The Resolution Enhancement in the Distributed Temperature Sensor with the Extremal Filtration Method

High resolution DTS research has sufficiently advanced. However further progress will be achieved due to mathematical techniques and algorithms. Novel mathematical algorithms can expand measurement traceability. Outcomes from discrete wavelet transformations are not sufficient for many practical applications. We propose the extremal filtration method as an analog of the empirical mode decomposition (EMD) approach. Advantage of the extremal filtration is the simplification of mathematical calculations. The essence of the method is finding the moving average value of extremums with the continued removing of the highest frequency components from the signal, smoothing of the curve, and subsequent transformations. Difference between the modeling reference signal and signal with Gaussian noise as error of the method is very small: most of the samples (90 %) are within an interval of ±0.003. It is an excellent result for the low signal-to-noise ratio (SNR). However, significant short-term splashes of error (which can reach the value of 0.038) occurred during the transition process.The method can filter low-SNR DTS signals. We expect that the extremal filtration of one target signal is much more effective than a simple averaging of many target OTDR pulses often used in practice. Also, this method can be customized for specific problems connected with subtracting the high-frequency components from the signal. This expands the field of use for the method.