摘要
在煤用重选设备评定的国际标准（ＩＳＯ９２３）和以之为蓝本的国家标准（ＧＢ／Ｔ１５７１５）中的第一项评定指标“可能偏差”及和其有关的“不完善度”的确定是由以下步骤决定：１．由仅仅一次浮沉试验取得一组数据（一个观察值）并由此计算两段分离所得产品的分配率；２．用手工凭想象中的“Ｓ型曲线”把分配率中的两组６ 至８ 个点联起来得到两条“分配曲线”；３．由这两条粗糙的曲线“量出”各自的２５％ 和７５％ 分位点，以此得到能反映重选设备分离能力好坏的两个指标：“可能偏差”Ｅ和“不完善度”Ｉ，很难想象，这样由一个观察值得到的指标并通过没有模型的手工绘图而得到的结果会有任何实际意义，本文建议利用尽可能多的观察值来拟合ｌｏｇｉｓｔｉｃ回归模型，并依此得到分配曲线和计算出参数Ｅ和Ｉ。
Summary Evaluating the performance of a jig,a gravity separating equipment in coal-cleaning process,concerns many fundamental statistical issues,such as population,sample,model-fitting,outliers,etc.We will see how these statistical issues were misunderstood by many,not only field engineers but also industrial standard draftsmen.This study mainly discusses problems surrounding how to fit the partition curves,related to criteria of evaluating the equipment,with various data,including those“unqualified”data sets.We suggest a sensible logistic regression model in contrast with commonly improperly believed “normal-like”curves.Through examples,we describe our process of model fitting with“unqualified”real data sets containing outliers.
出处
《数理统计与管理》
CSSCI
北大核心
2000年第1期34-38,共5页
Journal of Applied Statistics and Management
基金
中国国家自然科学基金!# 19671046
关键词
LOGISTIC模型
煤
重力选矿机
跳汰机
统计模型
and phrases:Clean coal,density fraction,high-density cut,jigging,logistic regression,low-density cut,outlier,middlings,partition coefficient,reject.
