Approximation techniques were used instead of expensive computing analysis in a traditional parametric design optimization of a complex system. A Kriging meta‐model was utilized, which enabled the fit of approximated design characteristics for a complex system such as turbine blades that incorporate a large number of design variables and non‐linear behaviors. This paper aims to discuss these issues.
The authors constructed a Kriging meta‐model with a multi‐level orthogonal array for the design of experiments, which were used to optimize the fatigue life of turbine blades under cyclic rotational loads such as centrifugal force. By combining a seven‐level orthogonal array with the Kriging model, the non‐linear design space of fatigue life was explored and optimized.
A computer‐generated multi‐level orthogonal array provided a good representation of the non‐linear design space information. The results show that not only was the fatigue life of the leading edge of the blade root significantly improved, but also that the computing analysis was effective.
To maximize the fatigue life of the turbine blade, the three‐design variables with seven factor levels were optimized via a Kriging meta‐model. As with the optimization technique, a desirability function approach was adopted, which converted multiple responses into a single response problem by maximizing the total desirability.
