Monitoring of friction-related failures using diffusion maps of acoustic time series

Friction-related failures are a frequent cause of catastrophic damage in mechanical systems, resulting significant repair expenses and/or production delays industry worldwide. Scuffing is one the most concerning friction-related mechanisms. It characterized by an abrupt failure mechanism, which makes its prediction unresolved today. Therefore, any situ and real-time monitoring system such great demand this contribution supplement to enrichment existing studies on topic. We propose probabilistic analysis acoustic signals synchronously recorded with coefficient friction during real-life experiments. They, simulate specific industrial operating condition cam, cylinder pairs gear teeth. The experimental conditions reciprocating movement stainless steel (counter-body) against grey cast iron body. counter-body was under constant load had line contact from one-directional strokes selected substituted their wavelet spectrograms. Then, diffusion maps computed over several Their corresponding Fourier spectra used as features surface state so upcoming failures. real were identified dynamical behavior friction. Based our approach, it found that can be divided into four confidence intervals, three for steady-state regime scuffing regime. transients between intervals allow estimating proximity major achievement work. first transient took place 5 min 40 s. – 47 30 prior scuffing. second 50 13 proposed has potential since built client server architecture requires minimum modifications commercial machines.