Abstract
Martensitic grade stainless-steel is generally used to manufacture steam turbine blades in power plants. The material degradation of those turbine blades, due to fatigue, will induce unexpected equipment damage. Fatigue cracks, too small to be detected, can grow severely in the next operating cycle and may cause failure before the next inspection opportunity. Therefore, a nondestructive electromagnetic technique, which is sensitive to microstructure changes in the material, is needed to provide a means to estimate the specimen’s fatigue life. To tackle these challenges, this paper presents a novel magnetic Barkhausen noise (MBN) technique for garnering information relating to the material microstructure changes under test. The MBN signals are analyzed in time as well as frequency domain to infer material information that are influenced by the samples’ material state. Principal component analysis (PCA) is applied to reduce the dimensionality of feature data and extract higher order features. Afterward, probabilistic neural network (PNN) classifies the sample based on the percentage fatigue life to discover the most correlated MBN features to indicate the remaining fatigue life. Furthermore, one criticism of MBN is its poor repeatability and stability, therefore, analysis of variance (ANOVA) is carried out to analyze the uncertainty associated with MBN measurements. The feasibility of MBN technique is investigated in detecting early-stage fatigue, which is associated with plastic deformation in ferromagnetic metallic structures. Experimental results demonstrate that the magnetic Barkhausen noise technique is a promising candidate for characterizing.
