飛安查核工作排程與人員排班模式建立

Abstract

本研究針對民航局飛安查核工作排程與檢查員排班問題進行探討，由於查核人員排班與一般的人員排班問題具有不同特性，因此本研究在現有查核人力、查核作業需求及排班法規等限制下，考量排程班表之公平性與合理性，分別建立查核任務排程之限制規劃模式，及檢查員排班二階段求解演算流程之結合限制規劃與數學規劃模式，以解決查核任務排程與查核人員排班之組合最佳化問題與限制滿足問題。

在工作排程部分，本研究是以資源限制專案排程之專案網路結構為基礎，於OPL Scheduling排程系統中使用限制規劃方法，在現有檢查人力資源限制下，針對各家航空公司航務查核任務進行組合最佳化排程模式構建，以獲得規劃期之工作排程日程表；在檢查員排班部份，應用限制列舉式勤務組合產生架構，設計結合限制規劃與數學規劃之二階段求解演算流程，第一階段限制規劃模式在於產生符合排班法規限制的可行工作班集合，第二階段則應用集合涵蓋模式選擇最佳工作班組合，其間透過OPL Script程式將第一階段限制滿足問題的求解結果傳遞給第二階段的集合涵蓋模式，再透過第二階段集合涵蓋模式求解，得出滿足作業人力需求之最佳工作班組合。

本研究依據民航局之查核作業程序，以國內四家航空公司年度查核任務進行實例驗證。查核任務排程模式之問題規模為變數個數3,636個，限制式個數7,152條，求解運算時間為30秒；檢查員排班則分別對各家航空公司每月的查核任務進行排班求解，求解運算時間介於80~180秒，求解結果顯示各月份需由12~13位檢查員執行查核作業，平均檢查員每月總查核時數約為80小時。透過案例測試可知，本研究設計之求解模式，能有效協助排程管制人員快速獲得較佳且可行的工作班表。

The purpose of this study is to investigate the work and manpower scheduling problems of aviation safety audit of the Civil Aeronautics Administration. The characteristics of scheduling aviation safety audit are different from those of the general scheduling problem. Therefore, this study is aimed to design an algorithm which takes fairness and adequacy into consideration for solving this problem under constraints of the existing manpower, auditing operation, and the constraint of inspector pairing regulation. The work scheduling problem is formulated as a constraint programming problem and the manpower scheduling problem as a combinatorial optimization problem. For work scheduling, a resource-constrained project scheduling problem (RCPSP) model is applied. For manpower scheduling, a constraint satisfaction problem (CSP) model is employed to generate the feasible pairings that satisfy the inspector pairing regulation. Then, a set covering problem (SCP) model is applied to find out the minimum-cost pairings from the feasible solutions. Between the CSP and the SCP models, OPL Scripts is used as a bridge. The model is tested with the case of the four domestic airlines. According to the operation procedure of aviation safety audit, the problem size of the work scheduling (RCPSP) model has 3,636 variables and 7,152 constraints and it takes 30 seconds to solve. The manpower scheduling algorithm searches the solution of the inspector scheduling problem for every airline and it takes 80~180 seconds. The result shows that 12~13 inspectors are needed to execute the audit and each has about 80 working hours monthly. The result also suggests that the developed models can be applied effectively to obtain an improved work schedule.