Development of Genetic Neural Network System for Nonlinear Process Information

非线性过程信息遗传神经网络系统的开发

• 批准号: 10555263
• 负责人: KURODA Chiaki
• 金额: $ 8万
• 依托单位: Tokyo Institute of Technology
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B).
• 财政年份: 1998
• 资助国家: 日本
• 起止时间: 1998 至 2000
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-10555263/
• 关键词: Genetic Algorithm; Neural Network; Nonlinear; Process; Information Processing; Scheduling; Modeling; Reaction Separation System; 反応分離システム; (1)遺伝的アルゴリズム; (2)ニューラルネットワーク; (3)非線形; (4)プロセス; (5)スケジューリング; (6)モデル化; (7)反応システム; (8)分離システム; 非線形; 反応

This study aims to develop a new hybrid practical system "GANN" that is a new method using a three-layered neural network optimized by a genetic algorithm. This system was applied to operation and scheduling in batch processes with complicated constraints, and to modeling and control in nonlinear reaction-separation processes, and the following results were obtained.1. In a polymerization process and a fractionation process using liquid chromatography, nonlinear process data could be precisely modeled using an artificial neural network that was a basis of GANN.Through the above investigation, it was made clear that the number of units in a hidden layer was an important factor for flexibility of networks, and an adequate genetic coding method of networks was designed.2. A dynamic GANN scheduling system was developed for an integrated operational system that flexibly controlled fluctuating productions in real time. This system could appropriately cope with sudden changes of production plans, troubles of equipments and maintenances.3. The GANN system was applied to designing the structure of a control system for a mixing-tank reactor, and it was found that the system was a useful optimization method for design of process controllers.The above results made clear that this GANN system was a powerful and flexible tool for modeling, control and operation in nonlinear processes.

本研究的目的是开发一种新的混合实用系统“GANN”，这是一种新的方法，使用三层神经网络优化的遗传算法。将该系统应用于具有复杂约束条件的间歇过程的操作与调度，以及非线性反应分离过程的建模与控制，取得了如下成果.在聚合过程和液相色谱分离过程中，采用人工神经网络对非线性过程数据进行精确建模，并以此为基础建立了GANN网络，明确了隐含层单元数是影响网络灵活性的重要因素，设计了合适的网络遗传编码方法.为适应集成化生产系统对波动性生产进行真实的实时灵活控制的需要，开发了一个动态GANN调度系统。该系统能较好地科普生产计划的突然变更、设备故障和维修.将GANN系统应用于混合釜式反应器控制系统的结构设计，结果表明，该系统是一种有效的过程控制器优化设计方法，为非线性过程的建模、控制和操作提供了一种强大而灵活的工具。

项目成果

H.Matsumoto, C.Kuroda, S.Palosaari, K.Ogawa: "Neural Network Modeling of Serum Protein Fractionation using Gel Filtration Chromatography"J.Chem. Eng. Japan. 32, [1]. 1-7 (1999)

H.Matsumoto、C.Kuroda、S.Palosaari、K.Okawa：“使用凝胶过滤色谱法进行血清蛋白分级的神经网络建模”J.Chem。

Kim,Jinyoung: "Neural Network Modeling in Exothermic Polymerization Process"Proceedings of the International Conference on Engineering Applications of Neural Networks (EANN2000). 131-138 (2000)

Kim,Jinyoung：“放热聚合过程中的神经网络建模”神经网络工程应用国际会议（EANN2000）论文集。

安部雅彦: "遺伝的ニューラルネットを用いた外乱適応型再スケジューリングシステム"化学工学会第65年会研究発表講演要旨集. F121 (2000)

Masahiko Abe：“使用遗传神经网络的干扰自适应重调度系统”化学工程师学会第 65 届年会 F121 论文集（2000 年）。

C.Kuroda, S.Hikichi, K.Ogawa: "Application of Recurrent Neural Networks to Dynamic Predictions in Chemical Reactors"Proceedings of the International Conference EANN'98, 165-168, Gibraltar. (1998)

C.Kuroda、S.Hikichi、K.Okawa：“循环神经网络在化学反应器动态预测中的应用”国际会议 EANN98 论文集，165-168，直布罗陀。

M.Abe, H.Matsumoto, C.Kuroda: "An Artificial Neural Network Optimized by a Genetic Algorithm for Real-time Flow-shop Scheduling"Proceedings of International Conference KES'2000, 329-332, Brighton, UK. (2000)

M.Abe、H.Matsumoto、C.Kuroda：“通过遗传算法优化实时流水车间调度的人工神经网络”国际会议 KES2000 会议记录，329-332，英国布莱顿。