以整合型SMT方法論設計有效可靠的自動化製造系統

Abstract

自動化製造系統（AMS）設計的相關文獻中，曾分別探討監控、偵測異常 、錯誤診斷及回復。本研究旨在發展一套AMS設計法，以分析AMS所需的功 能、設計其控制、驗證系統的有效性、增進控制效率、偵測錯誤癥兆、找 出錯誤原因及接續因錯誤所中斷的程序。本研究以SMT為架構，發展出整 合型SMT設計法，該設計法是以四個階段設計AMS。在監控設計階段，得到 系統在正常狀況下的控制邏輯；在偵測設計階段，利用控制邏輯所導出的 系統動態，偵測錯誤癥兆；在維修設計階段，診斷出錯誤原因，並以回復 計畫接續中斷的程序；最後，在實作階段，由通訊協定連接實體製造系統 與控制電腦，以實現該系統。整合型SMT設計法提供系統化、一致性及實 用的步驟，以構建有效可靠的AMS。在整合型SMT設計法中，監控器可以由 Heuristic或Automaton-based兩種方式設計。Heuristic方式以擷取AMS的 物料流、資訊流及成本流知識，設計其監控器。由物料流及資訊流訂出製 程及其控制，而由成本流反應資源的耗用及顯示績效，如此可藉適當的調 整控制參數，以獲得較佳控制成效。Automaton-based方式依系統限制， 僅不允許禁止狀態發生。所設計的監控器具有最小控制限制，或意謂具有 最大控制彈性。此方式藉由建立層級式的監控器，以降低設計複雜度及減 緩狀態空間的擴增，並輔以時間控制機構決定何時下達控制命令。偵測器 係引用Modal Logic中World的觀念，利用World的系統動態及各World內的 第一法則，以偵測錯誤癥兆。一旦測得，偵測器立即將控制權轉交維修器 ，以保護系統，而維修器則立刻進行診斷與回復。診斷使用Bayes定理求 得各可能原因的機率，而回復計畫則儘可能以自動方式接續中斷的程序。 一實作的灌模系統成功的展示了整合型SMT設計法的應用。一電腦系統可 在遠方監控、偵測及維修實體灌模系統。其偵測器及維修器，可顯著延長 平均正常操作時間及減少維修時間，在錯誤發生時，執行回復計畫並可避 免人工回復的錯誤。 Previous literature has partially resolved AMS design issues, such as supervisory control, monitoring, fault diagnosis, and error recovery. This study aims to develop a method that supports analyzing required AMS functions, designing control, verifying system effectiveness, promoting control efficiency, detecting symptoms, locating faults, and recovering the interrupted process. The method, based on Supervisor-Monitor- Troubleshooter (SMT) framework, is termed as Integrated-SMT methodology that deploys four stages to design efficient and reliable AMSs. At supervisor design stage, normal control logic is obtained; at monitor design stage, the system dynamics derived from the control logic is utilized to detect anomaly; at troubleshooter design stage, fault sources are located through diagnosis and the interrupted process is continued through recovery plans; finally, at implementation stage, the physical manufacturing system is connected to the computer system via hand-shaking protocol. Through Integrated-SMT methodology, an efficient and reliable AMS can be constructed in a systematic, consistent, and practical way.In Integrated-SMT methodology, the supervisors can be designed through either heuristic approach or automaton-based approach. The heuristic approach acquires the knowledge regarding three flow systems in an AMS, i.e., material, information, and cost flows. Material and information flows specify the manufacturing process and process control, respectively. Cost flows reflect resource consumption and result in performance indication; thus, better supervisory performance can be achieved by appropriately setting control parameters. The automaton-based approach, base on Supervisory Control Theory, only eliminates those forbidden states according to system constraints. Therefore the obtained supervisor is the least restrictive one, which means it has the greatest control flexibility. In this approach, a hierarchy of supervisors is constructed to reduce the design complexity and alleviate state space explosion. Also a timing control mechanism is incorporated to resolve the temporal decision.Monitors are designed from world concept in modal logic. The system dynamics of worlds and first principles in each world are utilized to detect anomaly. Any monitor detecting a fault symptom immediately switches the control right to troubleshooters to protect the system. Then the troubleshooters perform fault diagnosis and recovery right away. Bayes* theorem is employed to compute the probability of fault sources; thus, the most probable fault is located. Also, the recovery plans either automatically fix the recoverable faults without human intervention or automatically continue the remaining recovery tasks after failure devices have been repaired/replaced.An implemented mold filling system has successfully demonstrated the application of the methodology. The computer system can remotely control, monitor, and troubleshoot the physical system. The designed monitor and troubleshooter significantly prolong mean time to failure and reduce mean repairing time. Also, manual recovery errors are avoided.