Abstract:Flotation production generates massive amounts of data, from which rules knowledge that people may not know in advance but are potentially useful can be mined. Empirical knowledge can be obtained by summarizing relevant literature or domain experts, expressed in a certain knowledge representation, and stored in a knowledge base or model base according to its representation. This paper mainly studies the theoretical and real-time prediction of foam grade based on the nature of raw ore and foam characteristics, establishes the prediction model, and summarizes a set of control logic based on "raw ore grade determines the yield control interval, concentrate grade difference is corrected, and foam pump pool level coordinated control", which realizes the stable control of concentrate yield; At the same time, research on foam characteristics and dosage based on big data mining algorithm was carried out to achieve intelligent regulation of dosage. Finally, a flotation expert system is built to match the real-time database data with the rule knowledge in the knowledge base and the data models in the model base under the action of the inference machine, according to certain inference methods and control strategies, thereby achieving intelligent adjustment of factors affecting flotation production such as flotation liquid level, aeration volume, and reagent dosage. The stable use of this expert system in a domestic beneficiation plant has greatly reduced the workload of frontline personnel, improved production stability, reduced the volatility of coarse and fine grades, and increased the recovery rate of metal minerals.