Better Quality Control: Stochastic Approaches to Optimize Properties and Performance of Plasma-Sprayed Coatings

Abstract

Statistical design of experiment (SDE) methodology applied to design and performance testing of plasma-sprayed coatings follows an evolutionary path, usually starting with classic multiparameter screening designs (Plackett-Burman), and progressing through factorial (Taguchi) to limited response surface designs (Box-Behnken). Modern designs of higher dimensionality, such as central composite and D-optimal designs, will provide results with higher predictive power. Complex theoretical models relying on evolutionary algorithms, and application of artificial neuronal networks (ANNs) and fuzzy logic control (FLC) allow estimating the behavior of the complex plasma spray environment through validation either by key experiments or first-principle calculations. In this review, paper general principles of SDE will be discussed and examples be given that underscore the different powers of prediction of individual statistical designs. Basic rules of ANN and FLC will be briefly touched on, and their potential for increased reliability of coating performance through stringent quality control measures assessed. Salient features will be reviewed of studies performed to optimize thermal coating properties and processes reported in the pertinent literature between 2000 and the present.