| Pressure vessels are commonly used in both daily life and industries.It is critically important to monitor the pressure inside the pressure vessel for the purpose of operational safety.As for traditional pressure measurement methods,operations such as drilling holes are usually required,which could damage the integrity of the pressure vessel.Non-invasive pressure measurement methods are promising due to their advantages of safety and portability,etc.The non-invasive pressure measurement method based on ultrasonics,basis of which are acoustoelastic effect and shell theory.Based on the literature and research results in the related fields,linear and nonlinear models of pressure measurement with critically refracted longitudinal wave(LCR)and reflected longitudinal waves have been established in this thesis.The selection of waves for models also has been researched.And different models have been compared and analyzed.The followings are the main contents and innovations of the thesis:(1)Algorithms based on best subsets regression and stepwise regression for selection of waves in the pressure measurement models have been proposed.Selection principles of waves for the pressure measurement models based on multiple waves were mainly the signal-to-ration ratio and the distinguishability of the waves.The selection of waves for establishing models is crucial:on one hand,selecting more waves is more likely to get complete information,which can possibly improve the predicting ability,but also may cause "overfitting" and raise the variance;on the other hand,less waves may cause "underfitting",lowering the predicting ability.The applications of best subsets regression and stepwise regression in selecting waves have been investigated in this thesis and better models can be established employing these methods.(2)Linear and nonlinear models with LCR and reflected longitudinal waves have been established.According to acoustoelastic effect and shell theory,the relationship between pressure and travel-time change of LCR and reflected longitudinal waves is linear.However,the experiment results showed that there is nonlinear relationship in the area of low pressure.Nonlinear methods including adding quadratic terms of travel-time changes to the linear model and neural networks have been employed to establish the pressure measurement models.The experiment results proved that models based on neural networks can improve the measurement precision,and models based on BFGS neural networks have better predicting ability than other nonlinear models.(3)The mechanism of amplitude changes of ultrasonics during the propagation in the pressure vessel wall has been theoretically analyzed.During the propagation between the transmitting probe and the receiving probe,mode conversion and energy partition of ultrasonics will take place due to reflections and refractions.The scattering and absorption also cause attenuation of energy,which contributes to the amplitude changes of ultrasonics.To determine the proper distance for receiving higher signal-to-noise ratio wa’ves,the thesis theoretically analyzed the amplitude changes of ultrasonics during the propagation in the cylinder pressure vessel wall. |