Multi-Objective Optimization of Liquid Silica Array Lenses Based on Latin Hypercube Sampling and Constrained...

Area of Science:

• Materials Science and Engineering
• Manufacturing Processes
• Computational Modeling

Background:

• Injection molding process parameters critically influence plastic product quality, cost, and efficiency.
• Optimizing these parameters is essential for enhancing manufacturing outcomes.

Purpose of the Study:

• To develop a multi-objective optimization method for injection molding processes.
• To shorten the time required for identifying optimal process parameters and boost production efficiency.

Main Methods:

• Utilized Latin hypercube sampling for experimental design.
• Integrated response surface models with a Constraint Generation Inverse Design Network (CGIDN).
• Focused on optimizing residual stress and volume shrinkage for LSR lens arrays in automotive LED lights.

Main Results:

• The CGIDN method achieved optimal residual stress of 8.47 MPa and volume shrinkage of 2.83%.
• Identified optimal parameters: 30°C melt temperature, 2.5s filling time, 40 MPa maturation pressure, and 15s maturation time.
• Demonstrated significant improvements compared to initial sampling results (11.96 MPa residual stress, 4.88% shrinkage).

Conclusions:

• The combined approach of Latin hypercube sampling and CGIDN offers a feasible data-driven method for injection molding optimization.
• The proposed technique effectively improves plastic part quality and manufacturing efficiency.