| Mineral is one kind of naturally formed simple substance or compound, with a fixed chemical composition and physical properties, is an important natural resource. However, minerals are present in the ore, and the ore just mined can not be used by the human. Before being smelted or used, minerals need to go through the process of beneficiation process to improve its quality, removing the unwanted gangue, while making useful minerals are relatively enriched. Beneficiation process is composed by the crushing, screening, grinding, classification, flotation and other processes, of which the grinding operations is the most important part which is in the nexus position.The purpose of grinding operation is to reduce the size of ore, making the useful minerals and useless gangue single separation mostly, in order to provide minerals in qualified size to the next processes. A large part of concentrator economic benefits depends on the product quality of grinding operation. At the same time, grinding equipment, the cost of investment in the mineral processing plant also has a large proportion. Therefore, effective and reasonable control on the grinding operation to improve product quality, reduce production costs is the inevitable choice. But the grinding operation is a continuous, cyclical process. operation mechanism is complex, with nonlinear, large delay and other characteristics. The whole process involves many parameters. It is difficult to achieve the desired effect only by manual adjustment. The focus of the research is finding an appropriate control strategy based on the development of control technology and a comprehensive and detail understanding on the grinding process.As time progresses, the computer industry has a rapid development. High-performance computer-based advanced automatic control technology is widely applied to industrial production process. Early application of traditional control methods, including classical control and modern control, both built on the basis of a precise mathematical model, lacking of flexibility and contingency. However, because of the delay, time variability and other feature, it is difficult to obtain an accurate mathematical model for the process of grinding operation. So the traditional control methods face many difficult problems in the practical application.The raise of intelligent control theory solved this problem. Intelligent control theory is put forward for the complexity, nonlinearity, uncertainty of the system, is an artificial intelligence, cybernetics, operations research, and information theory-based disciplines, is the advanced theory stage of development of the control theory. Intelligent control is used for the uncertain objects of control, solving the problem of nonlinear systems, and can meet the complex control requirements. Fuzzy control and neural network is an important branch of intelligent control. Fuzzy control method changes a clear and explicit system model into a vague, uncertain system model, making the control doesn't rely on the accurate value, which makes fuzzy control a powerful tool for uncertainty, nonlinear problems. Fuzzy control based on the experience of mankind, which promotes the research on machine simulation of human brain behavior largely. But there is great distance with the self-learning abilities of the human brain. Neural network uses the computer to simulate of the working mechanism of neurons from the perspective of physiology and psychology to achieve intelligent behavior of the machine. Relatively speaking, neural network has strong learning capacity and parallel processing capabilities. However, unlike the fuzzy control, neural network doesn't have the ability of expressing rule-based knowledge. So the experience and knowledge of experts cannot be made full use of. Therefore, combining the fuzzy control and neural networks, using the strengths of both to compensate for their weaknesses, we will get a better control method.This paper uses a practical project in a Mill as the background, through the analysis of the grinding and classification and the characteristics of them, combining with the characteristic of fuzzy control theory and neural network control, raising a fuzzy logic-based fuzzy neural network technology. The fuzzy neural networks based on fuzzy logic rule-meaning, and select the structure of BP neural network as structure basis. Which makes the network makes full use of expert rules, while also playing a neural network adaptive learning strengths. In addition, the network can adjust the membership function parameters while training itself to fine-tuning the expert rules. This paper is organized as follows:(1) At first, the paper introduces the background of the project. The detail introduction of the grinding process shows the important role of the grinding process in the whole process of the dressing process, which reflects the practical significance of the issue. Following is the introduction of the foreign an domestic research on grinding process automatic control studies and the development trends. And then the intelligent control technologies are introduced briefly.(2) Following is a introduction of grinding processes in detail, and pointed out the complexity of the grinding mechanism and the difficulty of the implement of the control. Through analysis of fuzzy control systems and neural network control system characteristics, we proposed a control strategy - logic-based fuzzy neural network technology. Next is the introduction of some key technologies, including the fuzzy neural network, MATLAB development environment and mixed technique, BP algorithm.(3) Next, based on the identified control strategy, the article focuses on the establishment of the fuzzy neural network, including the identity of fuzzy logic rules and BP network, the detailed derivation process of the learning algorithm.(4) Implement and test the fuzzy neural network in the MATLAB environment. And the test results meet the system requirements.(5)Implement the algorithm in the grinding control system by the technique of hybrid programming of MATLAB and C#. Design and implement the network training module of the system.(6) At last is the summary of the whole article and the later research work prospects.
|
PDF Full Text Request