| With the increasing difficulty of oilfield exploitation and the ascendant of digital communication technology,logging technology is moving towards the big data era.In order to accurately obtain large and complex information of downhole,it is indispensable to develop a high-speed and low-error logging data transmission system.However,the harsh underground environment with high temperature,high pressure and strong noise have brought great challenges to the narrow bandwidth and long-distance single-core cable data transmission system.Therefore,it is of great significance to develop a high-speed and low-error logging data transmission system.Focusing on the logging data transmission problems of downhole multi-sensor,this thesis has carried out researches on the performance optimization of the logging data transmission system based on dual-core digital signal processor(DSP).First,the overall architecture of the orthogonal frequency division multiplexing(OFDM)technology in data transmission system is introduced.Then,the transmission characteristics of the single-core cable are measured,and the synchronization problem and the peak-toaverage power ratio(PAPR)problem of the cable data transmission system are studied.Symbol synchronization and sampling synchronization are realized on a dual-core DSP cable data transmission system,and a PAPR suppression algorithm based on modeldriven deep learning is proposed.Finally,the data transmission speed and error rate of the multi-electrode conductance sensor are verified on the gas-liquid two-phase flow dynamic experimental device.The innovative research results of this thesis are as follows:1.Aiming at the synchronization problem of the OFDM system,non-data-assisted and data-assisted synchronization algorithms are simulated and implemented,and the synchronization is implemented on the receiving end of the dual-core DSP logging data transmission system.A synchronization algorithm based on the training sequence is used to accurately locate the starting position of the OFDM symbol and estimate the frequency offset.In the frequency domain,the pilot is used for phase compensation to achieve sampling synchronization.2.Combining the Tone Reservation(TR)model with a deep learning network,a PAPR suppression algorithm based on model-driven deep learning is proposed.The improved TR model is embedded into the deep learning network layer structure,and the clipping threshold and the weighted time-domain kernel function are optimized through training network parameters.This effectively overcomes the shortcomings of high network complexity,high training cost,and lack of interpretability,and has achieved a good PAPR suppression effect.3.The logging data transmission experimental platform is built to obtain the multielectrode conductance sensing test data of gas-liquid two-phase flow under different flow conditions,and verify the transmission speed and data error rate of multi-electrode conductance sensor data in single-core cable.The experiment shows that the logging data transmission rate is 36.36 Kbps and the transmission data error rate is 1.22 × 10,and good data transmission performance is obtained. |
