Unexpected failures in rotating machinery can cause costly downtime and safety hazards in industrial systems, highlighting the need for accurate and robust fault diagnosis. However, fault-related signals are often weak and easily obscured by noise, making reliable diagnosis challenging in real-world environments. To address this, we propose Time-frequency and time-series dual-branch fusion network(TFDFNet), a novel dual-branch deep learning model designed to improve fault classification performance under noisy and complex conditions. The model combines two complementary types of information: time-frequency representations derived from continuous wavelet transform and raw time-sequence data captured through sliding-window sampling. A Swin Transformer is used to extract deep features from time-frequency images, while a specially designed module called Gated attention block(GABlock) learns key temporal patterns from the sequence data. These features are fused using a cross-attention mechanism to enhance fault-related information. Extensive experiments on two public bearing fault datasets (CWRU and Ottawa) show that TFDFNet achieves outstanding accuracy, even under severe noise interference. The model reaches up to 100% accuracy on CWRU and 99.44% on Ottawa, and consistently outperforms existing convolutional neural network (CNN) baselines. These results demonstrate the practical potential and robustness of TFDFNet for intelligent fault diagnosis in industrial applications.
Object detection benefits greatly from multimodal image fusion, which integrates complementary data from different modalities like RGB and thermal images. However, existing methods struggle with effective inter-modal fusion, particularly in capturing spatial and contextual information across diverse regions and scales. To address these limitations, we propose the dynamic channel adjustment and multi-scale activated attention mechanism network (MNCM). Our approach incorporates dynamic channel adjustment for precise feature fusion across modalities and a multi-scale attention mechanism to capture both local and global contexts. This design improves robustness while balancing computational efficiency. The model’s scalability is enhanced through its ability to adaptively process multi-scale information without being constrained by fixed-scale designs. To validate our method, we used two multimodal datasets from traffic and industrial scenarios, which consisted of paired thermal infrared and visible light images. The results first demonstrate strong performance in multimodal fusion and then show state-of-the-art results in object detection, proving its effectiveness for real-world applications.