The invention relates to the field of
data processing, in particular to a hydraulic drive
hydrogen compressor fault monitoring method, which comprises the following steps: acquiring a vibration
signal and a pressure
signal of a compressor cylinder, and determining a dynamic window length through autocorrelation analysis of the pressure
signal. In the window, empirical mode
decomposition is carried out on the vibration signals,
instantaneous energy is extracted, kurtosis is analyzed, and correlation coefficients between IMF component pairs are calculated. And taking the product of the kurtosis and the
correlation coefficient as a correlation pulse index, and constructing a
hydrogen embrittlement possibility index in combination with the pressure ratio to evaluate the
hydrogen embrittlement risk. The
hydrogen embrittlement possibility indexes exceeding the
background level are further accumulated, the
hydrogen embrittlement damage accumulation degree is obtained, and whether the compressor has the
hydrogen embrittlement fault or not is judged after normalization. The dynamic window length is determined through self-
correlation analysis,
noise is effectively filtered, the signal-to-
noise ratio is improved, the hydrogen
embrittlement risk is comprehensively evaluated,
false alarm and missing alarm are reduced, meanwhile,
continuous monitoring and real-time hydrogen embrittlement development tracking are achieved, and the safety of equipment operation is improved.