The Implementation of Neural Networks for Polymer Mold Surface Evaluation

Area of Science:

• Materials Science and Engineering
• Manufacturing Technology
• Surface Metrology

Background:

• Additive manufacturing (AM) is increasingly used for mold production, presenting unique surface quality challenges.
• Maintaining mold surface integrity is critical due to degradation during use, necessitating regular evaluation.
• Direct scanning of injection mold surfaces is difficult, making surface replication essential for accurate analysis.

Purpose of the Study:

• To describe the production of surface replicas for in-house developed polymer molds.
• To determine suitable descriptive parameters and methods for comparing variances and mean values in surface evaluation.
• To present a novel summary of the replica evaluation process for polymer molds.

Main Methods:

• Development of a surface replication technique for polymer molds.
• Application of nonlinear regression to establish functional dependencies between parameters.
• Utilizing a neural network (Rosenblatt's perceptron) for statistical significance verification.
• Employing machine learning for comparative analysis between original surfaces and replicas.

Main Results:

• Successful production and characterization of polymer mold surface replicas.
• Identification of key parameters and statistical methods for replica surface evaluation.
• Validation of a neural network model for assessing surface quality.
• Demonstrated effectiveness of machine learning in comparing original and replica mold surfaces.

Conclusions:

• The developed replica production and evaluation method provides a reliable approach for assessing AM polymer molds.
• Nonlinear regression and neural networks offer robust tools for analyzing surface topography and quality.
• Machine learning enhances the comparison accuracy between original mold surfaces and their replicas, supporting quality control in additive manufacturing.