基于BP神经网络多用炉炉温预测控制

发布时间：2018-03-24 18:29

本文选题：多用炉　切入点：预测控制　出处：《辽宁科技大学》2015年硕士论文

【摘要】：近年来,随着工业控制技术的提高,航空航天技术、风电、交通运输、机械等行业迅速发展,对工件的精密热处理要求越来越高。热处理在一定程度上,直接反映我国的工业水平。温度、碳势、氮势等是影响产品热处理质量的重要指标。热处理炉的先进控制系统、可控气氛热处理、真空热处理得到了广泛关注和快速发展。可以在生产过程中,实现对各种参数的精确控制,使工件无氧化、无脱碳,表面质量好,畸变小。热处理炉是一个复杂的被控对象,存在着非线性、时变性、纯滞后因素和不确定随机干扰等因素,因此基于精确数学模型的控制方法受到很大限制,传统的PID控制难以满足温度控制精度和升温实时性要求。本文在这样的背景下,研究了热处理的温度优化控制方法。综述了预测控制和神经网络的理论和应用,并以某公司的可控气氛多用炉为对象,针对其多变量、非线性、大惯性和纯滞后的特点,结合神经网络和预测控制算法,提出了基于BP神经网络的预测控制算法。建立多用炉炉温预测模型,并且在模型的基础上,对BP算法进行优化,利用现场数据对模型训练,以预测未来的炉温输出值。结合预测控制算法,采用反馈校正,克服系统中其他干扰造成的模型误差,然后对控制量进行滚动优化,得到系统最优控制序列。通过MATLAB软件仿真,验证模型的预测输出,在可控误差范围内,验证模型有效性。对预测控制算法的主要参数进行调整,选取最优控制参数,取得了较快的响应速度,基本不存在超调。把该模型应用到某公司1号多用炉温度控制系统中,取得了较好的实践效果。
[Abstract]:In recent years, with the improvement of industrial control technology and the rapid development of aerospace technology, wind power, transportation, machinery and other industries, the requirements for precise heat treatment of workpieces are becoming more and more high. Temperature, carbon potential, nitrogen potential and so on are the important indexes that affect the quality of heat treatment of products. The advanced control system of heat treatment furnace, heat treatment in controlled atmosphere, Vacuum heat treatment has been widely paid attention to and developed rapidly. The precise control of various parameters can be realized in the process of production, so that the workpiece has no oxidation, no decarbonization, good surface quality and small distortion. The heat treatment furnace is a complex controlled object. There are many factors, such as nonlinear, time-varying, pure lag and uncertain random disturbance, so the control method based on accurate mathematical model is greatly restricted. Traditional PID control is difficult to meet the requirements of temperature control precision and real-time heating. In this paper, the optimal temperature control method of heat treatment is studied, and the theory and application of predictive control and neural network are summarized. Aiming at the characteristics of multivariable, nonlinear, large inertia and pure lag, combining neural network and predictive control algorithm, a controlled atmosphere multipurpose furnace of a certain company is taken as an object. A predictive control algorithm based on BP neural network is put forward. A multi-purpose furnace temperature prediction model is established. On the basis of the model, BP algorithm is optimized, and the model is trained with field data. In order to predict the output value of furnace temperature in the future, combined with predictive control algorithm and feedback correction, the model error caused by other disturbances in the system is overcome, then the control quantity is rolled optimized to obtain the optimal control sequence of the system. In the range of controllable error, the validity of the model is verified. The main parameters of the predictive control algorithm are adjusted, the optimal control parameters are selected, and the fast response speed is obtained. The model is applied to the temperature control system of No. 1 multipurpose furnace in a company, and good practical results have been obtained.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TK175;TP183

【相似文献】