建立營建工程產出模式之研究

Abstract

台灣地區地狹人稠，廢棄物之處理長期以來皆為主管機關與社會大眾所關切之問題。依據行政院環境保護署統計2012年營建廢棄物產生量為150萬公噸，但內政部營建署推估每年約產生1,000萬之營建廢棄物，因此營建廢棄物產出量之推估情形，應能即時更新掌握，並建立適切推估模式為本研究之重要議題。 研究中以行政院環保署事業廢棄物管制系統中，依「應檢具事業廢棄物清理計畫書之事業」擷取2012年營建廢棄物產生量為150萬公噸，計10,410家營造工程填報資料，針對佔營建廢棄物總量達80%以上之D-0599土木或建築廢棄物混合物（28%）及R-0503營建混合物(60%）進行產出量推估模式之確立。 研究以連續變項作為產量之影響因子，包括為工期、總樓地板面積、工程面積及工程契約金額等4項。經逐步迴歸分析後並透過非共線性以變異數影響因子（Variance Inflation Factor，VIF）檢視、獨立性以Pearson檢定後，確立之產量迴歸模式有：一、營建混合物(R-0503)以工程面積允差值為0.858進行分析，表示模式中總樓地板面積等其餘3個自變數對於工程面積解釋變異量為14.2%，解釋變異量越高代表容忍度越小，故迴歸分析方程式並無共線性問題，因此營建混合物產量（公噸/月）＝10.801+0.008＊工程面積（m2）+0.001＊總樓地板面積（m2）+0.3762＊工程契約金額（萬元）-0.01＊工期（日）；二、土木或建築廢棄物混合物(D-0599)以總樓地板面積及工地面積數值均為0.126，表示模式中工程面積對於總樓地板面積的解釋變異量為87%，恐有共線性問題。研究改擬以總樓地板面積、工程面積分別與D-0599土木或建築廢棄物混合物產量，進行迴歸分析模式建立，以降低共線性之影響。而得模式(一)土木或建築廢棄物混合物產量（公噸/月）=21.491＋0.001＊樓地板面積（m2）；模式(二)土木或建築廢棄物混合物產量（公噸/月）=24.193＋0.0005＊工地面積（m2）進行。 為進一步了解本研究迴歸模式之適用性，故以2013年產量推估後，與環保署2013年事業廢棄物申報及管理系統中實際申報量進行比對。驗證推算結果，顯示於推估營建混合物(R-0503)之MAP E<10%，係屬預測準確度極佳；於推估土木或建築廢棄物混合物(D-0599)中模式(一)之MAPE介於10%≦MAPE <20%中，係表示預測準確度優良，而模式(二)之MAPE<10%，係屬預測準確度極佳，建議可以模式(二)優先使用。研究結果可提供未來廠商自主性管理或主管機關管制應用之參考。

Waste management in Taiwan has long been an issue catching the attention of not only the government authority but also the general public because of Taiwan’s limited space and the dense population. Even though the Environmental Protection Administration (EPA) of Executive Yuan reported a total of 1,500,000 tonnes of waste produced from construction, the Construction and Planning Agency of Ministry of the Interior, on the other hand, estimated that about 10,000,000 tonnes of construction waste is likely to be generated each year. Therefore, it is important to establish an appropriate estimation model for the construction waste output quantity to correctly reflect the current situation, and that is the objective of the study. From the waste management system of EPA of Executive Yuan and based on the classification for industries mandated to submit the industrial waste report, the study found that in 2012, the construction waste output was 1,500,000 tonnes. According to information from a total of 10,410 construction companies, the investigators verified the output estimation model for D-0599 civil engineering or building waste mixture (28%), which accounts for more than 80% of the total construction waste, and R-0503 construction and demolition debris (60%). In this study, continuous variables were used as the impact factors of the output, and there were four of them: construction period, total floor area, construction area, and construction contract sum. The data were examined using regression analysis, followed by the analysis of variance inflation factor (VIF) for non-collinearity and the Pearson’s test for independence. The output regression model has the following features. First, construction and demolition debris (R-0503) was analyzed using 0.858 as the construction area tolerance value. According to the simulated results, the total floor area and other three independent variables explained 14.2% of the variance of the construction surface area. As an increase in the variance explained indicates a smaller tolerance, collinearity was not an issue in the regression equation. As a result, the output of construction and demolition debris (tonnes / month) = 10.801+0.008 x construction surface area (m2) + 0.001 x total floor area (m2) + 0.3762 x construction contract sum (10,000 NTD) - 0.01 x construction period (days). Secondly, for civil engineering or building construction waste mixture (D-0599), the tolerance value of the total floor area as well as the construction surface area was 0.126, indicating that in the model, the construction surface area explained 87% of the variance of the total floor area. Collinearity therefore was a concern here. Then, the study used either the total floor area or the construction surface area to build the regression analysis model with D0599 civil engineering or building waste mixture output to reduce impacts from collinearity. The acquired models are as follows. For model 1, civil engineering or building waste mixture output (tonnes / month) = 21.491＋0.001 x floor surface area (m2). For model 2, civil engineering or building waste mixture output (tonnes / month) = 21.193＋0.0005 x floor surface area (m2). To further understand the suitability of the regression model of the study, the output of 2013 was estimated and compared with the actual reported volume from 2013 Industrial Waste Report and Management System of EPA. The result indicates that the MAPE of construction and debris mixture (R-0503) was less than 10%, meaning excellent estimation accuracy. For Model 1 estimating civil engineering or building waste mixture (D-0599), the MAPE was more than 10% but less or equal to 20%, indicating good estimation accuracy. As for Model 2, the MAPE was less than 10%, indicating excellent estimation accuracy. Therefore, Model 2 is preferred. Taken together, the study results are useful for companies interested in autonomous management as well as for government authorities searching for relevant control and management.