Electromagnetic Assessment of Fatigue Degradation in Ferromagnetic Steel in View of Statistics and Monitoring
Christian Boller, Iman Ahadi Akhlaghi
Fatigue in metallic materials leads to progressive degradation driven by a sequence of microstructural mechanisms occurring over the life cycle. While fracture is typically the most obvious and critical damage state, it only appears at the end of life. However, when no fracture is monitored a material’s degree of fatigue degradation becomes difficult to be quantified. During that period a variety of other mechanisms act in a consecutive way starting from dislocation movements, phase transformation and leading over to plastic deformation. Electromagnetic techniques such as eddy current testing (ECT) are an interesting option to be considered for monitoring ferromagnetic materials. The attraction is the complexity of these ECT signals and the question arises: What can be read out of those? To address this question, fatigue experiments were conducted on structural steel S235 (ASTM A36) under strain-controlled constant-amplitude loading to determine what is generally known as an S-N curve. All tests were continuously monitored using a commercial ECT system. The signals recorded were processed in terms of the real and imaginary part and the resultant impedance and phase were determined as further parameters characterizing ECT. These were then plotted in a specific 3D format displaying loading over normalized life (in the sense of the traditional S-N curves) added by one of the forementioned ECT parameters as the third dimension. The resulting topography is shown as an example for ECT impedance in Fig. 1. Such a visualization allows the monitoring capability of the respective technique (here ECT impedance) to be discussed. Beyond this visualization, two complementary statistical processing frameworks were employed to analyze the time-domain signal response recorded. First, statistical thresholding techniques—both linear and moving standard-deviation based—were applied to identify significant deviation points that might be correlated with transitions in the material state. Second, a weighted scoring method was introduced leveraging nine time- and frequency-domain features including slope, roughness, RMS, high/low frequency ratio, local variance, kurtosis, zero-crossing rate, envelope rate, and the acceleration of the second-order slope. Robust normalization schemes were incorporated to reduce noise sensitivity and enhance tracking stability. Initial findings have been analyzed regarding these statistical approaches to possibly highlight critical transitions during the fatigue life and exhibit sensitivity to early state changes. The weighted scoring approach in particular shows improved stability compared to single-parameter tracking, making it a possible candidate for future automation. In the broader perspective, this work establishes a data-driven electromagnetic monitoring framework that with possible potential to support future machine learning-based damage classification even to be used in structural health monitoring. Ongoing work focuses on refining feature selection, enhancing thresholding robustness, and evaluating performance across multiple material grades and loading conditions.