论文部分内容阅读
针对岩石抗剪强度确定中图解法和最小二乘法存在的问题:①只适用于实验数据相关性较高的情况,不适用于数据离散并有异常值存在的情况;②在处理由试验法获得的试验数据时,最小二乘法由于采用残差平方和,容易夸大试验数据中异常值的影响,提出了岩石抗剪强度参数的稳健回归分析方法。该方法在实验数据相关性差、数据离散并有异常值存在的情况下,具备削弱数据离散和对异常值进行定位的能力,提高了估计参数的稳健性和可靠性。该方法以残差的绝对值之和代替残差平方和,并通过复形法求得力学参数,避免了异常值的二次项,可有效地减少异常值的影响。通过工程实例表明,在试验数据的相关性较好时,两种方法的计算结果相差不大,但当试验数据的相关性较差,存在异常值时,稳健回归方法的计算结果要优于最小二乘法。
In order to solve the problems of the graphic method and the least square method in the determination of rock shear strength, ① it is only suitable for situations with high correlation of experimental data, not suitable for data discretization with abnormal values; , The least squares method is easy to exaggerate the influence of outliers in the test data due to the square sum of residuals. A robust regression analysis method of rock shear strength parameters is proposed. This method has the ability of weakening data discretization and locating outliers under the condition of poor correlation of experimental data, discrepancy of data and outliers, and improves the robustness and reliability of the estimated parameters. In this method, the residual sum of squares is replaced by the sum of absolute residuals, and the mechanical parameters are obtained by the complex method to avoid the quadratic term of abnormal values, which can effectively reduce the influence of outliers. The engineering examples show that the results of the two methods have little difference when the correlation of the experimental data is good, but the results of the robust regression method are better than the minimum when the correlation between the experimental data is poor and there are abnormal values Two multiplication.