Karlsson Model先導性研究

Abstract

Karlsson Model為瑞典防火工程學者Dr. Bjorn Karlsson所研發出的火災模式。此模式利用小尺寸實驗數據，包括圓錐量熱儀所量得的熱釋放率和側向引燃延燒測試儀所得到的引燃溫度、火焰延燒參數及熱慣性等，去模擬大尺寸火災房間的燃燒現象。本研究以內政部建築研究所先前所做過的相關實驗數據為依據：在小尺寸實驗方面，以圓錐量熱儀和側向引燃延燒測試儀所量得的材料基本燃燒性質，計有熱釋放率曲線、熱慣性、引燃溫度及火焰延燒參數，經換算為此模式的輸入資料形態，再代入Karlsson Model的數學計算模式中，以模擬材料在大尺寸房間火災實驗中的燃燒現象，並和建研所以往所做過的以國際標準ISO 9705為實驗依據的大尺寸火災房間實驗數據作一詳細的比較。在熱釋放率的模擬結果上，基本上趨勢和物理量是大致相符，但可以發現有熱釋放率集中及引燃延遲的缺點，此部份誤差的改進有賴於對火場現象的相關假設以及程式設計上須做更進一步的考慮，而在熱氣層氣體溫度及板材表面溫度的模擬上，發現有估算值過高的情形，此部份的修正有賴於經驗公式係數的選擇以及環境熱損失機制的詳細探討，便如輻射熱損及至牆內部的傳導熱損等。為了使Karlsson Model模擬數據和我國已有的實驗研究數據能有更好的配合性，本文在最後也對研究過程中所發現的誤差，提出幾點建議，希望能對日後國內防火計畫中火災模式的發展上，能有所幫助。

Karlsson Model was developed by Dr. Bjorn Karlsson, a fire safety engineer and scientist in Sweden. This model utilizes the measurements obtained from two standardized bench-scale tests, Cone Calorimeter (CONE) and Literal Ignition and Flame Spread Test Apparatus (LIFT), as the input data to predict the fire growth in the full or 1/3 scale room tests. These input data consists of the heat release rate as a function of time, kpc, ignition temperature and flame spread parameter. The thesis focuses on the theory development of the numerical method and its program structure. Also a set of computational works are carried out by using the bench-scale fire test data, measured from local products, supplied by the Fire Laboratory of Architecture & Building Research Institute (ABRI), Ministry of Interior. The predicted results are then compared with the ones obtained from the large-scale fire room/corner test, complied with ISO 9705 standard. In the simulation of heat release rate, the trend and quantities are consistent in a general point of view. However, some shortcomes of the model are identified from the observation of the predicted results, such as the occurrence of pulse in heat release rate and the ignition delay. These can be improved by modifying the assumptions used in the prescribed fire scenario in the model. The computed hot-layer and ceiling surface temperatures are found over-predicted. It indicates that we should consider to apply the correct coefficient according to the material fire performance and to include the extra heat loss mechanisms, such as radiation and conduction to the ceiling and walls. Finally, according to the understanding of theory and structure of the model, the experience by executing the program and the comparison between the experimental and computational results, this thesis makes several recommendations and suggestions to improve the performance and predictive capability of Karlsson Model in the near future. The main purpose is to provide a good guidance for the development of fire modeling in the next five-year fire research program.