Better Quality Control: Stochastic Approaches to Optimize Properties and Performance of Plasma-Sprayed Coatings
- PDF / 670,601 Bytes
- 14 Pages / 593.972 x 792 pts Page_size
- 48 Downloads / 189 Views
Robert B. Heimann (Submitted June 18, 2009; in revised form July 25, 2009) 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.
Keywords
Artificial Neuronal Networks, D-optimal designs, fuzzy logic control, statistical design of experiments, Taguchi designs
1. Quality Implementation Owing to the chaotic nature of the plasma spray process, the structure and properties of coatings produced by this technique are subject to stochastic fluctuations. Consequently, many intrinsic and extrinsic parameters and their generally complex interactions affect in a nondeterministic way the properties and hence the in-service performance of plasma-sprayed coatings. It has been good practice for a long time to select those sets of parameter that are known to influence coating structure and properties to the largest extent, and to run a series of experiments with these statistically distributed parameters. As a result, the directions and magnitude of trends can be estimated with confidence, and a protocol established to manufacture coatings on an industrial scale that will adhere to stringent quality control measures and be monitored by statistical process control (SPC).
Robert B. Heimann, Professor Emeritus of Applied Mineralogy and Materials Science, Am Stadtpark 2A, 02826 Go¨rlitz, Germany. Contact e-mail: [email protected].
Journal of Thermal Spray Technology
An important aspect of industrial plasma spray technology is the development and implementation of quality control and assurance protocols to ensure consistency of properties. Because a multitude of spray parameters can potentially influence coating properties in a nonlinear way, parameter optimization involves statistical experimental design procedures. Such procedures provide a maximum of information on the anatomy and behavior of a system with
Data Loading...
