基于BP神经网络的PSD非线性校正算法研究

发布时间：2018-05-31 06:31

本文选题：PSD + 非线性校正　；参考：《南京理工大学》2017年硕士论文

【摘要】：PSD作为一种新型的光点位置敏感传感器,由于其具有响应速度快、位置分辨率高、不受光斑的约束、能同时测量位置和光强、可靠性高且配套信号处理电路结构相对简单等特点,它在位置坐标的精确测量方面受到了广泛的应用。然而,由于PSD在制造过程中其表面扩散层和底层材料电阻率的均匀性、有效感光面积等多种因素导致PSD器件的输出存在非线性变化,这一缺陷使得整个器件所测数据存在不小误差。针对这一问题,本文提出了基于改进BP神经网络的新型PSD非线性校正算法。本文首先研究了 PSD的结构和特点,介绍了其作为一个新型光电位置传感器的优点,并分析了 PSD的工作原理,阐释了非线性产生的原因和进行非线性校正的必要性。随后对目前常用的PSD非线性校正算法进行研究,包括插值算法和BP神经网络算法,对比分析其优劣。重点研究了基于BP神经网络的二维PSD非线性校正算法的原理,分析了它的不足,并借鉴国内外对于标准BP算法的最新改进动态,在此基础上从提高收敛速度和降低算法复杂程度出发,探讨一种实用性较强的新型改进BP神经网络校正算法。搭建了 PSD的位置测量平台包括信号处理电路和基于ARM和CPLD的控制系统。通过实验平台进行实验获取数据并通过计算机编程仿真实验完成了对新算法的验证,并将其与现有非线性校正算法进行对比,结果显示:新型改进BP神经网络校正算法收敛速度快、精度高,且算法程序容易实现,实用性较强。
[Abstract]:As a new type of spot position sensor, PSD can measure position and intensity at the same time because of its high response speed, high position resolution and no restrictions of spot. Because of its high reliability and relatively simple signal processing circuit, it has been widely used in the precise measurement of position coordinates. However, due to the homogeneity of the surface diffusion layer and the resistivity of the bottom layer, and the effective photosensitive area, the output of the PSD device has a nonlinear variation due to many factors, such as the uniformity of the surface diffusion layer and the resistivity of the underlying material. This defect makes the whole device measure the data to have not small error. To solve this problem, a new PSD nonlinear correction algorithm based on improved BP neural network is proposed in this paper. This paper first studies the structure and characteristics of PSD, introduces its advantages as a new photoelectric position sensor, analyzes the working principle of PSD, and explains the causes of nonlinearity and the necessity of nonlinear correction. Then, the common nonlinear correction algorithms of PSD, including interpolation algorithm and BP neural network algorithm, are studied, and their advantages and disadvantages are compared and analyzed. The principle of two-dimensional PSD nonlinear correction algorithm based on BP neural network is studied, and its shortcomings are analyzed, and the latest improvement trends of standard BP algorithm at home and abroad are used for reference. In order to improve the convergence speed and reduce the complexity of the algorithm, a new practical improved BP neural network correction algorithm is discussed. The position measurement platform of PSD includes signal processing circuit and control system based on ARM and CPLD. The new algorithm is verified by computer programming and simulation, and compared with the existing nonlinear correction algorithm. The results show that the new improved BP neural network correction algorithm has the advantages of fast convergence speed, high precision, easy realization of the algorithm program and strong practicability.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP183;TP212

【相似文献】