The Gas Leak Detection Based on a Wireless Monitoring System

Abstract

Industrial gas leaks cause accidents and pose threats to the environment and human life. Thus, it is essential to detect gas leaks in time. Usually, the abnormal concentration signals are defined by a fixed concentration value, such as 25% of the lower explosive limit. However, it is difficult to accumulate to the fixed point quickly when the leak is small. In addition, the actual leak signals are seldom available, making many data classifications inoperable. To solve these problems, this paper proposes a detection approach using the auto-correlation function (ACF) of the normal concentration segment. The feature of each normal segment is obtained by calculating the correlation coefficients between ACFs. According to the features of statistical analysis, a nonconcentration threshold is determined to detect the real-time signals. In addition, the weighted fusion algorithm based on the distance between the sensors and virtual leak source is used to fuse multisensory data. The proposed method has been implemented in a field by building a wireless sensor network. It is confirmed that the system detection rate reaches as high as 96.7% and the average detection time delay is less than 30 s on the premise of low false alarm rate.