Numerical correction for spark mapping analysis for large samples of metal materials

Spark Mapping Analysis for Large Samples (SMALS) is a powerful tool for characterizing the elemental distribution in large-sized metallic components. However, during long-duration and large-scale continuous scanning, the raw data is often affected by issues such as systematic spectral drift, abnormal row interference, and edge anomaly data. These problems lead to insufficient quantitative accuracy, which restricts the application of SMALS in the industrial field.

To address these issues, this study constructs a comprehensive numerical correction model consisting of three core steps. First, through matrixization and bidirectional row-column linear correction, the overall systematic errors caused by factors such as spectral drift are effectively eliminated. Second, a Hampel filtering algorithm with an adaptive window is introduced, utilizing the local median and Median Absolute Deviation (MAD) to dynamically identify and correct abnormal data rows, thereby filtering out local interference. Finally, to address numerical anomalies in the edge regions, a novel dynamic weighted fusion strategy combining linear fitting and quartic polynomial fitting is innovatively adopted for correction. Using cross-sections of low-alloy and alloy steel ingots as validation objects, the maximum relative deviation between the corrected SMALS data and the single-point measurements from spark optical emission spectrometry decreased from 6.65% to 3.08%, with the relative deviation distribution approximately following a normal distribution. Meanwhile, the corrected SMALS surface scanning results demonstrated good consistency with the μ-XRF surface distribution data; additionally, simulation verification studies under controlled error conditions quantitatively evaluated each calibration step, showing that the overall Root Mean Square Error (RMSE) of the calibrated simulated data approaches the theoretical limit.

The comprehensive correction model established in this study not only significantly enhances the precision and quantitative accuracy of SMALS surface scanning data, providing a systematic and reliable data processing workflow, but also successfully overcomes the limitations of traditional methods in handling complex systematic errors. This method lays a solid technical foundation for the precise characterization of macro-segregation in large-sized metallic materials and holds significant importance for promoting the practical application of SMALS in metallurgical quality control and materials science research.