【作者】 张勋；

【导师】 唐文彦； 许文海；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2008， 博士

【摘要】 浅海航道底部在风、浪、流等天气水文情况下淤积严重,由于缺乏实时的监测手段,经常出现因船舶装载量过少造成经济上的损失或装载量过多导致船舶搁浅、港口全面阻塞等问题。本文根据浅海航道的特点分析了航道中声传播特性及衰减特性,在此基础上提出了一种适合对浅海航道淤泥淤积位置和高度进行实时监测的技术。论文所进行的主要研究工作如下:基于水声学领域描述声波传播规律的射线声学理论,在简化声速分布的基础上,分析研究了浅海航道中小掠射角声传播特性,结果表明:航道中小掠射角声波声线轨迹近似直线传播。基于Snell定律和能量守恒定律研究了海面对声波的反射和透射特性,由研究结果可知:声波传播到海面时会发生反射和透射,但透射的能量远远小于反射能量,声波在海面传播时可以看作全反射。声波在浅海航道淤泥层进行传播时,通过比较散射和透射系数得出:高频声波在海底传播时,绝大部分声能都透射到了沉积层中,散射声能可以忽略。基于Bouguer-Lambert-Beer-Law理论,建立了多分散颗粒悬浊液声传播衰减损失模型,在实验室利用配制的不同粒径的超细玻璃微珠和二氧化钛的水悬浊液对五个频率的超声进行了声衰减测量,验证了该模型是有效的。以该模型为基础,结合声波在海水中的扩展损失和吸收损失,建立了浑浊海域中声传播衰减模型,通过在渤海某航道中的声传播衰减测量,结果表明浅海航道浑浊海域中声传播衰减损失与颗粒粒径、颗粒体积浓度以及声频率有关。根据现在浅海航道淤积测量或预报方法不能够实时监测的问题,在浅海航道声传播规律的基础上结合地震CT层析结构特点,提出了一种对浅海航道淤积位置和淤积高度进行实时监测的方法——基于换能器阵列的航道侧面多传感器淤积监测法,并设计了相应的系统原理样机。该方法是在航道一侧安放一条发射换能器阵列,另一侧安放一条接收阵列,监测时依次开启发射换能器发射声波信号,同时接收阵列中的所有换能器均接收信号,淤积位置和淤积高度不同时相应接收声纳信号的功率谱峰值与传输时间也不同,根据这个差异就可以判别监测区域的淤积情况。通过无线数传模块将监测数据传输到岸上的监控中心,应用客户端即可利用Internet实现对浅海航道淤积位置和淤积高度的实时监测。在判别浅海航道的淤积位置和淤积高度时,提出了基于决策导向无环图支持向量机(Decision Directed Acyclic Graph Support Vector Machine,简称DDAGSVM)算法的多类问题的航道淤积位置判别法和基于支持向量机回归的航道淤积高度判别法。利用离散K-L(Karhunen-Loeve)变换提取的航道淤积特征向量与实际淤积位置和淤积高度分别组成淤积位置训练样本和淤积高度训练样本,在淤积位置判别时,将航道宽度方向平分成若干区域,并将每个区域作为一类淤积,每两类之间各构造一个分类器,利用淤积位置训练样本对分类器进行训练,将淤积特征向量带入训练好的分类器结合DDAGSVM算法,判别出淤积位置;在淤积高度判别时,利用淤积高度训练样本对淤积高度回归模型进行训练,将淤积特征向量带入训练好的淤积高度回归模型,判别出航道的淤积高度。通过消声水池和浅海航道的淤积判别实验可知:在消声水池中的淤积位置判别精度达到了3m,在航道的淤积位置判别精度达到了实际航道的需求;对航道淤积高度判别的精度可达到60 mm。与传统的神经网络方法判别结果相比,该方法具有判别效果好、判别速度快、通用性强等优点。更多还原

【Abstract】 Generally, shallow sea channel has the severe deposition under the weather and hydrology conditions of wind, wave, flow etc. Because of lacking real-time siltation monitoring method, it often appears the economical loss caused by less load or ship grounding and port obstruction caused by excess load. In the paper, a new technology of real-time channel siltation monitoring is presented based on the analysis of characteristics of the shallow sea channel, acoustic propagation and attenuation in the channel. Main contents of this paper are as follows:Based on Ray Theory describing acoustic propagation law in underwater acoustics field, the characteristics of the small grazing angle acoustic propagation in the shallow sea channel is researched on the basis of simplified acoustic velocity distribution. The results show that the sound ray trace can be regarded as a straight line. According to the law of Snell and Conservation of Energy, the characteristics of acoustic wave scattering and transmission on sea surface is studied. The acoustic wave will scatter and transmit when it propagates to the sea surface and the energy of transmission is far less than that of scattering. By comparing scattering coefficient with transmission coefficient, when the high-frequency acoustic wave propagates on the bottom, most of the acoustic energy transmits into the deposition layer and the scattering acoustic energy can be neglected.Sound attenuation model of the polydisperse particle suspension liquid acoustic propagation attenuation loss is derived from Bouguer-Lambert-Beer Law. In the laboratory, the prepared aqueous suspension of ultrafine glass beads and aqueous titanium dioxide of different particle diameters are applied to measure the acoustic attenuation of five ultrasonic frequencies. The result shows that the model is effective. Based on the model and combined with the empirical formula of spreading loss and absorption loss of the acoustic wave in the sea , the acoustic propagation attenuation model in the turbidity sea area is established. The measurement results of the acoustic propagation attenuation in one Bohai sea channel show that the acoustic propagation attenuation loss in shall sea channel relates to particle diameter, particle volume concentration and the acoustic frequency.According to the problems that present siltaion measuring methods and forecasting method in the shallow sea channel can’t realize real-time monitoring, a new siltation real-time monitoring method in the shallow sea channel is proposed based on the research results of acoustic propagation in the shallow sea channel and seismic CT structure. It is called Multi-Sensor siltation monitoring method in the sides of channel based on transducer arrays. At the same time, a set of monitoring system is designed. In the method, one set of transmitting transducer array is fixed on one side of the channel and one set of receiving transducer array is fixed on the other side. On monitoring, the transmitting transducer is opened to transmit acoustic wave in turn and receiving transducers are always working. And when the siltation position and the height are different, the power spectrum’s peak and the transmission time of the sonar signal are different too. In terms of the difference, the siltation situation in the monitoring area can be judged. Through wireless data transmission module, the monitoring data can be transmitted to the monitoring center in the shore. And applying the Internet, the client can realize real-time monitoring on the siltaion position and height in the shallow sea channel.In order to judging the siltaion position and height, the multi-class problem method of judging the channel siltaion position based on Decision Directed Acyclic Graph Support Vector Machine(DDAGSVM) algorithm and the method of judging the channel height based on support vector machine regression are put forward. Firstly, applying discrete K-L transform, the siltation features vector is extracted. And it can compose siltation location training samples and height training samples respectively with the siltation location and height in the monitoring area. When judging the siltation location, the channel is bisected into some areas along channel width direction, and each area is one type siltation. Between every two classes, one classifier is constructed. With the siltation location training samples, the classifiers are trained. And when the siltation feature vectors are brought in the well trained classifiers and combined with DDAGSVM Algorithm, the siltation location is judged. When judging the siltation height, with the siltation height training samples, the siltation height regression model is trained. And when the siltation feature vectors are brought in the well trained model, the siltation height is judged. The experiment results in anechoic tank and Dalian channel show that in anechoic tank, the judging siltation position precision can reach 3m, and in Dalian channel, it is attained the actual demand of the channel. And in Dalian channel, the judging siltaion height precision can reach 60 mm. Compared with conventional methods of neural network, the methods have better effect, stronger generalization ability and fast decision speed.更多还原