| 姓名 | 陳柏鈞(Po-Chun Chen) | 電子郵件信箱 | E-mail 資料不公開 | |
| 畢業系所 | 營建工程系碩士班(Department and Graduate Institute of Constrction Engineering) | |||
| 畢業學位 | 碩士(Master) | 畢業時期 | 91學年第2學期 | |
| 論文名稱(中) | 隨機模擬最佳化之研究 | |||
| 論文名稱(英) | Optimization Stochastic Simulation Mechanism for Construction Operations | |||
| 檔案 |
請遵守中華民國著作權法之相關規定，切勿任意重製、散佈、改作、轉貼、播送，以免觸法。 |
論文使用權限 | 校內外完全公開 | |
| 論文語文/頁數 | 中文/97 | |||
| 摘要(中) | 摘要 營建作業電腦模擬技術發展的歷史已經超過三十年，不但可以根據營建作業施工的流程步驟建立模型，模擬工程施作過程，並且可以考量營建工作受到天候、地形或工作性質等因素，對作業時間之不確定性影響及分析作業項目其所需資源相關互動情形。電腦模擬工具雖然可用以分析作業流程，但是在解決營建作業資源組合最佳化問題，仍然以窮舉法一一的找尋最佳的資源組合，倘若遇到複雜的工程，作業項目眾多且資源組合數量爆炸時，以窮舉法求解則效率不彰。在過去研究中，營建領域有人以CYCLONE理論為基礎之電腦模擬技術與基因演算法結合，利用在求解多類型資源組合最佳化問題，且其求解效率及準確度亦獲得驗證。但前人的研究，僅考慮模擬作業時間為固定值，來進行模擬求解，使得所求得模擬結果之目標值為固定值。但是作業時間的不確定性對模擬結果影響甚大，導致所求得模擬結果之目標值為非固定值，如果僅以單一固定時間值做為模擬作業時間，則模擬所求得目標值，將無法符合現實。因此，本研究目的乃利用CYCLONE理論為基礎之電腦模擬技術配合統計觀念與區間數學運算方法進行目標值計算，以區間值域來代表適存值，並結合基因演算法在搜尋最佳解機制，來改善模擬演算系統在求解目標值不確定之多類型資源組合最佳化問題上缺失。本研究經由兩個案例驗證顯示，各世代之母體常態分佈圖及基因演算各世代最佳解收斂圖，確有其收斂之趨勢，且只要搜尋不到1%的解，即可迅速及精確地找到近似最佳解或最佳解，其求解效率較傳統之窮舉法提高很多。 | |||
| 摘要(英) | ABSTRACT Discrete event simulation has been used in the modeling construction operation for many decades. While running simulation, the system performance such as production rate and unit cost could be influenced by resource combinations. In order to allocate the resources distributions for optimizing system performance, different techniques have been applied in previous researches. Among them, Genetic Algorithms (GA) is recently proved to be successful in locating optimal resource combination with little computation efforts. The mechanism for the integration of GA and discrete event simulation is running simulation to test different resource combination for seeing their system performance and using the system performance as fitness values in GA to screen out the ones with poor fitness values. In previous research, the task duration is assumed to be constant. Therefore, it has no difficulty to build up the selection probability representing each resource combination since fitness value is always the same for running the same resource combination. However, simulation is powerful in the modeling the uncertainty of work tasks, thus once the stochastic type duration are used, fitness value will be fluctuated as the system performance obtained by running the same resource combination. Such fluctuation leads the difficulty for the establishment of the selection probability in GA. This research proposes a new mechanism to tackle this problem. Case studies show the mechanism can locate those resource combinations with have better mean value of system performance but little variance. | |||
| 關鍵字(中) |
|
關鍵字(英) |
| |
| 指導教授 |
| |||