Improved Estimation of End-Milling Parameters from Acoustic Emission Signals Using a Microphone Array Assisted by AI Modelling.
- Resource Type
- Academic Journal
- Authors
- Sio-Sever A; Grupo de Investigación en Instrumentación y Acústica Aplicada, Departamento de Ingeniería Mecánica, Universidad Politécnica de Madrid, 28040 Madrid, Spain.; Lopez JM; Grupo de Investigación en Instrumentación y Acústica Aplicada, Departamento de Telemática y Electrónica, Universidad Politécnica de Madrid, 28040 Madrid, Spain.; Asensio-Rivera C; Grupo de Investigación en Instrumentación y Acústica Aplicada, Departamento de Teoria de la Señal y Comunicaciones, Universidad Politécnica de Madrid, 28040 Madrid, Spain.; Vizan-Idoipe A; Departamento de Ingeniería Mecánica, Universidad Politécnica de Madrid, 28006 Madrid, Spain.; de Arcas G; Grupo de Investigación en Instrumentación y Acústica Aplicada, Departamento de Ingeniería Mecánica, Universidad Politécnica de Madrid, 28040 Madrid, Spain.
- Source
- Publisher: MDPI Country of Publication: Switzerland NLM ID: 101204366 Publication Model: Electronic Cited Medium: Internet ISSN: 1424-8220 (Electronic) Linking ISSN: 14248220 NLM ISO Abbreviation: Sensors (Basel) Subsets: MEDLINE
- Subject
- Language
- English
This paper presents the implementation of a measurement system that uses a four microphone array and a data-driven algorithm to estimate depth of cut during end milling operations. The audible range acoustic emission signals captured with the microphones are combined using a spectral subtraction and a blind source separation algorithm to reduce the impact of noise and reverberation. Afterwards, a set of features are extracted from these signals which are finally fed into a nonlinear regression algorithm assisted by machine learning techniques for the contactless monitoring of the milling process. The main advantages of this algorithm lie in relatively simple implementation and good accuracy in its results, which reduce the variance of the current noncontact monitoring systems. To validate this method, the results have been compared with the values obtained with a precision dynamometer and a geometric model algorithm obtaining a mean error of 1% while maintaining an STD below 0.2 mm.