Fig 1.
Flowchart of the overall framework.
Table 1.
Summary of representative prior works in EMC/EMI analysis.
Fig 2.
Typical interference scenario.
Fig 3.
Pulse modulation generator design.
Fig 4.
Heterogeneous pulse neural network.
Table 2.
Comparison of pulse feature extraction performance under different interference types.
Table 3.
Comparison of resource consumption with traditional FFT methods.
Fig 5.
Resource-efficiency comparison of pulse-sparse convolution versus traditional FFT-based EMC analysis across industrial platforms.
(a) Tractioner performance. (b) Industrial robot performance. (c) Drone related performance.
Table 4.
Extraction performance of near-field radiation characteristics for different cable configurations.
Table 5.
Comparison of resource consumption with traditional methods.
Table 6.
Implementation steps of pulse sparse convolution on FPGA.
Table 7.
Parameter table for maximum overlap length of single-ended signal (VVs = 5V, Z0Z=50 Ω).
Table 8.
Parameter table for maximum overlap length of differential signals (Vs = 10V, Z0 = 100 Ω).
Table 9.
Parameters of industrial robot wire harness platform.
Table 10.
Pantograph-catenary interference parameters.
Table 11.
GAN-generated data quality indicators.
Table 12.
Pulse feature extraction performance.
Table 13.
EMC optimization effect.
Table 14.
Comparison of crosstalk prediction accuracy.
Fig 6.
Engineering-benefit summary of the crosstalk-constrained routing scheme versus empirical, GA and RL baselines.
(a) Comparison of computing resources. (b) Maintenance costs.
Table 15.
Comprehensive engineering indicators.
Fig 7.
Ablation-study evidence of component contributions to crosstalk-prediction accuracy.
(a) Single end error. (b) Differential prediction error.
Table 16.
Ablation study of multi-scale discriminator and gradient penalty.
Fig 8.
Waveform comparison: (a) Measured waveform; (b) Measured waveform; Traditional method waveform; (c) The optimized waveform.
Table 17.
Comparison of key technical indicators.