Research Article

Safety Prediction Using Vehicle Safety Evaluation Model Passing on Long-Span Bridge with Fully Connected Neural Network

Table 5

Details of network training with different superparameter settings.

No.Hidden layerNetwork structureInitial learning rateBatch sizeActivation functionOptimization methodMax val acc (%)No. EpochsTest acc (%)

124-2-40.05100ReLUSGD + momentum92.643391.304
224-2-40.05250ReLUSGD + momentum93.318292.642
324-2-40.05500ReLUSGD + momentum91.649190.97
424-2-40.05500LeakyReLUSGD + momentum91.307593.98
524-2-40.05500LeakyReLUAdam92.986791.304
624-2-40.01500LeakyReLUAdam92.648291.973
744-4-40.05100ReLUSGD + momentum93.984394.649
884-8-40.05100ReLUSGD + momentum93.311093.98
984-8-40.05250ReLUSGD + momentum93.651493.98
1084-8-40.05500ReLUSGD+momentum93.981394.649
1184-8-40.05500LeakyReLUSGD + momentum93.652194.314
1284-8-40.05500LeakyReLUAdam93.654194.649
1384-8-40.01500LeakyReLUAdam93.317394.983
14124-12-40.05100ReLUSGD + momentum94.65593.645
15164-16-40.05100ReLUSGD + momentum93.65393.98
163/34-3-3-40.05100ReLUSGD + momentum92.982794.314
173/34-3-3-40.05250ReLUSGD+momentum92.982794.649
183/34-3-3-40.05500ReLUSGD + momentum93.316192.977
193/34-3-3-40.05500LeakyReLUSGD + momentum93.659094.314
203/34-3-3-40.05500LeakyReLUAdam93.315094.983
213/34-3-3-40.01500LeakyReLUAdam92.318693.311
227/74-7-7-40.05100ReLUSGD + momentum93.653694.314
2310/104-10-10-4 (Dropout)0.05100ReLUSGD + momentum92.984894.983
2415/154-15-15-4 (Dropout)0.05100ReLUSGD + momentum93.983194.983
2515/154-15-15-4 (Dropout)0.05250ReLUSGD + momentum92.981193.98
2615/154-15-15-4 (Dropout)0.05500ReLUSGD + momentum93.311694.314
2715/154-15-15-4 (Dropout)0.05500LeakyReLUSGD + momentum93.652094.983
2815/154-15-15-4 (Dropout)0.05500LeakyReLUAdam94.981194.983
2915/154-15-15-4 (Dropout)0.01500LeakyReLUAdam91.973593.645