在线X射线荧光分析仪建模及模型校正方法的研究

BOXA型载流X荧光品位分析仪采用X射线荧光技术对选矿流程中矿浆内金属元素品位值做在线实时检测。目前常用最小二乘法进行线性拟合建模,但是此方法在各个通道存在线性相关性时会失效,同时BOXA分析仪配置波长+能量双通道,当测量通道发生故障或者测量通道升级更换之后,原模型不再适用,重新标定时间过长。本文提出使用Bisquare建模的方法,模型精度以及对异常点的敏感性均优于最小二乘法,同时提出一种模型校正的方法,将模型校正到与原模型相近的精度,可以大大地节约时间,尽快恢复使用。通过原矿和锌尾矿进行实验验证,结果表明校正模型的精度与原模型基本相当,对指导选矿厂工艺流程有着重要的作用。

Modeling and model correction of on-stream X-ray fluorescence analyzer

BOXA on-stream X fluorescence grade analyser adopts X-ray fluorescence technology to conduct online real-time detection of the grade value of metallic elements in beneficiation flow. Ordinary Least Square is commonly used for linear regression analysis and fitting modelling, but this method will fail when there is linear correlation between each channel. When the measurement channel get replaced on account of malfunction or upgraded, the original model is no longer applicable, and it takes too long for re-calibration and modelling, during which the beneficiation flow cannot be guided, which will cause a long vacuum period for users. In this paper, In this paper, Bisquare is proposed, and the model accuracy and sensitivity to outliers are better than the least square method. And a model correction method is proposed, which can calibrate the model to the similar accuracy to the original model. It can greatly save time, and restore the analyser as soon as possible by using the model correction method. Through the experiment verification of raw ore and zinc tailings, the results show that the accuracy of the correction model is basically the same as the original model, which can avoid the vacuum period caused by the replacement of detectors and plays an important role in guiding the technological process of beneficiation flow.