Detecting Degradation using Structure-Borne Sound in Flame Torch Cutting

Created by W.Langdon from gp-bibliography.bib Revision:1.9039

@Article{Falkner:2026:procs,

• author = "Dominik Falkner and Christoph Seiringer and Leo Savernik and Markus Steindl and Evans Doe Ocansey and Alexander Kinast and Florian Bachinger and Michael Affenzeller",
• title = "Detecting Degradation using Structure-Borne Sound in Flame Torch Cutting",
• journal = "Procedia Computer Science",
• year = "2026",
• volume = "277",
• pages = "2247--2260",
• note = "7th International Conference on Industry of the Future and Smart Manufacturing (former International Conference on Industry 4.0 and Smart Manufacturing)",
• keywords = "genetic algorithms, genetic programming, industrial AI, structure-borne sound, flame cutting, predictive maintenance",
• ISSN = "1877-0509",
• URL = "https://www.sciencedirect.com/science/article/pii/S1877050926003820",
• DOI = "10.1016/j.procs.2026.02.262",
• abstract = "Flame cutting is a widely used metal processing technique whose performance depends critically on the condition of the cutting nozzle. Nozzle degradation can impair flame quality, compromise safety, and reduce process efficiency. This work investigates the feasibility of using structure-borne sound sensing for automated detection of nozzle wear. In a controlled laboratory setting, we compare alternative sensor placements, acquire an annotated dataset, and extract spectral features for classification. Support Vector Classifier, Symbolic Classification, and Multilayer Perceptron models are evaluated to assess diagnostic potential. Additionally, symbolic regression is employed to validate the findings. Within the constraints of a limited dataset, results indicate that intact and degraded nozzles can be distinguished with high accuracy under stable conditions. While these findings are not yet generalisable to full-scale industrial deployment, they provide a reproducible methodology, practical sensor placement insights, and a foundation for future studies incorporating greater data diversity and operational variability.",
• notes = "Also known as \cite{FALKNER20262247}",

}

Genetic Programming entries for Dominik Falkner Christoph Seiringer Leo Savernik Markus Steindl Evans Doe Ocansey Alexander Kinast Florian Bachinger Michael Affenzeller

Citations