基于BP神经网络饱和砂土液化判别方法_李方明.pdfVIP

14 2 V o.l 14, No. 2 2005 4 JOURNA L OF NATURAL D ISA STER S A pr. , 2005 : 1004- 4574 ( 2005 02- 0108- 07 BP 李方明, 陈国兴 ( , 2 10009 : , ( , 11, BP LM BP , : ( 1 14 BP , ; ( 2 ( BP , BP ; ( 3 BP , LM BP, , LM ; ( 4 BP , ; ( 5 , 56 , BP M 4 M 5a, , 5 6 : ; BP ; LM ; BP; : P642 : A Saturated sand liquefaction potential estimation m ethod based on BP neural network L I F ang-m ing, CHEN Guo-x ing ( Inst itu te of G eotechn ical Engin eering, N anjing U n iversity of T echnology, N anjing 2 10009, Ch ina Abstract: Based on a lot o f sandy so il liquefaction data obtained during T angshan earthquake in 1976, se lect ing di-f ferent com b inations of 11 indexes as input variations, the faster algo rithm of BP and LM a lgorithm are adopted to construct the BP neural newt ork m odel for liquefact ion po tentia l estmi at ion o f saturated sand. T he 11 indexes are earthquake intensity, the epicenter d istance, peak ground acce lerat ion for descr ib ing the earthquake ground m otion character istics, and the depth of underground w ater, the depth o f standard penetration test point( so il layer , the th ickness o f the covered non- liquefied so il layer, effectively overlay ing pressure describ ing the em bedd ing env iron- m ent of soil layers, and standard penetrat ion b low- count, m ean d iam eter, nonun iform ity coefficient, m od ified standard penetration b low-counts for denoting the sandy soil feature. Through the pract ice and verification of the m ode,l