Abstract:

It is obvious, that many students have to carry out some experimental research during their PhD studies at technical universities. So, in technical practice, it is very important to understand how factors and responses relate to each other, and to reveal which factors are influential for which responses. If some processes are very complex, it is necessary to recognize and identify the relationships between considered variables only experimentally. In order to minimize costs and time and maximize the reliability and objectivity of information obtained about studied system during experimental work, it is necessary to know how perform experiments with a minimum number of experimental runs and consider interaction between factors. That is why the analysis of these processes by classic methods appears to be non-efficient and many times leads to incorrect conclusions. The change of only one selected factor in given time is considered at COST approach within the frame of experimental work, which is inefficient approach, because it does not provide necessary information in order to reach real optimum, experimental work is overpriced. Design of experiments (DOE) methodology is very useful for this purpose, whereby it enables us to obtain the maximum amount of information with high statistical and numerical correctness at minimum number of performed test runs during experiment and authors have a lot of experience with it. The authors of the paper present the usefulness of DOE methodology and programme system R on own experimental research. The paper deals with the application of mathematical and statistical methods to identify and analyse physical and chemical factors (the electrolyte temperature, anodizing time, amount of sulphuric acid H2SO4 in electrolyte, concentration of aluminium cations Al in electrolyte, etc.) acting during the process of aluminium anodic oxidation on the resulting layer thickness at defined current density. Based on DOE methodology the set of 26 experimental runs was carried out, data were analysed and mathematical prediction model was developed. This article is not only helpful for students as inspiration how to design experiments for doctoral thesis. Moreover it can be stated that the results obtained by this experimental work have important benefits for technical practice, because they were practically verified under conditions of real production.

Keywords:

Design of Experiments (DOE), mathematical-statistical model, R programming system.