Raziskava metod strojnega učenja za napovedovanje zvočnega tlaka

Anko, Matej (author), Slavič, Janko (mentor)

Velike količine podatkov, ki se beležijo tekom proizvodnega procesa in končne kontrole nam omogočajo, da z uporabo metod strojnega učenja izdelamo model, ki napove vrednost meritve na končni kontroli še preden je bila ta izvedena. Raziskali smo vpliv masne neuravnoteženosti na amplitudo odziva in kako nastane hrup. Prikazali smo celoten proces izdelave modela z uporabo Python knjižnice Scikit-learn. Izdelali smo model za napoved amplitud zvočnega tlaka pri posameznih lastnih frekvencah in modele za napoved hitrosti vibracij ter jih ocenili.

• Language: Slovenian
• Keywords: strojno učenje, scikit-learn, masno uravnoteževanje, zvočni tlak, vibracije
• Work type: Final paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FS - Faculty of Mechanical Engineering
• Place of publishing: Ljubljana
• Publisher: [M. Anko]
• Year: 2022
• Number of pages: XIV, 45 f.
• PID: 20.500.12556/RUL-139126
• UDC: 004.83+534(043.2)
• COBISS.SI-ID: 121344003
• Publication date in RUL: 31.08.2022

Secondary language

• Language: English
• Title: A research on machine learning for the estimation of sound pressure level

Abstract

Large amounts of data recorded during manufacturing process and final inspection enable us to use machine learning methods to create a model that predicts value of the measurement at end of line inspection even before it has been carried out. We researched the effects of mass imbalance on the amplitude response and how noise is generated. We showed the entire process of making the model with the use of Python library Scikit learn. We created a model for prediction of sound pressure level and model for prediction of vibration velocity and evaluated them.

• Keywords: machine learning, scikit-learn, mass balancing, sound pressure, vibrations