Customized Acoustical-Analysis Programs continued...
The Task: A Lengthy Process
The recommended data analysis procedure for AE testing involves the evaluation of linear source location clusters. Significant clusters of located events are evaluated against criteria such as whether the activity correlates with pressure or temperature excursions when higher stresses would be present in the piping system, and whether the features of the clusters fall within the characteristic range of typical emission from previously identified problem areas.
The process of performing a cluster analysis begins with the collection of on-line AE data using a computerized data acquisition system. The sensing part of the system consists of specialized AE sensors mounted on waveguides that are welded to the piping. The sensors convert the sound waves to voltage signals that are amplified and transmitted by cables to the data acquisition system. The data is filtered by the instrument to ensure that most of the noise encountered in the steam line is ignored. The raw data is then converted from binary format to ASCII for analysis on a PC, where it can be handled through the use of a spreadsheet program such as Excel. When AE Consulting was using this approach, importing the raw data resulted in an enormous spreadsheet with 21 columns of features and other information about each acoustic event. Events were presented on the rows of the spreadsheet, which could be 50,000 or more rows deep.
Figure 2- Event Density contour plot, Peak Amplitude vs Location. Shows entire pipe section under test.
The first step in making the data understandable was plotting events versus their linear locations along the pipe. Next, engineers scaled temperature and pressure data, obtained from the plant monitoring system, to the engineering units they were using. Then they created a histogram of signal amplitude versus location and looked for areas of significant activity on which they would perform the cluster analysis. When they found these areas, they had to go back into the raw data and extract the appropriate cells to prepare the cluster plot. It was taking AE Consulting's engineers several hours of working with the data before they could generate a cluster analysis.