Determination of chemical elements assay plays an important role in mineral processing operations.This factor is used to control process accuracy,recovery calculation and plant profitability.The new assaying methods including chemical methods,X-ray fluorescence and atomic absorption spectrometry are advanced and accurate.However,in some applications,such as on-line assaying process,high accuracy is required.In this paper,an algorithm based on Kalman Filter is presented to predict on-line XRF errors.This research has been carried out on the basis of based the industrial real data collection for evaluating the performance of the presented algorithm.The measurements and analysis for this study were conducted at the Sarcheshmeh Copper Concentrator Plant located in Iran.The quality of the obtained results was very satisfied;so that the RMS errors of prediction obtained for Cu and Mo grade assaying errors in rougher feed were less than 0.039 and 0.002 and in final flotation concentration less than 0.58 and 0.074,respectively.The results indicate that the mentioned method is quite accurate to reduce the on-line XRF errors measurement. Determination of chemical elements assay plays an important role in mineral processing operations. This factor is used to control process accuracy, recovery calculation and plant profitability. The new assaying methods include chemical methods, X-ray fluorescence and atomic absorption spectrometry are advanced and accurate. This paper, an algorithm based on Kalman Filter is presented to predict on-line XRF errors. This research has been carried out on the basis of based the industrial real data collection for evaluating the performance of the presented algorithm. These measurements and analysis for this study were conducted at the Sarcheshmeh Copper Concentrator Plant located in Iran. The quality of the obtained results was very satisfied; so that the RMS errors of prediction. obtained for Cu and Mo grade assaying errors in rougher feed were less than 0.039 and 0.002 and in final flotation concentration les s than 0.58 and 0.074, respectively. These results that the mentioned method is quite accurate to reduce the on-line XRF errors measurement.