請用此 Handle URI 來引用此文件: http://hdl.handle.net/11455/47285
標題: 資源限制下之線性專案排程最佳化研究
Optimization of Linera Scheduling under Resource Constraint
作者: 謝孟勳
關鍵字: 技術發展
土木水利工程類
重複性排程
動態規劃
遺傳演算法
多資源限制
最佳化排程
LSM
摘要: 『重複性專案』係指該專案中之作業具有高度相似性、重複性,且作業資源之使用可因應作業重複特性作連續性之安排。綜觀現今國內外學者針對重複性排程最佳化之研究,其計算模式皆採『距離』為切割軸。例如:工項主觀採用某種長度(例如:100M)切割,在同一切割單元內,派遣相同工班組合施作。如此之假設,目的是為了降低運算之複雜度,以利模型建立與求解。然而經由本計畫之前導研究實地探查,在「重複性工程」中,若採用『距離』為切割軸,常導致工班之更換派遣發生在奇怪的時間點。例如某段工項完成於某時間點(例如,中午11:30),則在當日剩餘的工作時間,必須對工班進行調整(例如:將2部挖土機,減少為1部挖土機)。此點假設所衍生的Schedule非常不符合工地之實際運作,且將形成工班派遣之浪費,這是造成多年來重複性排程不易施行於實際工程之主因。為克服上述缺點,本計畫創新提出以『日期』為切割軸之排程最佳化模式(Time-driven Linear Scheduling Method, 簡稱 T-LSM)。T-LSM以『時間-日』為一個切割單元,確保同一天內,工班組合為相同,以期盡量貼近廠商實際派遣工班之方式。T-LSM針對單一工項列舉所有可能排列方式,並能考慮工項與工項之間有『時間限制』(例如 灌漿後2天才可拆模)、『距離限制』(例如 怪手作業範圍30m內不可以有其他工項施作),以及多元『資源限制』(鋼筋工班、模板工班、土方工班、挖土機…)。本計畫擬以2009年發表之進化遺傳演算法求解最佳化排程模式為根基,積極改良研發更快速、更高品質的演算模式,並確保求解的『最佳性』。目前針對採用『動態規劃』學理,如何快速獲得最佳重複性排程已有具體而可行的構想。俟本計劃之執行與完成撰寫核心運算程式,擬將本計劃研發成果,進一步開發成為Internet Web網站,提供營建業求解『重複性排程最佳化』的網路平台。透過此一Web遠端服務,達到學術研究成果『回饋』業界的理念。本計畫預期可為營建業之重複性專案,規劃出優質可行之排程,並達到縮短工期,節省成本的效益。
Repetitive project scheduling indicates that operational procedures of a project have high similarities and the procedural resources, in response to its repetitive qualities, can be arranged continuously and thus become more effective. Most of the previous scholars proposes the optimization research of repetitive project scheduling adopt “distance” as segment-axis. For instance, labor items are cut subjectively by certain length unit (such as 100 Meter); within this cutting unit, the same work crew is assigned. Such hypothesis is created to decrease the complexity so as to build a model and locate the optimal solution.However, after the actual investigation for this proposal, repetitive project scheduling based on “distance” as the segment axis is found to be flawed. That is, crew has to be stopped at a certain time (e.g., 2:00 pm) to undergo crew change shift. Therefore, the repetitive project scheduling is hardly applied to the actual construction spot due to its impractical effect and its trait of waste of work crew assigned.To overcome this problem, this proposal creatively proposes a Time-driven Linear Scheduling Method (T-LSM). T-LSM regards “time - day” as unit to ensure that the work crew is the same within one day in order to match up the actual method the contractor assigns. T-LSM lists all possible scheduling methods toward a single labor item and takes into account time constraints between labor items (e.g., the form works are removed two days after grouting), distance constraints (e.g., other construction works are not allowed within 30 meters from excavator work), and multi-resource constraints (e.g., reinforced concrete crew, model crew, earthwork crew and so forth).Based on our paper published in 2009, an optimal scheduling model which employs evolution genetic algorithm to solve the problem of "different length of genes" was developed. However the genetic algorithm for repetitive scheduling still has the drawbacks of long-time convergence and its optimization is not guaranteed. An improved approach, Dynamic Programming, will be developed to overcome the above drawbacks. Then the results will be build into an internet web service.
URI: http://hdl.handle.net/11455/47285
其他識別: NSC100-2221-E005-082
文章連結: http://grbsearch.stpi.narl.org.tw/GRB/result.jsp?id=2324392&plan_no=NSC100-2221-E005-082&plan_year=100&projkey=PB10007-0531&target=plan&highStr=*&check=0&pnchDesc=%E8%B3%87%E6%BA%90%E9%99%90%E5%88%B6%E4%B8%8B%E4%B9%8B%E7%B7%9A%E6%80%A7%E5%B0%88%E6%A1%88%E6%8E%92%E7%A8%8B%E6%9C%80%E4%BD%B3%E5%8C%96%E7%A0%94%E7%A9%B6
顯示於類別:土木工程學系所

文件中的檔案:
沒有與此文件相關的檔案。


在 DSpace 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。