Session: 03-15-02: Smart Manufacturing and Robotics for the Future II

Paper Number: 166699

AI-Driven Automation for Optimized Injection Molding: Case Study With
Polypropylene (PP) ASTM D638-14 Testing Specimen 

The primary goal of this paper is to leverage AI to automate injection molding manufacturing processes. In plastic manufacturing, injection molding stands out as one of the most widely employed production techniques, remaining in active use today to create numerous products. Optimized plastic products require accurate adjustment of several parameters, such as injection pressure, curing time, cooling duration, and mold cooling. Historically, these processes have depended on the extensive knowledge and experience of skilled professionals to operate injection molding machines effectively. However, with the emergence of the AI era and the evolution of Industry 4.0 within the manufacturing sector, the development and integration of AI-enabled injection molding machines have become an inevitable advancement. Notable companies that manufacture and develop injection molding machines on production lines include ENGEL, ARBURG GmbH, Sumitomo, and BOY . These companies and researchers are accelerating the development of AI-based automation systems in response to the era of Industry 4.0. For example, they are actively researching and developing systems to incorporate AI to solve issues such as product quality and mold cooling.

Quality prediction for injection molded products is crucial for economic advantages. Research indicates a strong link between product quality and temperature field distribution, yet quantitatively predicting quality through temperature analysis remains difficult. A quality prediction method has been introduced that combines infrared thermography with a conventional neural network (CNN) model. An online system using an infrared camera was developed to measure temperature fields in injection molded products. Mass, tensile strength, and warpage deformation were chosen as quality indicators, and various injection molding experiments were conducted to gather thermal images for dataset creation. The images were preprocessed with segmentation and data augmentation to enhance model training. A CNN model was designed to capture the complex nonlinear relationships between temperature fields and product quality, with hyperparameters optimized to boost predictive performance, evaluated using five-fold cross-validation. The extraction and visualization of temperature features were also achieved. Notably, the temperature in the product's boundary region significantly affects its quality. This method advances quality prediction and monitoring in injection molding.

The integration of conformal cooling systems has dramatically improved the efficiency and quality of the injection molding process. Although automated methods exist for designing these cooling channels, optimizing their design parameters remains challenging due to labor-intensive thermal simulations and reliance on human expertise. A novel machine-learning approach that combines nonlinear regression and neural networks has been introduced to assess the thermal performance of conformal cooling systems. Utilizing a logarithmic regression model for temperature prediction and employing a neural network to estimate model coefficients enables designers to evaluate and optimize thermal efficiency with greater accuracy and efficiency, thereby minimizing the reliance on manual simulations.

 

In this context, we propose a data-driven AI solution for injection mold manufacturing that enables real-time monitoring and optimizes control parameters to produce the highest quality products. For the evaluation study, we will collect actual data using Polypropylene (PP) ASTM D638-14 testing specimens, utilizing high-resolution and thermal imaging cameras. We will analyze the thermal images through deep learning techniques to assess product quality characteristics, including morphology, warping, and cooling time. This approach will reinforce the concept of a cyber-physical system, bringing advanced AI applications in injection mold manufacturing closer to reality.

Presenting Author: Jun Han Bae Rochester Institute of Technology

Presenting Author Biography: Jun Han Bae is an Assistant Professor in the Department of Manufacturing and Mechanical Engineering Technology at Rochester Institute of Technology. He obtained his B.S. degree in Mechanical Engineering from Yonsei University, Republic of Korea, in 2011, and M.S. in Mechanical Engineering Technology and Ph.D. in Technology from Purdue University, West Lafayette, IN, USA, in 2014 and 2021, respectively. His research interests are prototyping, robot design and control, field robotics, and manufacturing automation.

Authors:

Jun Han Bae Rochester Institute of Technology
Bell Muthukumaran Rochester Institute of Technology
Massimo Tozzi Rochester Institute of Technology
Maheen Madhi Rochester Institute of Technology
Kyle Wat Rochester Institute of Technology
Jonathan Curran Rochester Institute of Technology
Gilchan Park Brookhaven National Laboratory
Spencer Kim Rochester Institute of Technology