傳統劇場為了隔音與音場，牆體加內裝反射飾版，常常要做到 2、3 公尺厚。臺中國家歌劇院把結構、隔音、音場反射和造型全塞進同一面牆，厚度只有 40 公分的實心混凝土。這個數字出自歌劇院官方的十週年專訪，讀來像規格表上的筆誤，其實是整個專案最難的地方。
一張被判定「蓋不出來」的圖
伊東豊雄的「美聲涵洞（Sound Cave）」，不要傳統梁柱，改以三度空間的曲牆連續體撐起整棟建築。同一個概念，伊東先拿去比利時根特的音樂廳國際競圖，評審的評語是：這根本是蓋不出來的紙上建築。落選。
到了臺中，營造副總經理黃明晴說，剛開始連設計圖都看不懂就被叫去開會。看不懂不丟臉，圖上的東西，前面確實沒有人蓋過。
噴凝土行不通，然後呢
最初的設計考慮過鋼骨加噴凝土，在臺灣的法規與現實條件下走不通。接著摸索出中空混凝土牆加金屬網模板；結構顧問奧雅納（Arup）借用汽車外殼的設計概念，把曲牆化為實心牆，這條路才算打通。
卡關的地方更小，也更要命：金屬網模板拆不掉。留在牆裡，日後粉刷層可能脫落。
黃明晴在某次考察的回程路上冒出一句：能不能用螺絲鎖起來？後來的「珠牙螺桿工法」，用的就是臺灣工地最日常的東西，鎖模板用的螺母。螺母先和主筋焊接，螺桿再鎖進定位鋼網。金屬網的問題解掉了，粉刷層也能和混凝土咬在一起。全世界最難蓋的建築，關鍵一步長得像五金行貨架上的零件。
正式動工前，團隊還按 1 比 1 做了一座實體模型，取名「序曲」，先在地上驗證工法與強度。
五十三本筆記與週三晚上的會
建築師楊立華為這個案子寫了 53 本筆記，一個困難一個困難拆。預算不足、工期難估、細節又不停有人想再優化，楊立華的對策是建立「收斂機制」：每週三晚上開「施工窒礙難行檢討會」，疑義全部攤在桌上。
有位日本年輕設計師，花了好幾個月研究怎麼遮住曲牆上一段露明的消防灑水管線。請示伊東之後，答覆只有一句：「沒關係，露明就好，結構美感更真實。」幾個月的預算與變更設計問題，一句話收掉。
計畫性與現場感
1987 年就在日本參與過設計監造的楊立華，形容日本團隊極度重視「計畫性」：每一步設目標、事先模擬，細緻嚴謹，缺點是容易被既定框架綁住。楊立華夾在日方與麗明營造之間，當橋。黃明晴的評語很準：一般設計師愛「追究」，楊立華勤於「協調」。
建築師和營造商通常是監督與被監督的關係，這案子裡兩邊卻建立了信任。伊東是否說過這案子只可能在臺灣實現，原文是楊立華轉述的，轉述就照轉述算；擺在眼前的事實是，一張被評審退掉的圖，在 2016 年變成了立在臺中惠來路上的房子，牆裡鎖著一排臺灣工地慣用的螺母。
楊立華最愛的時刻是傍晚，室內燈光穿過曲牆與帷幕牆透出來。楊立華說，整棟建築像在呼吸。牆裡的混凝土與螺桿看不見，這點倒也沒什麼好遺憾。
— 洪俊傑
The Wall Judges Called Unbuildable Was Bolted Down
Traditional theaters often need walls two or three meters thick, structure plus interior reflector panels, to get their sound insulation and acoustics. The National Taichung Theater squeezed structure, insulation, sound reflection and form into a single wall just 40 centimeters of solid concrete. That figure comes from the theater’s own tenth-anniversary interview, and it reads like a typo on a spec sheet. It is the hardest part of the whole project.
A drawing judged unbuildable
Toyo Ito’s “Sound Cave” concept drops conventional beams and columns and lets a three-dimensional continuum of curved walls hold the building up. Ito first entered the same idea in an international competition for a concert hall in Ghent, Belgium. The jury’s verdict: paper architecture that simply could not be built. It lost.
In Taichung, Huang Ming-ching, vice president of the contractor Liming, says Huang was called into meetings before the drawings could even be read. No shame in that. Nobody before them had built what the drawings showed.
Sprayed concrete failed. What next?
An early scheme considered steel frame plus sprayed concrete, which did not work under Taiwan’s regulations and site realities. The team then groped toward hollow concrete walls with metal-mesh formwork. Arup, the structural consultant, borrowed the logic of a car body shell and turned the curved wall into a solid one, and that opened the door.
The remaining snag was smaller and nastier: the metal mesh formwork could not be removed. Left inside the wall, it risked making the plaster flake off later.
On the drive back from a site visit, Huang blurted out a question: could it just be screwed on? That became the “bead-thread bolt method.” It uses the nuts Taiwanese crews already use to lock formwork. Nuts are welded to the main reinforcing bars, and threaded rods then lock the positioning mesh in place. The mesh problem went away, and the plaster could bond to the concrete. The key move on the world’s hardest building looks like something off a hardware-store shelf.
Before real construction began, the team also built a 1:1 full-scale mock-up called “Overture” to test the method and the strength on the ground.
Fifty-three notebooks and a Wednesday-night meeting
Architect Yang Li-hua filled 53 notebooks on this project, breaking each difficulty down step by step. Facing a thin budget, an unpredictable schedule and a team that kept wanting to refine details, Yang set up a “convergence mechanism”: a weekly Wednesday-night review of construction problems, with every doubt put on the table.
One young Japanese designer spent months working out how to hide an exposed fire-sprinkler line on a curved wall. After the question went to Ito, the answer was one line: “沒關係，露明就好，結構美感更真實” (it’s fine to leave it exposed; the structural beauty is more honest). Months of budget and redesign worries ended there.
Planning versus the site
Yang, who had worked on design supervision in Japan back in 1987, describes Japanese teams as intensely devoted to planning: goals set at every step, everything simulated beforehand, precise and rigorous, but sometimes boxed in by the framework. Yang sat between the Japanese side and Liming as the bridge. Huang’s verdict is sharp: most designers like to chase blame, Yang preferred to coordinate.
Architects and contractors usually stand as watcher and watched. Here the two sides built trust. Yang relays that Ito once said the project might only be realizable in Taiwan; relayed is relayed, so it stays that way. What stands on the ground is a drawing a jury once rejected, finished in 2016 on Huilai Road in Taichung, with a row of ordinary Taiwanese site nuts locked inside its concrete.
Yang’s favorite hour is dusk, when interior light passes through the curved walls and curtain wall and the building seems to breathe. The bolts and nuts inside the wall can’t be seen. That is fine.
— 洪俊傑
延伸閱讀
https://justfly.idv.tw/s/YM63xE7