Objective <p>Multi-source signal recognition is a prevalent challenge in engineering vibration control. Traditional methods heavily depend on prior knowledge and professional expertise, which constrain both efficiency and accuracy. This study aims to propose a high-performance vibration recognition model to address this limitation.</p> Methods <p>A vibration recognition model integrating signal processing and deep learning was developed. First, Continuous Wavelet Transform (CWT) was applied to convert one-dimensional original vibration signals into two-dimensional time-frequency representations, which contain richer feature information. Subsequently, the time-frequency images were input into a Convolutional Neural Network (CNN) for automatic feature extraction. Finally, the Softmax layer was used to complete the vibration recognition task.</p> Results <p>Twenty sets of measured vibration data were used to evaluate the model’s performance. Experimental results demonstrate that the proposed model achieves a recognition accuracy of 99%. It exhibits excellent performance in both component recognition and vibration signal separation.</p> Conclusion <p>This study provides an effective solution for multi-source vibration signal recognition. The proposed model is of great significance for engineering vibration diagnosis, the front-end design of vibration control systems, and the analysis and optimization of control effectiveness.</p>