Small leakage detection technology of environmental protection gas based on deep learning
Leakage detection of environmentally friendly gases has always been a challenging issue in the industry.This technology utilizes deep learning to identify leaks.In order to collect acoustic signals of leaks,an adjustable pressure gas pipeline system was established to control the gas.Convert the collected acoustic signals into the frequency domain for preprocessing.Unlike traditional two-step detection techniques(feature extraction and classification),a CNN based model with multiple convolutional and pooling layers was established.In this model,the convolutional layer and pooling layer aim to learn the representation of the transformed acoustic signal,while the Softmax layer in the final layer is developed to further identify the representation from the previous layer.The experiment has proven the effectiveness of this model in real-world data.Compared with traditional CNN models,this model takes less time and has higher efficiency.
NETL
NSTL
万方数据
