Evaluation of the liquefaction potential of soil deposits based on SPT and CPT test results

Transcription

1 Earthquake Resistant Engineering Structures V 43 Evaluation of the liquefaction potential of soil deposits based on SPT and CPT test results A. M. Hanna 1, D. Ural 2 & G. Saygili 1 1 Department of Building, Civil and Environmental Engineering, Concordia University, Canada 2 Department of Building and Civil Engineering, Istanbul Technical University, Turkey Abstract In the literature, predictions for the occurrence of nonlinear soil liquefaction in soil deposits have been investigated through numerous empirical methods. These methods which are also known as conventional techniques were derived from several in-situ tests, laboratory tests and case records. An alternative general regression neural network (GRNN) model that addresses the collective knowledge built in a simplified procedure is proposed. To meet this objective, a total of 3895 case records including twelve soil and seismic parameters driven mostly from the cone penetration test (CPT) results are introduced into the model. The data includes the results of field tests from the two major earthquakes that took place in Turkey and Taiwan in 1999 and some of the desired input parameters are obtained from correlations existing in the literature. The soil liquefaction decision in terms of seismic demand and capacity is determined by recognized simplified approach, namely a stress-based method and a strain-based method. Furthermore, the liquefaction probability of soils with significant fines is tested with the so-called Chinese Criteria. The proposed GRNN model is developed in four phases, mainly: the identification phase, collection phase, implementation phase, and verification phase. An iterative procedure was followed to maximize the accuracy of the proposed model. The case records were divided randomly into testing, training, and validation datasets. The proposed GRNN model effectively explored the complex relationship between the introduced soil and seismic input parameters and validated the liquefaction decision. Keywords: liquefaction, general regression neural networks, GRNN, Kocaeli earthquake, Taiwan earthquake, cone penetration test results, standard penetration test.

2 44 Earthquake Resistant Engineering Structures V 1 Introduction Monotonic and/or dynamic loading can incur liquefaction in soil deposits at any saturated soil under undrained conditions. By definition, soil liquefaction is a loss of shear strength due to the transfer of intergranular stress from grains to pore water [8]. The partial and complete loss of shear strength associated with liquefaction can be explained by a sudden increase in the pore water pressure as the effective stress reduces or becomes zero. In-situ based empirical methods are economical, enabling geotechnical engineers throughout the world to easily employ them. In the last two decades, artificial intelligence (AI) has been used in several applications in civil engineering because of its heuristic problem-solving capabilities. The Artificial Neural Network (ANN) is one of the AI approaches that can be classified as machine learning. It has the ability to simulate the learning capabilities of the human brain by automating the process of knowledge acquisition and data mining. Further work regarding CPT based soil liquefaction assessments using neural network methodology have been presented by Goh [3,4]. Goh [3] developed a back propagation neural network (BPNN) to assess liquefaction potential from CPT data. He reported that neural networks were proven to be feasible tools for soil liquefaction assessments, simpler to apply, and yield more reliable results when compared to conventional methodologies. It is noted that Goh had used a relatively small data set in developing the BPNN model, which is considered the main limitation of ANN approaches. However, these models could be improved as further field case records became available. In later studies, Goh [4] developed probabilistic neural network (PNN) models to analyze the databases based on cone penetration tests and shear wave velocity data. In this research, soil particle-size information was introduced as an input variable and the liquefaction phenomenon was considered a classification problem. Unlike the aforementioned research, in this study the methodology of using general regression neural network, which is a well recognized technique in solving and understanding many engineering problem, is applied to develop a predictive model for liquefaction triggering for soil deposits under seismic loading. In this proposed GRNN model, liquefaction decision is supported by related standard penetration test results. Accordingly, the viability of the SPT-to- CPT data conversion, which is the main limitation of simplified techniques, is verified. 2 Field data In 1999, two major earthquakes, namely Chi-Chi, Taiwan earthquake (magnitude M w =7.6) and Kocaeli, Turkey earthquake (magnitude M w =7.4) triggered ground failure principally induced by soil liquefaction throughout the city of Adapazari (Turkey) and the cities of Wufeng, Nantou and Yuanlin (Taiwan). The dataset is interpreted from the 43 CPT profiles and 38 SPT borings carried out for the Turkey Earthquake region and 27 CPT profiles and 25 SPT from the Taiwan

3 Earthquake Resistant Engineering Structures V 45 earthquake region. Additionally, conducted seismic CPT and spectral analysis of surface waves test (SASW) results are utilized [7, 11] 3 Artificial neural networks ANN ANNs are advanced tools stimulated by the physical and computational characteristics of the human brain. Like biological neurons, they consist of interconnected information processing neural elements (neurons) working in union to make decisions, classifications, and predictions. Neural networks are capable of learning linear and nonlinear functions that make them influential tools in the analysis of complex relations. Interconnections among neurons are established by weights, which are applied to all values passing through one neuron to another. Changing weights improves adaptabilities and prediction capabilities of neural networks. Neural networks are arranged in three or more layers according to their connection to the outside world. Through the learning process, input and output data of a specific engineering problem are given, and the aforementioned weights among neurons are updated without requiring human development of algorithms. The Neural Network performs well, as long as large volumes of data covering all possible governing parameters and field conditions are provided. 3.1 The general regression neural network (GRNN) GRNNs were first introduced by Specht [10] as an alternative to feed-forward Neural Networks. Unlike feed-forward neural networks, GRNN requires neither time-consuming trials nor over-training conditions. Based on nonlinear regression theory, the GRNN has special prediction capabilities for systems including sparse and noisy data. Unlike other kinds of neural networks, the GRNN does not require initial setting of learning parameters. Instead, a smoothing factor or bandwidth of all the parameters is calculated. The success of the network depends on the smoothing factor. As opposed to using a PNN to categorize data, the GRNN is a universal tool used to approximate smooth functions and to produce continuous outputs with only one pass through the training set [10] GRNN Model The proposed GRNN model was developed in four phases: identification, collection, implementation, and verification. An iterative procedure was followed to maximize the accuracy of the proposed model Identification phase In this phase, in order to provide necessary information with reasonable accuracy for site characterization, soil and seismic parameters affecting soil liquefaction potential were identified. Soil behaves as a deformable body with increasing depth; hence depth is a site condition parameter and has a huge impact on soil response against its capacity to resist liquefaction. For liquefaction to occur, ground water conditions should be sufficient to create

4 46 Earthquake Resistant Engineering Structures V saturated soil conditions. Thus depth to the groundwater table is an important consideration in identifying soils that are susceptible to liquefaction. Representing miscellaneous locations of soil deposits, depth of soil specimen and ground water table, which are real world parameters than can be obtained from various globally practiced field tests, are introduced into the proposed GRNN model. CPT penetration resistance has been commonly used for characterization of liquefaction resistance and CPT procedures have been developed by researchers to obtain information for liquefaction assessment. Thus, CPT test parameters, namely CPT tip resistance, CPT sleeve friction resistance, and CPT friction ratio are used as an index of liquefaction assessment model. It has been proven that increase in overburden stresses increases susceptibility of soils to cyclic liquefaction [9]. Accordingly, total and effective overburden stresses are brought into the model. Shear wave velocity values, earthquake magnitude and maximum horizontal acceleration at ground surface characterize nature of loading, intensity of seismic ground shaking induced by the earthquakes. Obtained from strong ground networks, they are real world parameters that are crucial to introduce into the seismic model. Conventional liquefaction potential assessment procedures profoundly rely on empirical correlations such as cyclic stress ratio for stress-based methodology and threshold acceleration for strain-based methodology. Accordingly, parameters derived from field tests are used directly through empirical calculations. These correlations derived from case histories and laboratory experiments yield reliable results. Dobry et al. [1] showed that liquefaction resistance of saturated undrained specimens against soil liquefaction can also be quantified by threshold shear strain, which is not dependent on the method of sample preparation, and it is approximately 0.01%. By utilizing this value, acceleration corresponding to threshold shear strain value is evaluated by the strain-based assessment procedure for liquefaction occurrence. Threshold acceleration value together with strain-based factor of safety states the exceedance of the threshold strain, which is required for liquefaction to occur. In order to evaluate the threshold acceleration, easily obtained field tests variables and earthquake characteristics are utilized. Therefore, threshold acceleration can be considered as a real world parameter, which contributes strain-based procedure knowledge into the complex liquefaction problem. Moreover, formulated by Seed and Idriss [8], cyclic stress ratio is a normalized measure of cyclic load severity and it represents the seismic demand on soil to liquefy. In that respect, threshold acceleration and cyclic stress ratio values are introduced into the model Collection phase The data set consists of 3895 case records: 1812 for the Kocaeli earthquake and 2083 for the Taiwan earthquakes. The data set represents 1665 cases that liquefied and 2230 cases that did not liquefy. The authors believe that better predictions can be achieved by employing all the governing soil and seismic parameters in one model. Therefore, in this study, a comprehensive approach that consists of all independent soil and seismic parameters proven to be influential for decision making are taken into account.

5 Earthquake Resistant Engineering Structures V 47 In that respect, the authors believe that it is valid to include input variables coming from different concepts. On the other hand, presenting alternative parameters to the model improves prediction capabilities of neural networks that have thought-provoking parameter recognition capabilities. The database, summary of which is given in table 2 covers a wide spectrum of soil and seismic parameters. In the first column of table 2, the depths of the soil specimens are given, and were extracted at depths of up to 20 meters from the ground surface. Total and effective vertical stresses are presented in columns 2 and 3, respectively. The depth of the ground water table is given in column 4. In table 2, CPT test outcomes for cone tip resistance, sleeve friction resistance, and friction ratio values are presented in columns 5, 6 and 7, respectively. In column 8, the shear wave velocities, which were actual measured values from the related seismic cone penetration test (SCPT) and spectral analysis of surface wave test (SASW) was presented. The threshold acceleration values were calculated following the procedure given by Dobry et al. [1] and are given in column 9. In column 10, cyclic stress ratios required to generate soil liquefaction were tabulated [8]. Column 11 presents the moment magnitude of the Turkey and Taiwan earthquakes. Maximum horizontal accelerations at ground surface are given in column 12. The distribution of the peak accelerations recorded by strong motion stations are given in table 1 below. In table 2, the closest measurements of all available strong ground motion recorders are taken into account in order to characterize the region realistically. Table 1: Maximum horizontal accelerations recorded by strong motion stations. AREA TAIWAN TURKEY LOCATION Nantou Wufeng Yuanlin Adapazari City VALUE 0.38 (g) 0.67 (g) 0.18 (g) 0.4 (g) STATION TCU076 TCU065 TCU110 NSMP* * National Strong Motion Network Liquefaction decision Column 13 in table 2 presents the model s output for the liquefaction resistance of the soil deposits. The continuous soil profile obtained by CPT results lacks the detailed definition of soils as proposed by Youd et al. [12]. Furthermore, Seed et al. [9] concluded that SPT based correlations have been better defined, and have provided lesser levels of uncertainty than CPT based simplified procedure due to impracticality of sampling in the CPT. In accordance with the aforementioned resources, the authors believe that SPT test results supported by shear wave velocity measurements are capable of accurately identifying and predicting liquefaction potential of liquefiable soils. Accordingly, liquefaction potential of soil deposits has been evaluated by means of an assortment of empirical criteria including SPT based simplified method. Liquefaction decision is evaluated by using SPTbased simplified method in addition to strain-based approach and Chinese

6 48 Earthquake Resistant Engineering Structures V Criteria. To meet this objective, the SPT based liquefaction prediction is transferred to all case records in the CPT data set through the profile. Based on the three following criteria; (i) stress-based liquefaction triggering analyses developed for SPT tests [12] (ii) strain-based procedure [1], and (iii) Chinese criteria [2] the occurrence of soil liquefaction results was evaluated. Based on the Youd et al. [12] SPT-based triggering correlations, relationship between cyclic stress ratio and corrected standard penetration resistance for sands and silty sands with varying peak ground acceleration and earthquake magnitudes are evaluated. Cyclic resistance ratios for 7.5 magnitude earthquakes are determined from the results of corrected SPT (corrected N values) boreholes close to investigated CPT soundings. Moreover, soils with significant plasticity are evaluated for liquefaction susceptibility by the Chinese Criteria proposed by Finn et al. [2]. This criteria specifies that soils with plastic fines can liquefy if all the four following conditions are met: (1) percent finer than mm 20%, (2) (liquid limit + 1%) 35%, (3) water content +2%) 0.9, (4) liquidity index (based on liquid limit + 1% and water content +2%) For strain-based liquefaction susceptibility evaluation, as given in eqn (1), the threshold acceleration can be calculated as follows; 2 G Vs t a G γ t max = (1) g 0.65gzr where V s is the shear wave velocity, γ t is the threshold strain, (G/G max ) is the modulus reduction factor at the threshold, r d is the stress reduction coefficient, and M is the magnitude of the earthquake. The modulus reduction factor at the threshold, (G/G max ) t is assumed to be 0.8 with a strain level of an order of 0.01% [5]. Factor of safety in threshold acceleration criteria is given in eqn (2) below; at F = (2) a d a max F a implies that there is no risk, F a 1 not necessarily mean that liquefaction will occur; instead the value predicts that there will be gross sliding of the grain-tograin contact surfaces which is essential for excess pore water pressure generation and therefore crucial for liquefaction Implementation phase In this study, the neural network software (NeuroShell 2) developed by the Ward Systems Group (United States) was used for training, testing, and validating the GRNN model. The data set was randomly categorized into three subsets: training, testing and validation. The training set generates the algorithm with the best smoothing parameter. The testing set is used for observing the capabilities of the generated algorithm to assess the intricate relationships amid input and output values. The validation set is used for applying the trained algorithm to a separate data set that was not previously introduced to the network. Therefore, the validation set can be regarded as the true/genuine test for the performance of the model. In this investigation, 2922 of these case records were randomly selected for the development phase, 521 for the testing phase, and the remaining 452 records were selected for forecasting.

7 Earthquake Resistant Engineering Structures V 49 Table 2: Summary of CPT dataset for GRNN model. # Parameters z σvo σ'vo dw qc fs Rf Vs at τav/σ'vo Mv amax Liquefaction GRNN Unit (m) (kpa) (kpa) (m) (kpa) (kpa) (-) (m/s) (g) (-) (-) (g) (-) (-) Summary of testing phase Summary of training phase Summary of validation phase no no yes yes no yes* yes yes no no yes yes no no yes yes yes yes yes yes no no no no yes yes no no yes yes no no no no yes yes yes yes no no no no no no yes no* *Indicates error in the dataset

8 50 Earthquake Resistant Engineering Structures V The proposed GRNN model includes 12 neurons in input layer, which also, corresponds to the total number of input parameters. The output layer includes one neuron to predict liquefaction. The liquefaction potential is characterized by output neuron 1, where the binary number 1 represents the occurrence of liquefaction and 0 represents the non-occurrence of liquefaction. The GRNN approach used for soil liquefaction potential estimations with a smoothing factor of σ= The error limit for the analysis is taken as 0.3 (i.e., 30%). This means that with a difference between the target output and network result greater than 0.30, the result is considered incorrect. In the parameter sensitivity study, most influential parameters impacting liquefaction assessment are outlined and given in fig. 1. Figure 1: Parameter sensitivity study Validation phase In this phase, model accuracy and efficiency were examined by making predictions based on the case records that were not used during model training and testing. The validation data was randomly selected from the Kocaeli, Turkey, and Chi-Chi, Taiwan earthquake, and corresponds to approximately 12% of the data set. In this phase, the proposed algorithm does not require human development; instead, it confirms the architecture s prediction capabilities of the model. The results of this phase are also given in table 2 and summarized in table 3. Table 3: Results of GRNN model. Overall Taiwan Earthquake Adapazari Earthquake # of Perfor # of Perfor # of Perfor Error Error Error Case mance Case mance Case mance Train % % % Test % % % Forecast % % % Total % % % As depicted in fig. 2., the results produced by the proposed GRNN model compare well with the evaluated results. It provides a viable liquefaction potential assessment tool that assists geotechnical engineers in making accurate and realistic predictions.

9 Earthquake Resistant Engineering Structures V 51 Figure 2: Network s performance on soil liquefaction potential. 4 Conclusion The impact of soil and seismic parameters on soil liquefaction potential in soil deposits has been investigated. An artificial intelligence computational approach General Regression Neural Network (GRNN) was developed to predict liquefaction potential in soil deposits. A total of 3443 case histories were used for the analysis and 452 case histories were used to make predictions. As case records, the proposed model utilizes reliable soil test results that were obtained following the two critical earthquakes in Turkey and Taiwan. In this proposed model, determination of SPT based liquefaction potential, is incorporated into CPT based soil and seismic data. Therefore, the model verifies the feasibility of an SPT-to-CPT data conversion throughout the liquefaction potential analysis. In other words, the model gives way to addressing soil and seismic parameters based on SPT and CPT into one Neural Network analysis, which believed to be the primary limitation of the simplified techniques. Furthermore, a sensitivity analysis was conducted to determine the parameters that govern this behavior. Twelve soil and seismic parameters that characterize soil type and material properties, seismic characteristics, magnitude and nature of loads and other site conditions including stress, strain, strength, saturation and seismological aspects were incorporated into the network. Contrary to conventional methods for analysis of liquefaction, this study has integrated input parameters that account for all possible variations in the field. It is believed that this study, which integrates and transfers knowledge between two severe earthquakes will contribute to the ongoing development of soil liquefaction analysis. Acknowledgements The financial support from the National Science and Engineering Research Council of Canada (NSERC) and Concordia University are acknowledged.

10 52 Earthquake Resistant Engineering Structures V References [1] Dobry, R., Ladd, R.S., Yokel, F.Y., Chung, R.M. & Powell, D., Prediction of pore water pressure buildup and liquefaction of sands during earthquakes by the cyclic strain method NBS Building Science Series 138, U.S. Dept. of Commerce, 152 p, [2] Finn, W.D.L., Ledbetter, R.H., & Wu, G., Liquefaction in silty soils: Design and analysis, Ground Failures under Seismic Conditions. Geotechnical Special Publication 44, ASCE, New York, 51-76, 1994 [3] Goh, A.T.C., Neural Network modeling of CPT seismic liquefaction data, J. Geotech. Eng., 122(1), 70-73, 1996 [4] Goh, A.T.C., Probabilistic neural network for evaluating seismic liquefaction potential. Canadian Geotechnical Journal, 39, , [5] Hardin, B.O. & Drnevich, V.P., Shear modulus and damping in soil: measurement and parameter effects. Journal of the Soil Mechanics and Foundations Division, ASCE, 98, , [6] NeuroSHELL 2 user s manual version 3.0, Frederick M.D., Ward Systems Group, Inc., USA, [7] PEER, Pacific Earthquake Engineering Research, [8] Seed, S.B. & Idriss, I.M., Simplified procedure for evaluating soil liquefaction potential. J. Soil Mech. Found. Div., 97(9), , [9] Seed, R.B., K.O. Cetin, R.E.S. Moss, A. Kammerer, J. Wu, J.M. Pestana, M.F. Riemer, R.B. Sancio, J.D. Bray, R.E. Kayen & A. Faris, Recent Advances in Soil Liquefaction Engineering: A Unified and Consistent Framework, Keynote Address, 26th Annual Geotechnical Spring Seminar, Los Angeles Section of the GeoInstitute, American Society of Civil Engineers, H.M.S. Queen Mary, Long Beach, California, USA, 2003 [10] Specht, D.F., A general Regression Neural Network, IEEE Transactions on Neural Network, 2(6), , [11] Stewart, J.P., Chu, D.B., Lee, S., Tsai, J.S., Lin, P.S., Chu B.L., Moss R.E.S., Seed R.B., Hsu S.C., Yu M.S., & Wang M.C.H., Liquefaction and non-liquefaction from 1999 Chi-Chi, Taiwan, earthquake, Technical Council on Lifeline Earthquake Engineering, Monograph No. 25, J.E. Beavers (ed.), pp , [12] Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Finn, W.D., Harder, L.F., Hynes, M.E., Ishiara, K., Koester, J.P., liao, S.S.C., Marcuson, W.F., Martin, G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed, R.B., & Stokoe, K.H., II., Liquefaction resistance of soils: Summary report from the 1996 NCEER and 1998 NCEER/NSF Workshops on evaluation of liquefaction resistance of soils, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 127(10), , 2001.