基于纹理单元分布的硫浮选过程病态工况检测  

Fault Condition Detection for Sulfur Flotation Process Based on Texture Unit Distribution

作　　者：凌弈秋[1] 阳春华[1] 何明芳[1] 桂卫华[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《计算技术与自动化》2013年第1期28-31,共4页Computing Technology and Automation

基　　金：国家杰出青年科学基金项目(61025015);国家自然科学基金重点项目(61134006);国家科技支撑计划项目资助(2012BAF03B05)

摘　　要：针对硫浮选过程难以及时发现病态工况的问题,提出一种基于泡沫图像纹理单元分布函数的硫浮选过程病态工况检测方法。通过计算纹理单元值提取泡沫图像的纹理特征,避免传统灰度共生矩难以准确描述纹理特征的问题,设计核函数估计算法逼近泡沫纹理单元概率密度函数(PDF),使得纹理单元分布函数转化成一组动态权系数。建立基于纹理单元分布动态权系数的主元模型,并基于该模型的T2统计量得到检测阈值,从而实现硫浮选过程病态工况的及时检测。湿法炼锌直接浸出硫浮选工段现场运行结果表明此算法的有效性。Considering the difficulty to detect the fault condition of sulfur flotation in real-time, a new fault detect meth- od based on probability density function （PDF） of froth texture unit is proposed. The froth texture unit is used to describe the froth texture features to avoid that the gray level dependence matrix can not describe the froth texture features accurate- ly. By means of the designed kernel method,the PDF is transformed into dynamic weight coefficients, based on which a prin- cipal component analysis （PCA） model is established. And the T2 statistical quantities based on PCA model can be used as a threshold criterion to realize the fault condition detection in sulfur flotation process. Simulation results show the effectiveness of this algorithm by using actual running data from the sulfur flotation process of zinc direct leaching hydrometallurgy.

关 键 词：纹理单元 核密度估计 主元模型 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]