传感器可靠性研究与信息处理技术

项目名称： 传感器可靠性研究与信息处理技术

项目编号： No.60876077

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 白艳萍

作者单位： 中北大学

项目金额： 25万元

中文摘要： 本课题通过分析MEMS与宏观设备可靠性要求的差异，从微结构本身和微系统的装配与封装等对寿命的影响，失效的主要原因等探讨MEMS传感器中特定可靠性因素和失效机理和失效模型。通过对电容式MEMS微型加速度计失效分析，结合数学统计和各项失效分析技术，掌握电容式MEMS微型加速度计在结构级失效的各种失效模式。在MATLAB7.0 环境下，先利用多尺度小波变换算法对实验所得的微悬臂梁加载力和挠度的原始数据进行了消噪预处理，接着建立了微悬臂梁加载力与挠度关系的BP神经网络模型，拟合和分析了微悬臂梁的加载力与挠度关系。用改进的BP神经网络与遗传算法对传感器测得的信号进行预测及模拟试验比较，经过一系列的解算处理，得到较可靠的预测结果。针对加速度型振动传感器工艺的不同结构进行分析,讨论工艺结构对传感器失效的影响, 提出各失效模式造成结构总体失效的因素。 研究了车牌识别系统， 包括基于颜色特征和纹理特征的车牌定位，基于灰度处理、二值化和去除车牌的边框和上下铆钉的图像处理、基于改进的水平投影的字符分割、基于人工神经网络识别车牌字符的字符识别。这个系统行之有效地解决了汉字的不连通性、字符的粘连问题。

中文关键词： MEMS可靠性；数据融合；信息处理；模式识别；神经网络

英文摘要： By analyzing the macro and MEMS device reliability requirements of the differences from the micro-and micro-structure of the system, such as assembly and packaging of life, such as the main reason for the failure to explore the reliability of MEMS sensors in a particular failure mechanism and establish a complete reliability assessment system and failure model. By using of cantilever mechanical properties and the theory of formula, we analysis of the cantilever structure of the size effect. When the micro-size components close to the grain-scale materials, due to surface effects, organizational structure and process, mechanical properties will change. we creat the BP model which studying the relationship of load and deflection in the environment of matlab7.0. Then,we simulate and predict the relationship of load and deflection of different parameters micro-cantilever beam by wavelet multi-scale arithmetic analyzing the experimental data. By the analysis of the yield and failure analysis, found deficiencies in the design and technology, in order to rationalize and optimize the design of an objective basis for improving the reliability of the product guide.We adopt the regression model with a genetic algorithm for sensor data fitting. The results show: the ARMA model with genetic algorithm can be adjusted through genetic algorithm's accuracy in different parameters The model have been fitting parameters, which according to different requirements of precision fitting curve provides an effective method. And we study the reliability predictable model of micro-cantilever beam for static load. We has studied a lot of method and theory about vehicle plate location, character divide and character recognition systemically，Which is Including color features and the texture characteristics based on gray level processing of binary and remove the frame of vehicle license plate and context of the rivet and image processing。This system solve Chinese characters don't connectivity adhesion of the characters effectively.

英文关键词： MEMS reliability;data fusion;information procession;pattern recognition; neural network