Developing a New Methodology for Characterization of In-SituSoil Properties

本研究旨在提高岩土工程师现场表征土性的能力。利用选定的国家岩土工程试验场(NGES)的几种现场试验数据，探讨了人工神经网络在场地表征中的潜力。提出了一种新的方法，它利用神经网络泛化的“非结构化”知识和以本地经验、工程判断和专家规则的形式获得的“结构化”知识。对神经网络方法和概率方法的计算结果进行了比较。利用所开发的神经网络进一步探讨了几个问题：1)设计土壤参数之间的相关性，2)不同现场试验获得的设计土壤参数的交叉校核，3)不同现场试验得到的土壤参数剖面的平均，4)可能减少所需的“硬”数据的数量，5)概率神经网络学习。然后将研究结果综合成场地特征描述的设计方法。建立并记录了应用此设计方法的指导方针和程序。/元？？┬║？╝╣_？╣└？%└╣？元╝╣║？┤▓/？%/╝└/？(？╝？角/┬/元╣%╣？/┤元？？│？元？║？╝？(？_╣？║？╣？%？╣┤`╣/║╝？┤/元？╝？/┤▓╝╣？？╝╣_║/┐║%╣╣á╝？？╣/？║/┤│？║？║？┤╣元？？┐╝元？╝/？║/？║╝？？└╣？╝▓/？┤？元╣_/元？？│？┐元？╝/原媛/│，/┤á┤╣？╝/元？┴╣%%▓？│？_║/╝？┤┴╣元？╣┤？║？/元┤？元╣_/元？┤？？//%`：？？元？？╣？/%╝元？║？║╝？┤/元？╝？┴╣元？╣？╣元？│？/_▓？╝

This research is aimed at improving the Geotechnical Engineer's capability of characterizing in situ soil properties. The potential of artificial neural networks in site characterization is explored using several types of in situ test data from selected National Geotechnical Experimentation Sites (NGES). A new methodology is developed that utilizes both the `unstructured` knowledge generalized with neural networks and the `structured` knowledge is available in the form of local experience, engineering judgment, and expert rules. Comparison is made between the results obtained from the neural network approach and the probabilistic approach. Several issues are further explored using developed neural networks: 1) correlations among design soil parameters, 2) cross-checking design soil parameters obtained from different in situ tests, 3) averaging the profiles of soil parameters from different in situ tests, 4) possible reduction of the number of `hard` data needed, and 5) probabilistic neural network learning. The research findings are then synthesized into a design methodology for site characterization. Guidelines and procedure for applying this design methodology are established and documented. /¥Á ÂÁÁ┤╣À ╝/¥Á¢ Â╝?_ └╣┤Á? ╝Á│?╝┤╣À¢ /┤ ┤ ó/ ┴Á /││┐╝/¥Á%` _Á/¢┐╝Á ║╝Á` ▓Á©/└╣?╝ ┴╣¥© ?┐╝ └╣┤Á? ¢`¢¥Á_ ¼©Á¢Á ╗┐Á¢¥╣?¢ ┴╣%% ▓Á /┤┤╝Á¢¢Á┤ ╣ / ¢Á╝╣Á¢ ? │?¥╝?%%Á┤ Á┬║Á╝╣_Á¥¢ ╣└?%└╣À ¢¥╝╣║Á┤ ▓/¢¢ %/╝└/Á (?╝?Á ¢/┬/¥╣%╣¢ /┤ ¥©Á │¥Á?║©?╝Á (Á_╣?║¢╣¢ %Á╣┤`╣ /¢ ║╝Á┤/¥?╝¢ /┤ ▓╝╣Á ¢©╝╣_║ /┐║%╣╣ á╝¥Á_╣/ ¢║ /┤ │?║Á║?┤╣¥Á¢ ? ñ┐╝¥Á_?╝/ ¢║ /¢ ║╝Á` íÁ©/└╣?╝ ▓/¢Á┤ Á¢¥╣_/¥Á¢ ? Á│?┐¥Á╝ /¥¥/│, /┤ ÂÁÁ┤╣À ╝/¥Á¢ ┴╣%% ▓Á │?_║/╝Á┤ ┴╣¥© ╣┤Á║Á┤Á¥ Á¢¥╣_/¥Á¢ ┤Á╝╣└Á┤ Â╝?_ //%`:Á¢ ? ¥©Á Â┐│¥╣?/% ╝Á¢║?¢Á¢ ? ║╝Á┤/¥?╝¢ ┴╣¥©╣ ¥©Á │©/_▓Á╝