This paper proposes a one-dimensional convolutional neural network (1DCNN) optimized via particle swarm optimization (PSO) for accurately monitoring the degenerate status of bonding wires in Insulated Gate Bipolar Transistor (IGBT) modules. IGBTs are critical components in power electronic systems, and bonding wire degradation is a common failure mode affecting device reliability. Through dual-pulse tests, gate-emitter voltage (Vge) and collector-emitter voltage (Vce) waveforms were collected from both upper and lower IGBTs under varying degradation conditions: intact wires, 4 broken, 6 broken, and 8 broken wires. The 1DCNN automatically extracts discriminative degradation-related features directly from the raw voltage signals without manual feature engineering. The PSO algorithm efficiently optimizes critical hyperparameters of the network, enhancing its learning capacity and generalization performance. Experimental results demonstrate that the proposed method achieves a high classification accuracy of 95% across the four degradation states, significantly outperforming conventional approaches. This validates the framework’s effectiveness and robustness in health monitoring of power devices, offering a promising tool for predictive maintenance in industrial applications.