Exploring A Two-Level, Control-Variate Monte Carlo Method for Uncertainty Quantification in Criticality Safety
Explore the source record for details and available documents.
SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Explore the source record for details and available documents.
A general form of the integral equation of exospheric transport on moon-like bodies is derived in a form that permits arbitrary specification of time varying physical processes affecting atom creation and annihilation, atom-regolith collisions, adsorption and desorption, and nonplanetocentric acceleration. Because these processes usually defy analytic representation, the Monte Carlo method of solution of the transport equation, the only viable alternative, is described in detail, with separate discussions of the methods of specification of physical processes as probabalistic functions. Proof of the validity of the Monte Carlo exosphere simulation method is provided in the form of a comparison of analytic and Monte Carlo solutions to three classical, and analytically tractable, exosphere problems. One of the key phenomena in moonlike exosphere simulations, the distribution of velocities of the atoms leaving a regolith, depends mainly on the nature of collisions of free atoms with rocks. It is shown that on the moon and Mercury, elastic collisions of helium atoms with a Maxwellian distribution of vibrating, bound atoms produce a nearly Maxwellian distribution of helium velocities, despite the absence of speeds in excess of escape in the impinging helium velocity distribution.
Monte Carlo methods have been applied to reactor problems in order to solve for adjoint-weighted quantities. The Iterated Fission Probability (IFP) method allows adjoint-weighted tallies to be recorded during a forward Monte-Carlo simulation. A key parameter in the IFP method is the number of generations that must pass before the progeny neutrons of a progenitor neutron are tallied. A convergence diagnostic for the IFP is developed using a deterministic code. In particular, a discrete ordinates code is used as a surrogate to demonstrate the theoretical properties of the proposed Monte Carlo convergence diagnostic. The diagnostic is applied to 1-D slab problems and quantifies the error associated with tallying the neutron progeny before an asymptotic distribution is achieved. The results show that the default value of 10 for the number of generations used in MCNP is acceptable for most classes of problems. (authors)
This course is intended for graduate students who already have a basic understanding of Monte Carlo methods. It focuses on advanced topics that may be needed for thesis research, for developing new state-of-the-art methods, or for working with modern production Monte Carlo codes.
Purpose: Beam flatness check has been proposed for beam energy check for photon beams with flattering filters. In this work, beam flatness change with beam energy was investigated quantitatively using the Monte Carlo method and its significance was compared with depth dose curve change. Methods: Monte Carlo simulations for a linear accelerator with flattering filter were performed with different initial electron energies for photon beams of 6MV and 10MV. Dose calculations in a water phantom were then perform with the phase space files obtained from the simulations. The beam flatness was calculated based on the dose profile at 10 cm depth for all the beams with different initial electron energies. The percentage depth dose (PDD) curves were also analyzed. The dose at 10cm depth (D10) and the ratio of the dose at 10cm and 20cm depth (D10/D20) and their change with the beam energy were calculated and compared with the beam flatness variation. Results: It was found that the beam flatness variation with beam energy change was more significant than the change of D10 and the ratio between D10 and D20 for both 6MV and 10MV beams. Half MeV difference on the initial electron beam energy brought in at least 20% variation on the beam flatness but only half percent change on the ratio of D10 and D20. The change of D10 or D20 alone is even less significant. Conclusion: The beam energy impact on PDD is less significant than that on the beam flatness. If the PDD is used for checking the beam energy, uncertainties of the measurement could possibly disguise its change. Beam flatness changes more significantly with beam energy and therefore it can be used for monitoring the energy change for photon beams with flattering filters. However, other factors which may affect the beam flatness should be watched as well.
The Random Ray Method (TRRM) is a recently developed approach to solving neutral particle transport problems based on the Method of Characteristics. While the method previously has been implemented only in closed-source or limited-functionality codes, this work describes its implementation in two open-source Monte Carlo codes: OpenMC and SCONE. The random ray implementations required small modifications to the existing Multigroup Monte Carlo (MGMC) solvers, offering a rare venue for redundant, fine-grained, "apples-to-apples" speed and accuracy comparisons between transport methods. To this end, TRRM and MGMC solvers are evaluated against each other using each code's native capabilities on reactor eigenvalue problems with different degrees of energy discretization. On the C5G7 benchmark (featuring only seven energy groups), TRRM achieves a maximum pin power error comparable to or lower than that of MGMC for a given run time. On a problem with 69 energy groups, MGMC is found to scale more efficiently, obtaining a lower pin power error for a given run time. However, the defining difference between the two transport methods is found to be their vastly different uncertainty distributions. Specifically, TRRM is found to maintain similar levels of accuracy and uncertainty throughout the simulation domain whereas MGMC can exhibit orders-of-magnitude greater errors in areas of the problem that feature low neutron flux. For instance, TRRM provided an up to 373 times speed advantage compared with MGMC for computing the flux in low-flux regions in the moderator surrounding the C5G7 core.
Recently the Monte Carlo (MC) adjoint-weighted perturbation (AWP) methods with k-adjoint have been developed for the k sensitivity calculations. In order to extend the MC AWP methods to estimate sensitivities of a ratio of two reaction rates, Perfetti and Rearden devised a MC sensitivity calculation method with the generalized adjoint expressed by the intra-generational and intergenerational effects. Aufiero et al. suggested the collision history-based approach to calculate sensitivities of responses in the form of ratios of linear or bilinear functions by applying the correlated sampling for multi-generational collisions. However these methods may require a huge size of computer memory to save inter-generational contributions proportional to the number of the sensitivity tallies. In order to reduce memories required for the k-adjoint estimation, the present authors proposed the AWP algorithm in the MC Wielandt method where multiple generations from a fission source neutron are simulated in single history calculations. In this paper, we present perturbation formulations to estimate sensitivities for general tallies such as volume flux, reaction rates, etc. in the MC Wielandt calculations. The proposed method has been implemented into the Seoul National University MC code, McCARD and tested in two-group infinite homogeneous problems. (authors)
Radiation transport modeling methods used in the radiation detection community fall into one of two broad categories: stochastic (Monte Carlo) and deterministic. Monte Carlo methods are typically the tool of choice for simulating gamma-ray spectrometers operating in homeland and national security settings (e.g. portal monitoring of vehicles or isotope identification using handheld devices), but deterministic codes that discretize the linear Boltzmann transport equation in space, angle, and energy offer potential advantages in computational efficiency for many complex radiation detection problems. This paper describes the development of a scenario simulation framework based on deterministic algorithms. Key challenges include: formulating methods to automatically define an energy group structure that can support modeling of gamma-ray spectrometers ranging from low to high resolution; combining deterministic transport algorithms (e.g. ray-tracing and discrete ordinates) to mitigate ray effects for a wide range of problem types; and developing efficient and accurate methods to calculate gamma-ray spectrometer response functions from the deterministic angular flux solutions. The software framework aimed at addressing these challenges is described and results from test problems that compare coupled deterministic-Monte Carlo methods and purely Monte Carlo approaches are provided.
Purpose: The aim of this study is to investigate the variation of bone dose on photon beam energy (keV – MeV) in small-animal irradiation. Dosimetry of homogeneous and inhomogeneous phantoms as per the same mouse computed tomography image set were calculated using the DOSCTP and DOSXYZnrc based on the EGSnrc Monte Carlo code. Methods: Monte Carlo simulations for the homogeneous and inhomogeneous mouse phantom irradiated by a 360 degree photon arc were carried out. Mean doses of the bone tissue in the irradiated volumes were calculated at various photon beam energies, ranging from 50 keV to 1.25 MeV. The effect of bone inhomogeneity was examined through the Inhomogeneous Correction Factor (ICF), a dose ratio of the inhomogeneous to the homogeneous medium. Results: From our Monte Carlo results, higher mean bone dose and ICF were found when using kilovoltage photon beams compared to megavoltage. In beam energies ranging from 50 keV to 200 keV, the bone dose was found maximum at 50 keV, and decreased significantly from 2.6 Gy to 0.55 Gy, when 2 Gy was delivered at the center of the phantom (isocenter). Similarly, the ICF were found decreasing from 4.5 to 1 when the photon beam energy was increased from 50 keV to 200 keV. Both mean bone dose and ICF remained at about 0.5 Gy and 1 from 200 keV to 1.25 MeV with insignificant variation, respectively. Conclusion: It is concluded that to avoid high bone dose in the small-animal irradiation, photon beam energy higher than 200 keV should be used with the ICF close to one, and bone dose comparable to the megavoltage beam where photoelectric effect is not dominant.
Over the last decade, multiphysics coupled calculations have become one of the most important research topics in the area of development and enhancement of reactor analysis and simulation tools. One of the main applications of these coupled simulations is transient analysis, which may involve modeling of experiments, normal plant operation or accident scenarios. Traditionally, the reactor physics side of coupled simulations has been represented by deterministic or hybrid Monte Carlo - deterministic methods. Monte Carlo techniques are limited by their high computational cost. However increased computing power makes their application to multiphysics modeling of transients feasible. Active research is underway to develop algorithms for time dependent calculations, implement them in Monte Carlo codes, and evaluate their performance. In this paper time dependent simulations of the SPERT experiments using the Serpent2 code are presented.
A portable neutron spectrometer, nSpec, has been developed for passive interrogation of fast neutrons ( 1 MeV) from neutron sources. The design of the system includes four EJ-301 liquid scintillation cells, a custom stand, and tungsten caps for gamma-ray shielding. Data acquisition settings were optimized to provide a dynamic range of up to > 15 MeV neutron energy. Additionally, the detector response matrix was characterized using a time-of-flight method based on gamma-ray tagged neutron emission from 252 Cf. The spectrum unfolding was characterized using measured and simulated spectra of AmBe and 252 Cf with the GRAVEL and MLEM methods. Monte Carlo methods were used to examine the uncertainty propagation and determine sufficient counts for a measurement.
The last decade has seen tremendous growth in cloud dynamical and microphysical models that are able to simulate storms and storm systems with very high spatial resolution, typically of the order of a few kilometers. The fairly realistic distributions of cloud and hydrometeor properties that these models generate has in turn led to a renewed interest in the three-dimensional microwave radiative transfer modeling needed to understand the effect of cloud and rainfall inhomogeneities upon microwave observations. Monte Carlo methods, and particularly backwards Monte Carlo methods have shown themselves to be very desirable due to the quick convergence of the solutions. Unfortunately, backwards Monte Carlo methods are not well suited to treat polarized radiation. This study reviews the existing Monte Carlo methods and presents a new polarized Monte Carlo radiative transfer code. The code is based on a forward scheme but uses aliasing techniques to keep the computational requirements equivalent to the backwards solution. Radiative transfer computations have been performed using a microphysical-dynamical cloud model and the results are presented together with the algorithm description.
A Monte Carlo method for simulating Compton scattering in high energy density applications has been presented that models the photon-electron collision kinematics exactly [E. Canfield, W.M. Howard, E.P. Liang, Inverse Comptonization by one-dimensional relativistic electrons, Astrophys. J. 323 (1987) 565]. However, implementing this technique typically requires an explicit evaluation of the material temperature, which can lead to unstable and oscillatory solutions. In this paper, we perform a stability analysis of this Monte Carlo method and develop two time-step limits that avoid undesirable behavior. The first time-step limit prevents instabilities, while the second, more restrictive time-step limit avoids both instabilities and nonphysical oscillations. With a set of numerical examples, we demonstrate the efficacy of these time-step limits.
Monte Carlo (MC) simulation is widely used to solve the eigenvalue form of the Boltzmann transport equation that mathematically represents the neutron transport process through complex multiplying (fissionable) systems. Monte Carlo eigenvalue simulation starts with an assumed fission source distribution and uses the fission sites from the previous iteration (cycle) as the starting source in the current iteration. Important system parameters (MC tallies) such as fuel pin-power distribution are estimated over several cycles after the convergence of the fission source distribution to a stationary distribution. However, the MC fission source iteration algorithm that uses fission source sites from the previous cycle introduces a cycle-to-cycle correlation. Monte Carlo simulations that do not account for the cycle-to-cycle correlation systematically underestimate the variance of the estimated system parameters (sample mean). This paper presents the relationship between the spectral density in the frequency domain at frequency zero and the variance of the sample mean. This paper introduces a novel method in the frequency domain for the MC variance estimation. For the three test problems used in this paper, researchers have observed that the new method results in an improvement of more than one order of magnitude to the standard deviation of the sample mean. The new method also compares favorably with the previously introduced batch, bootstrap, and covariance-adjusted methods when applied to the three test problems investigated in this paper. This new method does not require modification of the MC eigenvalue algorithm (power iteration), is code agnostic, and is therefore easy to use when implementing in any existing MC code. In conclusion, the new estimate can be calculated without saving tally results of all active/stationary cycles.
We present a moment-based acceleration algorithm applied to Monte Carlo simulation of thermal radiative-transfer problems. Our acceleration algorithm employs a continuum system of moments to accelerate convergence of stiff absorption–emission physics. The combination of energy-conserving tallies and the use of an asymptotic approximation in optically thick regions remedy the difficulties of local energy conservation and mitigation of statistical noise in such regions. We demonstrate the efficiency and accuracy of the developed method. We also compare directly to the standard linearization-based method of Fleck and Cummings [1]. A factor of 40 reduction in total computational time is achieved with the new algorithm for an equivalent (or more accurate) solution as compared with the Fleck–Cummings algorithm.
Nuclear criticality calculations with Monte Carlo codes are normally done using a power iteration method to obtain the dominant eigenfunction and eigenvalue. In the last few years it has been shown that the power iteration method can be modified to obtain the first two eigenfunctions. This modified power iteration method directly subtracts out the second eigenfunction and thus only powers out the third and higher eigenfunctions. The result is a convergence rate to the dominant eigenfunction being |k{sub 3}|/k{sub 1} instead of |k{sub 2}|/k{sub 1}. One difficulty is that the second eigenfunction contains particles of both positive and negative weights that must sum somehow to maintain the second eigenfunction. Summing negative and positive weights can be done using point detector mechanics, but this sometimes can be quite slow. We show that an approximate cancellation scheme is sufficient to accelerate the convergence to the dominant eigenfunction. A second difficulty is that for some problems the Monte Carlo implementation of the modified power method has some stability problems. We also show that a simple method deals with this in an effective, but ad hoc manner.
Diffusion Monte Carlo is the most accurate widely used Quantum Monte Carlo method for the electronic structure of materials, but it requires frequent load balancing or population redistribution steps to maintain efficiency and avoid accumulation of systematic errors on parallel machines. The load balancing step can be a significant factor affecting performance, and will become more important as the number of processing elements increases. We propose a new dynamic load balancing algorithm, the Alias Method, and evaluate it theoretically and empirically. An important feature of the new algorithm is that the load can be perfectly balanced with each process receiving at most one message. It is also optimal in the maximum size of messages received by any process. We also optimize its implementation to reduce network contention, a process facilitated by the low messaging requirement of the algorithm. Empirical results on the petaflop Cray XT Jaguar supercomputer at ORNL showing up to 30% improvement in performance on 120,000 cores. The load balancing algorithm may be straightforwardly implemented in existing codes. The algorithm may also be employed by any method with many near identical computational tasks that requires load balancing.
The unitarity limit describes interacting particles where the range of the interaction is zero and the scattering length is infinite. We present precision benchmark calculations for two-component fermions at unitarity using three different ab initio methods: Hamiltonian lattice formalism using iterated eigenvector methods, Euclidean lattice formalism with auxiliary-field projection Monte Carlo methods, and continuum diffusion Monte Carlo methods with fixed and released nodes. We have calculated the ground-state energy of the unpolarized four-particle system in a periodic cube as a dimensionless fraction of the ground-state energy for the noninteracting system. We obtain values of 0.211(2) and 0.210(2) using two different Hamiltonian lattice representations, 0.206(9) using Euclidean lattice formalism, and an upper bound of 0.212(2) from fixed-node diffusion Monte Carlo methods. Released-node calculations starting from the fixed-node result yield a decrease of less than 0.002 over a propagation of 0.4E{sub F}{sup -1} in Euclidean time, where E{sub F} is the Fermi energy. We find good agreement among all three ab initio methods.