Xác định ứng suất do phá hoại kéo trong bê tông bằng Sóng phát xạ

26 T„P CHŠ KHOA H“C KI¦N TR”C - XŸY D¼NG KHOA H“C & C«NG NGHª Identification of tensile damage in concrete by Acoustic Emission Xác định ứng suất do phá hoại kéo trong bê tông bằng Sóng phát xạ Nguyễn Tất Tâm, Narintsoa RANAIVOMANANA, Jean-Paul BALAYSSAC Tóm tắt Phương pháp Sóng phát xạ (Acoustic Emission) đã được áp dụng để xác định một số dạng phá hoại điển hình trong kết cấu bê tông. Đề cập đến trong RILEM TC 212-ACD, kĩ thuật xác định dạng phá hoại của bê tông gây ra do ứng suất kéo hay ứng suất cắt được đặt tên là “phương pháp RA”, tuy nhiên phương pháp này chưa định lượng tỷ lệ phần trăm các ứng suất nói trên. Đó là hạn chế của phương pháp RA, vốn được dựa theo thí nghiệm uốn dầm bê tông đơn giản chịu hai tải trọng tập trung và thí nghiệm cắt trực tiếp mẫu bê tông. Trong bài báo này, nhằm mục đích củng cố kĩ thuật phân loại phá hoại, các tác giả đã tiến hành thí nghiệm kéo mẫu bê tông để xác định các thông số của sóng và dạng sóng cũng như ảnh hưởng của sự không đồng nhất của vật liệu đến đường truyền sóng phát xạ. Sau khi lọc bỏ các tín hiệu không phù hợp, phương pháp RA được áp dụng và tỷ lệ các dạng ứng suất (Mode I và Mode II) cũng đã được xác định. Trong biểu đồ RA - AF phân loại ứng suất, các điểm trên biểu đồ được xác định từ hai nguồn khác nhau: từ các tín hiệu còn lại sau khi hoàn tất quá trình loại bỏ tín hiệu liên quan tiếng ồn, và từ việc xác định giá trị trung bình liên tiếp của 50 tín hiệu. Kết quả cho thấy phá hoại gây ra do hầu hết là ứng suất kéo. Tuy nhiên, việc phân tích sâu hơn về tín hiệu (thông số của sóng, dạng sóng, tương quan giữa các dạng sóng) cho thấy, các tín hiệu có chung một nguồn gốc phát xạ (event) nhưng các thông số này có giá trị khác nhau, do vậy phương pháp này cần thảo luận thêm. Abstract The existence of typical crack modes in general concrete structure have been determined thanks to Acoustic Emission (AE) technique. In RILEM TC 212-ACD, the classification namely RA method can determine the tensile and shear which occurred in concrete damaged objects but the proportion of these stresses are not clarified. These are the limitations of this method which is based on the four-point bending tests and the direct shear tests of concrete specimens. In this paper, aiming to improve this classification technique, authors have performed tensile test on concrete specimen in order to determine the signal parameters as well as waveforms and to assess the influence of material to wave propagation. After filtering the raw data, the usual RA-AF classification process is used to determine the proportion of each type of damage (Mode I and Mode II). The RA-AF on the classification graph is calculated from filtered hit and from average of 50 continuous hits. The results show dominant proportion of AE signals are associated with mode I damage. However, a further analysis of the signals (AE parameters, wave forms, Cross-correlation) that generated from the same event to check the relevance of this classification shows that it needs to be discussed. Keywords: Concrete, Modulus of elasticity, Homogenization, deformation Nguyen Tat Tam Faculty of Civil Engineering Hanoi Architectural University, Vietnam Email: Narintsoa RANAIVOMANANA Jean-Paul BALAYSSAC LMDC, Université de Toulouse, INSA, UPS, France 1. Introduction 1.1. Tensile damage in concrete specimens In general, concrete is considered to be a brittle material. Especially in the case of tensile loaded concrete a very brittle behaviour is expected, but in some cases, e.g. anchorage and pure bending, tensile loaded concrete exhibits a ductile behaviour [1]. By centric tensile tests of concrete samples, a load-deformation-curve is analysed. As observation by the authors, when loading from 0 to before the peak load, the stress-strain relation is linear elastic. The modulus of elasticity is determined by initial tangential of angle that formed by this stress-strain relation and the horizontal line. Then immediately before the peak load (the round-segment), an accumulation of micro-cracking occurs at the weakest part of the specimen, and this leads to an additional strain over this part. Having passed the peak load, the crack band localizes and the deformation within the crack band increases, and the final failure occurs due to one single crack. In the notched specimen subjected to traction, the micro-cracks are visible just before loading is reached to the peak, and the cracks are going to concentrate at notch location. Within this notches, material bridges transfer the tensile load, as similar to the crack band at weakest part as indicated in [1]. After the formation of a real single crack, the transferred of stress is possible due to aggregate interlock. In most cases, a crack will run along the interfaces between the aggregate grains and the cement paste and then grains are pulled out of the paste. Due to this, friction forces between grains and paste are occurred. The grains act like friction blocks and transfer friction forces over the crack. In some others cases, e.g. the case of weak cement paste and strong aggregate, the crack runs through the paste and at the interfaces between aggregate and paste. In addition case, when the cement paste reaches the strength of the aggregate, the crack will split most of the aggregates. In [2], normal weight concrete has been tested under monotonic and cyclic loading. The aim of these tests was to provide an accurate description of the tensile behavior of concrete and simple enough for application to numerical analysis. Looking at the on-going fracture analysis, the authors inferred that there is still a lack of knowledge in field of interaction of Tensile crack (Mode I) and Shear (Mode II), and it is an open field of research. By AE tests, we are partly resolving this field of study. 1.2. Acoustic Emission in concrete damage classification The RA method listed in Recommendation of RILEM TC 212-ACD [3] is one of the crack classification methods which is based on the results that have confirmed under the four-point-bending tests and the direct shear tests of concrete specimens. As definition, the Rise time and the maximum Amplitude are applied to calculate RA value, while the Average Frequency (AF) value is obtained from AE count and the duration time. The RA and AF value are 27 S¬ 28 - 2017 recommended to be calculated from the moving average of more than 50 hits [4]. And in [3], the NDIS 2421 has not defined the criterion to determine the proportion of the RA value and the AF for crack types, as presented by the floating dash-line in Figure 1, that mean the location of this diagonal line will be fixed by the users. In this graph, if the vertical axis is shown in kHz while the horizontal is in ms/V, the ratio K = A/B (ms/V×kHz-1) shall be determined depending on materials and structures. As reviewed by Ono [5] in the RC beam test, this K value was reported as 1/50, while another group gives the value of 1/8 in bending and shear tests of concrete, but further works are needed to develop. In parallel, in some recent researches, Aggelis et al. [6] have proposed method of collecting RA value and to locate the dash-line. The RA method has been widely applied in some papers [7] and the K ratio is also determined following the type of the testing and material, but its variation confirms that there is no rule on this ratio. 2. The experimental, loading machine and AE system setup 2.1. Material and specimen set up Type I Portland cement with 28 days strength of 52.5 MPa is used. Coarse aggregate is gravel, which is composed of unconsolidated rock fragments that have rough surface and general particle size range with maximum value of 16 mm. Fine aggregate is crushed fine sand of maximum size not greater than 4 mm. The mechanical properties of concrete were determined at 28 days on three ϕ118×225 mm cylinders with a compressive strength (fc’) of 51.0 MPa assessed through direct compression tests; the tensile strength (ft) of 3.3 MPa was assessed by splitting tests. The Elastic modulus of 37.5 GPa was determined based on RILEM CPC8 recommendation. One concrete specimen was subjected to traction test has dimension of 25×10×10 cm and a 10mm notch around the mid-span. The loading system was controlled by two COD1, 2 clip gauges locate across the notches. Due to the expected brittle response, the test was conducted by loading was applied with rate of 5 μm/min and 20 μm/min to the CODs for before and after peak load, respectively. Loading platens are glued to both ends of the specimen by epoxy. The upper one was glued first and connected to the actuator; whereas the lower one was adjusted its location to the central of the lower platen before it is fixed by the epoxy. This step intends to reduce the eccentricity of loading during the test. 2.2. Acoustic emission setup The AE activity recorded was performed using eight- channel PCI–8 acquisition device of the Physical Acoustic Corporation (PAC). For recording the characteristic parameters an AEwin for SAMOS version 2008 software was used. AE detection was performed by sensors, R15-α series of PAC whose specification: Operating frequency range 50 Figure 1. Damage classify using AF and RA value Figure 3. Loading (kN) and AE amplitude (dB) vs. time (s) in tensile specimen Figure 2. Sensors arrangement on Front and right-side (a), and on back and left-side (b) on specimen (Dimensions are in mm) (a) (b) Table 1. AE sensors arrangement on specimen Sensor no. X (cm) Y (cm) Z (cm) 1 5 15.7 10 2 5 9.3 10 3 5 15.7 0 4 5 9.3 0 5 0 8.3 5 6 10.5 16.7 5 28 T„P CHŠ KHOA H“C KI¦N TR”C - XŸY D¼NG KHOA H“C & C«NG NGHª – 400kHz, Resonant frequency 150kHz, Peak sensitivity 80dB. These sensors are mounted on the surface of the specimens with silicon grease as coupling agent, and they were placed close to the expected location of the future cracks path to minimize errors in the AE event localization (Figure 2). These sensors have a coordinate that indicated in Table 1 as 3D analyses perform. The PAC preamplifiers model 2/4/6 (gain selectable 20/40/60 dB + 5% dB) were fixed a gain of 40 dBs intend to eliminate the background noise. The acquisition system was calibrated before each test using a standard source pencil lead break procedure Hsu-Nielsen and to verify that nothing has changed on sensors sensitivity before and after the test, the Auto Sensor Test was performed. In these tests, the AE events are located by applying the wave velocity of 4,000 m/sec. 3. Crack classification applying RA value 3.1. AE raw data filtering After the time duration of 260s, the testing system stopped as a result of the Table 2. AE parameters in event 2 Record Channel di (cm) Rise time (μs) Amp (dB) AF (kHz) Counts Duration (μs) RA (ms/V) ABEN (aJ) 0 5 3.34 32 50 117 34 290 1.012 393.68 1 4 4.32 28 48 90 19 212 1.115 188.03 2 3 7.04 0 45 49 9 184 0.000 79.45 3 2 7.76 21 44 59 12 202 1.325 106.06 4 6 11.03 48 54 92 45 487 0.958 718.98 5 1 9.54 4 48 59 23 387 0.159 198.01 Table 3. Normalized Cross-correlation (NCC) of signals in selected events Event Number of records Group name Record 0-1 Record 0-2 Record 0-3 Record 0-4 Record 0-5 1 6 Concentrate -0.08 -0.01 -0.13 -0.08 -0.13 2 6 Scatter -0.09 -0.01 0.30 0.08 -0.20 Figure 4. Tensile specimen (a) and AE events at notch portion (b) Figure 5. Damage classification at Peak load Figure 7. Damage classification at Peak load Figure 6. Damage classification at failure Figure 8. Damage classification at failure 29 S¬ 28 - 2017 specimen was completely damaged and the number of AE hits that recorded thanks to six sensors is 30,607. The peak value of loading is 21.23 kN corresponding to the CMOD of 4.8 μm (Figure 3). After this peak point, the curve gradually dropped up to a brittle failure of the specimen. By observing the images (a) and (b) in Figure 4, the location of the crack on specimen is good agreement with the events which are localized by AEwin. The first observed AE signals are on the upper part of the beam and they concentrated beneath the loading-jack possible due to contact damages. The next hits are visible at the lower location and random in fracture process zone. The number of AE signals obtained in experiment tests is almost large with inconsistent shapes and either their parameters. Filtering work on AE hits may be associated to raw data with surround noise elimination. The hits with low magnitude (Duration less than 10 μs and Count less than 2) could be related to background noise [8]. And it is noteworthy in some studies [9] that the AE energy have a good correlation to the fracture energy. And as the comments in those papers, users are possible to cite that AE energy can be a feature to determine the fracture energy of concrete. They also confirm in the three-point-bending test with notched concrete beam, the high energy events are located above the tip of the notch. In addition to above filtering task, signals that have the Duration higher than the Frame-time that definite by AEwin before starting of signal recording will also be discarded. To determine the appropriate Duration value, the Hit Definition Time (HDT, μs) is calculated through the input parameters. According to [10], HDT is defined as follows Eq. 1. 1024 LHDT P S = × − Eq. 1 Where: L (μs) - Length in k (1 k = 1024 μs) of signal; S – Sample rate in MSPS (Millions of Samples Per Second), 1 MSPS = 106 Hz; P (μs) is Pre-trigger time. In this test, L = 2 k, S = 1 MSPS and P = 96 μs then HDT = 1952 μs. In this test, AE data filtering work has removed the signals with Count less than 2, zero of PAC energy and Duration higher than 1952 μs. Comparing to the raw data with 30,607 hits, the filtered data remaining 15,121 hits (49.4%), thus, 50.6% of inconsonant signals have been eliminated after filtering work. 3.2. Crack classification applying filtered RA value The result of damage classification performing to 275 hits which are recording from the beginning of the test to peak loading is indicated in Figure 5. It can be seen, the number of hits that resulting damage Mode I is occupied 97.1%, thus, the dominant damage mode is tensile. And at the failure (15,121 hits), Mode I is increased and accounted for 98.4% as shown in Figure 6. AE analysis confirmed that the damage in specimen is caused by tensile stress. The Shear mode exists but it contributes low proportion with 1.6%. 3.3. Crack classification applying RA value of average 50 continuous AE hits As indicated in Subsection 1.2, in NDIS 2421 [3] classification process, the RA and AF value are calculated (b) (c) Figure 9. 3D event localization (a), crack classification for six signals of event 1 (b) and event 2 (c) Figure 10. Signal waveforms of Record 0 to 5 of event 2 (a) 30 T„P CHŠ KHOA H“C KI¦N TR”C - XŸY D¼NG KHOA H“C & C«NG NGHª from the moving average of more than 50 hits [4]. In this subsection, RA and AF value of individual hit are determined and then the average value of group 50 continuous hits is created. At peak load, the plots show 100% tensile crack in the specimen as show in Figure 7. The dots in the graph represent the average value of RA and AF of 50 continuous hits. In the following process after peak, the result on the plot clarifies that 100% damage mode during this process is tensile (Figure 8). 4. AE events source discrimination The NDIS 2421 damage classification has been applied RA value as well as Average frequency of signals but without considering other independent parameters of those signals such as Amplitude, Count, Duration, Energy and etc. Thus, by generated from one event and having the similar damage mode, but the received signals at individual sensor have the differential shapes and parameters. Figure 9.a) depicts an event with the ranges to the sensors are di (i = 1 - 6). It can be seen, the different in travel distance from source to the sensors possible influence to the waveforms. To verify this, two events are extracted from the 3D event localization then classify by RA value and Cross-correlation. The Cartesian coordinate of event 1 is (2.21; 12.44; 7.18) and event 2 (1.91; 10.09; 2.92) cm. In Figure 9.b) and c), it is clearly seen that all signals of the event 1 and event 2 are classified in Mode I. Although having the similar mode I but the distribution of records on the RA - AF chart is different to the events and there are two trends of signal grouping. The first is ‘concentrate’, for example the signals in event 1 are closely located on the plot that represents the same RA and AF value. In contrast, the second Group is ‘scatter’ as event 2, the position of signals are varying in larger zone comparing to event 1 with AF from 50-120 kHz, RA from 0-2 ms/V. In terms of waveform, the Figure 10 presents the waveforms of Record 0 to 5 of event 2. It can be seen, the presence of high AF in the 6 signal waveforms improve that they are tensile mode. As indicated in [11], when the distance from sensors to event increase, the AF and energy decrease while RA increases. By observing the events that defined in the tensile test, authors recognized that these events are incompatible with above attenuation rule in [11]. For example, from Record 0 to 4, the distances from the sensors to the event rise from 3.34 cm (Record 0) to 11.03 cm (Record 4) while the Amplitudes reduce from 50 to 44 dB (in Record 0 to 3) but increase to 54 dB in Record 4. Similarity, the fluctuation of RA and ABEN (Absolute Energy) from Record 0 to 5 clarify that there is no exhaust regulation on these factors (Table 2). 5. Signal waveform Cross-correlation Another technique for AE sources discrimination consists in applying Cross-correlation method. Wave Cross- correlation aims to find the similarity between waveforms, thereby, it could help to evaluate if the received signals by sensor 1 to 6 are compatible or incompatible with each other. The correlation result reaches a peak at the time when the two signals have the best match. When the two signals are identical in terms of shape, this peak is reached at time t = 0 without delay. However, if one of these two signals has delay time and is possibly influenced by the travel distance then correlation is a good method to measure that delay. The Cross-correlation (CC) of discrete signal is defined as Eq. 2. ( ) [ ] [ ] 1 0 , N n n n CC x y X Y − = = ∑ Eq. 2 Where: N is number samples in the signal. In the AE signal acquirement system, N is determined by a rate of 1 point per μs. In this test, AE signals are recorded with N = 2048 samples (equivalent to 2048 μs) and it will stop at point which is zero Amplitude. And X[n] and Y[n] are function of physical quantity varies over time or spacy. In general, the CC is a measure of how similar signals are and the high CC indicates that the signals are quite the same. However, if two events that have high energy (or high amplitude) at some samples at different time, the CC value could be comparatively high but actually the signals are not quite similar. Thus, the CC value may cause the misleading to the users. Then the normalized of Cross-correlation (NCC) is necessary apply to the two signals to conclude that they are identical or not, as defined in Eq. 3. ( ) [ ] [ ] [ ] [ ] 1 0 1 12 2 0 0 , N n nn N N n nn n X Y NCC x y X Y − = − − = = = ∑ ∑ ∑ Eq. 3 To evaluate the correlation between signals in the two groups named ‘concentrate’ and ‘scatter’ that mentioned above, signals in some events will be selected to calculate the correlation and normalized value. The events in Group one is event 1 and Group two is event 2. The results of the Figure 11. Correlation Record 0 - Record 2 with NCC = -0.01, event 2 Figure 12. Correlation Record 0 - Record 3 with NCC = 0.3, event 2 31 S¬ 28 - 2017 Tài liệu tham khảo 1. Gert Konig and Herbert Duda, “Basic concept for using concrete tensile strength,” ETH Zür. Rämistrasse 101 8092 Zür. Schweiz Wwwlibraryethzch, 1991. 2. Hans W. Reinhardt, Hans A. W. Cornelissen, and Dirk A. Hordijk, “Tensile tests and failure analysis of Concrete,” Univ. Neb.-Linc. 060613, 2013. 3. Kentaro Ohno and Masayasu Ohtsu, “Crack classification in concrete based on acoustic emission,” Constr. Build. Mater., vol. 24, no. 12, pp. 2339–2346, Dec. 2010. 4. RILEM Technical Committee, “Recommendation of RILEM TC 212-ACD: acoustic emission and related NDE techniques for crack detection and damage evaluation in concrete: Test method for classification of active cracks in concrete structures by acoustic emission,” Mater. Struct., vol. 43, no. 9, pp. 1187–1189, Nov. 2010. 5. Kanji Ono, “Application of acoustic emission for structure diagnosis,” Diagn. ISSN 1641-6414, pp. 3–18, 2011. 6. D.G. Aggelis, “Classification of cracking mode in concrete by acoustic emission parameters,” Mech. Res. Commun., vol. 38, no. 3, pp. 153–157, Apr. 2011. 7. Arash Behnia, Hwa Kian Chai, and Tomoki Shiotani, “Advanced structural health monitoring of concrete structures with the aid of acoustic emission,” Constr. Build. Mater., vol. 65, pp. 282–302, Aug. 2014. 8. L. Calabrese, G. Campanella, and E. Proverbio, “Noise removal by cluster analysis after long time AE corrosion monitoring of steel reinforcement in concrete,” Constr. Build. Mater., vol. 34, pp. 362–371, Sep. 2012. 9. R. Vidya Sagar and B. K. Raghu Prasad, “An experimental study on acoustic emission energy as a quantitative measure of size independent specific fracture energy of concrete beams,” Constr. Build. Mater., vol. 25, no. 5, pp. 2349–2357, May 2011. 10. MISTRAS Group, Inc, SAMOS AE system User’s Manual, Rev. 3. 2011. 11. D. Polyzos, A. Papacharalampopoulos, T. Shiotani, and D. G. Aggelis, “Dependence of AE parameters on the propagation distance,” J Acoust Emiss, vol. 29, pp. 57–67, 2011. calculation are shown in Table 3. It can be seen that event 2 gives higher NCC value than event 1. For example, by assess the Record 0 and 3, the NCC value in event 2 is 0.30 while in event 1 has NCC = -0.13. Figure 11 presents the waveform of Record 0 and Record 2 of event 2 with the normalized cross-correlation between the two records is NCC = -0.01. Similarity, Figure 12 demonstrates the waveform of Record 0 and Record 3 in event 2 with NCC = 0.3. 6. Comments and conclusions The filtering plays an important role in eliminating the signals that could be related to surrounding noise (low of count, duration and energy). It is about 50% of the raw signals have been removed from the classification processes. On the crack classification chart, signals concentrated in high AF areas exhibit damage mode I, which is consistent with RILEM TC 212-ACD. By observing the crack shape and also section in Figure 13, it is identified that the almost mode I cracks have pulled out the gravels and divided them in to two parts. The occurrence of shear stress (1.6%) when determining RA-AF from individual hits can be caused by damage at the interface between the aggregate and mortar (de-bonding), slip damage between the two materials is possible to generate shear mode. However, by determining the RA-AF from the mean value of 50 continuous hits, mode II is noticeably dissipated. The possible reason is that the number of mode I is negligible compared to mode II, thus by applying the average, mode II was filtered out. There are significant differences when comparing the waveforms of the signals that generate from the similar event. Although the signals share the same damage zone (mode I or mode II) but the correlation between waveforms and parameters varies considerably. This could be due to the influence of the transmission distance and the heterogeneous of material to the waveform. In [11], when the distance between the sensor and the event increases, the RA value increase while the AF, Amplitude and Energy decrease. However, this attenuation rule was not observed in the signals that received from tensile experiment; instead, these values fluctuate without identify the trend./. Figure 13. Crack shape at the notch (a) and plan view (b)