Degree Type

Dissertation

Date of Award

2004

Degree Name

Doctor of Philosophy

Department

Industrial Education and Technology

First Advisor

Joseph C. Chen

Abstract

Three in-process, mixed material-caused flash monitoring systems in the injection molding process have been developed in this research. They are: (1) the in-process, mixed material-caused flash monitoring (IPMFM) system under fixed processing parameter settings; (2) the artificial neural networks-based, in-process, mixed material-caused flash monitoring (ANN-IPMFM) system under a different combination of processing parameters settings; (3) the fuzzy neural networks-based, in-process, mixed material-caused flash monitoring (FNN-IPMFM) system under a different combination of processing parameters settings.;The vibration signals during the mold opening, closing, and injection filling stages were captured in real-time by the PCB accelerometer sensor, then transmitted, amplified, collected and converted into digital data by DagBook 100 in the personal computer. These data were analyzed to generate a process characteristic indicator and then compared with the flash prediction threshold value based on pure material feeding to determine whether or not flash will flash occur. The IPMFM system was shown to monitor flash with at least 94.7% accuracy. The ANN-IPMFM system consists of two sub-systems: the vibration monitoring sub-system and the ANN-based threshold prediction sub-system. The developing (training) procedure of the threshold prediction sub-system includes four major steps: (1) Construct an experimental design to collect data for ANN training; (2) Determine the input and output variables in the ANN model; (3) Scale and prepare the data set before ANN training; (4) Implement PCN-based training procedure to determine the optimal ANN model. The ANN-IPMFM system was developed by integrating the two sub-systems and showed 92.7% accuracy in monitoring flash. The FNN-IPMFM system integrates two sub-systems: the vibration monitoring sub-system and the FNN-based flash prediction sub-system. The developing procedure of this FNN-based flash prediction sub-system consisted of five steps: (1) Divide both the input and output domains into fuzzy regions and create membership functions; (2) Generate fuzzy rules for the given data; (3) Solve conflicting rules; (4) Develop fuzzy rule bank; (5) Defuzzification. Integrating both sub-systems, the FNN-IPMFM showed 96.1% accuracy in monitoring flash.;The efficient performance in these systems indicates that the ANN-based or FNN-based, in-process, mixed material-caused flash monitoring system can be applied to monitor flash and to help identify and correct mixed material problems in real-time and in an on-line fashion in the injection molding process.

DOI

https://doi.org/10.31274/rtd-180813-14319

Publisher

Digital Repository @ Iowa State University, http://lib.dr.iastate.edu

Copyright Owner

Jie Zhu

Language

en

Proquest ID

AAI3158391

File Format

application/pdf

File Size

112 pages

Share

COinS