Improved leak detection quality by automatic signal filtering

Abstract

Constant improvement of leak detection techniques in pipe systems used for liquids transportation is a priority for companies and authorities around the world. If a pipe presents leakage problems, the liquid which is lost generates specific signals which are transmitted in the material of the pipe. These signals can be recorded using accelerometers. They are analyzed with the purpose of identifying the location of the leak. An important data analysis tool which helps in the process of leak position detection is the Cross Correlation Function (CCF). It is calculated between two simultaneously recorded specific leak signals with the purpose of time delay estimation. Time delay estimation calculations lead to the identification of the leak position. However, a traditional implementation of the CCF may not be satisfactory. Recorded signals are affected by noise coming from pipe elbows, junctions or traffic. All these unwanted noise sources can influence the aspect and accuracy of the CCF. Further improvements are necessary. This paper presents a Labview 8.5 implementation of an algorithm which improves the quality of the calculated CCF. Being part of a more complex software application, the algorithm uses the calculation of the coherence function. By analyzing the coherence the application will automatically select two frequency intervals (narrow bands) for filtering the leak signals. After the filtering process, the calculation of the CCF shows an increased quality. The application can be accessed from distance if a remote user needs to study or save the CCF results. No filtering settings are required and the method assures good results especially in the case of recorded signals which are highly affected by unwanted noise.