Development of an in-process Pokayoke system utilizing accelerometer and logistic regression modeling for monitoring injection molding flash

Today, using recycled materials is a common practice in plastic industries for the sake of saving material cost and pursuing sustainable manufacturing. The recycled materials may have some properties (for example, fluidity and viscosity) significantly different from the primary plastic resin, which may lead to quality problems. An in-process Pokayoke system was developed in this research to monitor injection molding parts’ flash caused by adding a foreign polymer in the lab test, which was used to simulate the recycled plastic. The proposed system employed an accelerometer to capture the injection molding vibration signals. The featured injection molding vibration signals were identified through data analyses, and they were then used as input variables through logistic modeling to predict flash in an injection molding process that utilizes pure polystyrene (PS) mixed with a small portion of low-density polyethylene (LDPE). The testing results indicated that this Pokayoke system could monitor the injection molding flash status caused by the mixed material with approximately 95 % accuracy while the injection molding is in process. This Pokayoke system can help the injection molding machine take immediate actions to avoid wastes caused by flash.