Dynamic Control Model For The Secondary Cooling Of Slab Casting Based On Neural Networks And Genetic Algorithm

Secondary cooling water controlling is closely related to the quality of continuous casting slab in the process of high efficiency continuous caster. The quality of slab such as the formation cause of internal cracks,surface cracks,shape defect,bulging and central segregation etc is unreasonable and inhomogeneous cooling in the secondary cooling zone.Secondary cooling zone in slab caster is usually composed of spraying segments. For each spraying segment target temperature and controlling water is different from those in other segments. Moreover,And the process of production is often of complexity and uncertainty. Heat transfer mathematical model is the foundation of controlling and optimization in the process of continuous casting. The secondary cooling water controlling based on heat transfer mathematical is widely used, and it can simulate the producing in the process of continuous casting. However, when considering many factors affecting the production, significant input changing, the real-time response of the model becomes difficult. Therefore, application of the artificial intelligence optimization algorithm such as Genetic Algorithm(GA),neural networks, fuzzy controlling to dynamic distribution and optimize of water is necessary for reasonable water distribution and dynamic controlling in secondary cooling zone in continuous casting.In this thesis,secondary cooling water controlling and optimization based on heat transfer mathematical model in secondary cooling zone of slab caster is investigated. Genetic algorithms is used for optimizing each segment water in secondary cooling zone which integrated with heat transfer mathematical model on the basis of the criteria for the metallurgical constraints, determining the speed-water experience parameters of static distribution water, in order to offer important reference to the further optimization of water distribution. Fuzzy logical technology integrated with heat transfer mathematical model is used to optimize the water distribution under given speed and control water distribution in variable speed. Back-propagation neural network is used to identify the heat transfer mathematical model and predict the surface temperature. Fuzzy neural networks is used to realize the dynamic distribution and controlling of water and solve the problems when fuzzy logic controlling water distribution cannot learn. Back-propagation neural networks is used to predict the surface temperature in the slab casting and fuzzy neural network is used to dynamically control water in secondary cooling zone to decrease the target temperature and calculate temperature by back-propagation neural networks.Interface design of dynamic distribution model of casting slab gives prominence to availability, flexibility and reliability based on the windows software platform using Microsoft Visual Basic 6.0 programming. Software testing consists of online and offline testing. The object is secondary cooling water controlling in slab caster No.2 in Vanndium-extracting and Steel-making Plant of Panzhihua Iron and Steel Corporation(PISCO) and controlling target is the surface temperature from the secondary cooling zone exit. The offline testing data is extracted from scene of history database in PISCO. The online testing data is extracted from real-time database in PISCO and carry through the suface temperature testing of slab measurement simultaneously from the scene, in order to complete the methods testing and software testing.The testing and application results of the software of secondary cooling dynamic distribution water model of casting slab show that: the model which based on genetic algorithms can decrease differences between the target surface temperature and surface temperature, and heat and cooling rate also tends to be more reasonable, satisfy the requirement of improving slab cooling process and improve product quality under metallurgical criteria. Fuzzy neural network can quickly response to the requirement of distribution water in secondary cooling zone when speed and target temperature changes. The model of intelligent control based on the back-propagation neural network and fuzzy Gaussian neural network can overcome the problems such as heat transfer coefficients which is difficult to determine, muti-variable, external interference in heat transfer model calculations. The measured temperature can be used to amend the back-propagation neural network forecasting temperature and the model error can be limited within 5℃. The model system for improving dynamic distribution water in secondary cooling of slab caster and improving the secondary cooling control of the intelligent level, further enhancing the quality of slab are realistic significance as well as later application.