研究目的
To predict the surface properties of laser processed stainless steel 316 samples using artificial intelligence-based models, specifically ANN and ANFIS, to improve the precision of laser texturing and surface interference.
研究成果
The ANFIS model proved to be more accurate in predicting the results of laser surface texturing compared to the FFNN model, with a 48% improvement in prediction accuracy. This suggests that ANFIS can serve as a precise machine learning method for the LST process, offering better control over the laser texturing and surface interference properties.
研究不足
The study is limited to the use of ANN and ANFIS models for predicting the surface properties of laser processed stainless steel 316 samples. The accuracy of the models is dependent on the quality and quantity of the experimental data used for training and testing.
1：Experimental Design and Method Selection:
The study used ANN and ANFIS models to predict the characteristics of laser surface texturing on 316L cylindrical pins. The models were developed based on a 3^3 factorial design with laser power, pulse repetition frequency, and percentage of laser spot overlap as main processing parameters.
2：Sample Selection and Data Sources:
Cylindrical 316L stainless steel pins of 10 mm in diameter and 60-mm length were laser textured. Data was obtained from design of experiment (DoE) study on laser surface texturing for interference fit technique.
3：List of Experimental Equipment and Materials:
A computerised numerical control (CNC) CO2 laser, Rofin DC-015, with 1.5-kW maximum average power was used. The focal position was set at a distance of 1 mm below the sample surface to achieve 0.2-mm diameter focal spot size over the sample surface.
4：5-kW maximum average power was used. The focal position was set at a distance of 1 mm below the sample surface to achieve 2-mm diameter focal spot size over the sample surface.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The laser process was carried out with varying laser power, pulse repetition frequency, and overlap between each laser scan. The diameter increase, insertion force, and removal force were measured as output parameters.
5：Data Analysis Methods:
The accuracy of the models was measured by calculating the root mean square error and mean absolute error. The reliability of the ANFIS and FFNN models was investigated by comparing percentage error prediction.
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
