A Muon Detection Method For Heavy Nuclear Materials Based On Prior Knowledge

Muon tomography(MT),based on the multiple Coulomb scattering of cosmic-ray muons when traversing an object,has recently been a novel,promising technique for non-destructive detection of high-Z materials,especially the special nuclear materials(SNM).Compared to other radiographic techniques,MT has many adequate advantages,such as no artificial radiation source,extraordinarily penetrative against shielding,sensitive to high-Z materials and direct 3-D imaging.It has been widely used in applications related to nuclear safety,nuclear waste surveillance and treaty verification for nuclear disarmament.However,it has been fraught with difficulties in image reconstruction,regarding of the limited flux of natural cosmic muons and the complication of their scattering due to the differences in aspect of energy and zenith angle.Furthermore,the present algorithms used for imaging reconstruction are generally designed and developed towards a common object,without further consideration on any prior knowledge,if existed,about its structure and composition.Thus,such algorithms always require a huge amount of input data and consume comparative computing resources.These problems have limited,to some extent,the application range and the performance of MT.Two types of prior knowledge of target objects,the geometric symmetry and the given reference,were focused on in order to solve such problems.Pathways of deducing the target structure from muon scattering by utilizing the prior knowledge were proposed in the respective scenarios.They were proved robust by the verification of several typical target scenarios built by the Geant4 simulation framework.Additionally,a modified multi-group(MMG)model was proposed and examined as an approximation of muon source,to deal with the characteristic distinction of cosmic muons.The main content and results were listed in detail as follows:1)Numerical simulation for the characteristic of cosmic muons and its approximate mathematical model.The Cosmic-RaY muon generator(CRY)was used to obtain the distributions of energy and zenith angle and the flux of cosmic muons at different geographical sites on the Earth surface on different observation dates.Then,an MMG model that approximately describing such characteristics was proposed and ameliorated to accommodate an effective adjustment to the consequent diffusion of muon scattering.According to the tests based on the simulated scenario of a slab model,the MMG model could help to improve the precision of material identification and thickness evaluation of a slab,resulting in a considerable elevation for MT on the detection performance of high-Z materials.2)MT implementation based on the prior knowledge of geometric symmetry.Developed on the fundamental of the conventional MT method of image reconstruction,this method regarded the geometrical symmetry of some specific objects as the prior knowledge and attempted to incorporate it into the algorithms.As a result,the MT performance for scenarios related to detection of such objects was improved,since the data amount required for the same image resolution were slashed by the reduced image dimensionality and the simplified computation.The implementation of this method was introduced in detail by taking example for several typical configurations,including an axially symmetric one,a spherically symmetric one and a cylindrically symmetric one as well.The structural images were successfully obtained for all the configurations.3)MT implementation based on the prior knowledge of given reference.Regarding that the reconstructed images for some extraordinary structures,especially those with hollow-like features,must be disturbed or even distorted by artifacts,a novel method for detecting some local features of the target was proposed.It discriminated the target from its analogs derived from the reference and searched for the most alike one,by directly comparing the muon scattering properties(e.g.the scattered angle and/or the scattered displacement)of both objects.Three typical scenarios focusing on a slab with a slit,nested spherical shells with the interior deviated and a cylindrical container with a high-Z target hidden inside were introduced as examples to demonstrate the implementation of this method.The results showed that such local features,including the slit,deviation,the material and location of the hidden target for the respected scenario,could be clearly identified.