【作者】 王杰；

【导师】 马本俊；

【作者基本信息】 哈尔滨工程大学 ， 海洋科学， 2025， 硕士

【摘要】 沉底雷及未爆弹等静态小目标在近岸波流联合冲蚀作用下产生泥沙掩埋现象。掩埋状态预测是海底小目标探测等任务规划的重要参考依据,但由于掩埋过程机理复杂,预测难度大,目前尚无精细化预测方法。鉴于此,本文以圆柱形沉底雷为典型海底小目标,通过海洋环境模式数据与机器学习算法的协同机制,实现了近岸海底圆柱形小目标冲蚀掩埋深度的短周期序列精细化预测,主要研究内容如下:首先,针对海底圆柱形小目标冲蚀掩埋机理问题,开展了海底圆柱形小目标冲蚀掩埋数值模拟研究。基于Flow-3D构建海底圆柱形小目标冲蚀掩埋数值计算模型,重点分析了圆柱体周围流场特征和冲蚀坑深度的动态演化规律,并研究了多影响因子条件下的冲蚀坑深度大小变化。结果表明:(1)圆柱体周围流场呈现显著的三维流动分离现象和尾流涡旋结构,导致局部流速增大和湍流作用增强,进而影响沉积物输运与堆积;(2)冲蚀坑深度在初期发展迅速,后期因目标下沉暴露体积减小而趋于动态平衡;(3)流速、圆柱体直径与冲蚀坑深度呈正相关关系,沉积物粒径与冲蚀坑深度呈负相关关系,圆柱体密度与冲蚀坑深度相关性较弱,相关系数仅为-0.082;(4)该数值模拟方法虽然能够描述这一物理过程,但因其计算资源消耗大、预测效率低,不适用于探测任务规划等应用场景的快速响应预测。其次,针对基于传统数值模拟的掩埋预测方法计算效率低的问题,提出了基于堆叠泛化框架的海底圆柱形小目标冲蚀掩埋预测模型。基于数值模拟结果及掩埋机理分析,构建了以冲蚀掩埋率为输出,希尔兹θ数和KC数为输入的初级预测模型。通过贝叶斯优化算法提升堆叠泛化框架内初级预测模型的精确度,实现基于堆叠泛化框架的海底圆柱形小目标冲蚀掩埋预测方法优化,验证结果表明:该方法具有较高的预测精度,相较于本文最优单一机器学习预测模型,决定系数提升了3.6%,均方误差以及均绝对误差则分别降低了15.7%、1.2%。充分验证了该方法的优越性,为海底圆柱形小目标探测任务规划提供了高效的预测工具。最后,针对海底圆柱形小目标冲蚀掩埋方法适用性提升需求,基于所构建的堆叠泛化框架的海底圆柱形小目标冲蚀掩埋预测模型,提出了面向海底小目标探测任务规划的热力图表征方法。以黄海海域为案例,开展实例应用,利用海洋环境模式数据,实现了海底圆柱形小目标冲蚀掩埋深度的24小时短周期序列的精细化预测。结合热力图技术、掩埋控制线(Burial Dominance Line,BDL)理论和底质分级知识(Doctrinal Bottom Type,DBT),形成圆柱形沉底雷冲蚀掩埋状态热力图,为圆柱形沉底雷探测任务规划方案的快速判别提供了辅助决策支持。更多还原

【Abstract】 Static small targets such as bottom mines and unexploded ordnance are buried by sediment under the combined effects of nearshore waves and currents.The prediction of the burial state is an important reference for mission planning such as the detection of small underwater targets.However,due to the complex mechanism of the burial process and the difficulty of prediction,there is currently no refined prediction method.In view of this,this thesis takes cylindrical bottom mines as typical small seabed targets,and realizes the short-period series fine prediction of the scour burial depth of small cylindrical targets on the nearshore seabed through the synergistic mechanism of marine environment model data and machine learning algorithms.The main research contents are as follows:Firstly,aiming at the scour burial mechanism of small cylindrical targets on the seabed,a numerical simulation study on the scour burial of small cylindrical targets on the seabed was carried out.Based on Flow-3D,a numerical calculation model of scour burial of small cylindrical targets on the seabed was constructed.The flow field characteristics around the cylinder and the dynamic evolution law of the scour pit depth were analyzed in detail,and the changes in the depth and size of the scour pit under the conditions of multiple influencing factors were studied.The results show that:(1)The flow field around the cylinder exhibits significant three-dimensional flow separation and wake vortex structure,which leads to increased local flow velocity and enhanced turbulence,thus affecting sediment transport and accumulation;(2)The depth of the scour pit develops rapidly in the early stage,and tends to a dynamic equilibrium in the later stage due to the reduction of the exposed volume of the target sinking;(3)The flow velocity and cylinder diameter exhibit a positive correlation with the scour pit depth,while the sediment grain size shows a negative correlation.The correlation between cylinder density and scour pit depth is relatively weak,with a correlation coefficient of only-0.082.(4)Although this numerical simulation method can describe this physical process,it is not suitable for rapid response prediction in application scenarios such as detection mission planning due to its high consumption of computing resources and low prediction efficiency.Secondly,to address the problem of low computational efficiency of burial prediction methods based on traditional numerical simulation,a prediction model for scour burial of small cylindrical targets on the seabed based on a stacking generalization framework was proposed.Based on the numerical simulation results and burial mechanism analysis,a primary prediction model was constructed with the scour burial rate as output and the Shields number and Keulegan-Carpenter number as input.The accuracy of the primary prediction model within the stacking generalization framework was improved through the Bayesian optimization algorithm,achieving the optimization of the prediction method for the scour burial of small cylindrical targets on the seabed based on the stacking generalization framework.The verification results show that this method has high prediction accuracy.Compared with the optimal single machine learning prediction model in this thesis,the determination coefficient is increased by 3.6%,and the mean square error and mean absolute error are reduced by 15.7%and 1.2%respectively.This fully verifies the superiority of this method and provides an efficient prediction tool for the planning of submarine cylindrical small target detection tasks.Finally,to improve the applicability of scour burial prediction methods for small cylindrical seabed targets,a heatmap representation method for mission planning in seabed small target detection was proposed,based on a prediction model developed using a stacked generalization framework.Taking the Yellow Sea as a case study,a practical application was conducted,utilizing marine environmental model data to achieve refined 24-hour short-term sequence predictions of the scour burial depth of small cylindrical seabed targets.By integrating heatmap techniques,the Burial Dominance Line(BDL)theory,and Doctrinal Bottom Type(DBT)classification,a heatmap depicting the scour burial states of cylindrical bottom mines was generated,providing auxiliary decision support for rapid identification and planning of detection missions targeting cylindrical bottom mines.更多还原