基于多方法集成的复杂工艺优化设计研究

针对流程工业复杂的连续生产工艺的特点,提出了基于模式识别、神经网络、遗传算法、非线性回归等多种智能技术集成的复杂工艺过程优化系统的设计思想、体系结构、关键技术和实现方法,主要解决多因子、高噪声、非线性、非高斯分布和非均匀的复杂工艺系统难题.采用代理技术设计系统的体系结构,用偏最小二乘法和Chelyshev多项式建立领域模型,通过演化计算进行最优问题求解,并用正交实验取得模型学习的样本数.实际应用证明,利用这些方法可以在很少的实验情况下,使所建立的模型能在较大误差范围内指导生产实践.

Research on optimal design of complex process based on multi-method integration

Aiming at characteristics of complex production process in process industries, an intelligent software system on optimal formula of production processing with multivariate factors based on pattern recognition, artificial neural network, genetic algorithm and nonlinear regression methods was introduced. Design philosophy, architecture, key technologies and implementation methods of the proposed system were narrated. The proposed system was designed to solve problems on multi-factor , high noise, nonlinear, non-Gaussian distribution and non-uniform distribution in optimal industrial complicated craft process. The architecture of the system was designed by using agent technology, and the domain model was constructed by the least square method and Chebyshev polynomial. Then the optimal solution of model was solved by the evolution algorithm, and learning sample data was gained by the orthogonal test, Applications indicated that these methods could provide guidelines for the industrial production with allowable error even though a few experiments were made.