CLASCO announces a new peer-reviewed open access publication in Advanced Intelligent Systems on real-time surface quality monitoring in Direct Laser Interference Patterning (DLIP). The paper, titled “Entropy-Guided Convolutional Neural Network Classification of Sensor Signals for Real-Time Surface Quality Monitoring in Direct Laser Interference Patterning”, is authored by Marcelo Daniel Sallese, Ignacio Tabares, Wei Wang, Marcos Soldera, and Andrés Fabián Lasagni.
The study addresses the challenge of achieving consistent surface quality in DLIP by linking offline surface topography characterization with in situ sensor monitoring. White light interferometry (WLI) images are analyzed in the frequency domain to compute spectral entropy as a quantitative indicator of texture order. Based on entropy values, unsupervised clustering assigns OK and NOK quality labels without manual annotation. These labels are then transferred to time-resolved photodiode signals recorded during processing, together with key laser parameters. A supervised 1D convolutional neural network is trained to predict surface quality using only the sensor signals and process inputs, enabling predictive assessment during fabrication.
Key outputs of the paper include:
An unsupervised entropy-based labeling method that classifies DLIP surface quality from WLI images without manual scoring.
A sensor-driven 1D-CNN model that predicts OK/NOK quality from photodiode signals and laser parameters.
A reported classification accuracy of 90%, supporting real-time quality assurance without post-process metrology.