Parameter Estimation Of Magnetic Resonance Response Signals

Nowadays,the population is increasing rapidly,industry and agriculture are developing increasingly as well,all of which lead to the increase of human water consumption and about 1 billion people around the world are living in areas lacking fresh water resources.Groundwater stored in crustal pores for a long period of time accounts for 97% of the global fresh water resources,which can be used as a quality water source for domestic use.Therefore,groundwater detection is particularly important for human society.Because of its high efficiency,quantitative and nondestructive characteristics,NMR detection technology has been widely used in groundwater resource exploration,tunnel and high-speed railway hydrological information assessment,underground hydrological information investigation and early warning of hydrological disasters such as landslides.In the practical application,the response signal of magnetic resonance is extremely weak(only for nanofilts)and the acquisition environment has strong mixed noise,including peak noise,power frequency harmonic noise and random environment noise.Low signal-to-noise ratio leads to large error of characteristic parameters extracted from the magnetic resonance response signals,which affects subsequent inversion interpretation and thus makes the obtained subsurface hydrological information inaccurate.Therefore,studying how to estimate the parameters of magnetic resonance response signals under the background of mixed noise has great practical value and is of great engineering significance.To deal with it,high-precision magnetic resonance signal parameter extraction method based on maximum likelihood function is proposed in this paper.The problem of parameter extraction is transformed into that of finding the extreme of the multivariate function by maximum likelihood function,which can omit the traditional denoising process before extracting the corresponding signal parameters of magnetic resonance,and provides a new idea for the parameter extraction method of NMR detection technology.To solve the problem of long operation time for the extreme search of multivariate function,hybrid particle swarm optimization(PSO)is used to optimize the maximum likelihood function and reduce the running time of the algorithm.In order to solve the problem that the maximum likelihood function method is sensitive to power frequency harmonic noise and the algorithm takes a long time,this paper proposes a parameter extraction method of magnetic resonance response signal based on cyclic correlation,and deduces the cyclic correlation function of magnetic resonance response signal containing noise.According to this method,a specific cycle frequency is selected to filter out the interference of background noise,the rotation invariant technique is used to estimate the parameters of the cycle correlation function,and the appropriate matrix dimension is selected to further reduce the operation time.Simulation is used to verify the feasibility of the algorithm.Firstly,by comparing the parameter estimation accuracy of the maximum likelihood function method and the cyclic correlation method under different noise backgrounds,the sensitivity of the two methods to different environmental noises is obtained.Secondly,the effects of different matrix dimensions in ESPRIT algorithm on the parameter estimation accuracy and operation time of the cyclostationarity method are analyzed.Finally,according to the actual test situation,the influence of the Larmor frequency deviation of the geomagnetic instrument on the algorithm is verified and analyzed.