@article { author = {Manoochehri Kian, Mohammad and Khandan Bakavoli, Mohammad}, title = {Site Classification Based on H/V Response Spectra, Using Image Processing and Neural Networks}, journal = {Bulletin of Earthquake Science and Engineering}, volume = {8}, number = {2}, pages = {99-112}, year = {2021}, publisher = {International Institute of Earthquake Engineering and Seismology}, issn = {2476-6097}, eissn = {2476-6100}, doi = {10.48303/bese.2021.245590}, abstract = {In order to estimate the seismic hazard of a specific site, the classification of that site is of particular importance.On the other hand, in order to interpret and analyze the ground motion data in different parts of the world, it isnecessary to know the site conditions in seismic stations. In some countries, including Iran, there is insufficientinformation on the geotechnical and geological status of many seismic stations. The conventional methods tocharacterize the site are based on shear wave velocity measurement such as SCPT measurement, downhole testing,and seismic refraction. These methods have some limitations such as costs, maximum depth, execution problems,etc. This research is a new and efficient approach in site classification using the data recorded from the seismicnetworks, image processing techniques, neural networks and set of 5% damping spectral ratio reference curves ofhorizontal to vertical component (H/V) for the four different site classifications. These reference sets, that includefour separate H/V curves for four different site conditions labelled as rock, dense soil, medium soil and soft soil andclassified as site I, II, III and IV, have been selected from the study of Zhao et al. [1]. The reference curves are basedon K-net seismic network data. The adopted soil classifications are based on Japan Road Associationrecommendations. For the periods of interest, which were not presented in the Zhao et al. [1], the curves wereinterpolated to come up with the values at the missing periods.In this research, two types of basic radial functions (RBF) are called "probabilistic neural network (PNN)" and"general regression neural network (GRNN)", as well as "convolutional neural network (CNN)" have been used. Forneural network input, the data from 182 seismic stations have been incorporated. The site condition at the location ofeach station has been fully characterized. The horizontal to vertical spectral ratio for each recorded seismic eventwas calculated. The ratio for each data was smoothed using the moving average function. Then, the smoothed H/Vratio was normalized to match the sigmoid transfer function upper and lower range, which could minimize thenetwork training time. For the CNN network, the input H/V spectral ratio images were first unified using the exactdimension of 150×300 pixels and then compared to the reference H/V spectral ratio using image processingtechniques implemented in MATLAB software.To verify the proposed technique, H/V spectral ratio was calculated for all events recorded at all 182 stations andthen used as input for training the PNN, GRNN and CNN networks and then compared to the reference curvesproposed by Zhao et al. [1]. Two normalization methods were incorporated; in the first method, all the H/V spectralratios normalized to the maximum amplitude, and the second was to normalize the maximum to one and minimumto zero. The results confirmed that the second normalization method could produce more accurate results due to abetter matching the sigmoid function.According to the obtained results incorporating the second method of normalization and all 790 ground motiondata, which were recorded at 182 different stations., the PNN, GRNN and CNN networks have succeeded inaccurately predicting the site conditions in 73%, 71% and 81% of the stations, respectively. The results could provethe applicability of the proposed approach, using neural networks, in site characterization.References1. Zhao, J.X., Irikura, K., Zhang, J., Fukushima, Y., Somerville, P.G., Asano, A., Ohno, Y., Oouchi, T., Takahashi, T. and Ogawa, H. (2006) An Empirical site-classification method for strong-motion stations in japan using H/Vresponse spectral ratio. Bulletin of the Seismological Society of America, 96, 914-25.}, keywords = {Site Effect,Neural network,image processing,Site Classification}, title_fa = {طبقه‌بندی ساختگاه به کمک پردازش تصویر و شبکه‌های عصبی و بر اساس طیف‌های پاسخ H/V}, abstract_fa = {به‌منظور برآورد خطر لرزه‌ای یک ساختگاه مشخص، طبقه‌بندی آن ساختگاه از اهمیت ویژه‌ای برخوردار است. از سوی دیگر به‌منظور تفسیر و تحلیل داده‌های ثبت شده از حرکت زمین در مناطق مختلف جهان، شناخت شرایط ساختگاه در ایستگاه‌های لرزه‌نگاری ضروری می‌باشد. در برخی از کشورها از جمله ایران اطلاعات کافی از وضعیت ژئوتکنیکی و زمین‌شناسی در بسیاری از ایستگاه‌های لرزه‌نگاری وجود ندارد. از این‌رو شرایط ساختگاه در این مناطق در دسترس نمی‌باشد. این پژوهش به رویکردی جدید و کارآمد در طبقه‌بندی ساختگاه بر اساس داده‌های ثبت شده از شبکه لرزه‌نگاری و با استفاده از تکنیک‌های پردازش تصویر و شبکه‌های عصبی و به‌کارگیری مجموعه‌ی مرجع از منحنی‌های نسبت طیفی 5 درصد میرا شده مؤلفه افقی به عمودی (H/V) برای چهار نوع ساختگاه می‌پردازد. این مجموعه‌ی مرجع که شامل چهار منحنی H/V برای چهار نوع ساختگاه مختلف با نام‌های سنگ، خاک متراکم، خاک متوسط و خاک نرم و با طبقه‌بندی I، II، III و IV می‌باشد، از مطالعه ژائو و همکاران [1] انتخاب شده است. در این پژوهش از دو نوع تابع شعاعی پایه (RBF) به نام‌های «شبکه عصبی احتمالی (PNN)» و «شبکه عصبی رگرسیون عمومی (GRNN)» و همچنین «شبکه عصبی کانولوشنی (CNN)» استفاده شده است. با توجه به نتایج به‌دست‌آمده مشاهده می‌شود که شبکه‌های PNN، GRNN و CNN در پیش‌بینی درست شرایط ساختگاه با استفاده از داده‌های زلزله در بهترین حالت به‌ترتیب در 73، 71 و 81 درصد ایستگاه‌ها موفق عمل کرده‌اند.}, keywords_fa = {اثر ساختگاه,شبکه عصبی,پردازش تصویر,طبقه بندی ساختگاه}, url = {http://www.bese.ir/article_245590.html}, eprint = {http://www.bese.ir/article_245590_9f69c4d9ab4eec5ac4a19155013d9e42.pdf} }