Modelling of Fatigue Failure for Plasma Coated Members Using Artificial Intelligence Technique

Abstract

Coating materials in form of powder such as Magnesium Zirconate, Aluminum Bronze and Molybdenum were mixed in different portions and sprayed on steel specimen to find the fatigue properties of steel using plasma technique. The effect of coating mixture on the number of cycles needed for failure under different loads was done experimentally. A cyclic loading was applied to it repeatedly until failure occurs. The results were compared with those for the same specimen without coating. The results were then modelled using Artificial Intelligence Technique then optimized for maximum cycles of coated substance failure. The results showed significant improvement to the specimen’s resistance to failure with coating. Further, models were developed out of the experimental data and tested for accuracy and gave satisfactory results. However, the time consumed by the GA method was greater than that consumed by the same software for the ANN model development.Also, sensitivity analysis showed that the key effect for the variables studied was for the load while the least effect was for the Molybdenum mixture. On the other hand, using GA method, the importance of variables was maximum for the load and minimum for Magnesium oxide and Zirconate oxide mixture Further, using the correlation method, there was strong negative (i.e. inverse relationship) correlation between the number of cycles and load and weak with Magnesium oxide and Zirconate oxide mixture   while strong positive correlation was shown with Molybdenum and least positive for  Aluminum Copper Balance.

Keywords: Artificial neural network, modeling, Plasma coating, fatigue failure.