Bandwidth Selection of Normal Information Spread Estimation Method for Geotechnical Parameter Probability Distribution

Abstract

Normal information spread estimation method (NISEM) is an effective method to infer the geotechnical parameter probability density function. However, it is found that the optimal bandwidth calculated by the existing method is not consistent with the actual situation in engineering application. In order to determine the matching bandwidth of NISEM, the histogram grouping number and group distance according to the sample quantity and range are first determined in this paper. Then the bandwidth h of NISEM is calculated, taking h as the benchmark, 0.1h as the step to change bandwidth. And combining with judgment coefficient, the optimal matching bandwidth of NISEM is finally obtained. The validity and rationality of the proposed method are verified by geotechnical test data of two groups of large samples and four groups of small samples in practical engineering. The results show that: under the condition of limited geotechnical samples, the best matching bandwidth is related to the number of histogram groups, which is relative. And the bandwidth of each group of test samples is different. The research work in this paper perfects the application of NISEM in inferring geotechnical parameter probability density function.