Improved Analyses of Construction Baseline Productivity

[[abstract]]營建績效或營建生產力的評估，最大的癥結在尋求合適的生產力量測模式和合理的評鑑標準，而關鍵是如何確定工人基準生產力（代表某一專案工程的最佳生產力），基準生產力是評估效率水準，評鑑工程預算和工期，或確定生產力的損失等等的基礎。 本文彙整目前廣被運用的評估營建基準生產力的分析方法，包括實測哩程（Measured mile）分析法、集群分析（ Cluster analysis）法、Thomas的基準分析（Baseline analysis）法、以及單一量測值管制圖分析（Analysis by individuals control chart）的統計性分析方法。上述方法各以主觀或客觀的規定，確認施工期間不受干擾的基準生產力集合，再比較工程中受變更影響與不受變更影響期間之間的生產力差異，進而求得基準生產力。 本文另外提出一項客觀的基準生產力的評估方法，是以資料包絡分析工具求得每天的生產效率前緣，而據以決定基準生產力，可以處理不限於單一產出或投入的作業項目的工作，是一項全新和綜合的方法。本研究也綜合所有評估方法，建構一套以專案計畫為基礎的基準生產力量測與評估的標準模式。 本研究為比較及驗證各種不同的營建基準生產力之評估方法的差異，蒐集八個鋼結構建築工程之吊裝工作實際案例，共含409 工作天(677日曆天)之工程觀測資料，運用上述五種評估方法分別求得不同的基準生產力；研究結果顯示各種方法之間有相當差異，因此建議，在選擇與應用這些評估方法時應慎重。 本研究建議，若是為了工作排程和成本預算計劃的需求，應就不同的工程特性和影響工程執行的內外環境因素，斟酌採用不同營建基準生產力之評估方法分析，或採取綜合成果比較後決定採用生產力標準；若應用於生產力的折損評估和工程爭議的仲裁或求償，除了上述的衡量外，則適用的關鍵，還在業主與承包商的協調與共識，以及裁判者的判斷能力和對評估方法的理解程度。 根據實際案例驗證之結果，可發現本文所發展的「以DEA分析評估營建基準生產力」的方法，其決定基準生產力之方式，最符合根據生產前緣衡量最佳效率以確認標竿學習目標的常理；加以本方法適用於多項產出或投入的生產績效評估，對於較複雜（較多作業）的工作，評估基準集合可能較容易。若針對各種分析方法的特點、基本原理、和操作方式，實際案例的分析運作的應用，以及分析系統的「簡易性、正確性、低操作成本、可檢核性、和適時性」等比較，顯然資料包絡分析法的結果最為客觀、確實和實用。 本研究也認為，根據實際案例分析成果之差異，五種基準生產力評估方法可分為以下兩種不同的等級：（1）從排除影響因素的積極程度，以及從各方法期望求得最佳工作績效典範的嚴謹性來判斷，可發現Thomas的基準分析法和資料包絡分析法兩者，屬要求比較高者（以每天生產力平均的百分比值衡量），可用於訂定營建工程最佳績效之標竿。（2）實測哩程分析、集群分析、和2σ單一量測值管制圖等計算所得到的基準生產力值，是為警告界限，可做為工程施工績效出問題之早期警訊，用以檢討和調整施工管理系統。但不管如何，工程進行都必須維持的基準生產力水準，應不低於3σ單一量測值管制圖所規範的行動界限，否則必須調查影響工作的可歸屬原因或採取修正措施。[[abstract]]The purpose of this study was to find standard and reasonable measures to derive labor baseline productivity which is the best productivity of constructor in any project. The BP could be used as criteria to forecast schedule and cost, as well as to evaluate loss of productivity, for resolving conflicts between constructor and owner. Some widely accepted methods to quantify labor baseline productivity are measured mile approach, cluster analysis, Thomas baseline analysis, and analysis by individuals control chart. A major problem with these methods is that they apply somewhat subjective methods to identify the un-impacted reference periods to infer the baseline productivity. This article introduces another general objective, non-parametric, frontier evaluation approach which was established with data envelopment analysis (DEA) tools. The DEA method which was concerned with the estimation of technical efficiency is sufficiently powerful to accommodate the measurement challenge and can be applied to analyze multiple outputs and inputs without pre-assigned weights. The paper outlines all baseline productivity analysis methods and constructed a project-oriented model for measuring baseline productivity. Five baseline productivity analysis methods were tested and verified by a productivity database resulted from 8 numerical examples of steel structure construction projects with 409 workdays (677 calendar days) for the erection of frame members. It is found that there are differences among these methods. It is noted that methods must be utilized carefully. For the project scheduling and costing, this research recommended that the method should be applied on the different project identity and the influence of environmental factors. For the calculation of productivity loss and resolving the project disputation and arbitration, the owner and contractor's coordination and the mutual recognition may be necessary. According to the results of actual case confirmation, the DEA method that is most objective, truly and practical is recognized to be the best method of efficiency surveying for bencmarking goals and target. To fully evaluate the physical characteristics, the five methods may divide into following two different ranks: (1) The Thomas baseline analysis method and DEA method are available for establishing efficiency criteria for construction engineering. (2) The data that are measured by the measured mile approach, cluster analysis, and analysis method of 2σ individuals control chart are the warning boundary to examine and adjust construction system. But no matter how, the project does carry on the baseline productivity standard which should not be lower than the action boundary of the measurement of 3σ individuals control chart