Author Affiliations
1CAS Nanjing Astronomical Instruments Research Center, Nanjing, Jiangsu 210042, China2University of Science and Technology of China, Hefei, Anhui 230026, China3CAS Nanjing Astronomical Instruments Co., LTD., Nanjing, Jiangsu 210042, Chinashow less
Fig. 1. Framework of DLCM algorithm
Fig. 2. Framework and parameters of kinetic model
Fig. 3. Loss decline curve of strategy network
Fig. 4. Framework and parameters of strategy network
Fig. 5. Convergence process of evolutionary strategy algorithm
Fig. 6. Actuator distribution map. (a) Support method of standard spherical mirror; (b) ANSYS simulation model
Fig. 7. Running time of DLCM algorithm
Fig. 8. Comparison of effect of first correction. (a) Before correction; (b) corrected results
Number of layers | 3 | 4 | 5 | 6 |
---|
Accuracy /% | 91.35 | 95.14 | 97.54 | 97.60 |
|
Table 1. Relationship between model network layers and accuracy
Dropout density | 0.3 | 0.4 | 0.5 |
---|
Accuracy /% | 98.41 | 98.37 | 98.18 |
|
Table 2. Relationship between model network dropout density and accuracy
FC layer | Convolutional layer |
---|
1 | 2 | 3 | 4 | 5 |
---|
2/% | 84.37 | 84.92 | 90.81 | 96.14 | 97.92 | 3/% | 86.41 | 88.59 | 93.01 | 98.42 | 98.51 | 4/% | 85.20 | 88.38 | 92.97 | 98.34 | 98.27 |
|
Table 3. Relationship between correction rate and convolutional layers and FC layers
Name | Parameter |
---|
Diameter /mm | 1000 | Thickness /mm | 80 | Radius of curvature /mm | 4000 | Material | K9 glass | Mass /kg | 174.445 |
|
Table 4. Main parameters of spherical mirror
Name | Parameter |
---|
CPU | Intel Core i7-4790 CPU @3.6 GHz | Memory | 16 GB | Graphics card | NVIDIA GeForce GTX980 Ti | System | Windows 7 professional | Environment | Python3.7, PyTorch 1.5.1-GPU |
|
Table 5. Hardware and software parameters of algorithm training platform
Method | Initialstate | Result | Numberof times | Singlepromotionratio /% |
---|
DLS | 0.27λ | 0.02λ | 2 | 46.26 | LS | 0.27λ | 0.04λ | 2 | 42.60 | DLCM | 0.27λ | 0.01λ | 1 | 96.30 | DLS | 0.56λ | 0.03λ | 3 | 27.98 | LS | 0.56λ | 0.02λ | 4 | 24.11 | DLCM | 0.56λ | 0.02λ | 1 | 96.42 | DLS | 0.86λ | 0.02λ | 5 | 19.54 | LS | 0.86λ | 0.05λ | 6 | 15.70 | DLCM | 0.86λ | 0.03λ | 1 | 96.51 | DLS | 1.21λ | 0.05λ | 5 | 19.17 | LS | 1.21λ | 0.04λ | 7 | 13.81 | DLCM | 1.21λ | 0.02λ | 2 | 49.17 |
|
Table 6. Comparison of calibration results of three algorithms