岩体力学参数Bayes推断中验前信息的数量与融合技术研究

针对岩体力学参数Bayes统计推断时存在的多源验前信息问题,将验前信息分为数值模拟信息、相似材料信息、专家意见和历史信息四类。对验前信息属同一总体或不同总体的情况进行均值和方差估计的推导;在试验样本数量一定的情况下对验前信息的最优数据量进行探讨,得出试验样本为6的情况下,仿真验前信息量在46～150之间最好,建议取50。用实例进行验前信息的加权融合计算,结果表明加权信息融合技术为Bayes推断提供了方便。

Study on the Prior Information Quantity and Fusion Technology Applied in Inference of Rock Mechanics Parameters with Bayes Method

In view of the multi-source prior information existing in statistical inference of rock mechanics parameters by the Bayes method,the paper divided the prior information into four categories: the numerical analog information,the similar material information,the expert advice and the historical information.Meanwhile,the paper made inference on the average value and the variance of prior information under identical population and different population.In the case that the experimental sample quantity was definite,the paper discussed the most superior data quantity of the prior information,and concluded that,when the experimental samples were 6 in number,the best data quantity of prior information was between 46 and 150,and 50 was proposed.Furthermore,the paper carried out weighted fusion calculation of prior information by example,which indicated that the weighted information fusion technology provided convenience for inference by the Bayes method.