Author Affiliations
1School of Optoelectronic Science and Engineering, Soochow University, Suzhou 215006, Jiangsu, China2School of Mechanical and Electrical Engineering, Soochow University, Suzhou 215021, Jiangsu, Chinashow less
Fig. 1. Comparison of single-track section in a single-layer and multi-layers
Fig. 2. Principle of powder feeding in annular laser beam
Fig. 3. Structural representation of closed-loop control of cladding height system
Fig. 4. Variable angle thin wall deposition experiment. (a) 30°; (b) 60°; (c) 90°; (d) 135°
Fig. 5. Comparison of forming effect at the same inclination angle (partially)
Fig. 6. Trend of measurement data by layer height control system. (a) Trend of thin wall layer height; (b) cross section of thin wall
Fig. 7. Variation of measured cladding layer height with process parameters
Fig. 8. Variation of measured cladding layer width with process parameters
Fig. 9. BP neural network topology
Fig. 10. Prediction result of test set. (a) Prediction result of layer width; (b) prediction result of layer height
Fig. 11. Prediction effect of cladding layer height and width models. (a) Layer height prediction variance; (b) layer width prediction variance
Element | Mass fraction/% |
---|
Fe | Bal. | C | 0.1 | Cr | 15.0 | B | 1.0 | Si | 1.0 | Ni | 1.0 |
|
Table 1. Chemical composition of F314 iron-based alloy powder
Angle /(°) | Fixed scanning speed Vs /(mm·s-1) | Processing window of laser power P /W |
---|
0 | 8 | [600,1600] | 30 | 8 | [600,1600] | 60 | 8 | [750,1500] | 90 | 8 | [800,1500] | 120 | 8 | [800,1400] | 150 | 8 | [800,1200] | 180 | 8 | [800,900] |
|
Table 2. Single factor experiment table of laser power
Angle/(°) | Fixed power P /W | Processing window of scanning speed Vs /(mm·s-1) |
---|
0 | 800 | [3.5,8] | 30 | 800 | [3.5,8] | 60 | 800 | [3.5,8] | 90 | 800 | [3.5,8] | 120 | 800 | [3.5,8] | 150 | 800 | [4,8] | 180 | 800 | [6.5,8] |
|
Table 3. Single factor experiment table of scanning speed
Network parameter | Width-BP | Height-BP |
---|
Learning rate | 0.1 | 0.1 | Maximum number of iterations | 5000 | 5000 | Training target error | 0.01 | 0.01 | The number of hidden neurons | 4 | 4 | Nodes of each hidden neurons | [6,10,10,6] | [6,10,10,6] |
|
Table 4. Neural network parameter of angle-varied cladding height and width