Research On Soft-sensor Of Oxidation Reduction Potential Based On Data Driven

Biological oxidation of gold is the main technology of refractory gold ore with arsenic and sulfur, and is an important developing trend of green metallurgy with low production cost and less discharge of harmful substances. During the biological oxidation pretreatment process, oxidation-reduction potential(ORP) reflects the bacterial activity and the degree of oxidation reactions between pulp and bacteria in oxidation tanks. High ORP means high activity of bacteria and full oxidation degree of gold ore, which finally obtains a relatively high gold extraction rate. Therefore, many devices are controlled and adjusted by the ORP in the production site. Due to the nonlinear, hysteresis and other characteristics of bacterial oxidation pretreatment process, variation trend of ORP can’t be judged and then failed to guide the industrial production effectively.In order to solve the above problems, this paper have taken the research background of biological oxidation pretreatment process of a gold mine in Xinjiang and put forward the method of soft sensor to measure ORP in order to provide guidance for the process control. The major contents in this paper include:1) The research status of biological oxidation pretreatment and ORP were summarized. After exposition of mechanism of soft sensor, least squares support vector machine(LSSVM) was proposed to achieve ORP during the oxidation pretreatment process.2) The classification of support vector machine(SVM), the theory of support vector regression machine(SVR) and LSSVM were introduced. Artificial bee colony algorithm(ABC), a new artificial intelligence algorithm, was used to optimize parameters of LSSVM model.3) For ABC algorithm, Euclidean distance was introduced to make employed bees and observed bees in the colony take different search strategies. The performance test between the improved and standard algorithms showed that the improved algorithm had a faster convergence speed. So, the improved ABC-LSSVM model was built and tested by nonlinear function and UCI data. The result indicated that the model can obtain a good prediction result.4) The mechanism of bacteria oxidization ore, technological process and the influence factors of biological oxidation pretreatment process were studied. The improved ABC-LSSVM was used to build soft sensor model of ORP. The simulation of the model with field data showed a good result. ORP soft sensor simulation platform of biological oxidation pretreatment process was accomplished by the software of MATLAB.