研究目的
To predict the relationship between the experimental variables (speed, feed, laser power and beam approach angle) on surface roughness Ra (μm) using response surface method (RSM) and artificial neural network (ANN).
研究成果
ANN models are found to be capable for better prediction of surface roughness within the training range. ANN model is found to be more accurate in prediction as compared to RSM model with a limited number of experiments. ANN prediction model provides a maximal benefit in term of precision 10% for Ra compared with RSM prediction model. The approach angle is the significant parameter for Ra roughness followed by the speed from Pearson correlation analysis.
研究不足
The study is limited to the predictive modeling of surface roughness Ra (μm) in laser assisted turning of Inconel 718 alloy using RSM and ANN approaches. The comparison is based on a limited number of experiments.
1：Experimental Design and Method Selection:
The experiments are planned according to the second order central composite rotatable design (CCRD) based L31 (43) on the principle of response surface methodology (RSM). This design permits to estimate all the 1F: main factor effects, 2FI: two-way, and 3F: quadratic. CCRD design has 31 trials which comprise of 16 factorial (trials 1-16), 6 axial with α= ±2 for approximation of curvature (trials 17-23), and 8 center (trials 24-31) at zero level for replication to determine the pure error.
2：Sample Selection and Data Sources:
Precipitated age hardened Inconel 718 has been turned with laser assistance for a depth of cut of
3：50 mm. A cutting length of 60 mm was machined during each experiment. List of Experimental Equipment and Materials:
A solid state continuous wave 2 kW Nd:YAG laser source is used for heating the workpiece. The surface roughness Ra (μm) value was measured using the Mahr surf test (Model GD120).
4：0). Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The laser was irradiated at the workpiece and once the yield temperature of workpiece material is reached the feed is given to the tool in the axial direction. The measurements of surface roughness are taken at three different locations around the circumference of turned workpiece and the average value is taken for the analysis.
5：Data Analysis Methods:
The relationship between the cutting parameters and objective functions of surface roughness, Ra, are modeled using RSM based quadratic regression equations. The optimal neural network architecture was chosen according to higher correlation coefficient (R2) and lower root mean square error (RMSE) with varying the number of neurons and number of iterations in the hidden layer.
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
