- 學位論文
進步型主控制室的作業型態研究與客觀心智負荷評估工具發展
Research on Operators’ Behavior and the Development of the Objective Mental Workload Measure Tool for Advanced Main Control Rooms
摘要
當今複雜工業的進步型主控制室中，數位科技與網路的通訊技術已被廣泛地運用，所有重複性高以及容易犯錯的作業，都以自動化的形式取代了原本的人員操作。在這個作業型態的轉變歷程中，運轉員的行為、內在的認知歷程以及工作負荷來源都迥異於過往。部份研究人員認為，運轉員的角色已經由第一線的執行者，轉變為系統功能的監督者與決策者，知識層面的認知運作才是主要的內在歷程，也是其主要的工作負荷來源。也有研究人員認為，在複雜多變的動態環境中，操作的自由度是人為誤失的主要來源，需要由標準作業程序來加以規範，行為的分析是許多人因領域的研究基礎，而工作負荷反應的是內在資源與作業需求之間的關係，在資訊負擔龐大的進步型電廠中，都是相當重要的議題。進步型主控制室運轉團隊的角色與分工不同，用「認知」或「規則」的二分法來描述行為以及工作負荷並不恰當，對於進步型主控制室中運轉員的作業型態與工作負荷，需要更精緻的描述。 本研究的第一部份蒐集了大量運轉員在核電廠模擬室中演練的行為資料，透過Rasmussen的「技能-規則-知識」行為分類法，分別對資深運轉員、運轉員、助理運轉員的行為頻率、花費時間與工作負荷三個變項進行統計分析。在第一部份的研究過程中，我們發現目前最常用於現場工作負荷量測的方法，就屬生理指標取向的量測，由生理指標的來推論工作負荷雖然能夠持續獲取資料，但訊號感測器對於作業的干擾，運轉員接受程度不高，以及數據的分析困難，都是難以克服的問題，目前仍缺乏客觀的工作負荷量測方法，能夠對於運轉員的工作負荷進行監督或預測。於是，本研究的第二部份，透過Rasmussen的「技能-規則-知識」行為分類法，建立了一個適用於進步型主控制室的工作負荷預測模型，該模型包含了「注意力需求指標」與「不確定性指標」，並以NASA-TLX的量測結果驗證其效度。 研究一的結果顯示，對資深運轉員而言，主要工作型態及工作負荷來源，來自於「知識」與「規則」為基礎的行為。對運轉員而言從事此三種類型行為的頻率及工作負荷差異不大，但其從事「規則」為基礎的行為的時間顯著大於其他二者。對於助理運轉員而言，從事「技能」為基礎的行為的時間顯著大於其他二者，但就行為頻率與工作負荷而言，「規則」為基礎的行為才是其主要的類型。研究二的結果顯示客觀的工作負荷預測模型，能夠以Y=4.79+0.07x1+2.56x2表示(其中x1與x2分別表示「注意力需求指標」與「不確定性指標」)，經NASA-TLX驗證其效度為R2=0.51。 運轉員的操作行為被視為電廠深度防禦的一環，對於電廠安全運轉扮演著極為重要的角色，而客觀工作負荷的預測模型的發展，提供了電廠管理者在人為失誤產生之前，預先介入及矯正的機會，兩者不論是在理論或是實務層次都有相當重要的貢獻。
並列摘要
Nowadays, digital technology and multiplexing network techniques are generally adopted to help operators to perform tasks that are repetitive and error-prone in the advanced main control rooms of complex industries. During this transformation, the cognitive processes, behaviors, and workload source in advanced MCRs are very different from those in conventional environments. Some researchers posit that, with automation increasingly taking over tasks, the operator’s behavior is shifting from mainly operation to mainly supervision and diagnosis, and as a result, cognitive operation will become the major behavior in the operating processes. Knowledge-based workload is the main stress sources during operation. Scholars also believe that the natural variability of complex human agents must be controlled by establishing strict procedures, reducing the variations of human response and individual autonomy, which is considered potential sources of human error, is essential to safe operations. And rule-based workload is the main stress sources during operation. Different job positions connote different responsibilities, authorities, and abilities that are required to perform the job successfully. The application of a dichotomy in operators’ behavior and workload may be inappropriate. In study one, large quantities of behavioral data were gathered systematically from operators during training courses and human factors validation activities in a full-scale Advanced Boiling Water Reactor (ABWR) plant simulator in Taiwan. We clarified these controversial points through the analysis of the time, frequency, and workload based on Rasmussen’s skill-rule-knowledge classification according to the different positions of the operating crew (supervisor operator, reactor operator, and assistant reactor operator). It is worth noting that, physiological measures were widely used in measuring real-time workload since it could record the data continuously. Undeniably, physiological measurements often require the attachment of certain form of physical sensors, and they may disturb operators’ behaviors. Furthermore, the physiological response pattern is complex and high noise-to-signal ratio data is hard to analyze to be applied to practical fields. In study two, a performance-based workload predictive tool was developed based on Rasmussen’s skill-rule-knowledge framework, and validated with NASA-TLX. The results showed that, for the SRO, rule- and knowledge-based behaviors occurred more often than skill-based behavior in terms of time and frequency, and knowledge-based behavior was the main source of workload. For the RO, no significant differences were found among the three behavior types in terms of frequency and workload, but more time was spent on rule-based behaviors than on skill- and knowledge-based behaviors. The ARO spent more time performing skill-based behaviors than rule- and knowledge-based behaviors, but in terms of frequency and workload, rule-based behavior was the predominant type. Furthermore, the objective workload predictive model was proposed and validated with NASA-TLX (R2 =0.51), which can be applied during the operation phrase. Operators’ behaviors contribute to a plant’s defense-in-depth approach to safety and serve a vital function in ensuring its safe operation. And performance-based workload predictive tool is expected to provide managers and supervisors with opportunities to intervene and improve tasks before error occurred. Researches on both behavioral taxonomies and workload have many significant benefits in both scientific-theoretical and applied practical fields.