force between fiber and matrix is decreased within a reasonable range. Since the fiber tends to be pulled out rather than broken in the crack propagation, and the bridging stress of fiber to matrix can be maintained, making it easier to exert strain hardening effect. Zhanget al[21]of Tsinghua University proposed through experiments that the maximum length of fiber (Lc) is described as equation (5) in order to ensure that the fiber was pulled out during the loading process instead of breaking:

where, σfis the fiber tensile strength,dfis the fiber diameter, and τ is the cohesive force between fiber and matrix. Length-diameter ratio was satisfied equation (6):

Liet al[22]found the properties of the fiber interface with and without surface treatment. The surface treatment could significantly reduce the frictional bond and chemical bond effect between the fiber and the matrix, which means the value of τ is greatly reduced.When the tensile strength was similar, the length-diameter ratio (length was 6 mm, and diameter was 17 μm) of the chopped PVA produced in China in this research was slightly larger than that of the mainstream PVA from Japan (length is 12 mm, and diameter is 39 μm). Since it is reasonable that multiple cracking and strain hardening effects were not fully exerted. However, in this research, the peak tensile strain of random specimens chopped PVA produced in China could reach 0.27% or more, which is much higher than that of ordinary concrete materials. It indicates PVA produced in China has certain practical value in engineering.

3.5.2 Future work

This research found that the random chopped fiber in the matrix could directly promote the toughness of matrix and improve the stress and strain at the initial cracking to a certain range, while the directional continuous filament fiber bundles brought stronger deformation and continued bearing capacity after cracking.If ECC with random chopped fiber can be combined with pre-tensioned and surface-treated directional continuous filament fiber bundles, the performance of the material may be further improved. It needs further experimental verification and research.

4 Conclusions

a) The cracking stress, peak stress and deformation capacity of specimens with random chopped fiber increased with the increase of the fiber content, and failure form was single-point cracking. The cracking stress of specimens with directional continuous filament fiber bundles was lower, but the deformation and bearing capacity were obviously improved, and the peak stress increased with the increase of the fiber content; When the fiber content was high, there was not obvious stress decline when specimens reached the initial cracking point and specimens performed multiple cracking; The integrity of the filament inside the fiber bundle was insufficient, and the performance was not good.

b) The uniaxial tensile constitutive relations of ECC was studied. Double-fold line model was adopted for the ascending section; linear model and parabolic model were adopted for the descending section corresponding to different fiber distribution forms. The fitting curves and the test curves fitted well indicating the model had certain applicability and could be used as the reference of subsequent finite element analysis and structural design.

c) Since PVA produced in China used in this test did not be specially surface treated, and the bonding effect with the matrix was strong, meanwhile the length-diameter ratio was too large, multiple cracking and strain hardening effects were not fully exerted.Deformability of structure containing it was still better than that of ordinary concrete indicating there was certain practical value in engineering.