We consider the problem of calibrating an imperfect computer model using experimental data. To compensate for the misspecification of the computer model and make more accurate predictions, a discrepancy function is often included and modeled via a Gaussian stochastic process (GaSP). The calibrated computer model alone, however, sometimes fits the experimental data poorly, as the calibration parameters become unidentifiable. In this work, we propose the scaled Gaussian stochastic process (S-GaSP), a novel stochastic process that bridges the gap between two predominant methods, namely, the $L_2$ calibration and the GaSP calibration. It is shown that our approach performs well in both calibration and prediction. A computationally feasible approach is introduced for this new model under the Bayesian paradigm. Compared with the GaSP calibration, the S-GaSP calibration enables the calibrated computer model itself to predict the reality well, based on the posterior distribution of the calibration parameters. Numerical comparisons of the simulated and real data are provided to illustrate the connections and differences between the proposed S-GaSP and other alternative approaches.
5 December 2022 in The Journal of Chemical Physics
A statistical emulator can be used as a surrogate of complex physics-based calculations to drastically reduce the computational cost. Its successful implementation hinges on an accurate representation of the nonlinear response surface with a high-dimensional input space. Conventional “space-filling” designs, including random sampling and Latin hypercube sampling, become inefficient as the dimensionality of the input variables increases, and the predictive accuracy of the emulator can degrade substantially for a test input distant from the training input set. To address this fundamental challenge, we develop a reliable emulator for predicting complex functionals by active learning with error control (ALEC). The algorithm is applicable to infinite-dimensional mapping with high-fidelity predictions and a controlled predictive error. The computational efficiency has been demonstrated by emulating the classical density functional theory (cDFT) calculations, a statistical-mechanical method widely used in modeling the equilibrium properties of complex molecular systems. We show that ALEC is much more accurate than conventional emulators based on the Gaussian processes with “space-filling” designs and alternative active learning methods. In addition, it is computationally more efficient than direct cDFT calculations. ALEC can be a reliable building block for emulating expensive functionals owing to its minimal computational cost, controllable predictive error, and fully automatic features.
Differential dynamic microscopy (DDM) is a form of video image analysis that combines the sensitivity of scattering and the direct visualization benefits of microscopy. DDM is broadly useful in determining dynamical properties including the intermediate scattering function for many spatiotemporally correlated systems. Despite its straightforward analysis, DDM has not been fully adopted as a routine characterization tool, largely due to computational cost and lack of algorithmic robustness. We present statistical analysis that quantifies the noise, reduces the computational order, and enhances the robustness of DDM analysis. We propagate the image noise through the Fourier analysis, which allows us to comprehensively study the bias in different estimators of model parameters, and we derive a different way to detect whether the bias is negligible. Furthermore, through use of Gaussian process regression (GPR), we find that predictive samples of the image structure function require only around 0.5%–5% of the Fourier transforms of the observed quantities. This vastly reduces computational cost, while preserving information of the quantities of interest, such as quantiles of the image scattering function, for subsequent analysis. The approach, which we call DDM with uncertainty quantification (DDM-UQ), is validated using both simulations and experiments with respect to accuracy and computational efficiency, as compared with conventional DDM and multiple particle tracking. Overall, we propose that DDM-UQ lays the foundation for important new applications of DDM, as well as to high-throughput characterization.
This article proposes a specifically designed parallel partial Gaussian process (PPGP) for large-scale probabilistic power system transient stability assessment (TSA). The differential and algebraic equations with uncertain resources are defined and reformulated regarding uncertainty quantification. The challenges of probabilistic TSA in large-scale systems are investigated and the necessity of introducing efficient modeling approaches is emphasized. PPGP inherits the advantages of Gaussian process modeling, including data-driven modeling that does not require the knowledge of input distribution and can effectively quantify the confidence interval of the estimate. It accelerates the model construction and evaluation by reducing the number of parameters to be estimated. In addition, thanks to the enhancement techniques via isotropic kernel, reparameterization, and composite likelihood, PPGP is able to achieve probabilistic TSA under high uncertainty conditions while maintaining very high computational efficiency. A theoretical analysis of the time complexity of PPGP is provided. Comparison results with the Latin hypercube sampling method, many single Gaussian processes, and sparse Gaussian process methods on the Texas 2000-bus system with 471 PVs and 470 wind generations highlight the significantly improved model efficacy and efficiency of the proposed method. Furthermore, experimental validation demonstrates the proposed method's efficacy in handling the co-existence of both stable and unstable cases.
Evolution of composition, rheology, and morphology during phase separation in complex fluids is highly coupled to rheological and mass transport processes within the emerging phases, and understanding this coupling is critical for materials design of multiphase complex fluids. Characterizing these dependencies typically requires careful measurement of a large number of equilibrium and transport properties that are difficult to measure in situ as phase separation proceeds. Here, we propose and demonstrate a high-throughput microscopy platform to achieve simultaneous, in situ mapping of time-evolving morphology and microrheology in phase separating complex fluids over a large compositional space. The method was applied to a canonical example of polyelectrolyte complex coacervation, whereby mixing of oppositely charged species leads to liquid–liquid phase separation into distinct solute-dense and dilute phases. Morphology and rheology were measured simultaneously and kinetically after mixing to track the progression of phase separation. Once equilibrated, the dense phase viscosity was determined to high compositional accuracy using passive probe microrheology, and the results were used to derive empirical relationships between the composition and viscosity. These relationships were inverted to reconstruct the dense phase boundary itself, and further extended to other mixture compositions. The resulting predictions were validated by independent equilibrium compositional measurements. This platform paves the way for rapid screening and formulation of complex fluids and (bio)macromolecular materials, and serves as a critical link between formulation and rheology for multi-phase material discovery.