I have been arguing for validation procedures of computational methodologies for years. Even talking a process proposed for one system and run that process through an evaluation of another system that already exists and is fielded. Not a single engineer, certainly no manager, ever considers that to a be value added concept. It is "I have been doing FEA for 20 years therefore I am right" approach. Even as systems fail over and over and over.
This is the failure mode exactly. Validation by seniority instead of validation by correlation. Running a proposed method against a system that's already fielded is nothing more than asking the question the old lab notebooks answered by default: does the prediction match what the hardware actually did? That conservatism baked into the allowables was earned — decades of prediction-versus-test.
And here's part of why it's gone: people aren't staying in jobs, or at companies, long enough to offer a historical perspective. The notebook system worked because an engineer watched hardware succeed and fail over years and carried that forward. Nobody's keeping that ledger anymore — partly because nobody's in the seat long enough to keep it.
I have been arguing for validation procedures of computational methodologies for years. Even talking a process proposed for one system and run that process through an evaluation of another system that already exists and is fielded. Not a single engineer, certainly no manager, ever considers that to a be value added concept. It is "I have been doing FEA for 20 years therefore I am right" approach. Even as systems fail over and over and over.
This is the failure mode exactly. Validation by seniority instead of validation by correlation. Running a proposed method against a system that's already fielded is nothing more than asking the question the old lab notebooks answered by default: does the prediction match what the hardware actually did? That conservatism baked into the allowables was earned — decades of prediction-versus-test.
And here's part of why it's gone: people aren't staying in jobs, or at companies, long enough to offer a historical perspective. The notebook system worked because an engineer watched hardware succeed and fail over years and carried that forward. Nobody's keeping that ledger anymore — partly because nobody's in the seat long enough to keep it.