It is widely known that unexpected foreign objects, such as glue residues and plastic fragments, may be left inside electronic products during production and assembly. These foreign objects may cause electronic products to malfunction. This paper employs a vibration test bench to investigate electronic products and collect vibration signals of foreign objects inside them. These signals are typically weak fault vibration signals, often obscured by strong background noise. To recover the periodic pulses from this noise, we apply the maximum second-order cyclostationarity blind deconvolution (CYCBD) method. However, the selection of the key parameters of CYCBD, cyclic frequency and filter length, is very challenging. To address this issue, we propose an adaptive CYCBD (ACYCBD) method. The proposed method utilizes an improved envelope spectrum via candidate fault frequency optimization-gram (IESCFFOgram) to efficiently identify information frequency bands.The cyclic frequency is estimated from the enhanced harmonic product spectrum (EHPS) of these bands. The proposed power dispersion entropy (PSE-DE) is used to determine the optimal filter length, enhancing sensitivity to signal vibrations. The feasibility and superiority of the proposed method are validated through simulated signals and actual data analysis on a vibration test bench, and experimental results indicate its robustness and resistance to strong background noise.