【作者】 陈晓智；

【导师】 李蓓智；

【作者基本信息】 东华大学 ， 机械制造及其自动化， 2006， 硕士

【摘要】 过程监控是一门由多学科相互交叉相互渗透而形成的新兴学科。它利用各种传感器对设备或系统的运行状态进行监测和控制,借助现代信号处理和分析技术对传感器输出信号进行观测和综合处理,从而获得系统的运行状态信息,由此而进一步对系统进行状态控制、故障诊断和预测。 本文在详细概述与分析过程监控和信号处理技术的特点及应用的基础上,介绍了制造过程信号的采集与信号预处理方法,针对刀具切削过程中声发射信号的特点,提出了基于小波分析的声发射刀具状态评价方法,建立了基于神经网络和模糊理论的多传感器估算模型,实验表明这些模型能有效的对表面粗糙度进行预测,利用灰色关联分析对影响表面粗糙度的因素进行了关联度计算,以此为依据借助神经网络技术提出了基于过程监控信号信息融合的切削参数优化策略,其有效性在实验中得到了验证。 在详细介绍了小波分析的基础理论后,针对信号奇异性的特点与小波变换可以写成函数卷积形式的特性,分析了小波对信号奇异性检测原理与小波分解系数模极大值之间的关系,运用小波多尺度分析的方法,对车削过程中刀具的声发射信号能量频带分布进行了分析,由此提出了以主能量频带分解尺度下的小波系数模极大值均值和均方根作为特征量的评价策略,并通过实验验证了这个方法的有效性。更多还原

【Abstract】 Process monitoring is a new subject which intercrosses and infiltrates with other knowledge. Monitoring the system operation conditions by different kinds of sensors, it can obtain the system operation informations by modern signal processing and analyzing technology, and then it can give the system a diagnosis and prediction.Based on the introduction for the research work about process monitoring and signal processing technology, the engineering signal acquisition and pretrement methods were presented. Aimming at the character of the acoustic emission signals came from the cutting tools, an acoustic emissioin method for tool wear monitoring based on wavelet analysis was gave out, and the estimate models for roughness prediction which based on artificial neural network and fuzzy theory were set up. The experiments proved that the models for roughness predictoin in turning are workable and precise. Therefore, the strategy, by using ANN, for cutting parameter optimization based on process monitoring signals was gave out, and its value is shown in the experiments.The wavelet analysis theory was introduced in detail. Combing the characteristic of the signal’s singularity with wavelet’s specialty thatwavelet transfer can be represented by roll-integral, the relationship between the wavelet singularity detection theory and the wavelet coefficient mudule-maxima was deduced. By wavelet multi-resolution, AE signals’ energy distribution was analyzied. Then extract the character paramecter, MEAN & RMS, by the wavlet coefficient module-maxima of the scale which contained most of the signals’ energy. This method was proved by the experiments.In signal processing technology’s application research, the datas from different sensors were collected and used to discuss the influence from different datas selection models. It guarantees the real signal’s character can be represented by the caculated parameter. Therefore, a multi-sensor signal character fusion model was built up. Regarding the roughness as the target, the suitable character parameters and net work structure were found by experiment comparision. The relevancy between cutting parameters and roughness were analyzed by gray theory respectively. Then the cutting parameter optimization strategy based on process monitoring signals was given out. And it passed the experiments’ detection.更多还原