【作者】 靳建明；

【导师】 谢康和； 王奎华；

【作者基本信息】 浙江大学 ， 岩土工程， 2004， 博士

【摘要】 本文在大量的室内试验基础上，较系统地对土密实度(包括干密实度和含水量)瞬态振动测试信号进行了分析，给出了土密实度振动测试的定量计算方法。主要工作如下： 1．在实验室内对制作成一定密实度的试样进行了大量瞬态振动测试，获得试样在锤击作用下的加速度和力信号。 2．对比了几种常用信号降噪方法在土密实度瞬态振动测试信号中的适用性，得到了小波阈值降噪法最适用于土密实度瞬态锤击测试信号的结论，并给出了适用于土密实度测试信号降噪的小滤阈值确定方法。 3．从时间域和频率域两个角度对土密实度瞬态振动测试信号进行了分析，提取了信号的特征参量。分析了各特征参量同土密实度之间的相关关系，并给出了相应的经验关系式。 4．根据室内试验结果对提出的几个经验关系式进行了验证，分析了各经验关系式的可靠性和精确度，并对各个关系式在土密实度定量计算中的性能进行了对比分析。 5．根据人工神经网络的基本原理，建立了土密实度定量分析的BP神经网络模型，编制了相应的分析程序，并对其准确性和可靠性进行了验证和讨论。 6．针对标准BP神经网络的缺点，提出了两种改进方法：①基于信息优化的网络学习算法；②基于遗传算法的网络学习算法。通过与标准BP算法的比较，表明这两种改进方法都能有效地提高神经网络模型的精度。 7．对本文提出的各种土密实度定量分析方法进行了对比分析，分析结果表明：基于遗传算法和信息优化的神经网络模型的分析效果最佳。 本文为土密实度的瞬态振动测试建立了整套分析方法，并通过室内试验证明了该法的可行性。这为在实际工程土密实度的测试提供了一种新的无损、可靠、快速的测定方法。更多还原

【Abstract】 Based on a great amount of laboratory test, the instant vibration signal from soil compactness (include dry density and water content) testing is analyzed systemically and the corresponding quantitative analysis method is put forward for the determination of soil compactness. The main work is as follows.1. By using the soil samples prepared with different dry density and water content, a great amount of instant vibration test is performed and the signals of acceleration and force are acquired.2. The applicability of several commonly used de-noising methods is examined in signal analysis of instant vibration testing for soil compactness. The analysis result indicates that the method based on wavelet thresholds is the best. The method for determination the threshold of wavelet de-noising is also proposed to satisfy soil compactness test signal.3. The signal of instant vibration testing for soil compactness is analyzed both in time domain and frequency domain. Some feature parameters reflecting the characteristics of the signal are extracted. The relationship between soil compactness and feature parameters is then investigated and the corresponding experiential formulas are established.4. The experiential formulas are verified by experimental results and then-correctness and reliability is discussed. Furthermore, the capability in determining soil compactness of these formulas is compared and analyzed.5. Based on the fundamental of ANN, the model of BP neural networks is built for soil compactness quantitative analysis, and the corresponding computer program is developed. The correctness and reliability of the model and the program are then validated and discussed.6. To overcome the disadvantages of standard BP networks, two unproved algorithm are put forward. One is information entropy based, and the other is genetic algorithm based. By comparing with standard BP model, it shows that both the two improved methods can improve the precision of ANN efficiently.7. Comparisons are made among the all quantitative analysis methods for the determination of soil compactness proposed herein, and the result shows that the models based on information optimization and on genetic algorithm (GA) are the best.The dissertation established a set of analysis method for the determination of soil compactness by instant vibration testing, and the feasibility of the method was verified by laboratory test. This provides a new non- destructive, reliable and fast testing method for determining soil compactness in engineering practice.更多还原