Thursday, September 18, 2014

I'm being overly dramatic, right?

So my last statement was rather extreme: "...there is one rule that MUST NOT be forgotten..." OK, I admit, it does seem rather extreme, but before you rush to judgment, consider two cases:

Case 1: Several years ago, I was involved in modeling a pollution control system. The system was part of a gas turbine power plant and was located after the gas turbine and was designed to remove NOx from the exhaust gas stream. The company where I worked was selling one component of the system and I was using CFD to demonstrate the effectiveness of our equipment. At the same time, the engineering firm designing the system had hired an independent consulting firm to use CFD to model the entire system to help ensure the effectiveness of the finished design. I supplied 3D models of our equipment to the consultant who included my equipment in his model and performed the analysis. When I reviewed the results of the modeling, I discovered that the consultant had incorrectly used a "symmetry" boundary condition. The use of this incorrect boundary condition completely changed the results of the analysis.

OK, the results of this error weren't too bad; it resulted in more time to fix the boundary condition and to rerun the analysis and modify the design. The only cost in this case was money. But what if more than money were at stake?

Case 2: In Chapter 15 of his book To Engineer Is Human, Henry Petroski relates the following incident:

"The two and a half acres of roof covering the Hartford Civic Center collapsed under snow and ice in January 1978, only hours after several thousand fans had filed out following a basketball game..."

Subsequent investigations revealed that the FEA model used to design the structure did not accurately represent the combination of the dead weight of the roof plus the live loads (snow and ice load and other loads) acting on the roof and its supporting structure. This was the result of an incorrect boundary condition being used to model the loading. In the post accident analysis, a computer model showed the failure when the correct boundary conditions were used. Petroski puts in this way: "The computer provided the answer to the question of how the accident happened because it was asked the right question explicitly..." (italics mine). Unfortunately the right question was only known after the accident. If the failure had happened a few hours earlier with the arena filled with spectators, the cost of the failure would possibly have included deaths and injuries to many people.

So once again, I will repeat my statement: There is one rule that MUST NOT be forgotten:

GABAGE IN = GARBAGE OUT


No comments:

Post a Comment