In situ process monitoring using acoustic emission and laser scanning techniques based on machine learning models

Ke Xu, Jiaqi Lyu, Souran Manoochehri

Research output: Contribution to journalArticlepeer-review

24 Scopus citations

Abstract

Fused Filament Fabrication (FFF) is a widely used additive manufacturing method for obtaining prototypes with complex structures. On the other hand, Commercial FFF machines have limits on process dependability and product quality. Numerous parameters can impact the printing process's accuracy and stability. This study created a real-time monitoring system based on acoustic emission (AE) and laser scanning technology to monitor the warpage defect in fabricated parts throughout the printing process. The AE signal generated during the printing process indicates variations in vibration caused by changes in the part's internal tension. It is capable of precisely displaying real-time part characteristics. After amplifying and filtering the AE signal, the frequency and time domain characteristics are retrieved and trained in the machine learning model. Laser scanning depth image data could be utilized to quantify the dimensional changes of the part. After noise reduction and processing, the point cloud data collected by the laser scanner can represent the specific percentage of warpage. The experimental results demonstrate that the machine learning models trained by AE sensory data can accurately identify if warpage of various sizes is formed with great accuracy in real time. This study examines the support vector machine, naïve bayes classifier, and decision tree data-driven machine learning models.

Original languageEnglish
Pages (from-to)357-374
Number of pages18
JournalJournal of Manufacturing Processes
Volume84
DOIs
StatePublished - Dec 2022

Keywords

  • Acoustic emission (AE)
  • Additive manufacturing (AM)
  • Fused filament fabrication (FFF)
  • Naïve bayes classifier (NBC)
  • SVM (support vector machine)

Fingerprint

Dive into the research topics of 'In situ process monitoring using acoustic emission and laser scanning techniques based on machine learning models'. Together they form a unique fingerprint.

Cite this