杭长客专浙江段沉降预测精确度评价建议模式

Model proposed for assessing the accuracy of the settlement forecast of Zhejiang section of Hangzhou-Changsha passenger dedicated line

在杭长客专浙江段沉降评估实践中发现,用常规的4种拟合算法得出的相关系数普遍偏低,绝大部分测点小于0.92。分析了相关系数普遍偏低的原因,并引入了3种新的评估指标:平均绝对误差MAD、均方误差MSE和平均绝对百分误差MAPE。通过与相关系数的对比分析表明,即使在相关系数很低的情况下,沉降预测模型也可以获得足够高的精确度。因此,采用MAD、MSE和MAPE的组合模式来代替相关系数,从而建立针对沉降"小量级、大波动"特点的沉降预测精确度评价建议模式。对于沉降预测结果同时满足MAD≤0.11mm、MSE≤0.015 mm2和MAPE≤8%的观测点,可以认为其沉降预测精确度满足要求。研究表明,此建议模式在杭长客专浙江段具有很好的适用性,并显著提高了满足沉降预测精确度要求的测点百分比。

This article is inclined to propose a simulated model for making accurate assessment of the settlement forecast of Zhejiang Section of Hangzhou-Changsha passenger dedicated line based on the field measurement. As is known,the settlement evaluation of the underline engineering project of the passenger dedicated line is the key engineering link before track-laying. In practice,the curve fitting correlation coefficients of the four conventional methods are generally thought of being low with most of the subsections being less than 0. 92. At the same time,attention has to be paid to the down-settling data which are characteristic with ＂small scale and large fluctuation＂and mix MAD,MSE and MAPE into a complicated structural system instead of the correlation coefficient. It is just for this reason that we have analyzed the rationality and the causes for the proposed design and corresponding measures to be taken in the engineering practice. In detailed manner,we have introduced the low correlation coefficient in the settlement evaluation and the three new-settled evaluation indicators,that is,the mean absolute deviation（ MAD）,the mean squared errors（ MSE） and the mean absolute percentage errors（ MAPE）. Comparing the said evaluation indicators with the correlation coefficient,the results of the evaluation demonstrate that the settlement forecast model can help to achieve high enough accuracy even in the case of the low correlation coefficient. Thus,the limited conditions of the three evaluation indicators in the proposed model have been given by the relationship between the three evaluation indicators and H value of T-test for the predicted make-up measures. It is supposed that,when MAD≤0. 11 mm,MSE≤0. 015 mm2 and MAPE≤8%,it would be possible to believe that the observed points can be taken as being able to meet the demands of the settlement forecasting accuracy. What is more,the engineering examples can also remind us to suggest the model proposed by us is available for the sown-settling forecast accuracy for the section of the railroad. In addition, Asaoka method and the hyperbolic curve method has also proved all the observed points in the section under testing,which can meet the demands of the settlement forecast accuracy when exponential curve method is used for the purpose. With 90. 4 percent of the observed points in the section being able to meet the accuracy demand of the settlement forecast,the Grey Theory also tends to prove to be realistically applicable.