San Bruno Blast: New light on welding flaws from the NTSB

Thanks to the NTSB's latest information about the San Bruno gas explosion, the Metallurgical Group Chairman Factual Report, now we know where the crack started. It was in a lengthwise seam (called a "longitudinal weld") rather than where the pipe segments were joined up, end to end (called a "girth weld"). And it started in a particular short segment of pipe (called a "pup").
As a reminder, here's a diagram from page 17 on how the various pups and two long sections joined up. 

Attention is focused on the fragment that extends from Pup 4 on the north, to a fracture in the long joint on the south. (It's marked by a bracket on the left side of the diagram.) 

Originally entrenched three feet below grade, this 27' 8" fragment split open and flew out of the ground, landing in the street a hundred feet away. Here's a photograph of the fragment, from page 19. 

Metallurgists have concluded from close study of chevron marks and fracture faces that the failure started in the wall of Pup 1. It's pictured on page 51:

In this December post I guessed that Board experts had found the initiation site in either Pup 1 or 2, so I was only half-right. 

The crack started not in a girth weld where it connected to another pup or a long section, but a longitudinal seam. This is surprising. At the time when the pipe used at San Bruno was fabricated (late 1940s) such longitudinal seams were normally made in a factory setting. Hydraulic rams and rollers bent a thick steel sheet into a tube shape, then automatic welders joined the edges.

I would have thought that failure was more likely to begin in a girth weld, since girth welds are made in the field and tend to have more quality problems than longitudinal seam welds done in a factory setting. 

To see what factory seams should look like, take a look at the weld shown on page 62. It was in a long section of pipe, not a pup.

The upper surface is the outside of the pipe. Notice that the full thickness of the pipe wall has been joined, with no gaps, cracks, or flaws visible. The only problem visible is a slight misalignment of the two sides when joined, about five degrees off kilter.

For a poorly executed longitudinal seam, see this photo on page 63:

Not coincidentally, this is a cross section from a seam in pup 1, and it's near where the crack first started. 

The welding was done from the outside only, and penetrated to just 55 percent of the depth. According to standards of the day an acceptable weld should have joined 90 percent of the depth. Anyone checking the inside of the pipe later would have seen that the seam hadn't been welded from inside. Instead he would have seen a gap, like the one on the inside of Pup 3 (page 26):

That's not all. Go back to the picture from page 63, and look at the outer surface of the pipe, along the top ... Notice how the pipe wall is rather flat, rather than gently rounded like it should be. 

Apparently somebody took a grinder to the outer surface of the pipe, making the steel wall about .05 inches thinner than when it came out of the mill. The thinning is most noticeable on the right side (the side labeled "counterclockwise").

Why would anybody grind so much steel off the surface of the pipe wall? It would take a fair amount of effort. One hint: the report points out a serious misalignment of the two edges, labeled "radial offset."

It's only one of many quality problems called out. See page 40 for a photo of a section of welding rod found embedded in a girth weld joining Pups 1 and 2:

I expect the Board's final report will dig into how such a sustained system fracture from pipe fabrication through installation could have happened, and gone undetected.