【作者】 李张明；

【导师】 练继建；

【作者基本信息】 天津大学 ， 水利水电工程， 2007， 博士

【摘要】 本文对锚杆锚固质量无损检测的理论及智能诊断技术进行了系统深入的研究。研究了锚杆围岩系统瞬态激励响应的数学模型并给出了其波动方程的数值解法;在研究声波反射检测法的原理和检测技术方法的基础上,研制了一种锚杆检测专用传感器;结合模型锚杆试验及实际工程锚杆检测实践,提出了锚杆锚固质量无损检测评价标准;对锚杆检测的反射波进行了小波分形维数的计算,得到了表征锚杆质量的特征向量,进而提出了基于BP神经网络的锚杆质量智能诊断模型并得到验证。为锚杆锚固系统无损检测而研究开发的一套完整实用、快捷方便的检测诊断方法和技术,对锚杆锚固工程的质量检测具有重要的应用价值。本文主要做了以下几点工作和创新:1研究了锚杆围岩系统瞬态激励响应的数学模型及其数值解法。2.研究了锚杆声波检测反射的机理、反射波时频域分析和信号拟合方法。3.研究了声波反射检测法的原理和检测技术方法,在锚杆检测实践的基础上,研制了一种专用传感器,具有独创性。该锚杆检测三分量传感器固定装置已获国家实用新型专利。4.结合模型锚杆试验及实际工程锚杆检测实践,首次建立了锚杆锚固质量无损检测评价标准,具有一定的创新性和实用价值。5.研究了应用强大的小波包分解工具,对反射波在不同频段的波形特征进行小波包分解,为反射波形检测提供了深层次分析信息,提高了锚杆缺陷分析的准确性。6.首次将分形维数的分析方法引入锚杆锚固系统的无损检测中,对锚杆缺陷的小波分形维数特征向量和锚杆缺陷的关系进行了初步的探讨,明确指出了分形维数的大小可以量度反射波在该小波分解频段能量的大小,具有创新性。7.针对获取的表征锚杆缺陷的特征向量与锚杆密实度之间的非线性关系,借助人工神经网络技术,引入BP神经网络模型。分析了对锚杆质量评价有影响的因素,确定了BP神经网络的输入输出参数,对锚杆的质量进行了预测,预测结果与实测值具有满意的吻合,说明该模型对于锚杆锚固系统的智能化无损检测有较高的准确性,满足锚杆检测的工程需求,具有推广应用的价值。更多还原

【Abstract】 A study on nondestructive detection theory and intelligent diagnosis technology of rockbolt anchorage quality has been carried out systematically in this paper.The mathematics model of the transient responds of rockbolt and wall rock system have been studied and its numerical value solution of wave equation have been analyzed.On the basis of study on the principle and technology of the sonic reflection detection method and detection practicing,a kind of special sensor for rockbolt detection has been developed,Combining with model rockbolt testing and actual project detection practice in site, an evaluation standard for rockbolt anchorage quality of nondestructive detection has been put forward.The reflected wave containing the information of rockbolt defect was obtained on the dissertation firstly by means of sonic reflection, decomposing of the wavelet packet for the reflected wave and calculation of the decomposed waveform fractal dimension were then conducted, the characteristic vectors indicating rockbolt quality were accordingly achieved;based on the overall analysis of the factors which influence the rockbolt quality diagnosis, intelligent diagnosis model of rockbolt quality has been put forward based on the BP neural network,which has created a complete set of practicable, shortcut and convenient detection method and technology for the anchorage system, and has significant application value on the quality detection for the rockbolt works.The main achievements of study works and some innovations are as follows.1.The mathematics model of the wave equation of transient responds of rockbolt and wall rock system have been studied and its numerical value solution of wave equation have been analyzed.2.The principle of the sonic reflection of rockbolt detection;reflection analyse in time and frequency field and the method of signal fitting have studied.3.Principle and technology of the sonic reflection detection method have been studied. A kind of special sensor has been developed on the basis of detection practice, this sensors device has original creation, it has won the national practical and new-type patent.4.Combining with model rockbolt testing and actual project detection practice in site, an evaluation standard for rockbolt anchorage quality of nondestructive detection has been put forward.It has definite original creation and utility.5.Using the powerful decomposing tools of the wavelet packet, decomposing of the wavelet packet is carried out for the reflected wave according to the waveform nature of various frequency band , which has provided in-depth analysis of information for the reflected waveform detection, and also increase the accuracy of analysis by broadening and deepening the evaluation scope of rockbolt defect.6.It is the first time to apply the analysis method of fractal dimension to nondestructive detection of anchorage system. As the dimension calculation can better show the total information of the wave form,the defect of the anchorage system can be more accurately and comprehensively analyzed by employing the wavelet fractal dimension as the characteristic vector of rockbolt defect,this has original creation.7.As for the nonlinear relation between the obtained characteristic vector representing rockbolt defect and rockbolt density,BP neural network model is introduced by means of artifical neural network technique.The factors which would affect quality evaluation of the rockbolt have been analysed, the input and output parameters of the BP neural network finalized, and the computation method and formulas for learning and training purpose given on this dissertation.Upon learning, trying and verifying of this model, forecasting of the rockbolt quality has been conducted, and the forecasted results satisfactorily coincide with the measured data,it is illustrated that the developed model has better accuracy on intelligentized nondestructive detection of the anchorage system,can cater to the requirements of rockbolt detection,and thus is valuable to extensively apply on the engineering practice.更多还原