تخمين ظرفيت باربري نوك شمع بر مبناي اطلاعات CPT با استفاده از شبكه‌هاي عصبي GMDH

Estimation of Point Bearing Capacity of Single Pile from CPTu Data Using GMDH Type Neural Networks

آزمايش نفوذ مخروط (CPT) يك مدل كوچك‌‌مقياس شمع است كه براي تعيين ظرفيت باربري شمع‌هاي واقعي مورد استفاده قرار مي‌گيرد. (GMDH) يك نوع شبكه‌ي عصبي است كه ساختار آن توسط الگوريتم ژنتيك بهينه‌سازي شده است. در اين تحقيق داده‌هاي 29 آزمايش بارگذاري استاتيكي و ديناميكي شمع و اطلاعات CPT مجاور آنها جمع‌آوري شد. با كمك اين داده‌ها، رابطه‌يي بر مبناي GMDH براي تخمين ظرفيت باربري نوك شمع ارايه و تاثيرات qE (مقاومت موثر نوك مخروط) و fs (مقاومت غلاف مخروط) در مقاومت واحد نوك شمع بررسي شده است. مقايسه‌هاي انجام‌شده با روابط مستقيم تعيين ظرفيت باربري شمع بر مبناي CPT و CPTu نشان‌دهنده‌ي دقت بسيار مناسب رابطه‌ي جديد پيشنهادي براي تخمين ظرفيت باربري نوك شمع است.

Piles have been used for many years as a type of structural foundation. The design of pile foundations and the estimation of static pile capacities based on measured soil properties have improved considerably over the years. However, due to the inherent soil variability and the disturbance, there is always an element of uncertainty about the design capacity. Therefore, most theoretical approaches have mainly been based on simplifications and assumptions. The cone penetration test (CPT) is considered one of the most useful in situ tests for the characterization of soils. Due to the similarity between the cone and the pile, the determination of pile capacity from the CPT data is among the earliest applications of the CPT. The measured cone resistance (qc) and sleeve friction (fs) usually are employed for estimation of pile unit toe and shaft resistances, respectively. Over the last few years or so, the use of artificial neural networks (ANNs) has increased in many areas of engineering. In particular, ANNs have been applied to many geotechnical engineering problems and have demonstrated some degree of success. Group method of data handling (GMDH) type neural networks optimized using genetic algorithms (GAs) are used to model the effects of effective cone point resistance (qE) and cone sleeve friction (fs) as input parameters on pile unit toe resistance, applying some experimentally obtained training and test data. 29 pile case histories have been compiled including static and dynamic loading tests performed at sites including CPT sounding. The pile embedment lengths range from 9 m through 31 m. The pile unit toe resistances range from 0.4 MPa through 29.4 MPa. Sensitivity analysis of the obtained model has been carried out to study the influence of input parameters on model output. According to the sensitivity analysis results the pile unit toe resistance (rt) is considerably is influenced by the effective cone point resistance (qE), and the value rises by increasing qE value. Also, for a constant value of the effective cone point resistance (qE), by decreasing the cone sleeve friction (fs) the pile unit toe resistance increases. Pile toe capacities calculated by the proposed method are compared to toe capacities calculated by five other direct methods. The proposed method gives values that are more consistent and closer to the measured than the current methods. The results demonstrate that the proposed method gives values that are consistent and close to the measured.