Horseshoe Flame Glass Furnace Key Technology

Regenerative horseshoe flame glass furnace is used widely in our daily glass industry. Compared with other furnaces, it has a pair of regenerators, smaller kiln volume, relatively low investment,a small amount of furnace heat. Glass furnace as a high-energy-consuming equipment, high-yield, high-quality, high life, low energy consumption and low pollution problem in the production process have been concerned by the domestic and foreign manufacturers. To achieve the "three high and two low" target, ingredients and furnace controlling must be done. Therefore, the theoretical analysis and research of the process of mixing of the ingredients, the mathematical model of the furnace and the method of the controlling is necessary which can provide a reliable theoretical basis and technical support for different furnace design theory and production applications in different technological requirements.The regenerative horseshoe flame glass furnace as the research object in this paper, structure and working principle of the kiln are researched. Under the premise of the estimated and the actual working conditions of the measured temperature distribution, three-dimensional mathematical model of the turbulent flow field is represented by the turbulence equation, momentum equation, continuity equation. On the basis of the convergence of the flow field, heat temperature field model calculations is completed by gas flow data in different locations of regenerator, further operation of the regenerator is analyzed. The parameters are interrelated and influenced by each other during furnace operation, which will increase the difficulty of controlling. Temperature and pressure coupling system is decomposed into two independent single-variable system by invariance principle decoupling algorithm. The temperature system decoupled as the research object, the system can be represented as controlled auto-regressive model, whose input and output are the fuel flow rate,temperature measurements. In order to prevent one-sidedness of a single identification method, batch least squares, recursive least squares method, forgetting factor recursive least squares parameter estimation method and the gradient correction is applied separately during the identification of the model. To verify the accuracy of the model,cross validation is used to test the model, in other words, the new data collected is used for the model. In addition, the merits of the model identified is determined by the the difference between the model output and the actual output data. At the same time depending on the application of different models, model parameters is eventually decided. Since the new model form identified is differential equation of discrete system,to facilitate the study of the role of the system in response to the control system,first of all, according to the real number theorem of displacement, the differential equations are transformed into a mathematical model in the z domain(Tsez ?), and then be converted to mathematical model in the complex domain.Furnace temperature mathematical model is analyzed in real field and the complex domain respectively. The results show that self- balancing objects have steady-state error under the action of the unit step function. At the same time due to the lack of controlling action of the regulator, regulation time, rise time, steady-state time of the system is too long. Therefore, the regulator must be added to make the response of the system to achieve the desired results. The different aspects of PID controllers are applied to the controlled object, the dynamic performance of the system has been improved, but adjustment of the controller parameters is still relatively complex. In parameter adjustment, IFT adjustment method and the traditional PID parameter adjustment method are compared. Matlab simulation results show that performance indicators of the system are better with IFT regulation. The IFT method, improved PID controlling method, two-dimensional fuzzy controller temperature adaptive controlling method and fuzzy neural controlling method based on the Mamdani model being applied to the controlled object respectively, simulation results show that intelligent controlling method based on FNN have more advantages in the adjustment of controller parameters, because it can significantly shorten the settling time, rise time,reduce the overshoot and enhance the stability of the system.In design of batching system, aiming at the complex batching processin, low level of automation in the ingredients industry, PLC, intelligent ingredients instruments andCX-Protocol are applied in batching process. At the same time, the message between PLC and intelligent ingredients instruments is expressed by CX– Protocol. Practice shows that new type of automatic batching system can reduce labor, improve work efficiency and save the costs for the enterprise. Aiming at the problem of low accuracy of ingredients in the process of batching, nonlinear function extremum optimization method based on neural network and genetic algorithm has been applied in the optimization of mixing uniformity. The function relationship between mixing uniformity and its influence factors is established through the neural network. In the genetic algorithm, individual fitness value can be expressed by BP neural network predictive value, in addition, the best mixing uniformity value can be obtained by global optimization of genetic algorithm. Matlab software simulation analysis show evolutionary BP neural network has high prediction accuracy, at the same time,extremal optimization method can improve the mixing uniformity of the ingredients,which can create considerable economic benefits for the glass furnace enterprise.