Operational Feedback Control Approaches For Complex Grinding Processes And Their Applications

In metallurgical mineral processing, grinding circuit is the prerequisite procedure for many kinds of metal beneficiation. The grinding process is used to reduce the particle size of the ore such that the valuable mineral constituent is exposed and can be recovered in the subsequent beneficiation operation. As a typical process with high energy consumption and low efficiency, the grinding process accounts for 45-70% power consumption and 40-60% production cost of the whole concentrator. The grinding operational indices, such as the product particle size and the grinding production rate, are the key metrics indicating the grinding product quality and process efficiency, and directly affect the performance of the whole concentrator in terms of the product concentration grade and the production capacity. For a long time, because control and optimization are an important means to improve grinding quality and production efficiency, as well as to enhance the economic profit of the concentrator, they have been attracted much attention by scholars. Many popular control technologies, such as PI/PID control and model-based predictive control, are widely used to regulate grinding equipment in accordance with the desired goals. However, from the standpoint of hierarchical control in process engineering, The automation of grinding process has not only made the outputs of the controlled processes best follow their setpoints, but also achieved that the operation of the whole plant is well controlled so that the operational indices (such as grinding particle size and efficiency during production phase) are well controlled within their targeted ranges. This means that the optimal process operation should be realized. Therefore, it demands operational feedback control (OFC) for the whole plant composed by the complex grinding production and the running grinding equipments, so as to achieve integrated control and optimization for the control indices, the operational indices, and the economic performance index. This is also an important way to promote the competitiveness of mineral processing companies.Mineral grinding is a typical process with long processing units and large time delay. Moreover, it is affected by many factors which are coupled to each other. The grinding operational indices are closely related to the key process variables, such as fresh ore feed rate, water addition rate, and so on. Moreover, the dynamic relations between these variables are generally very intricate and affected by composition and nature of the processed ore, production technical specifications and operating specifications, operating conditions, and various unknown dynamic interferences. All these factors vary with time. Therefore, OFC for complex grinding processes is challenging. At present, the relevant theoretical approaches are still few. Most of these methods adopt the operational control mode of higher-level open-loop setting plus lower-level basic feedback control, thus in no real sense they achieves higher-level feedback control for the grinding operational indices. Moreover, this operational control method with open-loop optimization setting at higher-level can not adjust the operating process according to the change of working condition in-time, thus the optimization operation of grinding process can not be achieved. In this case, it is necessary to use the real-time or intermittent feedback information on operational indices to explore OFC approaches for grinding processes so as to achieve closed-loop feedback on operational indices at higher-level.Motivated by the urgent needs for improving product quality and production efficiency as well as energy saving, this dissertation will study on OFC for complex grinding processes. The studied approaches adopt hierarchical feedback control structure, which include the higher-level operational feedback control and the lower-level basic feedback control. This dissertation is supported by the National Natural Science Foundation of China (61104084), the National Basic Research Program of China (2009CB320 601), and the Fundamental Research Funds for the Central Universities of China (N090608 001). The main contributions of the work are summarized as follows:1) For the grinding circuits that the processed ore has relative stable composition, and approximate dynamic process models can be established, the model-based OFC method are studied and tested on a two-stage grinding circuits composed of a rod mill in open-circuit at the first-stage, and a closed-circuit with a ball mill and a hydrocyclone at the second-stage.A. Many industrial processes (such as the grinding system) are characterized by multiple inputs (fresh ore feed rate, water addition rate, etc.) and multiple outputs (grinding production particle size and circulating load, etc.). Moreover, it generally has large time delay in output measurements. Here,2-DOF analytical decoupling is first improved to deal with the large time delay in output measurements. Then the improved 2-DOF analytical decoupling method is extended to take OFC for the above mentioned grinding process. Because the model of controlled operational process is usually of multivariable and multiple I/O delays, a multiple-point step response fitting based model approximation method is proposed to simplify control system design and to help improve the stability and physical realizablity of the operational control system.B. As the model predictive control (MPC) based OFC does not handle the disturbances directly by control system design, it cannot achieve satisfying performance in controlling the complex grinding processes in the presence of strong disturbances and large uncertainties (such as the variations of ore hardness and feed particle size, and process coupling effects). Hence, by introducing the disturbance observer (DOB) technology, the integrated OFC based on DOB and MPC is investigated, which mainly includes:◇ Considered that the existing DOB design technique only suits for the minimum phase delay or undelay system, an improved DOB design with the considerations of time delay and non-minimum phase is presented. Then an integrated OFC framework based on the improved single-variable DOB (IDOB) and MPC is proposed for process operation. Simulation examples show that the grinding system with the proposed control scheme can reject the disturbances more effectively than that with the single MPC based operational feedback control scheme.◇ The common feature of the conventional DOBs is that they are designed for the SISO systems. In this thesis, an approximation inversion based multivariable disturbance observer (MDOB) is proposed for the MIMO delay systems. Based on it, an MDOB-MPC based compound OFC is presented. Compared to the conventional single MPC scheme, the IDOB-MPC scheme, and the improved 2-DOF decoupling scheme, the designed MDOB-MPC scheme can not only achieve much better performance of disturbance estimation and rejection, but also make that the performance of the setpoint tracking and decoupling of the proposed scheme be not debased compared to the single MPC scheme.2) For the complex industrial grinding processes, the composition and nature of the processed ore is of complex mineral composition and fine dissemination particle size, and the operational indices are generally difficult to be measured online. The associated dynamic characteristics are very intricate, involving complicated nonlinearities and strong cross couplings. Furthermore, accurate dynamic models are difficult to be obtained due to unknown mechanism. Therefore, the data-driven intelligent OFC approaches are investigated in this dissertation for this kind of complex grinding processes.A. The typical two-stage closed-circuit grinding process is widely used in China. Here, a novel OFC framework, consisting of a CBR-based loop pre-setting controller, an ANN-based PPS soft-sensor module and a fuzzy dynamic adjuster, is established for such a complex grinding process. This OFC control strategy integrates process data and operation knowledge, and uses hierarchical feedback control structure. The proposed approach has been successfully applied to a typical two-stage grinding process of a hematite processing plant in China. As a result, the grinding product quality is optimized, the mill throughput per unit and the operative ratio of mill are increased by ～2.78% and ～4.42% respectively, and the grinding economic benefit is improved more than ～930 million RMB per year.B. During the operation of the closed-circuit grinding process consisting of a ball mill and a spiral classifier, overload is a commonly occurred faulty condition that affects process safety. Focus on these practical problems, a novel intelligence-based OFC is proposed for optimal and safe operation of complex grinding processes. This control strategy includes setpoint optimization of the process control system, CBR-ANN based online soft-sensor for operational indices, multivariable intelligent feedback adjustment under normal working conditions, and diagnosis and adjustment of mill overload faulty condition using data-driven SPC and knowledge-driven RBR. The proposed approach has been applied to the grinding circuit unit of a hematite processing plant in China. As a result, the overall operational efficiency of the grinding production in terms of optimization of operational indices and its reliability and stability for minimum operational breakdowns has been improved significantly.