Research On Soft Sensor Methods For Component Concentration In Hydrometallurgy Extraction Process And Its Applications

There are lots of deposits of low-grade non-ferrous metal resources in our country. With the rapid growth of the national economy, and steady progress of industrialization, it is very important for sustainable development strategy of our country to use such resources effectively and economically. As one of the two extractive metallurgy technologies, the remarkable advantages of hydrometallurgy are high comprehensive recovery rates of valuable metals in raw materials, benefit for environmental protection, and easy fulfillment of continuous and automated production processes. Therefore hydrometallurgy is more suitable for recovering low-grade metal resources. Solvent extraction technology simplifies or cancels a large number of costly solid-liquid separation procedures in hydrometallurgical processe and promotes the rapid development of hydrometallurgical technic. Although extraction technic in hydrometallurgy of our country meets advanced world standards, the control of extraction production process still remains off-line analysis, experience adjustments and manual control, which lead to low efficiency, high resourses comsumption and unstable product quality. And it becomes a bottleneck for hydrometallurgy industrial development in our country.This thesis aims at the difficulty of on-line measursing component concentration in hydrometallurgy extraction process. Based on deeply analyzing the characteristics of extraction production process, and using hybrid modeling method, which is composed of mechanism and data modeling methods, research on soft sensor methods for component concentration in hydrometallurgy extraction process and its applications was carried out comprehensively and systematically. The main researches are summarized as follows:1. Through analyzing the basic principle of solvent extraction, the major factors affecting the extraction equilibrium were briefly analized firstly. Based on the relationship of material balance, the dynamic first principle model of multi-stage counter-current extraction process was established by using the serial structure. On the basis of multi-stage counter-current extraction process model, dynamic modeling of fractional extraction process was also proposed. Data modeling method was employed to describe the extraction equilibrium in process model, with which, the dynamic model of fractional extraction was composed by utilizing the relationship of material balance. By model simulation, the dyncamic and steady state features of the fractional extraction process was revealed. The major affected factors of the process were found. The secondary variables of the soft sensor model were identified. And the foundation of soft sensor model was established.2. To aim at the difficulity of applying the dynamic first principle model to the industrial field directly, soft sensor model of fractional extraction process was established adopting parallel hybrid model. The model was composed of the simplified first principle model and compensated model in parallel, therefore the advantages of different modeling methods can be exerted. The purpose of simplification was to reduce the unmeasured variables in the first principle model, and improve the computational efficiency of the model. The compensated model, which employed the process data, was used to remedy the problem of the decline in prediction accuracy brought about by model simplifying. For building compensated model, a nonlinear data modeling method based on PLS, fuzzy system and subtractive clustering algorithm was also proposed, which can restrain normal distribution noise in the process data, and simplify the architecture of fuzzy system model. By simulation experiment, the efficiency of the hybrid modeling method was verified.3. To aim at the difficulty of process data usually with outliers, nonlinear robust modeling method was used to calibrate compensated model. Due to it is difficult to use traditional methods removing outliers in multidimensional data one by one, a robust nonlinear data modeling methods based on RBF neural networks were proposed firstly. The nonlinear problem in the low-dimensional space was change into linear one in the high-dimensional space by RBF transforming. Then the different robust PLS algorithms was used to train the parameters in the model. On this basis, another robust KPLS algorithm was proposed. Since the algorithm adopted the kernel method, neither involving the dimension of feature space, nor needing to know the specific form of nonlinear mapping. Therefore it can be used to build compensated model. Additionally, when the built soft sensor model is applied to practical industrial production process, the model correction needs to be introduced to expend its using range. Thus a model correction strategy composed of long-term and short-term correction was presented in this paper. By alternately using the two correction methods, the prediction accuracy and practicality were futher enhanced. Employing the proposed robust hybrid method and correction strategy to predict the industrial data of copper extraction process, satisfied results were obtained.4. To aim at the problem of optimal operation of extraction process in hydrometallurgy, based on the soft sensor model, the optimal operating guide was proposed to improve product quality and reduce the consumption of the accessories by using particle swarm optimization algorithm. Taking an extraction section in a cobalt hydrometallurgy plant as a. research example, since the original automation level of the production conditions was low level, it is required to design basic automation system, which composed of executive layer, control layer and management layer. Under the support of basic automation system, based on the above theory studies, optimal operating system software of extraction section was designed and developed. Soft sensoring for component concentration of extraction process was achieved. And the optimal operating guide for this process was provided. Applying the software to practical extraction section of some hydrometallurgy plant, good economic returns were yielded.