【作者】 胡曦；

【导师】 顾金良；

【作者基本信息】 南京理工大学 ， 仪器仪表工程（专业学位）， 2023， 硕士

【摘要】 弹丸在水下运动时,由于有空泡等形态的存在,另外受水下光照度等因素的影响,通过高速摄像等图像处理方法很难得到弹丸的真实位置。磁场信号凭借其较强的穿透力与不易受水下环境干扰的特点成为水下目标探测的重要物理信号,而我国水下磁探测技术发展较晚,因此发展对于弹丸、潜艇等水下磁性目标的定位与测速技术迫在眉睫。将弹丸等效为磁偶极子进行仿真,分析弹丸在经过传感器时磁场强度随距离的变化关系,在对比分析多种磁传感器的优缺点后,选择了合适的TMR磁传感器,结合TMR磁传感器的工作机理与信号特性,提出了一种基于TMR磁传感器阵列的水下弹丸位置及速度测试技术方法。针对TMR磁传感器输出信号驱动能力弱、距离较远时输出电压较小的问题,完成水下弹丸测试系统的软硬件设计,包含信号放大电路、低通滤波电路、模拟数字转换电路、Flash存储电路与接口电路的器件选型与电路设计,并分别阐述了ADC数据采集的软件设计实现思想、Flash数据存储的软件设计实现思想及其读写擦除时序分析以及以太网数据传输的软件设计实现思想。探索了一种水下弹丸的定位与测速算法,利用FCN神经网络(Fully Connected Neural Network)模型直接近似和回归输出电压到距离的映射函数,并在此基础上受到RSS(Received Signal Strength)定位法的启发,结合TMR磁传感器阵列的安装位置信息联立距离方程组,使用最小二乘法进行弹丸位置坐标的求解。在单点定位的基础上选取穿靶点前后两个相同时间差的采样点完成位置估计,以两点之间的平均速度代替瞬时速度,实现弹丸的轨迹追踪与速度估计。进行了标定实验、定位与速度估计实验,在标定实验中首先通过弹丸在不同选定距离的磁场触发实验获取输出电压与距离之间的原始数据组,以此作为FCN神经网络模型训练集与测试集对网络进行训练与测试。在同一输出电压的情况下,将FCN神经网络模型的预测距离值与传统高斯函数的拟合距离值分别与原始数据值进行比较,验证FCN神经网络模型具有更高的拟合精度。随后进行定位与速度估计实验,使用传感器阵列测量并反演出弹丸坐标与速度,与实际位置进行对比,验证了所设计的磁传感测试系统具有测量精度高、可靠性好的特点,具有较高的工程应用价值。更多还原

【Abstract】 Due to the presence of bubbles and other forms of projectiles moving underwater,as well as the influence of underwater illumination,it is difficult to obtain the true position of the projectile through image processing methods such as high-speed camera.The magnetic field signal has become an important physical signal for underwater target detection by its strong penetrating power and less susceptible to interference in underwater environment,however,the development of underwater magnetic detection technology in China is late,and thus the development of positioning and velocimetry technology for underwater magnetic targets such as projectiles and submarines becomes imminent.The projectile is equated to a magnetic dipole to simulate and analyze the variation of the magnetic field strength with the distance when the projectile passes the sensor.After the comparison and analysis of the advantages and disadvantages of various magnetic sensors,a suitable TMR magnetic sensor is selected.Subsequently,a technological method for underwater projectile position and velocity testing based on the TMR magnetic sensor array is proposed by combining the working mechanism and signal characteristics of the TMR magnetic sensor.Aiming at the problems of weak driving ability of TMR magnetic sensor output signal and low output voltage at long distance,the software and hardware design of underwater projectile testing system is completed,including the device selection and circuit design of signal amplification circuit,low-pass filter circuit,analog-to-digital conversion circuit,Flash storage circuit and interface circuit,and separately elaborate the software design and implementation ideas of ADC data acquisition,Flash data storage and the Ethernet data transmission.An algorithm for localization and velocimetry of underwater projectiles is explored,using the FCN(Fully Connected Neural Network)model to directly approximate and regress the output voltage-to-distance mapping function.Inspired by the RSS(Received Signal Strength)localization method,a system of distance equations is combined with the information of the installed position of the TMR magnetic sensor array and the projectile position coordinates are solved using the least squares method.On the basis of single-point localization,two sampling points with the same time difference before and after the penetration point are selected for position estimation,and the instantaneous velocity is replaced by the average velocity between the two points.In this way,trajectory tracking and velocity estimation of projectiles can be achieved.The calibration experiments,localization and velocity estimation experiments were conducted.In the calibration experiments,the original dataset between output voltage and distance was first obtained through multiple distance experiments,which was used as the training set and test set of the FCN neural network model to train the network.In the case of the same output voltage,the predicted distance values of the FCN neural network model and the fitted distance values of the traditional Gaussian function fitting were compared with the actual values of the original data respectively,and it was verified that the FCN neural network model has higher fitting accuracy.Subsequently,the localization and velocity estimation experiments were conducted in which the magnetic field information was measured and the coordinates and velocity of the projectile were inverse performed using the sensor array.The comparison with the actual position verifies that the designed magnetic sensing test system is of high measurement accuracy and reliability,and possesses high engineering application value.更多还原