Application of Wavelet Characteristics and GMDH Neural Networks for Precise Estimation of Oil Product Types and Volume Fractions

Given that one of the most critical operations in oil and gas industry is to instantly determine volume type product passing through pipelines, this research, a detection system for monitoring pipelines proposed. The proposed works such way radiation from dual-energy source which symmetrically emits radiation, was received by NaI detector after shield window test pipeline. In pipe, four petroleum products—ethylene glycol, crude oil, gasoil, gasoline—were simulated pairs different fractions. A total 118 simulations were performed, their signals categorized. Then, feature extraction started reduce data, increase accuracy, learning speed neural network, better interpret data. Wavelet features extracted recorded signal used as GMDH network input. each divided into details approximation sections characteristics with names STD A3, D3, D2 extracted. This described structure modelled Monte Carlo N Particle code (MCNP). fact, precise estimation types fractions done using combination symmetrical asymmetrical network. Four networks trained estimate volumetric ratio product, maximum RMSE 0.63. addition high low implementation computational cost compared previous methods are among advantages present investigation, increases its application industry.