Background noise reduction using spectral subtraction for enhanced vibration analysis in precision manufacturing applications

Vibration is an important indicator of machine tool health and for prediction of the surface finish of a manufactured component. Most research has concentrated on the analysis of the dominant sources, usually the machine element under investigation or the process parameters being used on the machine. The quantification of the impact of inevitable background vibration in machine tools has received far less attention but is an underlying factor when determining the fidelity of predictions from vibration measurements. Sources of background vibration could be outside the workshop, such as nearby road vehicles, or within the workshop, such as air handling, operation of other machines, etc. Previous work has shown the characterisation of internal signal and mechanical noise arising from vibration sensors and its effect on vibration analysis. This paper presents a method for the characterisation of influence of background vibration, along with the identification of baseline background noise encountered in a typical machine shop in accordance with ISO 230-8:2010. Spectral subtraction is proposed to minimise the effect of background vibration noise in the analysis. The method can be effective for improving the quality of vibration signals affected by broadband and spectral background noise. This can lead to improved in-situ decision-making by determining the useful, and irrelevant, parts of the signal for different applications. Results from this study will provide viability and improved vibration analysis through the application of spectral subtraction on machine vibration signatures and careful consideration of contributory sources of vibrational background noise in metrology applications.