<p>Diagnosing faults in motor bearings is crucial for ensuring the reliability and efficiency of industrial machinery. Many traditional signal decomposition methods face challenges with computational efficiency, hindering their effectiveness in real-time fault detection scenarios. To address the challenge of motor bearing fault diagnosis, this study presents an innovative approach that integrates an enhanced fast signal decomposition technique with a deep learning architecture. The proposed approach first employs a fast iterative filtering (FIF) decomposition algorithm, where a key parameter (e.g., mask length) is adaptively optimized using the osprey-cauchy-sparrow search algorithm (OCSSA) to enhance decomposition accuracy. The OCSSA-FIF technique extracts intrinsic mode functions (IMFs) that effectively separate fault-related features from noise. These features are then input into a hybrid deep learning framework, which integrates convolutional neural networks (CNNs) for extracting local features and bidirectional long short-term memory (BiLSTM) networks for modeling sequences. This combination facilitates reliable fault classification in complex operational scenarios. Experimental results demonstrate that the enhanced FIF-CNN-BiLSTM model, with its hybrid neural network architecture, achieves superior diagnostic accuracy and enhanced real-time performance, as evidenced by similar studies in the field. Thus, the integration of optimized decomposition and deep learning not only enhances diagnostic precision but also provides a scalable framework for real-time industrial applications.</p>