长管道事故停泵水锤现场测试与信号分析

Field measurement of water hammer and signal analysis at pump-stop due to power failure in long water-supply pipeline

结合某泵站事故停泵现场测试数据,采用最大熵估计法分析压力、流量、转速等测试信号的时频特点,得出实测数据的主要污染是>50 Hz的高频噪声.利用小波包理论,采用自适应阈值方法获得无污染的去噪信号,并与快速傅里叶变换(FFT)去噪结果比较,同时进一步分析去噪后的实测信号,验证水锤仿真计算准确度以及单向调压塔防护效果.结果表明:两种去噪结果虽然波形相似,但FFT去噪同时明显削弱了峰值点的有效信号,而利用小波包理论,能够有效去除噪声污染,使去噪后的信号与原始信号保持相似性;空气阀-调压塔水锤防护措施效果明显;采用水柱分离模型,能够较为准确预测管线中出现的水锤压力峰值、最大倒流流量及最大倒转转速,为系统安全评估提供依据,但对第一次压力峰值以后的多次气穴溃灭与水柱弥合过程的模拟,存在明显误差,有待于建立更准确的数学模型.

The method of maximum entropy estimation was applied to analyze the transient pressure, rotational speed and discharge signals which were measured and recorded in a pump station for a pump at pump stop caused from power failure and to identify their time-frequency features. Before that, the wavelet packet theory was used to denoise the measured signals. It was found that the main pollutant in the signals was the noises with a higher than 50 Hz frequency. Then, both the wavelet pocket method and the fast Fourier transform （FFT） were used to remove the pollution. At last, the simulated water hammer signals were compared with the measured ones denoised to confirm effect of the surge tank combined with air valves on water hammer prevention. The results show that even waveforms of the two denoised signals respectively by the wavelet pocket method and FFT are similar. The latter method considerably reduces the useful information in pressure peaks, while the wavelet packet theory can e- liminate noises effectively and allow the denoised signal to have a similarity to the original one. A surge tank with air-valves takes a substantial effect on preventing from water hammer. The column separation model can correctly predict the maximum values of pressure, backflow discharge in a pipeline and the reverse rotational speed of pump. Those data provide evidence for pumping system safety assessment. Unfortunate, the existing models fails to accurately simulate the attenuation process of water hammer wave, thus more precise models are needed.