Mechanical fault diagnosis of high‑voltage circuit breakers relies on effective analysis of non‑stationary vibration

current

and travel signals. This paper proposes an IWOA‑VMD‑DBN method that integrates multi‑source time‑series features and employs an improved whale optimization algorithm (IWOA) to jointly optimize variational mode decomposition (VMD) and deep belief network (DBN) parameters. Experiments are conducted on an LW30‑126 breaker test platform under typical mechanical faults. The closing process is divided into six sequential stages

from which permutation entropy

envelope entropy

and energy entropy of vibration signals are extracted and fused with time‑domain current and travel features. IWOA

enhanced with elite opposition‑based learning and nonlinear inertia weight

improves feature separability and model generalization. Results show a diagnostic accuracy of 98.33% with limited samples

outperforming the comparison models

and verify the effectiveness and engineering applicability of the method.