Halogen detector is an important halogen gas leakage detection instrument. In
order to ensure that the upper and lower shells have the same quality, it is
necessary to use one mold and two pieces in production. Compared with the
conventional one-mold two-cavity process, it is easier to produce warpage and
volume shrinkage. To solve this problem, a multi-objective injection molding
process optimization method based on deep neural network (DNN) model based on
stochastic weight average (SWA) method and multi-objective evolutionary
algorithm based on decomposition (MOEA/D) was proposed. Melt temperature, mold
temperature, injection pressure, holding pressure, holding time, and cooling
time are the six parameters and important structure parameters (gate diameter)
as design variables, warpage, and volume shrinkage rate as the optimization
goal. The neural network model between variable and goal was established, and
the MOEA/D algorithm was used for global optimization. The multi-objective
decision method based on TOPSIS was used to evaluate the Pareto solution set.
Finally, the optimized warpage was 0.3031 mm and the volume shrinkage was
5.527%. Compared with the scheme before optimization, the warpage and volume
shrinkage are reduced by 0.8259 mm and 4.713%, which shows the effectiveness of
the multi-objective optimization method.