MemoryMamba is a novel memory-augmented state space model (SSM) designed to enhance defect recognition in manufacturing settings. It integrates state space techniques with memory augmentation to capture dependencies and intricate defect characteristics. The model incorporates coarse- and fine-grained memory networks to retain and access critical defect information, and a fusion module to integrate visual features and memory vectors. Optimization strategies based on contrastive learning and mutual information maximization are proposed for these memory networks. Experimental results across four industrial datasets (Aluminum, GC10, MT, and NEU) demonstrate that MemoryMamba outperforms existing methods, achieving high accuracy, precision, recall, and F1 scores, especially in challenging scenarios with limited or imbalanced defect data. The model's effectiveness is further validated through ablation studies and analysis of the impact of different components and parameters.MemoryMamba is a novel memory-augmented state space model (SSM) designed to enhance defect recognition in manufacturing settings. It integrates state space techniques with memory augmentation to capture dependencies and intricate defect characteristics. The model incorporates coarse- and fine-grained memory networks to retain and access critical defect information, and a fusion module to integrate visual features and memory vectors. Optimization strategies based on contrastive learning and mutual information maximization are proposed for these memory networks. Experimental results across four industrial datasets (Aluminum, GC10, MT, and NEU) demonstrate that MemoryMamba outperforms existing methods, achieving high accuracy, precision, recall, and F1 scores, especially in challenging scenarios with limited or imbalanced defect data. The model's effectiveness is further validated through ablation studies and analysis of the impact of different components and parameters.