摘要
为了解决神经网络逆系统软仪表对噪声敏感的问题,本文提出了一种数据预处理方法。该方法先采用两步判断法对数据进行一次处理,再用滑动平均滤波方法对数据进行二次处理,有效地滤除了噪声信号,比较精确地复现了原始信号。对红霉素发酵过程中的pH值过程变量进行了实验,结果表明了该算法的有效性和可行性。
A data preprocessing approach is presented to solve the problem of sensitiwty to noise in soft sensing by the artificial neural network （ANN） inversion based soft sensor. The data is first pretreated by the two-step-judgement method and then reprocessed by the moving average method. Thus the noise can be filtered effectively and the primary signal can be restored precisely. After making the experiment on the pH value in erythromycin fermentation, it shows that this method is a very effective one.
出处
《自动化与仪器仪表》
2006年第3期7-10,72,共5页
Automation & Instrumentation
基金
国家自然科学基金(60574097)
高等学校博士学科点专项科研基金(20050286029)资助
关键词
神经网络逆系统
软仪表
预处理
两步判断法
Artificial neural network inversion
Soft sensor
Preprocessing
Two-step-judgement method
