Research Of Stacking Velocity Picking Algorithm Based On Bayesian Estimation And HNN

The selection of the best superimposed velocity value is to use the seismic wave data to calculate the parameters in the velocity spectrum model,which is one of the inverse problems of data analysis.In the analysis of actual seismic wave data,in the face of hundreds of millions of data to be processed,the automatic pick-up algorithm for the optimal superimposed velocity becomes the key to the calculation of velocity spectrum parameters.Most of the existing methods are manual processing,but manual picking may cause the human eyes to not fully and accurately identify the channel peak arrangement and the wider and longer energy band of the velocity spectrum,and the accuracy of the picking and stacking velocity is not high,which directly affects the calculation of subsequent corresponding parameters and reduces the reliability of seismic wave data interpretation.And relying solely on experienced personnel to manually pick up the overlay speed will consume a lot of time and energy,and the seismic data processing efficiency is not high.Therefore,this paper introduces Bayesian estimation algorithm and Hopfield network to conduct an in-depth discussion on its mathematical principles and practical application,and realizes automatic picking of superimposed speeds under different models through experiments,which improves picking accuracy and efficiency.First of all,this paper deeply researches the basic principles of Bayesian estimation algorithm(BSPV)based on Markov process,and conducts mathematical derivation and research on the algorithm from the perspective of automatic extraction of superimposition speed,so that it can realize the joint estimation of candidate solutions by combining observation sequence and transition probability sequence.(State sequence),and backtrack to the optimal solution(channel sequence)through the maximum posterior probability.A two-dimensional simulation matrix is used to simulate the actual channel data,and the Bayesian estimation algorithm is used to simulate it.It is verified that the Bayesian estimation algorithm deduced and designed is suitable for the automatic pickup of stacking velocity,which is feasible and reliable for the question.Secondly,aiming at the "jumping" phenomenon of sequence values that may occur in the BSPV algorithm,a Hopfield neural network model that can solve this problem is established.By analyzing the distribution law and characteristics of channel peaks,three constraints that limit the solution space are given to optimize the initial sequence,so that the final result sequence eliminates part of the noise interference,which improves the accuracy and reliability of the automatic pick-up result of the BSPV method superposition speed.Finally,through the application of tunnel advance forecasting,the verification and analysis are carried out.The algorithm presented on the theory and the measured data to calculate the stacking velocity in the range of allowable error,more nearly accurate than manual human recognition speed,to calculate the interface location accurate,to explain the collected seismic data to make the best proof.