Damage to Pumps from Poor Upstream Pipe Welding: A Metallurgical Failure Analysis

Four, tri-lobe pumps were damaged and required extensive repair during the latter stages of construction for a modification in a pharmaceutical manufacturing plant. The economic loss was significant and the general contractor for the work paid the plant's owner for the loss. The renovated units were installed and are in use. However, the general contractor then sued one of its subcontractors - a welding contractor that fabricated and completed weld joints between piping sections on the inlet side of the pumps. The general contractor claimed poor quality welds in the piping permitted loosened weld fragments to enter the pumps during the required water flow verification of the system and this caused the damage.

I was retained by the plaintiff, the general contractor, as an expert witness in metallurgical engineering. I was asked to complete a root-cause failure analysis and specifically to determine if the tri- lobe pumps had been damaged as a result of loose weld fragments produced in the piping due to poor quality welding by the subcontractor defendant.

A tri-lobe pump consists of two, "three-leaf clover shaped", components that rotate on two, parallel drive shafts. Each tri-lobe is positioned so that one of its "leaves" or lobes contacts and slides in and out between two lobes on the adjacent tri- lobe. As the two shafts continue to rotate the desired pumping action is generated. This somewhat unique design is mechanically efficient and particularly suitable for high purity pumping operations as needed in this application. The dimensional clearance between interacting tri-lobes was necessarily very small. Each tri-lobe was approximately 12-inches in diameter, 2-inches in thickness, specified to be made of Type 316L stainless steel and weighed about 30 pounds. These two heavy, interacting components created considerable force in the close, sliding clearance space between them.

I visually examined the interacting surfaces of two representative tri-lobes that came from damaged pumps. I saw that small bits of metal had been crushed into the surfaces at random spots. I also saw loose pieces of metal that were not embedded in the tri-lobes but were found inside the pump casings. In addition there was a relatively large piece of metal (slightly less than 0.25-inch in diameter) that had been recovered from inside a valve on the upstream side of one of the damaged pumps. I later designated this larger piece of metal as Sample B.

I contacted the pump manufacturer to determine if the tri-lobes were cast or wrought products. The microstructure, i.e., the microscopic-scale crystalline structure, of wrought metal and cast metals of a given alloy are distinctly different. These lobes were wrought. The piping welded by the defendant had been confirmed as wrought, Type 316L stainless steel. However, weld metal always has a microstructure like that of a casting.

With this information in mind, metallurgical laboratory analyses were initiated to answer specific questions: What was the identity of the alloy (or alloys) of several of the alloy bits embedded in the contacting surface of one of the representative, damaged tri-lobes based on their chemical composition? Did these alloy bits have a wrought or a cast microstructure? What was the identity of the alloy of Sample B and did its microstructure indicate it came from a wrought or a cast product? The objective was to gain results that would either refute or confirm that the pumps' failures had come from poor quality pipe welding as claimed by the plaintiff.

A third-party internal inspection of the Type 316L stainless steel piping upstream of the damaged pumps had been completed before I started on the case. A borescope with a mounted camera was used for this purpose and multiple video tapes were generated. The primary objective was to view the condition and quality of approximately 300 welds completed by the defendant. I viewed all of these tapes and found several incidents of what, in my opinion were significant weld defects, e. g., undercuts, irregular weld bead profiles, concavity of beads and most importantly, weld bead convexity that protruded into the normal flow area of the pipe sections. The latter could have been broken off and carried by flow into the pumps during the required water flow system verification.

My opinion from the videos conflicted with that of an engineer retained by the defendant. He testified at deposition that the welds' quality was not inferior except at a few locations that had been cut out and re-welded. The engineer also testified that the metal pieces, embedded and loose, in the pumps could have been generated by other construction workers besides the welding contractor. He stated these pieces could have fallen into a partially opened tank that stored water used for the water verification test. He could not offer specific evidence of these claims.

I provided a written report and was deposed. As in my report, I testified that the metallurgical lab results were significant and in my opinion strongly supported the plaintiff's allegation regarding the cause of the failures. This opinion was further supported by my review of the videos of the condition of the welds inside the welded pipes.

The lab findings showed that all of the metal pieces found in the pumps were Type 316. Note: the analytical lab technique used (EDS) cannot distinguish Type 316 from its low percentage carbon version - Type 316L - but other significant alloyed elements clearly indicated the pieces were either Type 316 or Type 316L. Unfortunately the microstructures of multiple embedded Type 316 bits in the tri-lobe surfaces were distorted. It was impossible to determine if they originated from wrought or from cast products. I proposed that this inconclusive result occurred because the embedded Type 316 pieces had been deformed (during the process of damaging the pumps) because of the crushing forces between the close, sliding motion between tri-lobe rotors. Results for Sample B (also confirmed as Type 316 alloy) found inside the valve were conclusive and useful. This intact evidence clearly showed that the piece had a cast microstructure just like a weld. Thus it was very likely that it was a piece of broken off weld fragment that lodged in the valve without entering the pump. There was no welding around the valve.

The case settled in favor of the plaintiff and the defendant's insurance carrier had to pay the cost of repairing and re-installing the renovated pumps. The majority of my findings supported the plaintiff's allegation of the root-cause of the damage. All the hoped-for lab results were not available but, in my opinion, most of the findings were consistent and conclusive. This illustrates that failure analysis results are not always as complete as we desire.