- 學位論文
類神經模型預測控制的資訊安全建置與部署
Building and deploying the information security for the neural network model predictive control
摘要
在工業4.0概念下，工廠的電腦化、數位化與智慧化轉型是現今產業界之趨勢，透過巨量資料分析與雲端運算，使得工廠可於最佳化硬體及製程模組化下，進行自動控制。大量資料雲端化使電腦可掌握之資訊越來越多，因此資訊安全的問題逐漸受到重視，本報告比較OPC DA(Data Access)與OPC UA(Unified Architecture)國際通訊協定。針對現今化工廠通用之分散系統控制(Distributed Control System, DCS)、先進製程控制(Advanced Process Control, APC)及類神經模型預測控制(Neural Network Model Predictive Control, NN MPC)，本報告也整理其控制系統架構、控制方法、資料傳輸方式及硬體設備規格要求之差異，並探討現今化工廠中所使用之自動控制系統常因伺服器環境不一致或軟硬體設備更新導致系統斷線、延遲，使控制無法完全自動化仍須仰賴人工調整的問題。容器化的封包是一種可以縮短執行環境佈署時間並保護程式原始碼的技術。 以OPC UA、NNMPC與Docker為基礎，本報告提出一套智慧控制系統環境佈署與建立的架構與內容，並且以一個溶劑回收程序應用案例。針對應用案例，順利地完成了OPC UA的建立，透過Docker的容器化封包技術所需環境佈署時間僅需要2小時，以標準差作為製程穩定程度之指標，在傳統控制上製程標準差介於1.1~4.1間，而使用類神經模型預測控制製程標準差則可大幅降低至0.4~0.8。
並列摘要
Under the conception of Industry 4.0, the computerization, digitalization and smart transformation has become the trend of industrial transformation. By virtue of big data analysis and cloud computing, the factories are allowed to implement automation with the optimal hardware as well as the process modularization. The availability of large amount of data on cloud has raised the attention of information security. This report compares two international communication protocols, namely the OPC DA (Data Access) and OPC UA (Unified Architecture). With respect to the commonly adopted control approaches, i.e. Distributed Control System (DCS), Advanced Process Control (APC), and Neural Network Model Predictive Control (NN MPC), this report discusses the differences among them in terms of the control system architecture, control methods, data transmission methods, and hardware equipment specifications. Furthermore, the problem of disconnection or delay, which results in the unsuccessful automation, due to the inconsistence of server environment and updating of hardware or software is explored. For the implementation, containerized packaging is a technology that can shorten the time for environment deployment and protect the source code of programs. Based on OPC UA, NNMPC and Docker, this report proposes a set of framework and content for the deployment and establishment of a smart control system. A solvent recovery process is employed as a case study for demonstration. The OPC UA is successfully developed for the process and the deployment of the environment takes only 2 hours by applying the docker's containerized packaging technology. As an indicator of process stability, the standard deviation of the process control of the solvent recovery process using NNMPC is 0.4 to 0.8, which is significantly lower than the 1.1 to 4.1 by using traditional control.