建築物火災發生傷亡與都市空間使用類別關係之研究

Study on Relationship between the Casualties of Building Fire and Occupancies in Urban Space

本研究以高雄市為研究範圍，分析各種都市因子與建築火災之相關性，並建立都市避難空間、建築物使用類別對於火災傷亡事件之預測模式，探討都市不同之使用分區平均每年起火次數、都市中火災避難空間（停車場用地用地、公綠地、學校、道路）與建築火災傷亡人數之關係預測、並針對高雄市的火災事件，找出影響死傷事件發生與都市空間使用類別之特徵因素，建立一套建築物使用類別與火災死傷事件發生的預測模式，期能對都市內火災預測及對都市土地使用分區之配置作出貢獻，本研究主要以(1)迴歸分析(Regression)探討不同變數間之相關性，並針對顯著相關之因素進行迴歸分析，建立迴歸方程式以解釋或預測變數間之關係(2)羅吉斯分析(Logit Analysis)以建築物之使用類別及消防出水時間（開始灌救時間減去報案時間）標準，對已發生建築火災造成人命傷亡之可能性問題進行推估預測(3)區別分析(Discriminate Analysis)設定預測傷亡之區間，將複雜現象數量化以對該現象作系統之整理、分類進行預測。研究結果建立都市因子與火災發生次數之相關模型、都市避難空間與建築火災傷亡人數之區別函數及都市建築物使用類別與是否發生傷亡之預測模式三種，其中在都市避難空間與建築火災傷亡人數之區別函數預測，以死傷人數一人以下的正確判斷率最高達90%，在重大傷亡人數5人以上之傷亡之區別函數之正確判斷率達20%，而都市建築物使用類別與是否發生傷亡人數之預測模式中，以羅吉斯分析得到之模型其對死傷事件之判斷正確率為56%，對本研究所建立之三種火災預測模型而言，能以都市因子及建築物使用類別進行火災傷亡之區別預測，乃過去所沒有進行之研究，對未來火災問題有限度之單純化及適用性而言，將有助於此類研究之發展。

This research is in the city of Kaohsiung. To establish correlation among urban evacuation space, building use and fire casualty events, it analyze the relation between urban factors and building fires. Through the study of building fire casualty in average fire frequency per year in different zoning and fire evacuation space in the city (parking lot, woodland, school, road), the effective factors are found when a casualty event happened in city of Kaohsiung. Then, the forecasting model of occupancies of urban space and fire casualty event is set up. The anticipating results hope to make contribution to predicting the possibility of lives lost before planning the land use control. The main research tools are in (1) using Regression to discuss correlation in different variables and analyze remarkable correlation factor proceed with Regression formula which explains and forecasts relation between them, (2) using Discriminate Analysis to dig out potential problems of casualty happened in building fire according with building use and time of starting the fire hydrant, (3) using Logit Analysis to set interval of casualty trying to form the quantified forecasting by systematically sorting the data from complex urban phenomenon. The result is to establish three models which are the model of urban factor and fire frequency, the model of interval function of urban evacuation space and building fire casualty, and the model of urban zoning and happened casualty. As the model of interval function between urban evacuation and building fire casualty, the preduction of casualty event under one person achieves 90%, the prediction of casualty event under above five persons achieve 20%. In forecasting urban zoning and casualty, by the model with Logit analysis, the prediction of casualty event achieves 56%. Establishing the three models above through the relationship between urban and building to forecast the casualty when building fire happens, related research is never done before. By limitedly simplifying the terms of building fire on urban space, it can be sure to make contribution to further study on this issue.