Abstract：
Water hammer pressure diagnostics is a novel technique for fracture diagnosis developed in recent years. This
advanced technique, with advantages of real-time monitoring and low cost, has potential for future applications. By analyzing the
spectrogram of the pressure signal generated by pump shut-down in the fracturing process, this technique determines the response
time corresponding to the pressure wave propagating from the fracture to the wellhead. Computing the product of the tube wave
velocity and reflection time, estimates the depth of the fracture. In the process of the fracturing pump shutdown, because of the
fluid compressibility, the water hammer pressure wave signal can be obtained at the wellhead. The pressure wave signal contains
a large amount of random noise and fixed frequency noise. This noise is generated by the pipeline vibration, hydraulic fractures
opening and other bottom wave events. The existence of a lot of noise affects the accuracy of spectrum analysis, which leads
to the challenge of determining the response time. The key problem of water hammer pressure wave diagnosis technology is to
remove all kinds of random noise and fixed frequency noise. It is necessary to improve the signal-to-noise ratio (SNR) through
reasonable filtering. In this paper, we analyzed the characteristics of the water hammer pressure wave signal and the effect of fil
tering. Firstly, noise is added to the purified water pressure wave signal, including random noise, fixed-frequency noise consisting
of frequency close to the purified water pressure wave signal and frequency greatly different from the purified water pressure
wave signal. The characteristics of the mixed signal after adding various noise is analyzed. After that, the superposition average
filtering method, the finite impulse response (FIR) low-pass filtering method, adaptive filtering method and adaptive noise
cancellation method are used to filter out the noise. The mechanism of these filtering methods and the effect of these filtering
methods on SNR are studied. The SNR gain and the cepstrum effect are selected to evaluate the effectiveness of these filtering
methods. In this paper, through theoretical analysis and simulation verification, the superposition average method can filter the
random noise. Besides, with an increase of stacking times, the degree of increase in SNR will gradually slow down. The adaptive
noise cancellation method is the best method for fixed frequency noise filtering in this study. The adaptive noise cancellation
method has the largest SNR gain and the best cepstrum recognition effect. This study can provide some technical guidance for
signal filtering in a fracturing operation.
