Machine-Learning Classification of N = 5 Orbital Families via Kelvin-Simplex Face-Activity Descriptors: Structural Parity and the Onset of Energy-Partitioning Diffusion

Read full text Summarise with VIRAT
Authors: Lee Michael John Rich
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The Kelvin-simplex face-activity descriptor framework, demonstrated for N = 4 in Paper VI of this series, is here extended to the five-body problem. Five N = 5 orbital families are defined and simulated: Pentagon (five equal masses in a near-regular force-balance orbit), Hierarchical (inner close binary + middle singleton + outer binary), Triple+Binary (equilateral triangle sub-system + distant bound binary - a novel family with no N = 4 analogue), Chaotic (random initial conditions, zero net momentum), and Near-Ejection (compact quadruple + flyby body). The face-activity descriptor vector is extended to 25 components: five face-exclusion (FE) components, ten pair-dominance (PD) components with the dominance threshold rescaled from 0.80 to 0.70 to match the lower pair fair-share fraction at N = 5, and ten triplet-dominance (TD) components introduced here for the first time. Using 400 trajectories (80 per family) and five classifiers implemented from first principles, the face-activity vector achieves a best cross-validated accuracy of 79.75% (Gaussian Naive Bayes), tying the best classical scalar performance (k-NN: 79.75%), with no classifier showing a statistically significant face-activity advantage (all Wilcoxon p > 0.14). This constitutes a qualitative change from N = 4, where face-activity descriptors outperformed classical scalars by +13.8 percentage points (p < 0.01) for all four classifiers tested. Three targeted recovery experiments - dominance-threshold rescaling, triplet-dominance feature extension, and PCA-based dimensionality reduction - all return null results for k-NN accuracy. The Triple+Binary family is classified near-perfectly (recall 98.75%), the strongest individual-family result. The Chaotic/Near-Ejection confusion (31.25% recall for Chaotic) is identified as the fundamental bottleneck and attributed to genuine phase-space overlap at N = 5 rather than methodological insufficiency. Classical scalars are competitive primarily through the extreme discriminative power of the simplex path length (ANOVA F = 747), driven by the anomalously large path length of the Hierarchical family. These results identify N = 5 as a transitional regime; the operating boundary of face-activity dominance appears to lie between N = 4 and N = 5 under the conditions studied.

Fields of Study & Subjects

Stellar planetary and galactic studies Space Satellite Systems and Control Spacecraft Dynamics and Control

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
27 March 2026
Language
EN
Document type
Preprint