(清華大學(xué) 化學(xué)工程系， 北京 100084)
摘 要: 當(dāng)前， 集過程實時監(jiān)測、 故障診斷、 模擬、 優(yōu)化、 控制以及調(diào)度等各層次功能于一體的過程工業(yè)生產(chǎn)過程綜合自動化成了過程工業(yè)界和學(xué)術(shù)界共同關(guān)注的熱點之一。與離散產(chǎn)品的制造業(yè)相比， 由于流程型工業(yè)過程具有強非線性的特點， 給實現(xiàn)流程工業(yè)綜合自動化造成很大的困難， 因此必須引入新的思路， 開發(fā)新的方法。 人工神經(jīng)網(wǎng)絡(luò)是一種模擬人類思維活動的并行分布式的信息處理系統(tǒng)，可用于映射任何連續(xù)函數(shù)及進(jìn)行模式識別， 同時還具有自學(xué)習(xí)功能， 實現(xiàn)知識的自動獲取， 自20世紀(jì)90年代以來已在過程系統(tǒng)工程領(lǐng)域內(nèi)受到廣泛的矚目。重點討論了人工神經(jīng)網(wǎng)絡(luò)在過程系統(tǒng)建模、 故障診斷以及在線優(yōu)化等方面的應(yīng)用， 以展示這種方法在流程工業(yè)綜合自動化中的良好應(yīng)用前景。
關(guān)鍵字: 人工神經(jīng)網(wǎng)絡(luò)； 過程系統(tǒng)建模； 過程系統(tǒng)故障診斷；遺傳算法
（Department of Chemical Engineering,
Tsinghua University, Beijing 100084, China）
Abstract:Recently, the computer integrated process systems which integrate the real-time detection, fault diagnosis, modeling, optimization, control, planning and other information techniques at different levels have become a topic of common interests in both industry and academy. Compared with the manufacturing industry for discrete products, the continuous process industry possesses the characteristics of strong non-linearity, which makes the computer integrated process systems very difficult to be realized. Therefore, there is an urgent need to introduce new ideas and new methodologies. Artificial neural network is a kind of parallel information handling system for emulating human’s thought. It could be used for mapping any continuous functions and carrying out pattern recognition, meanwhile, it could realize the acquaintance of knowledge automatically. Since 1990s ANN has attracted wide attention in the field of process systems engineering. The emphasis will be put on illustration of the applications of ANN in the process modeling, process fault diagnosis, on-line optimization and other aspects so that to fully demonstrate the good potential of applying ANN in developing the computer integrated process systems.
Key words: artificial neural network; process modeling; process fault diagnosis; genetic algorithm
