Repair job caused the accident, says investigator with FEA

But what may not be widely known is that engineering tools including finite-element analysis (FEA) are being used to investigate forensic evidence. For instance, companies need to know if it was product design, user error, or some other cause that produced a failure. The cause of the accident often determines whether a manufacturer incurs the cost of damages, recalls, and replacements — not to mention potential legal liability for injuries.

Forensic analysts at Herrera, Stafford and Associates LLC (HS&A) of El Paso are often called on to determine why products fail. For instance, after the failure of a compressor-head cover at a Texas oil well, HS&A used Algor FEA to evaluate why and how it happened.

HS&A principal Juan Herrera found the compressor had been used for many years prior to the accident. However, the cover failed after the compressor was restarted following repairs. The failure released natural gas, which found an ignition source and exploded. The ensuing fire seriously injured several workers and caused millions of dollars of property damage.

Herrera found that a pneumatic impact wrench had been used to tighten bolts on the head cover instead of the recommended torque wrench. Tests conducted in the HS&A laboratory using a similar impact wrench revealed that the recommended 80 lb-ft of torque could be generated with supply air as little as 20 psi.

After studying fragments from the failed cover, Herrera concluded several errors had been made assembling the compressor head cover: the bolt-andnut assembly had not been lubricated, torque values were not equal on all eight bolts, and each was screwed to different depths. Also, the recommended torque value was exceeded and the recommended torque sequence was not followed. As a result, the aluminum seal on the bottom of the cover deformed unevenly around the base.

With this information, Anselmo Najera, design and analysis engineer at HS&A, conducted a series of FE analyses to determine how different geometric features and loading conditions affected stresses. He wanted to test design variations with and without the groove on the cover's top surface to find whether it led to high-stress concentrations. There were also concerns that manufacturing variations might be causing high stresses. Casting an iron part, such as the cover, often twists the mold a degree or two. This results in a slight asymmetry in the cover's thickness and the location of the dome's apex. To see whether the asymmetry would affect stress results, Najera planned several variations of the model to represent the dome as designed, with a shift in the dome apex, and with both an apex shift and thickness asymmetry.

Najera created structured-mesh models within the program. "The structured mesh allows specifying exact locations of nodes for the loads and constraints," says Najera. "The asymmetry of the dome was created by incrementally modifying the angle while extruding the mesh. Once I had a quarter of the model, it was mirrored to complete the geometry."

Najera positioned a thin layer of elements between the cover and its aluminum seal to simulate friction and gave it weak material properties so it would easily deform. "This element layer functions much the same way contact elements do," he says.

For all models, the aluminum seal was completely restrained and material properties were applied based on the manufacturer's information. Two loading conditions needed consideration: the bolt load holding the cover in place and the pressure load in the compressor. A variety of bolt loads were considered ranging from 1,024 to 15,000 lb. Najera also considered different loads on each of the eight bolts. The maximum compressor pressure, 600 psi, was applied on the inside surface of the dome for most of the models and omitted for several analyses to isolate effects of bolt loads. In all, Najera analyzed 18 different model variations.

For each linear static stress analysis, Najera looked at the maximum principal stress and compared it to the material's yield stress. He also looked for stress concentrations. "We found neither the groove nor asymmetry of the dome was enough to cause a failure," he says. "The most significant factor was the torque used to tighten the bolts."

In the laboratory, Najera tested each of the 18 FEA cases to verify simulation results. "We found good correlation between FEA and laboratory results," he says. "Tests proved user error, not design or manufacturing flaws, caused the accident."

MAKE CONTACT:

Algor Inc.
(412) 967-2700
Algor.com
Herrera, Stafford and Associates LLC
(915) 534-3995
[email protected]