工程地质趋势面分析的智能方法及其应用

鉴于神经网络和遗传规划的自适应、自学习功能,将径向基函数神经网络方法与混合进化方法引入到趋势面分析当中,在一定程度上可以避免传统地质体数学模拟时需要事先选定未知模型类型的弊端,并通过它们特有的学习和进化机制,去寻求一个能较好地描述和解释地质体空间展布特征(趋势)的数学表达形式.通过实例研究,发现RBF神经网络方法拟合和外推插值的效果最好;多项式趋势面法和混合进化方法拟合效果相差不大,但后者的外推插值效果更好.

Intelligent Methods and Its Application for Geological Trend Surface Analysis

The traditional geological trend surface analysis methods have the demerits of low precise and selecting the simulated function without definitely guide. For the artificial neural networks and hybrid evolution methods have property of the self-adaptive and self-study, the RBF networks and hybrid evolution are brought into the geological trend surface analysis. These methods can overcome the shortcomings of selecting the functions in advance of traditional geological trend surface analysis methods, and through their specially study and evolution mechanism to search a better mathematical express to describe and express the geological body's space distributing character. Through case study, the RBF networks have the best effect of simulation and extrapolate, and the effect of simulation are similar between geological trend surface analysis and hybrid evolution, but the effect of extrapolate for hybrid evolution is better than geological trend surface analysis.