FALL, 2007 - Our 30th Year

A PUBLICATION OF THE ACCIDENT ANALYSIS GROUPNORTHAMPTON, MASSACHUSETTS (413) 586-2161

BOLT AND OTHER THREAD FAILURES

HOW TO CELEBRATE 30 YEARS?

By doing what we do well. For this issue wewere tempted to focus on our history, since muchhas happened to us and in the accidentreconstruction field since David Pesuit, Ph.D.,armed with engineering degrees from Lehigh andYale and corporate engineering experience, hungout his shingle on November 10th, 1977.

But we decided instead to focus on areas wherewe’ve made recent contributions. The threadfailure case in this issue is a good example, sinceit shows some clear encounters with junk sciencedeveloped by experts short on test results andlong on points of view.

Next year we will publish correlations thatimprove the overall accuracy of the tire frictionfactors reconstructionists use to estimate vehiclespeeds. We will also write about propane leaksand fires, using a number of cases in our files.

Contributions aside, we must admit to beingproud of our record. We are independent analystsworking for plaintiffs and defendants, with astrong belief in the ability of tests to buttresstestimony and the power of images over words.On the back page of this issue we include a listof some of our more interesting past newsletters.Copies are available at our website:accidentanalysisgroup.com

Martin Schnall and David Pesuit

THE MINNEAPOLIS BRIDGE COLLAPSE

Mechanical failures come in all shapes and sizes.Since this bridge survived for 30+ years, it isobvious to us that its strength degradedsubstantially before the failure occurred.

Bolted systems like the ones in this newsletterare designed with fourfold safety factors. Butsome parts are designed with much higher safetyfactors because they can fail in fatigue at a smallfraction of their ultimate strength. We plan toreview the reports that will follow this disaster,with an eye toward distinguishing between lackof maintenance on one hand and poor engineeringdesign on the other. So stay tuned . . . .

MANY THANKS

Many thanks to attorney Charles E. Gilbert III ofGilbert & Greif in Bangor, Maine who stayedthe course on the thread failure case as an obviousmanufacturing defect turned into a duel ofopposing experts. We are pleased to say that hereached a satisfactory settlement just before trial.

accidentanalysisgroup.com

Page 2

FALL, 2007

A LONG WINDING THREAD FAILURE YARN

Closeup of the area of failure.

Some of you know that we have beenreconstructing failures of threaded connectionsfor more than 25 years. We started out withexploding manhole covers, but soon branchedout to include gunnite cement spray equipmentand spray painting devices.

Early on, we worked with Bill Heronemus, thenProfessor of Mechanical Engineering at theUniversity of Massachusetts, to build B-CALC,a computer model that predicts failure pressurefor threaded connections using basic informationincluding bolt size and number, metal strengthand thread wear.

Most of our thread failure cases have been fairlystraight-forward. But in a recent case in Maine,we watched with wonderment as an obviousmanufacturing defect turned into a battle withopposing experts using bogus theories andmisquoted science that we had to counter everystep of the way.

THE MANUFACTURING DEFECT

It all started when a high-pressure pipe connectionin a large rock crushing machine failed, causinga length of pipe to swing out and seriously injurea worker standing nearby (Exhibit 1).

B-CALC thread strength program from the 1980's

Exhibit 1 – The rock crusher.

Page 3

FALL, 2007

Exhibit 3 – The threads viewed top down.

Just before the accident, the piping manifoldabove this supply pipe had been replaced,and workers had aligned and tightened thesupply pipe. A hydraulic pump was beingused to pressurize the system, and pressurewas being monitored because a valve hadto be shut off at 2,500 – 2,600 pounds persquare inch (psi). One of the workers hadjust looked at a gauge and saw that thepressure was 2,500 psi when the men hearda loud bang and saw oil come flying out.

Exhibit 2 shows the supply pipe, themanifold section to which its top elbow wasattached, and the method of attachment. Aferrule or collar was threaded onto the endof the manifold pipe to provide a stop forthe large threaded nut used to secure thesupply pipe to the manifold pipe. Whenthis connection failed, the ferrule pulled offthe manifold, and the supply pipe rotatedforward.

We studied this failure by comparing thefailed connection with an identicalconnection at the bottom of the supply pipethat did not fail. Exhibit 3 shows two edge-on photos comparing the failed manifoldpipe threads with the pipe threads at thebottom end of the supply pipe. Exhibit 4shows two side-on photos comparing thesesame threads.

Exhibit 2 – Supply pipe and manifold assembly.

Exhibit 4 – The threads viewed in profile.

Page 4

FALL, 2007

It was obvious to us that the tops of the newmanifold pipe’s threads were not sheared off asthe ferrule came off the manifold pipe. Themanifold pipe threads showed no plowing alongtheir axis; they were sharp edged, andcircumferential machining marks were still visibleon their tops. Exhibit 5 shows that duringmanufacture the diameters of both pipe ends werereduced before they were threaded to accept theferrules, but that much more metal was cut awayfrom the surface of the new manifold pipe. Wehypothesized that the ferrule pulled off the newmanifold pipe because it had much less thread tograb.

We checked our hypothesis by calculating thestrength of the failed connection using the sameequations we had developed for the B-CALCmodel in the 1980’s, in which the pressure forcepulling the connection apart was just balanced atfailure by the shear force bulldozing the ferrulethreads. With only about 29% of the ferrulethreads damaged, the calculated failure pressurewas almost exactly 2,500 psi.

The same calculation for the undamaged threadsin the like connection at the bottom of the supplypipe gave a failure pressure of 8,500 psi, almost3.5 times greater than the 2,500 psi systemoperating pressure. A 4:1 safety factor has beencommon in mechanical engineering hydraulicdesign practice since World War II.

Therefore, we were certain this accident happenedbecause the new manifold section was incorrectlymade. Too much of the metal near the end ofthe manifold pipe had been removed to reduceits outside diameter before the pipe was threadedto receive the ferrule. The resulting pipe threadswere truncated. This error created the situationwhere the manifold’s pipe threads met the ferrulethreads near their tips where they were weaker.The pipe threads bulldozed through the ferrulethreads, allowing the ferrule to slide off and letgo.

Exhibit 5 – The manifold pipe was turned down too muchbefore threading for the ferrule.

THE FIRST DEFENSE

Although the above analysis was solid andfully documented, we expected resistance. Butwe were surprised by the bogus analyses thatfollowed over the next two years.

An engineer employee of the manufacturerfocused on a small dent in the elbow at the topof the supply pipe (See Exhibit 6), saying thatthis dent, post manufacture, prevented the nutfrom tightening fully against the ferrule andallowed hydraulic oil to escape and pressurizethe annular cavity of the sealing nut (See Exhibit7), increasing the force pulling the joint apartand causing it to fail. He provided no evidencethat hydraulic oil had entered this annulus, andno calculations to support his position. And herefused to acknowledge that the flattened threadswere a manufacturing defect, even though themarks on the tops of the threads werecircumferential, not axial.

We used the evidence to refute his position. Weproduced photos like Exhibit 6, clearly showingthat the seal between the ferrule and the elbowwas not affected by the dent.

And we used the elbow geometry in Exhibit 7 torefute his position. If hydraulic oil did leak into

Page 5

FALL, 2007

this annulus, the greatest additional axial force it could create was the ratio of the increased areassince force is the product of pressure and cross-sectional area. The maximum diameter of theannulus was 2.51 inches; the inside diameter of the elbow 2.02 inches. Therefore, the maximumfractional increase in axial force due to such a leak was the square of the diameter ratio:

Fractional axial force increase = [cavity maximum diameter / elbow diameter]2

= [ 2.51 inches / 2.02 inches ] 2 = 1.54

which is a 54% increase. But industry codes and the 8,500 psi calculation above showed that thishydraulic system was designed with a 4-fold safety factor. Therefore, there was no way that a 54%increase in internal axial force would make the system fail.

Exhibit 7 – Scale drawing of the threaded joint.

Exhibit 6 – The dent in the elbow and seating area.

Page 6

FALL, 2007

THE SECOND DEFENSE

A PhD engineer hired as an expert witnessby the defense also focused on the dent in theelbow, but he came up with an entirely differenttheory. He said the dent distorted the ferrule,allowing it to slip over the manifold pipe threadsand make the joint fail. We placed the ferrule inthe elbow without the manifold pipe andtightened the lock nut over it (See Exhibit 9),then measured the inside diameter of the ferruleat many places and found no distortion;the ferrule actually shrank. Therefore, theferrule would have gripped the pipe threadseven more tightly as the elbow nut wastightened, not slipped over them as thisengineer had opined.

This engineer also suggested that the ferrulecould dilate due to the forces generated bythe ramping of its threads as axial forcesgenerated by internal pressure pulled thejoint apart. But reference to the literatureshows that the dilation he spoke of is onlyvalid for unconstrained nuts. In this case,the ferrule was constrained by the extremelythick backing nut (See Exhibits 6 & 9).

Finally, this fellow calculated a failurepressure for the flattened threads that wasmuch higher than the 2,500 psi at whichthe joint failed, using generalized equationsfound in the “Bolts and Nuts” section ofMachinery’s Handbook. Therefore, in hisopinion, the flattened threads wereirrelevant, and our 2,500-psi calculationwas wrong, even though it agreed with the failurepressure. We noted that a recent textbookreferencing these same equations warns that theyare not valid if the thread geometry gets too faraway from the ASME B1.1 standard to whichsuch threads are built. This warning is not in thehandbook. His calculations were not valid inthis situation because the threads were completelyout of spec.

Exhibit 9 – The ferrule installed in the elbow.

AND THE UPSHOT?

This case settled to our client's satisfaction justbefore trial, but it seems ridiculous to us that aseriously injured plaintiff had to wait two extrayears before a reasonable settlement offer wasmade. Junk science mouthed by hired guns maymake economic sense in some quarters, butfortunately, firms like ours don't buy it. Weprefer to use objective, supportable science to doa thorough, accurate job.

ENGINEERING CONFIRMATION

Those of you who want to see the calculationthat the joint failed at 2,500 psi may call us oremail us at accident.analysis@verizon.net . Wehave instead used the facing page to list some ofthe more interesting topics we’ve covered innewsletters over the years.