New Method of Generating Approximation Profile of Highly Noisy Wireline Logs through Utilizing Wavelet Non-Parametric Regression
Mohammad Bagher Shahvar, Riyaz Kharrat, N. Dashtbesh Badounak · Nigeria Annual International Conference and Exhibition · 2011
Abstract Although highly advanced technologies and instruments provide different types of wireline logs data in an oil field today, geologists and petrophysists are faced with the problem of interpreting highly noisy large datasets. In order not to waste the high expenses spent for applying new technologies, effective de-noising techniques are required to lead to the exact interpretation of the data. In the following study that utilizes wireline logs datasets containing more than 30000 data points of an oil field in Iran, new approach of approximating general trend of the data concealed in a heavy noise against depth is applied. In this new method at first the whole profile of the data and their associated depths are transformed as a two vector matrix to a time-scale space of discrete wavelet domain. By reducing the general regression problem to a fixed-design model, decomposing the signal, reconstructing the function that exists between the two vectors and finally re-scaling the resulting function, wavelet provides the general trend of the profile which is thoroughly free of noise and have all the critical information remained within. This advanced technique of reconstructing the wireline log profile is useful where numbers of the data points are large or the signal is heavily overlaid by noises. In addition to logs data, other signals such as seismic traces can be analyzed using this approach too.