High-resolution acoustic imaging of machine tools in reverberant workshops using dereverberation beamforming combined with idling noise subtraction

• D ereverberation beamforming combined with idling removal for machining noise imaging. • Efficient dereverberation beamforming reduces error at a lower computational cost. • Idling noise removal suppresses side-lobes and improves source map interpretation. • CLEAN- Spatial Coherence delivers high-resolution, unambiguous localization. • Acoustic imaging validated on drilling and lathe, corroborated by acceleration data. This study demonstrates a practical workflow that integrates background noise reduction algorithms with computationally efficient dereverberation beamforming (DBF) and deconvolution techniques to deliver high-resolution imaging of noise sources emitted from machine tools during machining by suppressing idling noise contributions. Acoustic measurements were performed on a drilling machine and a lathe in a reverberant workshop, with data recorded using a microphone array during both idling and machining. Source maps obtained from cross-spectral conventional beamforming (CBF) using machining data exhibited localization errors and large side-lobes, particularly at higher frequencies, making interpretation difficult. To remove reflections, DBF was implemented using a frequency-dependent Hanning window of optimized length on the cross-correlation matrix. While the DBF improved localization accuracy to within one-tenth of a wavelength during machining, side-lobes remained limiting clarity at high frequencies. When combined with background noise removal (BNR) or eigenvalue identification, organization, and subtraction (EIOS), DBF unambiguously localized the dominant source, while ensemble empirical mode decomposition (EEMD) was less effective. The BNR+DBF and EIOS+DBF maps showed that the strongest contribution in the low-to-mid frequencies was localized near the power transmission region, whereas at higher frequencies it shifted to the tool–workpiece contact region. These findings were corroborated by acceleration spectra measured at key locations on the drilling machine, along with a localization deviation metric. Applying deconvolution further suppressed regions coherent with the main lobe, enhancing the dominant source resolution. The workflow provides engineers with a validated tool for tracing machining noise to its physical origins, offering a pathway toward diagnostics and fault detection.