Abstract: Pulp grade measurement has always been the research focus and technical difficulties in mineral processing industry， the accurate measurement of pulp grade with online feedback directly influences the effect of automatic beneficiation process. In order to achieve more accurate measurement of element grade in pulp during ore dressing， the new generation nuclear-free X-ray fluorescence analyzer was modified and developed based on the last generation nuclear online grade analyzer. The new type X-ray fluorescence online grade analyzer is based on STM32F407 control chip and the measurment structure combine the X-ray light pipe and semiconductor detector. Its technical architecture and analysis principle are as follows: X-ray light pipe and semiconductor detector constitute measurement unit， and the X-ray excited by X-ray light pipe irradiates to the surface of the measured slurry. Meanwhile， the semiconductor detector receives the X-ray characteristic fluorescence excited by the measured element in the slurry. The main control unit receives and processes the data signal uploaded by the semiconductor detector， and sends count rate of the measured element to the upper software of industrial computer through serial communication. Finally， upper software calls Matlab mathematical analysis tool and adopts BP neural network for data modeling analysis and grade value acquisition. The field test results of a concentrator in Shaoxing showed that the measurement accuracy and operational stability of the new type of X-ray fluorescence online grade analyzer are significantly improved compared with the previous generation nuclear grade analyzer，and all the parameters meets the design requirements.

Key words: Grade measurement, STM32, X-ray light pipe, Semiconductor detector, Data modeling, BP neural network