Conservativeness in Failure Probability Estimate: Redesign Risk vs. Performance

Abstract

Being conservative is in the nature of engineering design in order to compensate for uncertainty. Traditionally in the deterministic design approach, the level of conservativeness has not been a major concern for manufacturers in aerospace since it is defined and strictly required by the regulators as factors of safety. However, once a probabilistic design approach comes into play, the selection of conservativeness of probability of failure estimate considering epistemic uncertainty is an inevitable issue because conservativeness penalizes system performance. This paper investigates the tradeoff between performance and risk in development (redesign after certification test), to which regulators do not pay much attention. We conduct the study in the context of risk aversion of expected utility theory. With a practical design problem, a thermal protection system, it is shown that the level of risk aversion of the decision maker is not the key driver of the conservativeness selection, but the severity of redesign cost is. On top of that, the study reveals that expected utility theory may give unreasonable solutions if the level of risk aversion is set too high.