In injection molding environments where the same product is molded in large quantities, there is a challenge that, even when the same molding conditions are set, defective products can still occur due to environmental changes and other factors. Predicting the occurrence of defects before they happen and adjusting the molding conditions accordingly can suppress the production of defective items.
MAZIN, Inc. is engaged in developing AI to solve production challenges at the process level. In the injection molding process, we focus on addressing issues such as skill succession and improving production efficiency. We are working on developing algorithms for the automation of molding condition adjustments, with various experiments and analyses aimed at predicting the occurrence of defective products.
As part of our efforts towards automating molding condition adjustments, we introduce some of our work related to experiments and analyses for predicting the occurrence of defective products.
The initiative introduced here aims to predict the occurrence of defective products by capturing in-mold pressure during molding using pressure sensors attached to the mold and analyzing the time-series waveforms.
Through injection molding using resin, we have clarified the possibility of predicting molding defects by analyzing MAZIN's original feature quantities, which correlate with the state of the resin flowing inside the mold.
By generating and analyzing MAZIN's original feature quantities, which correlate with the state of the resin flowing inside the mold, we found that good products tend to cluster in the center on a scatter plot, while defective products distribute towards the outer edges.
This finding indicates a high potential for predicting the occurrence of defective products.
The visibility into the potential for predicting the occurrence of defective products beforehand opens up the possibility of applying this knowledge to a system that can automatically adjust molding conditions before defects occur, thereby suppressing the production of defective items.
The key to reducing the rate of defective product ejection through an automated molding condition adjustment system lies in how accurately we can predict the occurrence of defects and how effectively we can adjust the molding conditions. Moving forward, we will continue to develop and refine our methods for high-precision defect prediction and condition adjustment.
For more detailed information about the technology or inquiries related to research and development, please contact us.