【作者】 王静；

【导师】 符策；

【作者基本信息】 大连海事大学 ， 工程硕士（专业学位）， 2021， 硕士

【摘要】 近年来随着长江沿江经济和航运业的快速发展,船闸货运量不断攀升,三峡、葛洲坝船闸成为长江中上游重要节点的水利枢纽。但由于船闸的满负荷、超负荷运转使用,其结构安全隐患也愈加明显。船闸人字门在复杂的深水环境中频繁启闭,受低速重载的工况条件、闸门两侧巨大水压及启闭机运行状态的影响,会引发门体小位移量的跳动,产生门体共振、应力集中释放的情况,损伤门体结构,严重时将会发生门体病变,影响通航效率并造成严重的经济损失。目前针对船闸人字门结构安全采取的是定期停航人工检修,无法实时监测且存在漏检的可能。因此建立船闸人字门远程监测系统,对结构振动、应变进行健康监测,及时发现结构损伤并对其进行检修维护,对保障通航安全、提高通航效率具有重要意义。本文以葛洲坝3#船闸下游人字门为研究对象,在船闸人字门远程监测系统的基础上,对各结构的振动、应力信号进行数据处理研究与分析,实现对船闸各结构的健康监测,为人字门运行状况的日常监测及检修维护提供参考依据。本文主要工作内容包括:(1)根据船闸人字门结构对其进行门体建模,并通过Ansys-Workbench软件对门体结构进行模态分析,对A杆结构进行瞬态响应分析。根据有限元分析结果,得到人字门运行过程中易产生共振及应力集中释放的位置,并搭建船闸人字门远程监测系统。(2)针对船闸信号分布复杂,无法用单一函数表征并预测变化趋势的问题,提出一种基于非参数核密度估计的应变分析方法,对船闸人字门进行应变监测。以A杆结构为监测对象,对采集到的应力数据进行奇异值分解的平滑处理,然后进行核密度估计,得到各工况下A杆应力的概率分布。先结合仅平滑处理的应力信号波形,验证了该方法的有效性;再结合瞬态响应分析结果,判断A杆的健康状态,即是否存在损伤风险。(3)针对传统共振分析方法存在的识别准确率低的问题,提出一种基于多传感器融合的人字门辅助诊断方法,采用RBF神经网络和D-S证据理论对振动与应力信号进行联合分析。以门体底部门中缝处结构为监测对象,首先对其振动、应力数据进行多域特征提取,然后分别构建振动、应力RBF神经网络系统,对人字门结构进行初步状态识别。最后将初步识别结果作为D-S证据理论的证据体,进行决策层融合,得到最终基于振动与应力的状态识别结果。经过实验验证了该方法的有效性,并提高了多传感器系统的状态识别准确率。更多还原

【Abstract】 In recent years,with the rapid development of the economy and shipping industry along the Yangtze River,the cargo volume of ship locks is rising.The Three Gorges and Gezhouba ship locks have become important water conservancy hubs in the middle and upper reaches of the Yangtze River.However,due to the full load and overload operation of the lock,its structural safety risks are more and more obvious.The miter gate of ship lock is frequently opened and closed in the complex deep water environment.Affected by the low speed and heavy load working conditions,the huge water pressure on both sides of the gate and the operation state of the hoist,the small displacement of the gate body will jump,the resonance of the gate body and the stress concentration release will occur,and the structure of the gate body will be damaged.At present,for the structural safety of miter gate of ship lock,it is necessary to stop navigation regularly for manual maintenance,which can not be monitored in real time and may be missed.Therefore,the establishment of miter gate remote monitoring system for structural vibration and strain health monitoring,timely detection of structural damage and maintenance,is of great significance to ensure navigation safety and improve navigation efficiency.In this thesis,the downstream miter gate of Gezhouba 3# shiplock is taken as the research object.Based on the miter gate remote monitoring system,the vibration and stress signals of each structure are processed and analyzed,so as to realize the health monitoring of each structure of the shiplock,and provide reference for the daily monitoring and maintenance of miter gate.The main contents of this thesis are as follows:(1)According to the miter gate structure of the lock,the model of the miter gate is built,and the modal analysis of the miter gate structure is carried out by using ANSYS Workbench software,and the transient response of the a-bar structure is analyzed.According to the results of finite element analysis,the position of resonance and stress concentration release during the operation of miter gate is obtained,and the remote monitoring system of miter gate is built.(2)In order to solve the problem that the distribution of lock signal is complex and cannot be represented and predicted by a single function,a strain analysis method based on nonparametric kernel density estimation is proposed to monitor the miter gate.Taking the a-bar structure as the monitoring object,the collected stress data are smoothed by singular value decomposition,and then the kernel density estimation is carried out to obtain the probability distribution of the a-bar stress under each working condition.Firstly,the effectiveness of the method is verified by combining the stress signal waveform only smoothed;Combined with the results of transient response analysis,the health state of a-bar is judged,that is,whether there is a risk of damage.(3)Aiming at the low recognition accuracy of traditional resonance analysis methods,a miter gate aided diagnosis method based on multi-sensor fusion is proposed.RBF neural network and D-S evidence theory are used to analyze vibration and strain signals.Taking the middle seam structure at the bottom of the door as the monitoring object,the vibration and stress data are firstly extracted from multi domain features,and then the vibration and stress RBF neural network systems are constructed respectively to identify the initial state of the miter gate structure.Finally,the preliminary identification results are regarded as the evidence body of D-S evidence theory,and the final state identification results based on vibration and stress are obtained by decision fusion.The effectiveness of the method is verified by experiments,and the state recognition accuracy of the multi-sensor system is improved.更多还原