Evaluation of Mechanical Properties of Materials Based on Genetic Algorithm Optimizing BP Neural Network

Area of Science:

• Materials Science
• Mechanical Engineering
• Computational Materials Science

Background:

• Lightweight materials like magnesium alloys are crucial for aviation, aerospace, automotive, and electronics industries.
• Accurate evaluation of mechanical properties is essential for effective utilization of magnesium alloys.
• AZ31 and AZ91 are widely used magnesium alloys demanding precise property prediction.

Purpose of the Study:

• To develop a predictive model for the mechanical properties of AZ31 and AZ91 magnesium alloys.
• To investigate the influence of processing parameters on alloy performance.
• To enhance prediction accuracy compared to traditional methods.

Main Methods:

• Utilized a dataset of 196 mechanical performance experiments for AZ31 and AZ91 magnesium alloys.
• Employed a 5-8-1 three-layer Backpropagation (BP) neural network optimized by a genetic algorithm.
• Input parameters included deformation temperature, rate, coefficient, solid solution temperature, and time.
• Output parameters were ultimate tensile strength (UTS), yield strength (YS), and elongation (ELO).

Main Results:

• The genetic algorithm-optimized BP neural network accurately predicted UTS, YS, and ELO.
• The model demonstrated reduced discreteness and higher fitness compared to a general BP network.
• Prediction errors for AZ31 magnesium alloy were significantly reduced: 0.88% for UTS, 3.3% for YS, and 8.1% for ELO.

Conclusions:

• The developed predictive model offers high accuracy for magnesium alloy mechanical properties.
• Genetic algorithm optimization enhances BP neural network performance for materials science applications.
• This approach facilitates better material selection and application in lightweight industries.