BP人工神经网络在青藏铁路南段地壳稳定性定量评价中的应用

Applicatin of BP Artificial Neural Network to Quantitative Assessment on Crust Stability along the Golmud-Lhasa Railway across South Tibetan Plateau

将BP人工神经网络方法引入区域构造活动性、区域地壳稳定性研究领域，对青藏铁路南段沿线的构造活动性进行定量分析。选用断层运动速率、地震震级、温泉温度及剪切应变4个关键影响因子作为BP人工神经网络的输入向量，构造活动强度（α）作为输出向量，以α为定量判据，将全区划分为相对稳定区（α〈0．22）、较不稳定区（α≈0．22～0．38）、不稳定区（α≈0．38～0．69）、极不稳定区或强烈构造活动区（α≥0．69）。在青藏铁路南段沿线划分出格仁错、崩错、当雄一羊八井、错那湖、唐古拉山口南、聂荣东北、聂荣西北、雅鲁藏布江断裂沿线、萨迦等不稳定区，在不稳定区内部进一步划分出申扎、蓬错、尼木、桑雄、羊八井5个极不稳定区。

BP artificial neural network is introduced to study tectonic activities and crust stability and applied into the quantitative analysis on tectonic activities along the Golmud-Lhasa Railway across south Tibetan Plateau. For key factors, slip rate of active fault, magnitude of earthquake, temperature of hotspring and tectonic strain for any unit in south Tibetan Plateau,are selected as the input vector in BP artificial neural network,and strength of tectonic activity （α） as the output vector. According to the values of α , south Tibetan Plateau can be classified into four regions.tectonically stable area with α〈0.22,tectonically relatively unstable region with α≈0. 22-0. 38, tectonically active region with α≈0. 38--0.69 and tectoni- cally quite active region with α≥0.69. Typical unstable regions include Gerencuo, Bengcuo, DangxiongYangbangjing,Cuonan lake,southeast Tangula Mts. ,Sajia unstable regions and those along Yangluzangbu River. While five extremely unstable regions, Shenzha, Pengcuo lake, Nimu, Sangxiong, Yangbajing, are identified along the Golmud-Lhasa Railway across south Tibetan Plateau.