Estimation of Contact Forces in Reactor Internals using Neutron Noise Data: A Comparison of ApproachesHarn, Nicholas J (2022) Estimation of Contact Forces in Reactor Internals using Neutron Noise Data: A Comparison of Approaches. Master's Thesis, University of Pittsburgh. (Unpublished)
AbstractA pressurized water reactor’s radial keys prevent the reactor core from colliding with its housing. These supports gradually degrade at their points of contact, requiring regular inspections and preventative maintenance during reactor outages. Reactor outages are costly and time consuming, yet they are necessary for continued safe operations. Outages could be expedited by monitoring the conditions of internal components during reactor operations. This is partially achieved by tracking the contact forces acting on the core barrel, which are altered by radial key degradation. The contact force at each radial key is modeled as a hyperbolic tangent function of the core barrel’s velocity at the point of contact, capturing the contact force’s transition from viscous damping at low velocities to constant Coulomb friction at high velocities. Two condition-dependent parameters control the contact force model: the maximum contact force, α, and the inverse of the characteristic velocity at which the model transitions between behaviors, β. The contact force dampens the core barrel’s vibrations, which previous research has related to the neutron radiation measured outside the reactor core. The contact force is monitored using these ex-core neutron noise measurements. This report compares two methods for approximating the contact force parameters. One method applies an unscented Kalman filter to the ex-core measurements. The other applies a grid search algorithm to several ensembles of core barrel simulations, each associated with a combinations of contact parameters. The grid search identifies the combination whose ensemble best approximates the ex-core measurements. These methods are applied to four synthetic datasets with known contact parameters to determine which approach, if any, produces more accurate parameter estimates. The Kalman filter’s estimates are more accurate than the grid search’s approximations for Datasets 3 and 4. These datasets have α values of 490N and 660 N, with β values of 70 s/m and 110 s/m respectively. The grid search method performs better than the Kalman filter for Dataset 1 where α = 100 N and β = 10 s/m, while both methods are inaccurate for Dataset 2 where α = 31.6 N and β = 316 s/m. Neither method accurately estimates the contact parameters for an arbitrary synthetic dataset. Share
Details
MetricsMonthly Views for the past 3 yearsPlum AnalyticsActions (login required)
|