Coupling Photon Monte Carlo Simulation and CAD Software. Application to X-ray Nondestructive Evaluation

Transcription

1 Coupling Photon Monte Carlo Simulation and CAD Software. Application to X-ray Nondestructive Evaluation Joachim Tabary and Alain Glière LETI - CEA Technologies Avancées, F Grenoble, France Abstract A Monte Carlo radiation transport simulation program, EGS Nova, and a Computer Aided Design software, BRL-CAD, have been coupled within the framework of Sindbad, a Nondestructive Evaluation (NDE) simulation system. In its current status, the program is very valuable in a NDE laboratory context, as it helps simulate the images due to the uncollided and scattered photon fluxes in a single NDE software environment, without having to switch to a Monte Carlo code parameters set. Numerical validations show a good agreement with EGS4 computed and published data. As the program's major drawback is the execution time, computational efficiency improvements are foreseen. Introduction A wide range of applications of numerical simulation have been pointed out in the field of X-ray Nondestructive Evaluation (NDE). These applications are as diverse as assistance to X-ray hardware designers, help in validation of inspection procedures, defect detectability assessment, operators education and training, etc. Designers of digital processing techniques, such as tomography, tomosynthesis and automated defect detection can also take advantage of realistic simulated images. The necessity to have fast and practical software tools at disposal lead, during the last decade, to the development of solutions based on Computer Aided Design (CAD) model of the examined part and Graphical User Interface (GUI). In a first stage, analytical models using ray tracing techniques were developed in order to address the straightforward problem of the uncollided flux (i.e. the flux of photons which have not been subjected to any interaction inside the examined part) computation [1-4]. However, a X-ray radiographic image is generated by both uncollided photons and photons scattered inside the examined part. The latter phenomena is a major source of unwanted noise which accounts, in usual NDE conditions, from a few percents to more than half the overall measured intensity. Quantitative evaluation of the scattered flux is therefore of primary interest, for instance to correctly evaluate flaw detectability. In order to address this problem, several approaches have been proposed, ranging from rough approximations, such as build up factors to sophisticated numerical procedures solving integral transport equations [5]. Solutions based on the Monte Carlo method have also been implemented [6] [7] but the lack of flexibility in the description of the examined part and experimental setup and the relatively long execution time have, to date, prevented a widespread use. As desktop computers power increases continually and variance reduction techniques can drastically improve calculation efficiency, we decided to couple, within a NDE simulation system called Sindbad, a Monte Carlo radiation transport simulation code with a CAD and ray tracing package. This provides the simulation of photon scattering inside the examined part with the established accuracy of the Monte Carlo method and the mechanical parts design flexibility of CAD solid modelers. This paper first presents an overview of the models used by the NDE simulation software in order to compute the radiographic image due to the uncollided flux. It focuses in a second part on the use of the CAD model of the examined part and a Monte Carlo simulation to compute the image due to

2 the scattered flux. A validation of the model is then established by comparison with published data. At last, an example of application is presented. Presentation of the pre-existing NDE analytical simulation package The physics of the radiographic inspection process can be divided into three separate parts, namely the photon beam generation in the X-ray source, the beam interaction with the examined part, and the imaging process (detection of the remaining photon flux and transformation into a measured signal). They have been addressed as follows in the Sindbad NDE simulation package. The X-ray tube model semi-empirically simulates the physical phenomena involved in bremsstrahlung and characteristic photons production. It takes into account the anode angle and composition and the inherent and additional filtration. It can be used in the range kV. Experiments, previously performed in CEA-LETI experimental facility, show an agreement usually better than 20% between calculated and measured doses. The uncollided flux analytical simulation relies on the computation of the attenuation of the incident flux, binned in narrow energy channels, by the examined part. This is performed by tracing rays from the source point to every pixel of the detector through the part CAD model, built with BRL-CAD [8], a Constructive Solid Geometry based CAD and ray tracing package, and by calculating the energy dependent attenuation due to the crossed materials. The attenuation coefficients are those described in Storm-Israël data tables [9]. Cross sections for compounds or mixed materials are obtained by linear combination of elemental data. Detectors are modeled in two successive steps. The first one, common to all types of detectors, computes the energy deposition in the sensing part of the detector, using the energy absorption attenuation coefficients. The second stage, specific to each type of detector, simulates the successive physical phenomena involved in the energy to signal transformation. For instance, in the case of a scintillating screen viewed by a CCD camera, it accounts for energy to light photons transformation, light photon absorption in the screen and optical coupling system, and photon to electron conversion in the CCD device. Noise and blurring, defined by the system Modulation Transfer Function, can be added on the final image. User's interaction with non computing intensive modules is performed by the means of a Motif based GUI. Monte Carlo radiation transport in CAD models Sindbad has recently been improved by the developement of a module dedicated to the Monte Carlo simulation of the radiation scattered in the examined part. In order to facilitate the software use in our laboratory context and to limit the developemnt task, it has been decided to keep using the CAD model of the examined part, as well as the usual NDE parameters set and thus, to couple, within our NDE simulation system, a Monte Carlo radiation transport simulation code, with the BRL-CAD CAD and ray tracing package. Because of its availability in the C programming language and the established reliability of the models used in EGS4 [10], EGS Nova [11], a functionally equivalent adaptation of the default EGS4 code, has been chosen as a basis for our application. The CAD coupled software major features are as follows: The incident photon energy is sampled from a cumulated distribution function obtained from modeled or measured X-ray tube spectra as well as from radioisotope source data. A cone beam point source or an evenly distributed circular focal spot can be taken into account to determine the photon initial position and direction. Once the incident photon parameters are defined, the particles interaction and particles stack are managed by the EGS Nova shower simulation. The medium cross section data are generated beforehand by PEGS4, the EGS4 preprocessor.

3 The examined part geometry interrogation function, whose goal is to compute the distance from the current particle position to the next boundary that will be crossed and the next region index, uses the ray tracing functions provided by BRL-CAD. Several CAD formats are accepted. As in EGS Nova, user has easy access to the physical models switches and cutoffs. However, the electron energy cutoff is defaulted to a high value in order to limit unnecessary computation. In the scoring function, every photon that reaches the front side of the detection system deposits, in a single detector pixel, an amount of energy related to the pre-calculated energy dependant detection efficiency of the detector material. History flags are used in order to distinguish between uncollided photons, once scattered photons and several times scattered photons. The pre-existing simulation software and its GUI manage the remaining tasks, namely the X-ray source modeling, the overall parameterization of the radiography setup and the detection system modeling. The knowledge of the analytically computed uncollided flux and the Monte Carlo computed uncollided to scattered ratio can be used to eventually scale the scattered flux image to the actual experimental conditions. Implementation validation and application example A first validation study has been performed on steel slabs of various thickness. The front side of the slab is placed one meter away from a 60 Co gamma rays source. The detection system is directly placed on the plate back side. The comparison with results obtained in the same modeling conditions using the EGS4 code shows a very close agreement (the difference is less than one percent). The computation time ratio, approximately longer by a factor of 1.5 in this slab geometry case, would probably improve for more complex part geometries. In the same configuration, the comparison between analytical and Monte Carlo computations of the uncollided flux exhibits a slight difference, which remains in the few percent range. This discrepancy is probably caused by differences between the Storm--Israël and PEGS4 cross sections databases, respectively used by the analytical and Monte Carlo simulations. A second validation set uses a study by Inanc [12] in which comparisons between a deterministic scattering computation code and MCNP [13] are carried out. The radiography setup is formed by flat aluminum plates irradiated by a 140 kv X-ray tube, placed one meter away. The plate thickness is varied between 1 and 5 cm. The detection system, assumed perfect, is placed on the plate back side. Normalized uncollided and scattered photon fluxes are plotted with respect to plate thickness in Fig. 1.

4 Fig. 1. Comparison with published data. Normalized uncollided (empty triangle: deterministic code, empty circle: MCNP, dashed line: present results) and scattered (solid triangle: deterministic code, solid circle: MCNP, solid line: present results) flux magnitude versus plate thickness The agreement between uncollided flux computations by the three codes is very good. However, our CAD coupled implementation underestimates the scattered flux for thin plates. The discrepancy, which remains acceptable for our applications, is probably caused by a combination of differences in X-ray tube spectra, which are nor exactly similar, and differences in the materials cross sections and models used by EGS Nova and MCNP. A realistic geometry example has been simulated in order to demonstrate the capabilities of the CAD coupled code. The test sample is a small mechanical part made of aluminum (cf. Fig. 2 left) placed 70 cm away from a 110 kv X-ray tube. The electron cutoff energy is set in order to discard any electron transport inside the sample photons histories were computed (at a 16,000 histories/s rate on a SUN Ultra workstation). Computed radiographs (top views) relative to the uncollided and scattered photon fluxes are presented in Fig. 2. In agreement with experimental data, the computed amount of scattered radiation represents a small percentage of the uncollided flux. Fig. 2. Application example. Test part geometry (left), simulated uncollided flux image (center) and scattered flux image (right) Conclusion A Monte Carlo radiation transport simulation program, EGS Nova, and a CAD software, BRL- CAD, have been coupled within the framework of Sindbad, a NDE simulation system. A first level set of validation has been performed. In its current status, the program is very valuable in a NDE laboratory context, as it helps compute the magnitude and spatial distribution of the scattered flux in a NDE software environment, without having to switch to a Monte Carlo code parameters set. The major drawback is the execution time, usually more than one hour on a SUN UltraSparc workstation, which remains at present too long for everyday production use. The foreseen computational efficiency improvements are as follows. Firstly, as scattering contributes only to the low frequency details of the radiographic image, the pixel size used for the Monte Carlo computation will be increased. An interpolation will then be performed on the low resolution Monte Carlo computed scattered flux image before addition to the high resolution analytically computed uncollided flux image. Secondly, we will take advantage of the computation of the uncollided flux with an analytical method to force interaction of all photons within the examined part and thus reduce variance. These improvements will be carried out in a near future.

5 References [1] F. Inanc, J.N. Gray: `A CAD interfaced simulation tool for X-ray NDE studies', Rev. Prog. Quant. Nondestructive Evaluation, 9A (Plenum Press, New York 1990) [2] C. Bellon, G.R. Tillack, C. Nockemann, L. Stenzel: `Computer simulation of X-ray NDE process coupled with CAD interface', Rev. Prog. Quant. Nondestructive Evaluation, 16A (Plenum Press, New York 1996) [3] P. Duvauchelle, N. Freud, V. Kaftandjian, G. Peix, D. Babot, `Development of a simulation tool for X-ray imaging techniques', In: Workshop on the application of X-ray tomography in materials science, Villeurbanne, Oct , 1999, ed. by J. Baruchel et al. (Hermes, Paris 2000) [4] A. Glière: `Sindbad. From CAD model to synthetic radiographs', Rev. Prog. Quant. Nondestructive Evaluation, 17A, (Plenum Press, New York 1998) pp [5] F. Inanc, J.N. Gray: `Scattering simulation in radiography', Applied Radiation and Isotopes, 48, (1997) pp [6] Z.W. Bell: `Monte Carlo simulation of shadow formation by planar objects illuminated by an extended source', Rev. Prog. Quant. Nondestructive Evaluation, 10A (Plenum Press, New York 1991) [7] C.J. Leliveld, J.G. Mass, V.R. Bom, C.W.E. Van Eijk: `On the significance of scattered radiation in industrial X-ray CT imaging', IEEE Trans. Nucl. Sci., 41, 1 (1994) [8] P.C. Dykstra, M.J. Muus: `The BRL-CAD Package: An overview', In: USENIX, Proceedings of the Fourth Computer Graphics Workshop (1987) [9] E. Storm and H.I. Israël, `Photon cross sections from 1 kev to 100 MeV for elements Z=1 to Z=100', Nuclear Data Tables, A 7 (1969), pp [10] W.R. Nelson, H. Hirayama, D.W.O. Rogers: `The EGS4 code system', SLAC Report 265 (1985) [11] J.C. Satterthwaite: [12] F. Inanc: `Analysis of X-ray and gamma ray scattering through computational experiments', Journal of Nondestructive Evaluation, 18, 2 (1999) [13] J.F. Briesmeister, Editor: `MCNP -- A General Monte Carlo N-Particle Transport Code', Los Alamos National Laboratory report LA M (2000).