A study of classical and parametric spectral techniques for tube leak detection using a visual programming language

Detecting incipient tube leaks in fossil-fueled boilers consistently and reliably has proven to be difficult. Traditionally, boiler operators have relied on indicators such as fluctuation in pressure or obvious abnormal noise during furnace operation. These methods are ineffective in that they often do not provide early enough detection to avoid a forced outage. An acoustic leak detection system was developed in Europe in the 1970s and introduced to U.S. utilities in the 1980s. Most of these systems are hardware-intensive and employ primarily time-domain signal processing in the detection method. This research represents the approach, analysis, and conclusions drawn by applying classical and parametric spectral techniques to the leak detection problem. It is shown that a software-intensive system utilizing spectral techniques holds great promise for earlier, more reliable leak detection over the current hardware-intensive system. Contrasts and comparisons between the classical and parametric techniques are examined. The techniques are programmed with a visual programming language, which has proven to be a powerful and flexible tool that has greatly reduced development time for program changes over traditional text-based programming environments.