Sam,

I offer the following horror story from the many I have witnessed over the years:

We designed a plant that had three, nearly-identical units. The units were to be built in sequence. Unit 1 was to be built and put into service. Unit 2 was to be constructed a year-or-so later, and Unit 3 was to follow a year after Unit 2. Each unit contained a hot line connected at each end to equipment nozzles. I don’t remember the temperature details but I recall that the line was a special alloy about 24” diameter. Analysis was performed for the line on Unit 1. For reasons that were never determined the analysis did not appear to have been checked.

By the time the equipment for Unit 2 was procured the design of some of the equipment had changed necessitating a minor alteration to the line in question. A second engineer made appropriate adjustments to the initial analysis and in doing so uncovered a major error: The nozzle moment allowables given by one of the equipment vendors had been given in inch-pounds; the analyst had taken these to be foot-pounds. The result, of course, was that the nozzle on Unit 1 was seriously overloaded. The engineering company had to confess the error to the client and, at its own expense, redo the line for Unit 1 and replace it during a shutdown.

Unit 3 turned out to be identical to Unit 2 so the line was constructed and installed per the Unit 2 design. The construction manager wanted approval to buy the spring hangers. Due to the confusion associated with the previous analyses and some dubious record keeping there was some doubt as to what springs were required. A third engineer was asked to review the earlier work and determine the appropriate hangers. In doing so he noticed an additional error: The vendor allowables for the second nozzle were resultant values. Both the previous analyses had assumed that these values were actual allowables for each degree of freedom. (To clarify: The vendor had provided maximum values of FR and MR; the analyses had assumed that the value of FR was the max allowable for FX, FY and FZ and the value of MR was the max for MX, MY and MZ.)

It was never determined how the original analysts made such a blatant mistake. In part the problem may be attributed to the modern trend to get the maximum amount of work from the minimum number of staff. Certainly, the engineering company in question had enlisted personnel from at least four different offices during the course of the various analyses. In fact, after the last error was discovered, the line was sent to yet another office for re-analysis! Another cause may have been late acquisition of vendor data or some confusion in the transmit ion of the information. There was some evidence that the first engineer (possibly, “engineers!” since it was unclear whether the initial work had been started and finished by the same person) had sent more than one email and a few phone calls requesting the nozzle allowables. The final result, however, was that the engineering company had to go back to the client a second time and replace both the Unit 1 and Unit 2 piping with a third design, incorporating new springs and expansion joints.

I conclusion I think the following lessons-learned can be drawn from this story:

(1) Inadequate staffing caused work overload and consequent errors.

(2) Record keeping was inadequate possibly due to the work overloads.

(3) Proper checking must always be performed before the work is approved for construction.

(4) The checking package should be retained along with any vendor information used.

(5) On critical systems a second check should be considered.

(6) Checkers should be appropriately trained and experienced individuals and the results of the check should be communicated to the original analysts.