研究目的
Investigating the prediction of two-dimensional morphology of laser cladding using neural networks to improve the quality of the cladding layer and reduce crack tendency.
研究成果
Both BP and ELM neural networks can predict the two-dimensional morphology of laser cladding, with ELM providing more stable and accurate predictions. Future work could focus on further reducing prediction errors.
研究不足
BP neural network prediction results are not stable with error rates between 10% and 40%. ELM neural network improves stability but still has error rates of 10–20%.
1：Experimental Design and Method Selection:
Utilized a single hidden layer feedforward neural network for predicting the two-dimensional morphology of laser cladding, comparing BP neural network and ELM neural network models.
2：Sample Selection and Data Sources:
Used a YLS-4000 fiber laser by IPG Corporation on 30CrNiMo substrate materials.
3：List of Experimental Equipment and Materials:
YLS-4000 fiber laser, 30CrNiMo substrate materials.
4：Experimental Procedures and Operational Workflow:
Trained neural networks with experimental data to predict cladding layer morphology.
5：Data Analysis Methods:
Compared prediction errors between BP and ELM neural networks.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
