用改进后的BP神经网络评价黄土质边坡稳定性

首先介绍了改进BP神经网络性能的几种方法；在此基础上，考虑影响黄土质边坡稳定性分析的各种自然因素，包括坡高，坡比，强度指数，土何内摩擦角，土体容重，空隙水压力系数以及地震烈度统方法的计算结果进行逐一对比，对比结果证明了该方法能够满足一般黄土质边坡稳定性评价的精度要求；另一方面，由于方法的改进大大减少了网络的计算时间，使得黄土质边坡稳定性的评价更为便捷迅速，从而证明该方法具有一定的推广应用价值。

ESTIMATION OF LOESS SLOPE STABILITY WITH IMPROVED BP NEURAL NETWORK

Several methods used in improving BP neural network are introduced in this paper. Based on the data collected gathered from projects in typical loess areas, the stability of typical loess slopes is assessed with consideration of various natural factors, such as slope height, grade, intensify index, friction angle, bulk density, pore water pressure and earthquake intensity. Comparing the assessment results with those gained by traditional calculation suggests that that this method of assessment is accurate and convenient, and worth being popularized.