The industrial electronics surface mount production line can develop a range of defects over time, primarily due to the aging of equipment. Traditional deep learning approaches for defect classification would require retraining the neural network with each new defect type, which is inefficient and resource-intensive. To overcome this challenge, we introduce an incremental learning model for surface mount defects classification with distillation learning (DIL-SDC). Our model’s backbone network is a fusion of the Adaptive Aggregation Network and GELANResNext, which is augmented with a composite loss function to optimize the incremental learning process. Furthermore, we implement distillation learning techniques to mitigate the catastrophic forgetting that often accompanies incremental learning. The empirical results on the real-world defect dataset substantiate that our strategy markedly outperforms existing incremental learning techniques in diminishing the effect of catastrophic forgetting.