基于超声波的压力测量模型研究

发布时间：2018-08-18 17:11

【摘要】：压力容器在日常生活和工业生产很多领域中都得到广泛的应用,为了使其安全、正常运行,对于其内部压力的检测至关重要。传统的压力测量方法往往需要进行开孔等可能破坏容器完整性的操作。非介入式压力检测方法具有安全、便携等优点,应用前景良好。基于超声波的非介入式压力检测,其主要的理论基础是声弹性效应和薄壳理论。在对相关领域的文献和成果的研究基础之上,本论文建立了基于临界折射纵波、反射纵波的线性和非线性压力测量模型,对模型的波形选择进行了研究,并对各个模型进行了分析和比较。本论文的主要研究成果和创新之处在于:(1)提出了基于最佳子集回归和逐步回归算法的多变量压力测量模型输入波形的选择方法。之前多变量压力测量模型输入波形的选择只是简单的根据输入波形的区分度和信噪比。用于建立模型的波形的选择至关重要:一方面,越多的解释变量更多的自变量可以使信息更加完整、全面,使预测更加精确,但也可能造成"过拟合"的现象,同时需要估计的参数个数增加会导致方差提高;另一方面,过少的解释变量可能造成"欠拟合"的现象,预测准确度降低。本文将最佳子集回归和逐步回归算法应用于输入信号波形的选择,可以建立更优、更符合选定准则的模型。(2)针对线性压力测量模型精度不高的问题,提出并建立了基于临界折射纵波、反射纵波的非线性压力测量模型。根据声弹性效应和薄壳理论,压力与临界折射纵波、反射纵波呈线性关系,但实验数据表明在某些压力范围内,二者存在非线性关系。本论文尝试使用非线性方法进行建模,包括将传播时延的二次项加入测量模型以及基于神经网络的测量模型,实验结果表明,非线性压力测量模型比线性压力测量模型的测量精度更高,非线性模型中基于BFGS神经网络的模型测量精度最高。(3)对超声波在容器壁中的传播过程的信号幅值变化机理进行了理论分析。超声波从发射探头到接收探头传播过程中,由于反射、折射等会引起波型转换和能量分配,另外,在传播过程中由于散射、扩散和吸收也会引起能量的衰减,综合表现为幅值的变化。为了确定合适的探头间距,以获取信噪比更高的波形,本论文对超声波在圆柱型压力容器壁的传播过程中的幅值变化进行了理论分析。
[Abstract]:Pressure vessels are widely used in many fields of daily life and industrial production. In order to make them safe and normal operation, it is very important to detect the internal pressure. Traditional pressure measurement methods often require operations such as opening holes that may damage the integrity of the vessel. The non-interventional pressure detection method has the advantages of safety and portability, and has a good prospect in application. The main theoretical basis of non-interventional pressure measurement based on ultrasonic wave is acoustic elastic effect and thin shell theory. Based on the research of literature and achievements in related fields, the linear and nonlinear pressure measurement models based on critical refraction P-wave and reflected P-wave are established in this paper, and the waveform selection of the model is studied. The models are analyzed and compared. The main achievements and innovations of this thesis are as follows: (1) an input waveform selection method based on optimal subset regression and stepwise regression algorithm for multivariable pressure measurement model is proposed. The selection of input waveform of multivariable pressure measurement model is simply based on the discrimination and signal-to-noise ratio of the input waveform. The choice of waveforms used to build models is crucial: on the one hand, more explanatory variables and more independent variables can make information more complete, more comprehensive, and more accurate predictions, but may also result in "overfitting". On the other hand, too few explanatory variables may cause the phenomenon of "under-fitting", and the accuracy of prediction will be reduced. In this paper, the optimal subset regression and stepwise regression algorithms are applied to the selection of input signal waveforms, which can set up a better model, which is more consistent with the selected criteria. (2) the accuracy of the linear pressure measurement model is not high enough. A nonlinear pressure measurement model based on critical refraction P-wave and reflected P-wave is proposed and established. According to the acoustoelastic effect and thin shell theory, the pressure is linearly related to the critical refraction longitudinal wave and the reflected longitudinal wave, but the experimental data show that there is a nonlinear relationship between them in some pressure range. This thesis attempts to use nonlinear method to model the model, including adding the quadratic term of propagation delay into the measurement model and the measurement model based on neural network. The experimental results show that, The accuracy of nonlinear pressure measurement model is higher than that of linear pressure measurement model. In the nonlinear model, the model based on BFGS neural network has the highest measurement accuracy. (3) the mechanism of signal amplitude variation in the process of ultrasonic wave propagation in the vessel wall is analyzed theoretically. In the process of ultrasonic wave propagation from transmitting probe to receiving probe, wave type conversion and energy distribution will be caused by reflection and refraction. In addition, energy attenuation will also be caused by scattering, diffusion and absorption during the propagation process. The comprehensive performance is the change of amplitude. In order to determine the appropriate probe spacing and obtain the higher signal-to-noise ratio (SNR) waveform, the amplitude variation of ultrasonic wave during the propagation of cylindrical pressure vessel wall is theoretically analyzed in this paper.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH49;TB559

【参考文献】