Hole-Quality Inspection Using Machine Learning Based on Temperature Images Obtained Using Two-Color High-Speed Video for Cu Direct Laser Processes of a Printed Wiring Board

In recent years, as electronic devices have become smaller and more powerful, printed wiring boards have been required to have higher densities. The method of simultaneously processing copper foil and insulation layer using a CO 2 laser to process the blind via holes that electrically connect the multilayered layers has become popular. However, laser processing is a noncontact process, and the board is a composite material, which makes it difficult to ensure quality. It is also difficult to observe the internal state of the processed holes from the outside, and the quality inspection of a large number of holes on a single board relies on destructive inspection via sampling. Therefore, we first propose and evaluate an inspection method using multiple machine-learning methods for multipulse machining. We then investigated whether the accuracy of the anomaly detection varied based on the machined hole parameters.