Serving as a critical element within rotating machinery, the health status of rolling bears directly affects the efficiency, stability and lifespan of mechanical equipment. In practical industrial applications, accurate bearing fault diagnosis is of vital importance to improve production efficiency and ensure the safe operation of equipment. Using vibration signals to analyze bearings is an effective method for bearing diagnostic analysis, with envelope demodulation techniques being particularly effective in extracting fault-related features. However, complex processing environments and severe noise interference are prevalent issues in actual industrial scenarios. The vibration information induced by faults often exhibits multi-frequency band distribution characteristics. This makes traditional envelope analysis methods, which rely on selecting a single frequency band, often ineffective. Therefore, an adaptive weighted envelope spectrum based on harmonic energy product (HEPAWES) is proposed to overcome the aforementioned limitations and effectively extract fault information from bearings. Firstly, the bi-spectral map is acquired. Then, a fault information evaluation indicator is constructed to numerically quantify the fault information content in each spectral frequency slice, while also performing adaptive threshold determination. Finally, based on the determined threshold, weights are assigned to all spectral frequency slices to form the weighted envelope spectrum. This method can effectively enhance fault features and suppress interference components. Simulation analysis and experimental data analysis verified the significant advantages of this method, demonstrating its capacity to effectively identify bearing fault characteristics even despite conditions of strong interference.

rolling bearing fault diagnosis, cyclic spectral coherence, multi-band distribution, harmonic energy product, adaptive weighted envelope spectrum