The repair failed because the torch could not reach the root
修過一輪又判退：焊槍伸不進去的根部
The first NDT round found root defects and they were repaired. The second round found more, along the same root. The cause was not the welders' skill, it was the angle the joint allowed.
第一輪 NDT 找到根部缺陷，挖除補焊。第二輪又在同一條根部找到更多。問題不在焊工的手藝，而在這個接頭只讓焊槍擺到那個角度。
A weld root that had already been inspected, repaired and inspected again was still being rejected, and the second-round defects were not at the repairs. They were new indications running along the same root.
The joint was a single-bevel groove on a heavy structural assembly, with the bevelled member landing on the flat face of a plate and a GTAW root pass. Fit-up was accepted and welding went ahead. Once the NDT results came in, the schedule started to slip.
What the two rounds showed
Round one reported two families of defects: planar defects at the root, a mix of lack of fusion and cracks, and volumetric defects higher in the weld, porosity and slag. After repair, round two reported only one family, root planar defects again. Most indications sat at the bottom of the joint, and many of the round-two ones had not been recorded in round one.
When the root was opened up, what came out matched the reports. There was an open void along the root line and a thin sliver of weld metal that lifted out by hand. Metal that comes out in a glove was never fused to the joint.
Why the root never fused
The lack of fusion was almost all on the bevel face, not on the plate face. That detail pointed straight at the cause.
- The plate was in the way. The adjacent plate limited how far the torch could be tilted. In the photos below the torch is at the two ends of its swing, and neither position aims the arc into the root of the bevel.
- The pool formed outside the root. With the cup unable to get in, the arc melted metal on the bevel slope, a short distance out from where fusion was needed.
- The welders pushed it in. They knew the root had to show through on the far side to pass, so they pushed the molten metal into the root. It filled the gap and looked like a root bead, but it was lying against the bevel edge instead of being fused into it.
- Low parameters and weaving made it thinner. Voltage and current were set too low, and weaving on the root pass thinned the bead further.
A weak, partly fused root is a planar flaw waiting for load. Every fill pass adds shrinkage across the joint, and the repair adds more. The reasonable reading of the two NDT rounds is that tight lack of fusion opened into cracks under that shrinkage, which is why round two found indications round one did not.
Nobody set out to hide anything. The welders did what the acceptance criterion rewarded, with a torch that physically could not point where it had to.
What changed on the job
- Voltage and current raised, within the WPS
- Torch aimed at the face being fused, so the arc melts both the joint face and the filler rod
- Where the gap is poor, build up from the bevel side as the WPS allows, and grind out any pass that looks doubtful instead of welding over it
- No pushing molten metal into the root
Where it could have been caught earlier
A WPS and a welder qualification are both tested on coupons with open access. Neither one tells you whether the torch fits the production joint.
- Check torch access at fit-up with the actual torch and cup, not on the drawing
- Weld a mock-up with the obstruction in place and section the root before releasing the joint
- Where the assembly sequence allows, weld the root before the obstructing member goes in
- Write down what to do when the gap is out of tolerance, so it is not left to technique at the root
- After a root repair, re-test the length on both sides of it, not only the repaired spot
When the same root is rejected twice, the review cannot stop at welding parameters, because a WPS is qualified under ideal conditions. Whether the actual welding position is workable, and whether the torch can reach the root at the right angle, can only be judged at the joint.
一條已經檢驗、挖修、再檢驗過的焊道根部還是被判退，而且第二輪的缺陷不在補焊的位置，是沿著同一條根部新出現的指示。
那是大型鋼結構組裝上的單邊開槽接頭，開槽的構件頂在另一片板的平面上，根部用氬焊（GTAW）打底。組對檢查通過後照常施焊，等 NDT 結果出來，進度就開始往後掉。
兩輪檢驗看到什麼
第一輪有兩類缺陷：根部的平面型缺陷，未熔合和裂紋都有；另一類是焊道中段的體積型缺陷，氣孔和夾渣。修補後第二輪只剩一類，還是根部的平面型缺陷。指示大多集中在接頭底部，第二輪裡有不少是第一輪沒有記錄到的。
把根部挖開，看到的跟報告對得上：沿著根部有一道開口的空隙，裡面還有一小片薄薄的焊道金屬，用手就能取下來。戴著手套就拿得出來的金屬，從頭到尾都沒有跟接頭熔在一起。
根部為什麼沒熔進去
未熔合幾乎都出現在開槽面，而不是板面，這個細節直接指向原因。
- 旁邊的板擋住了焊槍。相鄰的那片板限制了焊槍能傾斜的角度。下面兩張照片是焊槍擺到兩個極限的位置，哪一個都沒辦法讓電弧對準開槽的根部。
- 熔池落在根部外面。瓷嘴伸不進去，電弧熔化的是開槽斜面上的金屬，離真正需要熔合的位置還有一小段。
- 焊工把鐵水推進去。他們知道根部要從背面透出來才會過，所以把熔融的金屬往根部推。間隙填滿了，看起來像一道打底焊，實際上只是貼在開槽邊緣，沒有熔進去。
- 參數偏低加上搖擺，焊道更薄。電壓電流調得太小，打底時又搖擺運槍，讓根部焊道變得更薄。
一條結合很弱、只熔了一部分的根部，就是一個等著受力的平面型缺陷。之後每一道填充焊都會在接頭上多加一次收縮，補焊本身又再加一次。對照兩輪 NDT 的結果，比較合理的解讀是原本很緊的未熔合在收縮拉力下被拉開成裂紋，所以第二輪才看得到第一輪看不到的指示。
沒有人存心隱瞞。焊工只是照著允收標準鼓勵的方向去做，而手上那支焊槍實際上對不到該對的位置。
現場後來改了什麼
- 在 WPS 範圍內把電壓電流調高
- 焊槍對準要熔合的那一面，確認電弧同時熔到母材面和焊條
- 間隙不好時，照 WPS 從開槽側堆焊補肉；過程中任何一道有疑慮就磨掉，不疊在上面繼續焊
- 不再把鐵水推進根部
可以更早擋下來的地方
WPS 和焊工資格都是在沒有阻礙的試板上考出來的，兩者都不會告訴你焊槍進不進得了實際的接頭。
- 組對時拿實際的焊槍和瓷嘴到接頭上比，不是看圖面
- 把遮擋物一起做進模擬試件，焊完切斷面看根部，再放行正式施焊
- 組裝順序允許的話，先焊根部，再裝會擋住的構件
- 間隙超出公差時怎麼處理要寫下來，不要留給焊工在根部用手法解決
- 根部補焊後，重測範圍要包含補焊位置的兩側，不是只測補的那一段
同一條根部被判退兩次，該檢討的就不只是焊接參數，WPS 是在理想條件下評定出來的，實際的施焊位置合不合理、焊槍能不能以正確的角度到達根部，要站到接頭前面才看得出來。