半導體廠房無塵室煙控系統設計

Abstract

本論文主要可分為兩個部份。第一部份為利用火災模擬軟體FDS，針對無塵室內的煙控系統進行模擬分析，由於目前在最先進的十二吋晶圓廠因無塵室屋頂必須架設運送軌道來搬運晶圓，此軌道設計會穿過防煙垂壁使防煙垂壁效用受到破壞，且無塵室內部下吹式空調設計於火災發生時會影響煙流的流動，可能無法有效的阻止煙流，故本論文首先針對無防煙垂壁對於無塵室內煙流流場之影響進行探討。由於無塵室上方有廣大的區域（SAC區）可用來蓄煙，以及FFU濾網於火災發生時會受到高溫的影響而燒破，使煙流往SAC區竄升，故本論文利用以上特性進行設計，利用SAC區蓄煙，並於該區增設排煙口進行排煙，以減緩煙流於主要工作區（FAB區）的沉積與擴散。此外，為了於火災發生時，可以將煙流侷限於起火區域內，使煙流無法隨無塵室內部循環流場擴散，故本論文利用無塵室內部循環式空調設計造成起火點附近之高低壓力差，分別以將煙流所流至之區劃風機關閉、降載至20%與反轉進行設計，並利用不同大小之火載量進行模擬，分析該設計於大火及小火（不足以使濾網產生燒破現象）的情況下，煙流的控制效果。由模擬結果分析後發現以下結論：（1）火災所產生之煙流將受無塵室內部循環流場及防煙垂壁之開口影響，而直接往四周擴散，故防煙垂壁無法發揮效用。（2）將風機關閉與反轉之設計於大火及小火的情況下皆可將煙流侷限於起火點鄰近的區劃內，而將風速降載至20%之設計於大火的情況下無法抑制煙流的擴散。（3）將風機關閉與反轉之煙控效果相近，但欲使煙流所流至之區劃風機可以反轉，須將所有風機進行加工設定，故建議以關閉之設定進行煙控為較好的選擇。 第二部份為利用Simulex、FDS+Evac及建築物避難安全檢證技術手冊計算人員於無塵室火場中所需的避難時間進行比較，探討各計算方法之差異性，並與煙層下降的速度進行比較。由模擬結果分析後發現以下結論：（1）Simulex、FDS+Evac及建築物避難安全檢證技術手冊計算結果分別為176秒、180秒及448秒。（1）建築物避難安全檢證技術手冊在計算人員抵達逃生口的時間中，僅利用移動距離除以人員移動速度，此方法無法估量人員於火場中可能因為恐慌而產生推擠的現象，或不熟悉環境而延長的逃生時間。且該技術手冊以樓地板面積估算人員避難開始時間，而由於無塵室內樓地板面積達11402m2，故人員開始避難時間為394秒，占總逃生時間（448秒）的88%，不符合實際狀況。（2）FDS+Evac將逃生人員視為粒子，利用建立outflow的流場決定人員逃生的方向，在結構較為複雜的空間中，易發生寬度不足以通過的通道，仍被視為逃生路徑，使逃生人員於該通道入口卡住，大幅的拖延了逃生時間，也不符合實際狀況。（3）利用Simulex模擬人員避難所需時間，並以偵測器探測時間30秒與人員反應時間60秒，共90秒做為人員開始避難時間，為較保守且符合實際狀況的方法。

This study is divided into two parts. The simulation of smoke control system in clean room by FDS is provided in the first part. It is necessary to setup the automated guided vehicle to carry the wafer on the roof of the most advanced 12-inch wafer facility clean room. Such this design may not discourage the smoke flow effectively due to the destruction of the smoke screen. Besides, the vertical laminar air flow type design also has influences on the smoke flow while fire accident happened. The first work of this part supplies the influence on smoke screen in clean room. The smoke flow arises to supply air chamber (SAC) because the smoke storages in a large space and the filters are broken due to the high temperature during fire accident. The extra exhaust port are installed on SAC to reduce the effects of smoke flow in FAB. This study utilizes the design of FFU’s air velocity and makes the pressure differences near the origin of fire in order to gather the smoke flow in the firing range. The simulation includes the 100%-off, 80%-off and reverse of FFU’s air velocity under the big fire and small fire (FFU will not be damaged). The results show that: (1) The smoke screen cannot play a major role while the flow field, generated from the fire accident, diffuses to surroundings due to the effect of inner cyclic flow field and the opening of smoking screen. (2) The design of shut down and reverse the FFU in situation of big and small fire, can restrict smoke flow in the zones adjacent to origin of fire. (3) The effect between shut down and reverse FFU is similar, but the FFU should be processed that it can reverse operate to guide the smoke flow to the zones. Therefore, we suggest shut down the FFU is better chose for smoke control. In the second part, to compare and investigate the evacuation time, difference of calculation methodologies and descent velocity of smoke layer by Simulex, FDS+Evac and Handbook of Verification Method for Building Fire Egress Safety. The results show that: (1) The calculated results of Simulex, FDS+Evac and Handbook of Verification Method for Building Fire Egress Safety are 176, 180, 448 seconds respectively (2) Handbook of Verification Method for Building Fire Egress Safety only according to the ratio of staff moving distance and moving speed. This methodology cannot estimate the phenomenon of jostle which is due to frightened or unfamiliar to environment that postpone the evacuation time. Moreover, this manual utilizes the floor area to estimate the starting time of staff evacuation. However, the area of clean room is 11402 m2, so the starting time of staff evacuation is 394 seconds which occupied 88% of total evacuation time (448 seconds). The condition in this case is not close to the reality. (3) FDS+Evac recognize staff as particles and establish a flow field with outflow condition to decide staff evacuative direction. However, in some complicated spaces, the tunnel which is too small to pass through is still identified as evacuative route that cause staff accumulative in this tunnel and postpone the evacuation time. The condition in this case is not close to the reality (4) Utilize Simulex to simulate staff evacuation time and chose detecting time of detector 30 seconds with staff response time 60 seconds that totally 90 seconds for staff starting time of evacuation. This calculation methodology is more conservative and more close to the reality.