Cascaded and parallel stochastic resonance for weak signal detection and its simulation study

Considering that in engineering applications early faults are fairly weak and useful fault feature information is always submerged in strong noise, one of problems is how to effectively detect the weak impact signal(s) from a noisy signal for accurate fault diagnosis. By utilizing noise instead of suppressing noise, stochastic resonance (SR) has been successfully applied to weak signal detection in some fields. Aiming at solving one of open problems in the normal SR, i.e. the effective detection of a multi-frequency weak signal, this paper proposes a modified mechanism, in which the cascaded SR and the parallel SR systems are combined together to detect possible weak signal components in a complicated signal. The signal to be analyzed is firstly divided into multiple data segments and then concurrently processed by the parallel SR system to detect its weak signal components. Considering that the output of single enhancement may be not enough for the feature signal detection, this output signal is further enhanced by the cascaded SR system until the weak signals can be clearly identified from its spectrum analysis. The simulation result and comparison demonstrate that the proposed method is effective for detecting multiple weak signals. Its modification and expansion for detecting the weak impact signals would be considered in the future work.