摘要
提出用 ML RR法对矿石、焦炭分布进行图象重建。该方法借助于有限元分析软件 ,利用多元线性回归实现 ECT系统正向问题的求解 ,再利用正则化方法求解反向问题 ,获得成象矩阵。实验表明 ,在区域性分布对象的成象上 ,ML RR法的成象质量远远好于 L BP法。当矿石、焦炭的介电常数在± 2 0 %变化时 ,对 ML RR法所成图象质量影响不大 ,十分有利于热态下的高炉成象。
An algorithm, MLRR, for imaging the distribution of cokes and ore in a blast furnace is presented. The method used multi-linear regression to solve the forward problem and then regularized the solution of the forward problem to get the image reconstruction matrix. The experiment showed that using MLRR algorithm could get much better images than by LBP algorithm. Moreover, the quality of the images reconstructed by using MLRR algorithm is robust to the permitivity change of cokes and ore if the change is within ±20%, which is very beneficial to blast furnace imaging in hot state.
出处
《控制与决策》
EI
CSCD
北大核心
2000年第2期201-204,共4页
Control and Decision
基金
国家自然科学基金项目!(5 96 740 16 )
