文档详情

基于人工智能的硬化层深预测系统-检测技术与自动化装置专业论文.docx

发布:2019-03-27约3.54万字共51页下载文档
文本预览下载声明
摘 摘 要 金属表面硬化层深的检测一直是机械工业部门急需解决的问题,涡流 法作为一种无损检测方法相较于以往的破坏性检测有更高的可靠性和经济 效益。然而用涡流法对它进行检测的最大难度在于检测信号的处理。由于 涡流的检测层深和灵敏度受频率影响很大,单一频率检测不能满足不同层 深的需要,本文采用多频涡流检测的方法,以使信号中充分包含有用信息。 信息融合作为信息领域的一项高新技术,目前主要用于多传感器融合 中。它是把多个信息源所得到的冗余或互补信息,依据某种准则进行组合, 以获得对被检测目标的一致性解释或描述。信息融合对于多信息的处理能 力已得到了广泛应用,本文对于它在多频涡流信号处理中的应用作了系统 的研究,首次提出了不同数量级信号融合的方法。 本文利用信息融合将多个频率下测得的涡流信号进行融合。采用模糊一 遗传算法,信息融合过程中的推理由模糊聚合函数完成,模糊聚合比传统 的集合论中的并和交操作能更好地模仿人的推理。模糊聚合函数的连接参 数由遗传算法确定。该算法的优点是在信息源的可靠性、信息的冗余性/互 补性不确定的情况下能够以近似最优的方法对信息进行融合。 通过信息融合,本文最后得出了一个硬化层深的预测系统,并通过待 测试件验证了预测的可靠性。 关键词: 涡流检测 硬化层深 信息融合 模糊聚合 遗传算法 茎坚堡三查堂堡主堂堡垒兰—————————————一ABSTRACT 茎坚堡三查堂堡主堂堡垒兰—————————————一 ABSTRACT Nondestructive testing of metal hardened·depth is afl important problem to be solved in machinery industry.Eddy current is an effective method in which the most difficult task is the processing ofthe signal.Since the penetrative depth and the testing sensitiveness are influenced seriously with the testing frequency. So multi-frequency eddy current is used in place of single—frequency eddy current to involve fall information of the objects. Data fusion iS a method to combine the redundant and complemented data of the multi—source by some rules.As a result,the consistency interpretation and description is obtained.This paper presents a method to fusion the signal of the eight—frequency eddy current.A fuzzy-genetic data fusion algorithm is utilized. The reasoning is performed by means of fuzzy aggregation functions,the fuzzy aggregation being capable of better set theories.,蚤e parameters of the connectives ale found by genetic algofithms.The distinctive feature of this algorithm is its capability of fusing data in a near-optimal manner when there exists no information about the reliability of information sources,the degree of redundancy/complementarities of the information sources,and the structure of the hierarchy. At the end,all example is given to test the reliability of the forecasting system. Keywords:eddy current testing hmxlened-d
显示全部
相似文档