Research on leakage detection of steam pipeline in closed space of coal-fired boiler
In the current process of steam pipeline detection,relying on single-scale convolutional neural networks to detect pipeline leaks is prone to losing some feature information in the pooling layer,resulting in low accuracy of detection results.Therefore,a leakage detection method for steam pipelines in closed space of coal-fired boilers was proposed.Install sound wave sensors and sound signal e-mission devices into small space intelligent operating UAV,and control the UAV to conduct autonomous inspections in the closed space of coal-fired boilers,collecting steam pipeline leakage signals.Using an improved homogeneous narrow wave local feature scale decom-position algorithm to process the collected sound wave signal and remove noise information from the signal.Perform three-level wavelet decomposition in the denoised time-domain signal,and extract wavelet statistical features such as standard deviation,kurtosis,and skew-ness based on wavelet coefficients.Finally,a multi-scale convolutional neural network was applied to construct a steam pipeline leakage detection model,and the wavelet statistical features were inputted into it to obtain the final leakage detection results.The experimental results showed that the accuracy of the steam pipeline leakage detection results obtained after the application of the studied method was greater than 92％,demonstrating the superior performance of this method.
NETL
NSTL
万方数据
