Marco Aurélio dos Santos Bernardes
Marco Aurélio dos Santos Bernardes has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
Xinping Zhou
5 shared publications- Theodor Willem von Backström
Stellenbosch University, Stellenbosch, South Africa
3 shared publications A. Voß
1 shared publicationD. G. Kröger
University of Stellenbosch
1 shared publicationDetlev G. Kröger
1 shared publication- Gerhard Weinrebe
Schlaich bergermann und partner
1 shared publication
Latest Publications
- 2 November 2026
- Published by Elsevier in Computers & Mathematics with Applications
- Vol. 224, 369-387
Prior-preserving CCPFM on irregular point clouds: Analysis, Poisson benchmarks, and extended diffusion tests
- 2 November 2026
- Published by Elsevier in Computers & Mathematics with Applications
- Vol. 224, 369-387
Prior-preserving CCPFM on irregular point clouds: Analysis, Poisson benchmarks, and extended diffusion tests
- Open Access
- 5 September 2026
- Published by Elsevier in SSRN Electronic Journal
The Bernardes-Branch Construction for Arbitrary Triangles in Minkowski Spacetime: Lorentz-Invariant Area-Balance Loci, Null Stratification, and Reeb Topology
We study the Lorentzian extension of the \emph{Bernardes-branch construction}: an additive triangle-area locus defined for \emph{any nondegenerate triangle}, with no right-angle requirement. Right triangles are therefore a special subclass rather than the defining geometry. Let $A,B,C$ be an arbitrary fixed simultaneous spacelike triangle and let $P$ vary in Minkowski spacetime. The three face areas $PAB$, $PBC$, and $PCA$ are defined by Lorentzian Gram determinants, and the locus is the union of the three conditions in which one unsigned invariant area equals the sum of the other two. The construction is Poincar\'e invariant. In the rest frame of the base triangle, all dependence on the normal coordinate $z$ and time coordinate $\tau=ct$ occurs through $s=z^2-\tau^2$, giving an exact $SO(1,1)$ boost reduction to a three-dimensional semialgebraic master surface. For an arbitrary nondegenerate triangle, the side-parallel end problem reduces to the universal law $|\delta-1|=\sqrt{|\delta^2-\theta|}$, where $\delta=d/h$ and $\theta=T/h^2$. Every side therefore has a projective end fold at $T_\infty=h^2/2$ and an end exchange at $T_E=h^2$. For a generic scalene triangle, a finite Morse birth and an explicit semialgebraic spine connect each side channel to the null network; each side forces one independent end-completed Reeb cycle. A complete critical-stratum exhaustion reduces the quotient to a seven-vertex, nine-edge null connector core and proves $b_1(\Rhat)=3$. Symmetry changes this topology: non-equilateral isosceles triangles have $b_1=2$, while the equilateral triangle has $b_1=0$; rightness alone is not a phase boundary. A $3$--$4$--$5$ case study independently exhibits the predicted three-cycle quotient and resolves the local regular-null and triple-null singularities. The results identify a shape-controlled topological structure that is absent from the Euclidean locus and is distinct from the classical hyperbolic Pythagorean identities of relativistic velocity space.- 5 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Bernardes-Branch Construction for Arbitrary Triangles in Minkowski Spacetime: Lorentz-Invariant Area-Balance Loci, Null Stratification, and Reeb Topology
We study the Lorentzian extension of the \emph{Bernardes-branch construction}: an additive triangle-area locus defined for \emph{any nondegenerate triangle}, with no right-angle requirement. Right triangles are therefore a special subclass rather than the defining geometry. Let $A,B,C$ be an arbitrary fixed simultaneous spacelike triangle and let $P$ vary in Minkowski spacetime. The three face areas $PAB$, $PBC$, and $PCA$ are defined by Lorentzian Gram determinants, and the locus is the union of the three conditions in which one unsigned invariant area equals the sum of the other two. The construction is Poincar\'e invariant. In the rest frame of the base triangle, all dependence on the normal coordinate $z$ and time coordinate $\tau=ct$ occurs through $s=z^2-\tau^2$, giving an exact $SO(1,1)$ boost reduction to a three-dimensional semialgebraic master surface. For an arbitrary nondegenerate triangle, the side-parallel end problem reduces to the universal law $|\delta-1|=\sqrt{|\delta^2-\theta|}$, where $\delta=d/h$ and $\theta=T/h^2$. Every side therefore has a projective end fold at $T_\infty=h^2/2$ and an end exchange at $T_E=h^2$. For a generic scalene triangle, a finite Morse birth and an explicit semialgebraic spine connect each side channel to the null network; each side forces one independent end-completed Reeb cycle. A complete critical-stratum exhaustion reduces the quotient to a seven-vertex, nine-edge null connector core and proves $b_1(\Rhat)=3$. Symmetry changes this topology: non-equilateral isosceles triangles have $b_1=2$, while the equilateral triangle has $b_1=0$; rightness alone is not a phase boundary. A $3$--$4$--$5$ case study independently exhibits the predicted three-cycle quotient and resolves the local regular-null and triple-null singularities. The results identify a shape-controlled topological structure that is absent from the Euclidean locus and is distinct from the classical hyperbolic Pythagorean identities of relativistic velocity space. - Open Access
- 29 July 2026
- Published by MDPI AG in Clean Technologies
- Vol. 8 (4) , 115
Adaptive Energy Stations for Sustainable Transport Infrastructure: Real-Time Dispatch Optimization Using Marginal Grid Emissions and Low-Carbon Fuel Pathways
Transport decarbonization requires infrastructure that can use time-resolved carbon information without overstating the representativeness of short proof-of-method runs. This study introduces Adaptive Energy Stations (AESs), multi-fuel transport-energy nodes that integrate marginal grid-emission signals, fuel life-cycle carbon intensities, wholesale electricity prices, and vehicle operating constraints into a station-level dispatch optimization. The implemented case is a one-week winter proof-of-method for CAISO/CAISO_NORTH using 168 hourly service events over 1–8 January 2026 Pacific time, archived WattTime marginal operating emissions, CAISO locational marginal prices, eGRID CAMX annual-average factors, and declared vehicle and fuel-pathway parameters. In the audited CAISO scenario, the attached dispatch outputs report a reduction from 181.76 to 123.38 g CO2e/km relative to the specified static baseline, corresponding to a 32.12% reduction for the one-week winter service-event stream. The populated dispatch trace shows that the carbon-priority AES plug-in hybrid electric vehicle (PHEV) run selected cellulosic E85 for all 168 events and selected no electric events; this result is interpreted as an operational scenario result for the archived week, not as an annual fleet-average, smart-charging benefit, or deployment forecast. The revised analysis explicitly separates implemented CAISO evidence from ERCOT, MISO-MROW, and ISO–NE extension sensitivities, which remain hypothetical until equivalent marginal-emissions, price, and service-event data are supplied. Battery-production amortization is treated as a separate sensitivity because it can change battery electric vehicle (BEV)–cellulosic E85 equivalence conclusions: at 50–100 kg CO2e/kWh over 240,000 km, a 75 kWh BEV pack contributes 15.6–31.3 g CO2e/km and a 14 kWh PHEV pack contributes 2.9–5.8 g CO2e/km. Practical-equivalence claims are therefore conditional on the declared boundary, equivalence margin, and production-emissions treatment. Full deployment requires validated marginal-emission access, transparent dispatch-audit outputs, supply-chain verification, user-behavior characterization, cost sensitivity analysis, and cybersecurity safeguards.
- 29 July 2026
- Published by MDPI AG in Clean Technologies
- Vol. 8 (4) , 115
Open AccessAdaptive Energy Stations for Sustainable Transport Infrastructure: Real-Time Dispatch Optimization Using Marginal Grid Emissions and Low-Carbon Fuel Pathways
Transport decarbonization requires infrastructure that can use time-resolved carbon information without overstating the representativeness of short proof-of-method runs. This study introduces Adaptive Energy Stations (AESs), multi-fuel transport-energy nodes that integrate marginal grid-emission signals, fuel life-cycle carbon intensities, wholesale electricity prices, and vehicle operating constraints into a station-level dispatch optimization. The implemented case is a one-week winter proof-of-method for CAISO/CAISO_NORTH using 168 hourly service events over 1–8 January 2026 Pacific time, archived WattTime marginal operating emissions, CAISO locational marginal prices, eGRID CAMX annual-average factors, and declared vehicle and fuel-pathway parameters. In the audited CAISO scenario, the attached dispatch outputs report a reduction from 181.76 to 123.38 g CO2e/km relative to the specified static baseline, corresponding to a 32.12% reduction for the one-week winter service-event stream. The populated dispatch trace shows that the carbon-priority AES plug-in hybrid electric vehicle (PHEV) run selected cellulosic E85 for all 168 events and selected no electric events; this result is interpreted as an operational scenario result for the archived week, not as an annual fleet-average, smart-charging benefit, or deployment forecast. The revised analysis explicitly separates implemented CAISO evidence from ERCOT, MISO-MROW, and ISO–NE extension sensitivities, which remain hypothetical until equivalent marginal-emissions, price, and service-event data are supplied. Battery-production amortization is treated as a separate sensitivity because it can change battery electric vehicle (BEV)–cellulosic E85 equivalence conclusions: at 50–100 kg CO2e/kWh over 240,000 km, a 75 kWh BEV pack contributes 15.6–31.3 g CO2e/km and a 14 kWh PHEV pack contributes 2.9–5.8 g CO2e/km. Practical-equivalence claims are therefore conditional on the declared boundary, equivalence margin, and production-emissions treatment. Full deployment requires validated marginal-emission access, transparent dispatch-audit outputs, supply-chain verification, user-behavior characterization, cost sensitivity analysis, and cybersecurity safeguards.
- Open Access
- 3 June 2026
- Published by Elsevier in SSRN Electronic Journal
Conservative Benchmark Assessment of Prior-Preserving CCPFM Variants for 2D Poisson Problems on Irregular Point Clouds
This paper studies constrained correction point-flux methods (CCPFM) for manufactured 2D Poisson problems on irregular point clouds. The benchmark uses four geometries---circle, rotated ellipse, star, and annulus---and compares a weighted least-squares (WLS) reference, a neighbor-adaptive normalized moving least-squares (NAMLS) operator, a frozen CCPFM baseline, an experimental live-boundary gated CCPFM branch, and sparse deep-core enrichment ablations within one common computational harness. The CCPFM branch is built around a prior-preserving constrained correction formulation on a fixed graph, together with diagnostics that separate near-boundary and core behavior, patch-test consistency, instability incidence, and added-edge activation. The main conclusion is conservative. The frozen CCPFM baseline remains the main recommended CCPFM method overall. The live-boundary gated branch exhibits geometry-sensitive behavior rather than uniform improvement: it is helpful on the star, dormant on the circle and ellipse, and not justified on the annulus at the finest tested spacing. Sparse deep-core enrichment yields an even narrower conclusion. A forced one-edge sparsecore branch acts as a proof of mechanism, showing that sparse enrichment can be inserted safely. A rarity-based selective sparsecore trigger activates occasionally without force-flagging, particularly on annulus and ellipse clouds, but its practical effect remains negligible in both all-run and activated-run-only views. The frozen CCPFM baseline remained the main recommended method overall, while live-boundary gating and sparse enrichment were retained as informative but secondary ablations.
- 3 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessConservative Benchmark Assessment of Prior-Preserving CCPFM Variants for 2D Poisson Problems on Irregular Point Clouds
This paper studies constrained correction point-flux methods (CCPFM) for manufactured 2D Poisson problems on irregular point clouds. The benchmark uses four geometries---circle, rotated ellipse, star, and annulus---and compares a weighted least-squares (WLS) reference, a neighbor-adaptive normalized moving least-squares (NAMLS) operator, a frozen CCPFM baseline, an experimental live-boundary gated CCPFM branch, and sparse deep-core enrichment ablations within one common computational harness. The CCPFM branch is built around a prior-preserving constrained correction formulation on a fixed graph, together with diagnostics that separate near-boundary and core behavior, patch-test consistency, instability incidence, and added-edge activation. The main conclusion is conservative. The frozen CCPFM baseline remains the main recommended CCPFM method overall. The live-boundary gated branch exhibits geometry-sensitive behavior rather than uniform improvement: it is helpful on the star, dormant on the circle and ellipse, and not justified on the annulus at the finest tested spacing. Sparse deep-core enrichment yields an even narrower conclusion. A forced one-edge sparsecore branch acts as a proof of mechanism, showing that sparse enrichment can be inserted safely. A rarity-based selective sparsecore trigger activates occasionally without force-flagging, particularly on annulus and ellipse clouds, but its practical effect remains negligible in both all-run and activated-run-only views. The frozen CCPFM baseline remained the main recommended method overall, while live-boundary gating and sparse enrichment were retained as informative but secondary ablations.
- Open Access
- 9 May 2026
- Published by Elsevier in SSRN Electronic Journal
Adaptive Energy Stations for Sustainable Transport Infrastructure: Real-Time Dispatch Optimization Using Marginal Grid Emissions and Low-Carbon Fuel Pathways
Transport decarbonization requires not only low-carbon vehicle technologies but also infrastructure capable of selecting the lowest-emission energy pathway in real time. This study introduces Adaptive Energy Stations (AES), multi-fuel transport-energy nodes that integrate marginal grid-emission signals, fuel life-cycle carbon intensities, wholesale electricity prices, and vehicle operating constraints into a station-level dispatch optimization. A proof-of-method implementation is demonstrated for the California Independent System Operator (CAISO) zone using 168 hourly service events over 1--8 January 2026, combining WattTime marginal operating emissions, CAISO locational marginal prices, eGRID CAMX annual-average factors, and declared vehicle and fuel-pathway parameters. For plug-in hybrid electric vehicles (PHEVs), AES dispatch reduced mean emissions from 145 to 90~\gcoeqkm{} --- a 38\% improvement over a static annual-average baseline --- by shifting electric charging toward low-carbon grid intervals and substituting cellulosic E85 during high-carbon periods. Monte Carlo uncertainty propagation (10,000 draws) established practical equivalence between CAISO-zone battery electric vehicle (BEV) and cellulosic E85 pathways under a $\pm 25$~\gcoeqkm{} margin, though equivalence failed at $\pm 10$~\gcoeqkm{}. Sensitivity analyses for ERCOT, MISO--MROW, and ISO--NE extension zones revealed region-dependent emission rankings, with BEV advantages diminishing under higher grid carbon intensity. The framework supports performance-based, technology-neutral decarbonization policy by demonstrating that operational life-cycle carbon information materially alters station-level energy-pathway selection. Full deployment requires validated marginal-emission access, supply-chain verification, user-behavior characterization, and cybersecurity safeguards.
- 9 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAdaptive Energy Stations for Sustainable Transport Infrastructure: Real-Time Dispatch Optimization Using Marginal Grid Emissions and Low-Carbon Fuel Pathways
Transport decarbonization requires not only low-carbon vehicle technologies but also infrastructure capable of selecting the lowest-emission energy pathway in real time. This study introduces Adaptive Energy Stations (AES), multi-fuel transport-energy nodes that integrate marginal grid-emission signals, fuel life-cycle carbon intensities, wholesale electricity prices, and vehicle operating constraints into a station-level dispatch optimization. A proof-of-method implementation is demonstrated for the California Independent System Operator (CAISO) zone using 168 hourly service events over 1--8 January 2026, combining WattTime marginal operating emissions, CAISO locational marginal prices, eGRID CAMX annual-average factors, and declared vehicle and fuel-pathway parameters. For plug-in hybrid electric vehicles (PHEVs), AES dispatch reduced mean emissions from 145 to 90~\gcoeqkm{} --- a 38\% improvement over a static annual-average baseline --- by shifting electric charging toward low-carbon grid intervals and substituting cellulosic E85 during high-carbon periods. Monte Carlo uncertainty propagation (10,000 draws) established practical equivalence between CAISO-zone battery electric vehicle (BEV) and cellulosic E85 pathways under a $\pm 25$~\gcoeqkm{} margin, though equivalence failed at $\pm 10$~\gcoeqkm{}. Sensitivity analyses for ERCOT, MISO--MROW, and ISO--NE extension zones revealed region-dependent emission rankings, with BEV advantages diminishing under higher grid carbon intensity. The framework supports performance-based, technology-neutral decarbonization policy by demonstrating that operational life-cycle carbon information materially alters station-level energy-pathway selection. Full deployment requires validated marginal-emission access, supply-chain verification, user-behavior characterization, and cybersecurity safeguards.
- Open Access
- 1 May 2026
- Published by Elsevier in SSRN Electronic Journal
Branch-Aware Pythagorean Fields for Feature-Sensitive Mesh Envelope Design and Seam-Network Initialization
This paper studies a branch-aware Pythagorean residual eld as a conservative exploratory primitive for mesh-envelope visualization and seam-network initialization. The eld is not a signed-distance eld, not an o set method, and not a fabrication-ready panelizer. Its purpose is narrower: to expose area-balance compatibility structures induced by local mesh triangles and to test whether these structures can provide candidate contour cues for later, constraintbased panel-layout tools. The formulation reduces each single-triangle branch relation to a quadratic equation in u = z2, applies an admissibility test to remove roots introduced by squaring, and evaluates a locality-gated residual eld over procedural benchmark meshes. The implementation reports point-triangle, point-plane, and curvature-proxy comparisons, raw benchmark metrics, threshold sensitivity, grid/scale/rotation/noise checks, and failure cases. The results show that the eld can generate branch-compatible contour networks that di er from ordinary proximity elds, but they also expose signi cant limitations: nonnearest-triangle responses, coordinate-frame dependence of the projected query layer, meshresolution sensitivity, and lack of manufacturability guarantees. The contribution should therefore be read as a reproducible geometry-processing diagnostic and seam-initialization study, not as a validated replacement for distance elds, robust o sets, or production architectural panelization.
- 1 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessBranch-Aware Pythagorean Fields for Feature-Sensitive Mesh Envelope Design and Seam-Network Initialization
This paper studies a branch-aware Pythagorean residual eld as a conservative exploratory primitive for mesh-envelope visualization and seam-network initialization. The eld is not a signed-distance eld, not an o set method, and not a fabrication-ready panelizer. Its purpose is narrower: to expose area-balance compatibility structures induced by local mesh triangles and to test whether these structures can provide candidate contour cues for later, constraintbased panel-layout tools. The formulation reduces each single-triangle branch relation to a quadratic equation in u = z2, applies an admissibility test to remove roots introduced by squaring, and evaluates a locality-gated residual eld over procedural benchmark meshes. The implementation reports point-triangle, point-plane, and curvature-proxy comparisons, raw benchmark metrics, threshold sensitivity, grid/scale/rotation/noise checks, and failure cases. The results show that the eld can generate branch-compatible contour networks that di er from ordinary proximity elds, but they also expose signi cant limitations: nonnearest-triangle responses, coordinate-frame dependence of the projected query layer, meshresolution sensitivity, and lack of manufacturability guarantees. The contribution should therefore be read as a reproducible geometry-processing diagnostic and seam-initialization study, not as a validated replacement for distance elds, robust o sets, or production architectural panelization.
- Open Access
- 3 April 2026
- Published by Elsevier in SSRN Electronic Journal
Explicit Branch Surfaces for Bernardes’ 3D Pythagorean Construction and a Branch-Aware Compatibility Field over Triangle Meshes
Bernardes' three-dimensional extension of the Pythagorean theorem replaces the classical balance of squared side lengths by a family of area-balance relations generated from a base triangle and an auxiliary spatial point. The original construction is implicit. In this paper, each of the three Bernardes branches is reduced explicitly to a quadratic equation in the variable $u=z^2$, yielding a branch-wise sheet representation together with admissibility conditions required to remove spurious algebraic roots. This explicit reformulation reveals that the construction is intrinsically multi-sheeted, branch-dependent, and symmetric with respect to the reflection $z\mapsto -z$. We then define a branch-aware scalar field over triangulated surfaces. The field itself is evaluated from the original Bernardes residuals rather than by solving the quadratic sheets at every query point. That choice is deliberate: the explicit reduction provides analytic structure, clarifies admissibility and degeneracy, and supports visualization of the sheet geometry, while the residual definition supplies a simple branch-symmetric quantity on arbitrary 3D meshes. Numerical experiments on a reference STL-like patch and on isolated triangles show that the field is not a Euclidean distance surrogate. Instead, it behaves as a geometry-aware compatibility descriptor whose strongest alignment with point-to-triangle distance occurs over a finite offset band and whose spatial patterns respond to triangle anisotropy. The contribution of the paper is therefore threefold: an explicit branch-wise formulation of Bernardes' 3D construction, a clarified residual-based field interpretation over triangle meshes, and a reproducible reference implementation with documented numerical settings and figure-generation scripts. The method is positioned as complementary to signed-distance and level-set formulations rather than as a replacement for either.- 3 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessExplicit Branch Surfaces for Bernardes’ 3D Pythagorean Construction and a Branch-Aware Compatibility Field over Triangle Meshes
Bernardes' three-dimensional extension of the Pythagorean theorem replaces the classical balance of squared side lengths by a family of area-balance relations generated from a base triangle and an auxiliary spatial point. The original construction is implicit. In this paper, each of the three Bernardes branches is reduced explicitly to a quadratic equation in the variable $u=z^2$, yielding a branch-wise sheet representation together with admissibility conditions required to remove spurious algebraic roots. This explicit reformulation reveals that the construction is intrinsically multi-sheeted, branch-dependent, and symmetric with respect to the reflection $z\mapsto -z$. We then define a branch-aware scalar field over triangulated surfaces. The field itself is evaluated from the original Bernardes residuals rather than by solving the quadratic sheets at every query point. That choice is deliberate: the explicit reduction provides analytic structure, clarifies admissibility and degeneracy, and supports visualization of the sheet geometry, while the residual definition supplies a simple branch-symmetric quantity on arbitrary 3D meshes. Numerical experiments on a reference STL-like patch and on isolated triangles show that the field is not a Euclidean distance surrogate. Instead, it behaves as a geometry-aware compatibility descriptor whose strongest alignment with point-to-triangle distance occurs over a finite offset band and whose spatial patterns respond to triangle anisotropy. The contribution of the paper is therefore threefold: an explicit branch-wise formulation of Bernardes' 3D construction, a clarified residual-based field interpretation over triangle meshes, and a reproducible reference implementation with documented numerical settings and figure-generation scripts. The method is positioned as complementary to signed-distance and level-set formulations rather than as a replacement for either. - Open Access
- 2 April 2026
- Published by Elsevier in SSRN Electronic Journal
Task-Based Information Horizons in Inverse Heat Conduction: A Probabilistic Framework for Sensor Placement
This paper studies a steady inverse heat conduction problem from a task-based rather than a purely residual-based viewpoint. The practical question is not merely whether noisy interior temperatures can be fit, but how deep a sensor can be placed while still allowing reliable discrimination of two nearby boundary heat sources once regularization is imposed. We consider a linear forward map on a square domain with an isothermal lower boundary, recover the unknown top-boundary flux by adjoint-based inversion, and control the descent geometry through fractional Sobolev metric smoothing. Recoverability is quantified by the probability of resolving a two-source boundary pattern over repeated noisy trials, then summarized by information horizons $H_{10}$, $H_{50}$, $H_{90}$ and by the transition width $|H_{90}-H_{10}|$. The resulting phase diagrams show that recoverability is strongly surface-limited in the present reference configuration: for the case $\sigma=0.01$ and $d=0.25$, $H_{50}$ remains near $0.025$ for $s\le 0.5$, drops sharply at $s=0.6$, and vanishes for $s\ge 0.7$. Increasing source separation delays this collapse, whereas stronger Sobolev smoothing accelerates it. A modal attenuation coordinate combines depth and metric smoothing into a single spectral design variable and explains the observed collapse. The main contribution is therefore a reproducible framework that converts regularization strength into an experimentally interpretable sensor-placement constraint for a specified inverse task.- 2 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTask-Based Information Horizons in Inverse Heat Conduction: A Probabilistic Framework for Sensor Placement
This paper studies a steady inverse heat conduction problem from a task-based rather than a purely residual-based viewpoint. The practical question is not merely whether noisy interior temperatures can be fit, but how deep a sensor can be placed while still allowing reliable discrimination of two nearby boundary heat sources once regularization is imposed. We consider a linear forward map on a square domain with an isothermal lower boundary, recover the unknown top-boundary flux by adjoint-based inversion, and control the descent geometry through fractional Sobolev metric smoothing. Recoverability is quantified by the probability of resolving a two-source boundary pattern over repeated noisy trials, then summarized by information horizons $H_{10}$, $H_{50}$, $H_{90}$ and by the transition width $|H_{90}-H_{10}|$. The resulting phase diagrams show that recoverability is strongly surface-limited in the present reference configuration: for the case $\sigma=0.01$ and $d=0.25$, $H_{50}$ remains near $0.025$ for $s\le 0.5$, drops sharply at $s=0.6$, and vanishes for $s\ge 0.7$. Increasing source separation delays this collapse, whereas stronger Sobolev smoothing accelerates it. A modal attenuation coordinate combines depth and metric smoothing into a single spectral design variable and explains the observed collapse. The main contribution is therefore a reproducible framework that converts regularization strength into an experimentally interpretable sensor-placement constraint for a specified inverse task. - Open Access
- 28 March 2026
- Published by Elsevier in SSRN Electronic Journal
Conservative Benchmark Assessment of Prior-Preserving CCPFM Variants for 2D Poisson Problems on Irregular Point Clouds
This paper studies constrained correction point-flux methods (CCPFM) for manufactured 2D Poisson problems on irregular point clouds. The benchmark uses four geometries---circle, rotated ellipse, star, and annulus---and compares a weighted least-squares (WLS) reference, a neighbor-adaptive normalized moving least-squares (NAMLS) operator, a frozen CCPFM baseline, an experimental live-boundary gated CCPFM branch, and sparse deep-core enrichment ablations within one common computational harness. The CCPFM branch is built around a prior-preserving constrained correction formulation on a fixed graph, together with diagnostics that separate near-boundary and core behavior, patch-test consistency, instability incidence, and added-edge activation. The main conclusion is conservative. The frozen CCPFM baseline remains the main recommended CCPFM method overall. The live-boundary gated branch exhibits geometry-sensitive behavior rather than uniform improvement: it is helpful on the star, dormant on the circle and ellipse, and not justified on the annulus at the finest tested spacing. Sparse deep-core enrichment yields an even narrower conclusion. A forced one-edge sparsecore branch acts as a proof of mechanism, showing that sparse enrichment can be inserted safely. A rarity-based selective sparsecore trigger activates occasionally without force-flagging, particularly on annulus and ellipse clouds, but its practical effect remains negligible in both all-run and activated-run-only views. The frozen CCPFM baseline remained the main recommended method overall, while live-boundary gating and sparse enrichment were retained as informative but secondary ablations.
- 28 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessConservative Benchmark Assessment of Prior-Preserving CCPFM Variants for 2D Poisson Problems on Irregular Point Clouds
This paper studies constrained correction point-flux methods (CCPFM) for manufactured 2D Poisson problems on irregular point clouds. The benchmark uses four geometries---circle, rotated ellipse, star, and annulus---and compares a weighted least-squares (WLS) reference, a neighbor-adaptive normalized moving least-squares (NAMLS) operator, a frozen CCPFM baseline, an experimental live-boundary gated CCPFM branch, and sparse deep-core enrichment ablations within one common computational harness. The CCPFM branch is built around a prior-preserving constrained correction formulation on a fixed graph, together with diagnostics that separate near-boundary and core behavior, patch-test consistency, instability incidence, and added-edge activation. The main conclusion is conservative. The frozen CCPFM baseline remains the main recommended CCPFM method overall. The live-boundary gated branch exhibits geometry-sensitive behavior rather than uniform improvement: it is helpful on the star, dormant on the circle and ellipse, and not justified on the annulus at the finest tested spacing. Sparse deep-core enrichment yields an even narrower conclusion. A forced one-edge sparsecore branch acts as a proof of mechanism, showing that sparse enrichment can be inserted safely. A rarity-based selective sparsecore trigger activates occasionally without force-flagging, particularly on annulus and ellipse clouds, but its practical effect remains negligible in both all-run and activated-run-only views. The frozen CCPFM baseline remained the main recommended method overall, while live-boundary gating and sparse enrichment were retained as informative but secondary ablations.
- Open Access
- 27 March 2025
- Published by MDPI AG in Preprints.org
Variational Formulation of the Heat Equation with Adjoint Fields: Bridging Dissipation and Action Principles
This paper presents a variational formulation of the heat equation using adjoint fields to embed dissipation. By constructing a Lagrangian density \( \mathcal{L}=\tilde{T} (\rho c_p \frac{\partial T}{\partial t} - k \nabla^2 T) - \frac{\gamma }{2} \tilde{T}^2 \), we derive coupled forward and backward-in-time dynamics for the temperature \( T \)) and the adjoint \( \tilde{T}\ \). The action \( S= \int \mathcal{L} \,dV dt \) exhibits dimensional consistency kg·m²/sa nd non-monotonic behavior with respect to the number of time steps \( M \). A numerical example of heat diffusion on a 1D rod demonstrates the role of the adjoint field in stabilizing solutions and quantifying sensitivity. This work bridges irreversible thermodynamics with variational mechanics, offering insights into controlled dissipation in thermal systems.- 27 March 2025
- Published by MDPI AG in Preprints.org
Open AccessVariational Formulation of the Heat Equation with Adjoint Fields: Bridging Dissipation and Action Principles
This paper presents a variational formulation of the heat equation using adjoint fields to embed dissipation. By constructing a Lagrangian density \( \mathcal{L}=\tilde{T} (\rho c_p \frac{\partial T}{\partial t} - k \nabla^2 T) - \frac{\gamma }{2} \tilde{T}^2 \), we derive coupled forward and backward-in-time dynamics for the temperature \( T \)) and the adjoint \( \tilde{T}\ \). The action \( S= \int \mathcal{L} \,dV dt \) exhibits dimensional consistency kg·m²/sa nd non-monotonic behavior with respect to the number of time steps \( M \). A numerical example of heat diffusion on a 1D rod demonstrates the role of the adjoint field in stabilizing solutions and quantifying sensitivity. This work bridges irreversible thermodynamics with variational mechanics, offering insights into controlled dissipation in thermal systems.