MFO Tunned SVR Models for Analyzing Dimensional Characteristics of Cracks Developed on Steam Generator Tubes

Accurate prediction of material defects from the given images will avoid major cause in industrial applications. In this work, a Support Vector Regression (SVR) model has been developed Gray Level Co-occurrence Matrix (GLCM) features extracted Magnetic Flux Leakage (MFL) wherein length, depth, and width are considered response values data set, percentage for testing SVR model. Four parameters like Kernel function, solver type, validation scheme, its value % that affect model’s performance to select best Six different kernel functions, three kinds solvers as two schemes, 10% 30% set levels above parameters. The accuracy is by simultaneously minimizing measures both Root Mean Square Error (RMSE), Absolute (MAE) maximizing R2 values. Moth Flame Optimization (MFO) algorithm implemented four based on conflict converting multi-objectives into single object using Technique Order Preference Similarity Ideal Solution (TOPSIS) method. MFO compared statistically with Dragon Fly Algorithm (DFO) Particle Swarm (PSO).