Jacobian-Scaled K-means Clustering for Physics-Informed Segmentation of Reacting Flows

Shivam Barwey,Venkat Raman

2024-06-24

Abstract:This work introduces Jacobian-scaled K-means (JSK-means) clustering, which is a physics-informed clustering strategy centered on the K-means framework. The method allows for the injection of underlying physical knowledge into the clustering procedure through a distance function modification: instead of leveraging conventional Euclidean distance vectors, the JSK-means procedure operates on distance vectors scaled by matrices obtained from dynamical system Jacobians evaluated at the cluster centroids. The goal of this work is to show how the JSK-means algorithm -- without modifying the input dataset -- produces clusters that capture regions of dynamical similarity, in that the clusters are redistributed towards high-sensitivity regions in phase space and are described by similarity in the source terms of samples instead of the samples themselves. The algorithm is demonstrated on a complex reacting flow simulation dataset (a channel detonation configuration), where the dynamics in the thermochemical composition space are known through the highly nonlinear and stiff Arrhenius-based chemical source terms. Interpretations of cluster partitions in both physical space and composition space reveal how JSK-means shifts clusters produced by standard K-means towards regions of high chemical sensitivity (e.g., towards regions of peak heat release rate near the detonation reaction zone). The findings presented here illustrate the benefits of utilizing Jacobian-scaled distances in clustering techniques, and the JSK-means method in particular displays promising potential for improving former partition-based modeling strategies in reacting flow (and other multi-physics) applications.

Computational Physics,Machine Learning,Fluid Dynamics

What problem does this paper attempt to address?

The paper aims to address the issue where traditional clustering methods fail to effectively capture dynamically similar regions in reactive flows (such as explosive flows). Specifically, the standard K-means clustering method tends to cluster based on the features of the samples themselves (such as temperature, species concentration, etc.) when dealing with complex reactive flow data, while ignoring the physical dynamic information behind these features. This leads to clustering results that may not accurately reflect the dynamic similarity in terms of chemical reaction rates and other aspects in different regions. To overcome this limitation, the paper proposes the Jacobian-scaled K-means (JSK-means) clustering method. This method modifies the distance function by introducing a Jacobian matrix-based scaling, injecting underlying physical knowledge into the clustering process. Specifically, JSK-means uses the Jacobian matrix to scale the distance vector from the sample to the centroid, thereby making the clustering process more focused on the dynamic similarity between samples rather than just the similarity of sample feature values. In this way, the clustering results can better capture highly sensitive regions, such as areas with high heat release rates near explosive reaction zones, providing a new approach for local modeling of reactive flows. The paper validates the effectiveness of the JSK-means method on a complex channel explosion simulation dataset, demonstrating the potential of this method in improving local modeling of reactive flows, especially in multiphysics applications. This method not only enhances the physical consistency of clustering but also provides general methodological support for dynamically consistent segmentation of feature spaces in other multiphysics systems.

Jacobian-Scaled K-means Clustering for Physics-Informed Segmentation of Reacting Flows

Dynamic Mode Decomposition Analysis of Spatially Agglomerated Flow Databases

Equation-informed data-driven identification of flow budgets and dynamics

Information Theoretic Clustering for Coarse-grained Modeling of Non-equilibrium Gas Dynamics

Distributed Kernel K-Means for Large Scale Clustering

Capturing mesoscale structures in computational fluid dynamics simulations of gas‐solid flows

Two step clustering for data reduction combining DBSCAN and k-means clustering

Complex-Valued K-Means Clustering of Interpolative Separable Density Fitting Algorithm for Large-Scale Hybrid Functional Enabled Ab Initio Molecular Dynamics Simulations within Plane Waves

A fully coupled, fully implicit simulation method for unsteady flames using Jacobian approximation and clustering

Complex-valued K-means clustering of interpolative separable density fitting algorithm for large-scale hybrid functional enabled \textit{ab initio} molecular dynamics simulations within plane waves

Spatio-Temporal Surrogates for Interaction of a Jet with High Explosives: Part II -- Clustering Extremely High-Dimensional Grid-Based Data

Machine Learning K-Means Clustering in Interpolative Separable Density Fitting Algorithm: Advancing Accurate and Efficient Cubic-Scaling Density Functional Perturbation Theory Calculations within Plane Waves

Using Ripley's K-function to Characterize Clustering In 3-Dimensional Point Patterns With a Case Study in Atom Probe Tomography

An Improved Method for Coherent Structure Identification Based on Mutual K-nearest Neighbors

Dynamic Fuzzy c-Means (dFCM) Clustering and its Application to Calorimetric Data Reconstruction in High Energy Physics

A combined clustering/symbolic regression framework for fluid property prediction

Using SIMD and SIMT vectorization to evaluate sparse chemical kinetic Jacobian matrices and thermochemical source terms

A Unified Gas-Kinetic Scheme for Multiscale and Multicomponent Flow Transport

Physics-Informed Machine Learning Models for Predicting the Progress of Reactive-Mixing

k-Means NANI: An Improved Clustering Algorithm for Molecular Dynamics Simulations

Cluster-based reduced-order modelling of a mixing layer