Search NASA⌕ Search

SEARCH · Search NASA

Results for “dynamic baselines”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,171 records · Page 65

Intercomparison of cloud model simulations of Arctic mixed-phase boundary layer clouds observed during SHEBA/FIRE-ACE

An intercomparison of six cloud-resolving and large-eddy simulation models is presented. This case study is based on observations of a persistent mixed-phase boundary layer cloud gathered on 7 May, 1998 from the Surface Heat Budget of Arctic Ocean (SHEBA) and First ISCCP Regional Experiment - Arctic Cloud Experiment (FIRE-ACE). Ice nucleation is constrained in the simulations in a way that holds the ice crystal concentration approximately fixed, with two sets of sensitivity runs in addition to the baseline simulations utilizing different specified ice nucleus (IN) concentrations. All of the baseline and sensitivity simulations group into two distinct quasi-steady states associated with either persistent mixed-phase clouds or all-ice clouds after the first few hours of integration, implying the existence of multiple equilibria. These two states are associated with distinctly different microphysical, thermodynamic, and radiative characteristics. Most but not all of the models produce a persistent mixed-phase cloud qualitatively similar to observations using the baseline IN/crystal concentration, while small increases in the IN/crystal concentration generally lead to rapid glaciation and conversion to the all-ice state. Budget analysis indicates that larger ice deposition rates associated with increased IN/crystal concentrations have a limited direct impact on dissipation of liquid in these simulations. However, the impact of increased ice deposition is greatly enhanced by several interaction pathways that lead to an increased surface precipitation flux, weaker cloud top radiative cooling and cloud dynamics, and reduced vertical mixing, promoting rapid glaciation of the mixed-phase cloud for deposition rates in the cloud layer greater than about 1-2x10-5 g kg-1 s-1. These results indicate the critical importance of precipitation-radiative-dynamical interactions in simulating cloud phase, which have been neglected in previous fixed-dynamical parcel studies of the cloud phase parameter space. Large sensitivity to the IN/crystal concentration also suggests the need for improved understanding of ice nucleation and its parameterization in models.

54 ENVIRONMENTAL SCIENCES↗

Molecular diversity of dissolved organic matter reflects macroecological patterns in river networks

Deciphering dissolved organic matter (DOM) molecular complexity is crucial for understanding ecosystem function. Using the continental-scale Worldwide Hydrobiogeochemistry Observation Network for Dynamic Rivers Systems (WHONDRS) Fourier-transform ion cyclotron resonance mass spectrometry (FTICR-MS) dataset, we reveal fundamental scaling patterns of DOM chemodiversity with watershed characteristics. Analysis of 54 river sites shows local and regional watershed features significantly influence DOM chemodiversity (2500–8718 unique formulae), exhibiting consistent scaling patterns across compound classes and a novel latitudinal gradient (decreasing diversity with increasing latitude). Scaling relationships for DOM composition vary by compound class. Crucially, the scaling parameters (B, baseline chemodiversity; Z, sensitivity) are linearly interrelated. This B–Z relationship is most robust for potentially bio-labile carbohydrates (coefficient of determination R 2 ≈ 0.85), diminishing for recalcitrant, plant-derived molecules (such as lignin), and indicates (potential) biolability-dependent coupling between baseline diversity and environmental responsiveness. These quantitative scaling relationships, with scaling exponents ranging from − 2.1 to 2.2 across compound classes, enable prediction of DOM composition across watersheds, offering a framework to understand ecosystem responses to environmental change. This research bridges biogeochemistry and ecology, providing tools to anticipate molecular transformations across scales.

59 BASIC BIOLOGICAL SCIENCES↗

Examining the Impacts of Great Lakes Temperature Perturbations on Simulated Precipitation in the Northeastern United States

Most inland water bodies are not resolved by general circulation models, requiring that lake surface temperatures be estimated. Given the large spatial and temporal variability of the surface temperatures of the North American Great Lakes, such estimations can introduce errors when used as lower boundary conditions for dynamical downscaling. Lake surface temperatures (LSTs) influence moisture and heat fluxes, thus impacting precipitation within the immediate region and potentially in regions downwind of the lakes. For this study, the Advanced Research version of the Weather Research and Forecasting Model (WRF-ARW) was used to simulate precipitation over the six New England states during a 5-yr historical period. The model simulation was repeated with perturbed LSTs, ranging from 10°C below to 10°C above baseline values obtained from reanalysis data, to determine whether the inclusion of erroneous LST values has an impact on simulated precipitation and synoptic-scale features. Results show that simulated precipitation in New England is statistically correlated with LST perturbations, but this region falls on a wet–dry line of a larger bimodal distribution. Wetter conditions occur to the north and drier conditions occur to the south with increasing LSTs, particularly during the warm season. Additionally, the precipitation differences coincide with large-scale anomalous temperature, pressure, and moisture patterns. Care must therefore be taken to ensure reasonably accurate Great Lakes surface temperatures when simulating precipitation, especially in southeastern Canada, Maine, and the mid-Atlantic region.

54 ENVIRONMENTAL SCIENCES↗

The Discovery of the Long-Period, Eccentric Planet Kepler-88 d and System Characterization with Radial Velocities and Photodynamical Analysis

We present the discovery of Kepler-88 d (P{sub d} = 1403±14 days, Msini{sub d}=965±44 M{sub ⊕}=3.04±0.13 M{sub J}, e{sub d}=0.41±0.03) based on six years of radial velocity (RV) follow-up from the W. M. Keck Observatory High Resolution Echelle Spectrometer spectrograph. Kepler-88 has two previously identified planets. Kepler-88 b (KOI-142.01) transits in the NASA Kepler photometry and has very large transit timing variations (TTVs). Nesvorný et al. performed a dynamical analysis of the TTVs to uniquely identify the orbital period and mass of the perturbing planet (Kepler-88 c), which was later was confirmed with RVs from the Observatoire de Haute-Provence (OHP). To fully explore the architecture of this system, we performed photodynamical modeling on the Kepler photometry combined with the RVs from Keck and OHP and stellar parameters from spectroscopy and Gaia. Planet d is not detectable in the photometry, and long-baseline RVs are needed to ascertain its presence. A photodynamical model simultaneously optimized to fit the RVs and Kepler photometry yields the most precise planet masses and orbital properties yet for b and c: P{sub b}=10.91647±0.00014 days, M{sub b}=9.5±1.2 M{sub ⊕}, P{sub c}=22.2649±0.0007 days, and M{sub c}=214.0±5.3 M{sub ⊕}. The photodynamical solution also finds that planets b and c have low eccentricites and low mutual inclination, are apsidally anti-aligned, and have conjunctions on the same hemisphere of the star. Continued RV follow-up of systems with small planets will improve our understanding of the link between inner planetary system architectures and giant planets.

79 ASTRONOMY AND ASTROPHYSICS↗

Advanced solar receiver conceptual design study

High temperature solar dynamic Brayton and Stirling receivers are investigated as candidate electrical power generating systems for future LEO missions. These receivers are smaller and more efficient than conventional receivers, and they offer less structural complexity and fewer thermal stress problems. Use of the advanced Direct Absorption Storage Receiver allows many of the problems associated with working with high-volumetric-change phase-change materials to be avoided. A specific mass reduction of about 1/3 with respect to the baseline receiver has been realized.

Kesseli, J. B.↗

Changes in plasma volume and baroreflex function following resistance exercise

The dynamics of change in plasma volume (PV) and baroreflex responses have been reported over 24 h immediately following maximal cycle exercise. The purpose of this study was to determine if PV and baroreflex showed similar changes for 24 h after resistance exercise. Eight men were studied on 2 test days, 1 week apart. On 1 day, per cent change (% delta) in PV was estimated at 0,3, and 6 h after resistance exercise using haematocrit and haemoglobin. Baseline PV was measured 24 h after exercise using Evans blue dye. The carotid baroreceptor-cardiac reflex response was measured before, and 3, 6, 9, 12, and 24 h post-exercise. Each subject performed six sets of the bench press and leg press with 10 repetitions per set with a load that induced failure within each set. On a control day, the protocol was used without exercise. Plasma volume did not change during the control day. There was a 20% decrease in PV immediately post-exercise; the recovery of the PV was rapid and complete within 3 h. PV was 20% greater 24 h post-exercise than on the control day. There were no differences in any of the baroreflex measurements. Therefore, it is suggested that PV shifts may occur without altering baroreflex sensitivity.

NASA Program Space Physiology and Countermeasures↗

Design of Experiments for Both Experimental and Analytical Study of Exhaust Plume Effects on Sonic Boom

Computational fluid dynamics (CFD) analysis has been performed to study the plume effects on sonic boom signature for isolated nozzle configurations. The objectives of these analyses were to provide comparison to past work using modern CFD analysis tools, to investigate the differences of high aspect ratio nozzles to circular (axisymmetric) nozzles, and to report the effects of under expanded nozzle operation on boom signature. CFD analysis was used to address the plume effects on sonic boom signature from a baseline exhaust nozzle. Nearfield pressure signatures were collected for nozzle pressure ratios (NPRs) between 6 and 10. A computer code was used to extrapolate these signatures to a ground-observed sonic boom N-wave. Trends show that there is a reduction in sonic boom N-wave signature as NPR is increased from 6 to 10. As low boom designs are developed and improved, there will be a need for understanding the interaction between the aircraft boat tail shocks and the exhaust nozzle plume. These CFD analyses will provide a baseline study for future analysis efforts. For further study, a design of experiments has been conducted to develop a hybrid method where both CFD and small scale wind tunnel testing will validate the observed trends. The CFD and testing will be used to screen a number of factors which are important to low boom propulsion integration, including boat tail angle, nozzle geometry, and the effect of spacing and stagger on nozzle pairs. To design the wind tunnel experiment, CFD was instrumental in developing a model which would provide adequate space to observe the nozzle and boat tail shock structure without interference from the wind tunnel walls.

Castner, Raymond S.↗

Single Episode Policy Transfer in Reinforcement Learning

Transfer and adaptation to new unknown environmental dynamics is a key challenge for reinforcement learning (RL). An even greater challenge is performing near-optimally in a single attempt at test time, possibly without access to dense rewards, which is not addressed by current methods that require multiple experience rollouts for adaptation. To achieve single episode transfer in a family of environments with related dynamics, we propose a general algorithm that optimizes a probe and an inference model to rapidly estimate underlying latent variables of test dynamics, which are then immediately used as input to a universal control policy. This modular approach enables integration of state-of-the-art algorithms for variational inference or RL. Moreover, our approach does not require access to rewards at test time, allowing it to perform in settings where existing adaptive approaches cannot. In diverse experimental domains with a single episode test constraint, we report our method significantly outperforms existing adaptive approaches and shows favorable performance against baselines for robust transfer.

42 ENGINEERING↗

Simulating many-engine spacecraft: Exceeding 1 quadrillion degrees of freedom via information geometric regularization

We present an optimized implementation of the recently proposed information geometric regularization (IGR) for unprecedented scale simulation of compressible fluid flows applied to multi-engine spacecraft boosters. We improve upon state-of-the-art computational fluid dynamics (CFD) techniques in terms of computational cost, memory footprint, and energy-to-solution metrics. Unified memory on coupled CPU–GPU or APU platforms increases problem size with negligible overhead. Mixed half/single-precision storage and computation are used on well-conditioned numerics. We simulate flow at 200 trillion grid points and 1 quadrillion degrees of freedom, exceeding the current record by a factor of 20. A factor of 4 wall-time speedup is achieved over optimized baselines. Ideal weak scaling is observed on OLCF Frontier, LLNL El Capitan, and CSCS Alps using the full systems. Strong scaling is near ideal at extreme conditions, including 80% efficiency on CSCS Alps with an 8 node baseline and stretching to the full system.

Wilfong, Benjamin [Georgia Institute of Technology↗

Machine Learning Assisted Safety Modeling and Analysis of Advanced Reactors

With the advances in computational power and numerical methods, analysts can now rely on first-principle simulations to predict ultra-fine details in a variety of applications. Advances in machine learning (ML) have produced algorithms that can now learn high-level abstractions via hierarchical models. This project aims to leverage advances in ML techniques and the available high-resolution simulation data to develop a novel modeling and simulation (M\&S) methodology for reactor safety analysis. While application-agnostic ML techniques are available, complex physics constraints need to be incorporated into ML techniques to build ML-based closures for computationally efficient predictive simulations. This project intends to develop a physics-guided data-driven multi-scale methodology for M\&S of advanced reactors. The project focuses on thermal fluid (T/F) phenomena, which play major roles in advanced reactor safety. Specifically, we propose a data-driven coarse-mesh turbulence model based on local flow features for the transient analysis of thermal mixing and stratification in a sodium-cooled fast reactor (SFR). The model has a coarse-mesh setup to ensure computational efficiency, while it is trained by fine-mesh computational fluid dynamics (CFD) data with Reynolds-averaged Navier-Stokes (RANS) turbulence model to ensure accuracy. Three different neural networks are developed and tested for loss-of-flow transients in the hot pool of SFR, i.e. the densely connected convolutional neural network (DCNN), long-short-term-memory network based on proper orthogonal decomposition (POD-LSTM), and the DCNN informed by LSTM (DCNN-LSTM). The performances of these three neural networks are evaluated based on baseline models. The DCNN-LSTM model has been chosen for further hyperparameter optimization. Furthermore, based on a simplified two-dimensional case, uncertainty quantification (UQ) of the developed ML-based closure are investigated with three methods, i.e. Monte Carlo dropout, deep ensemble, and Bayesian neural network. The developed ML-based turbulent viscosity closure relation based on deep ensemble is then integrated into the system analysis module SAM and serves as a term in the conservation equations. Such a SAM-ML based procedure guarantees that the obtained results are consistent with the physical constraints of the thermal-fluid system. The SAM-ML simulation on the same loss-of-flow transient showed comparable accuracy with the CFD simulation but with a much coarser mesh setup. Last but not least, the ML-based closure improvement with the support of higher-fidelity data from large eddy simulation (LES) is discussed. As a first step towards this direction, a baseline LES simulation is performed to obtain comparable data with RANS results. Based on the early results, future investigation on further improving the ML-based closure is discussed. We believe the developed approach that combines scientific machine learning with nuclear system analysis code can benefit the advanced reactor community as more accurate safety analyses will better characterize reactor safety margins and reduce licensing efforts.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

OctoFAS: A Two-Level Fair Scheduler That Increases Fairness in Network-Based Key-Value Storage

We identified a fairness problem in a network-based key-value storage system using Intel Storage Performance Development Kit (SPDK) in a multitenant environment. In such an environment, each tenant’s I/O service rate is not fairly guaranteed compared to that of other tenants. To address the fairness problem, we propose OctoFAS, a two-level fair scheduler designed to improve overall throughput and fairness among tenants. The two-level scheduler of OctoFAS consists of (i) inter-core scheduling and (ii) intra-core scheduling. Through inter-core scheduling, OctoFAS addresses the load imbalance problem that is inherent in SPDK on the storage server by dynamically migrating I/O requests from overloaded cores to underloaded cores, thereby increasing overall throughput. Intra-core scheduling prioritizes handling requests from starving tenants over well-fed tenants within core-specific event queues to ensure fair I/O services among multiple tenants. OctoFAS is deployed on a Linux cluster with SPDK. Through extensive evaluations, we found that OctoFAS ensures that the total system throughput remains high and balanced, while enhancing fairness by approximately 10% compared to the baseline, when both scheduling levels operate in a hybrid fashion.

97 MATHEMATICS AND COMPUTING↗

Definition and preliminary design of the Laser Atmospheric Wind Sounder (LAWS) phase 1. Volume 2

The steps and engineering trades and analyses used in establishing the initial requirements and in developing a concept and configuration for the laser atmospheric wind sounder (LAWS) instrument. A summary of the performance anticipated from the baseline configuration, and a bibliography are presented. LAWS, which is a facility instrument of the Earth observing system (EOS), is the culmination of over 20 years of effort in the field of laser Doppler wind sensing and will be the first instrument to fly in space capable of providing global-scale tropospheric wind profiles at high spatial resolutions. Global-scale wind profiles are necessary for: (1) more accurate diagnosis of large-scale circulation and climate dynamics; (2) improved numerical weather prediction; (3) improved understanding of mesoscale systems; and (4) improved understanding of global biogeochemical and hydrologic cycles. The objective of phase 1 was to define and perform a preliminary design for the LAWS instrument. The definition phase consisted of identifying realistic concepts for LAWS and analyzing them in sufficient detail to be able to choose the most promising one for the LAWS instrument. Systems and subsystems configurations were then developed for the chosen concept. The concept and subsequent configuration were to be compatible with two prospective platforms: the Japanese polar orbiting platform (JPOP) and as an attached payload on the Space Station Freedom. After a thorough and objective concept selection process, a heterodyne detection Doppler lidar using a CO2 laser transmitter operating a 9.1 micron over a 2.1 micron solid state system was chosen. A configuration for LAWS that meets the performance requirements was designed at the conclusion of phase 1.

Source record↗

The Strategy for the Second Phase of Aerobraking Mars Global Surveyor

On February 19, 1999, the Mars Global Surveyor (MGS) spacecraft was able to propulsively establish its mapping orbit. This event followed the completion of the second phase of aerobraking for the MGS spacecraft on February 4, 1999. For the first time, a spacecraft at Mars had successfully employed aerobraking methods in order to reach its desired pre-launch mapping orbit. This was accomplished despite a damaged spacecraft solar array. The MGS spacecraft was launched on November 7, 1996, and after a ten month interplanetary transit was inserted into a highly elliptical capture orbit at Mars on September 12, 1997. Unlike other interplanetary missions, the MGS spacecraft was launched with a planned mission delta-V ((Delta)V) deficit of nearly 1250 m/s. To overcome this AV deficit, aerobraking techniques were employed. However, damage discovered to one of the spacecraft's two solar arrays after launch forced major revisions to the original aerobraking planning of the MGS mission. In order to avoid a complete structural failure of the array, peak dynamic pressure levels for the spacecraft were established at a major spacecraft health review in November 1997. These peak dynamic pressure levels were roughly one-third of the original mission design values. Incorporating the new dynamic pressure limitations into mission replanning efforts resulted in an 'extended' orbit insertion phase for the mission. This 'extended' orbit insertion phase was characterized by two distinct periods of aerobraking separated by an aerobraking hiatus that would last for several months in an intermediate orbit called the "Science Phasing Orbit" (SPO). This paper describes and focuses on the strategy for the second phase of aerobraking for the MGS mission called "Aerobraking Phase 2." This description will include the baseline aerobraking flight profile, the trajectory control methodology, as well as the key trajectory metrics that were monitored in order to successfully "guide' the spacecraft to its desired mapping orbit. Additionally, the actual aerobraking progress is contrasted to the planned aerobraking flight profile. (A separate paper will describe the navigation aspects of MGS aerobraking in detail.) Key to the success of the MGS mission is the delivery of the spacecraft to its final mapping orbit and the synergy the instrument complement provides to its scientific investigators when science data is returned from that orbit. The MGS mapping orbit is characterized as a low altitude, near-circular, near-polar orbit that is Sun-synchronous with the descending equatorial crossing at 2:00 AM local mean solar time (LMST).

Johnston, M. D.↗

The Stellar Imager (SI) Mission Concept

The Stellar Imager (SI) is envisioned as a space-based, UV-optical interferometer composed of 10 or more one-meter class elements distributed with a maximum baseline of 0.5 km. It is designed to image stars and binaries with sufficient resolution to enable long-term studies of stellar magnetic activity patterns, for comparison with those on the sun. It will also support asteroseismology (acoustic imaging) to probe stellar internal structure, differential rotation, and large-scale circulations. SI will enable us to understand the various effects of the magnetic fields of stars, the dynamos that generate these fields, and the internal structure and dynamics of the stars. The ultimate goal of the mission is to achieve the best-possible forecasting of solar activity as a driver of climate and space weather on time scales ranging from months up to decades, and an understanding of the impact of stellar magnetic activity on life in the Universe. In this paper we describe the scientific goals of the mission, the performance requirements needed to address these goals, the "enabling technology" development efforts being pursued, and the design concepts now under study for the full mission and a possible pathfinder mission.

Carpenter, Kenneth G.↗

Multi-Terrain Earth Landing Systems Applicable for Manned Space Capsules

A key element of the President's Vision for Space Exploration is the development of a new space transportation system to replace the Shuttle that will enable manned exploration of the moon, Mars, and beyond. NASA has tasked the Constellation Program with the development of this architecture, which includes the Ares launch vehicle and Orion manned spacecraft. The Orion spacecraft must carry six astronauts and its primary structure should be reusable, if practical. These requirements led the Constellation Program to consider a baseline land landing on return to earth. To assess the landing system options for Orion, a review of current operational parachute landing systems such as those used for the F-111 escape module and the Soyuz is performed. In particular, landing systems with airbags and retrorockets that would enable reusability of the Orion capsule are investigated. In addition, Apollo tests and analyses conducted in the 1960's for both water and land landings are reviewed. Finally, tests and dynamic finite element simulations to understand land landings for the Orion spacecraft are also presented.

Fasanella, Edwin L.↗

All Recent Mars Landers Have Landed Downrange - Are Mars Atmosphere Models Mis-Predicting Density?

All recent Mars landers (Mars Pathfinder, the two Mars Exploration Rovers Spirit and Opportunity, and the Mars Phoenix Lander) have landed further downrange than their pre-entry predictions. Mars Pathfinder landed 27 km downrange of its prediction [1], Spirit and Opportunity landed 13.4 km and 14.9 km, respectively, downrange from their predictions [2], and Phoenix landed 21 km downrange from its prediction [3]. Reconstruction of their entries revealed a lower density profile than the best a priori atmospheric model predictions. Do these results suggest that there is a systemic issue in present Mars atmosphere models that predict a higher density than observed on landing day? Spirit Landing: The landing location for Spirit was 13.4 km downrange of the prediction as shown in Fig. 1. The navigation errors upon Mars arrival were very small [2]. As such, the entry interface conditions were not responsible for this downrange landing. Consequently, experiencing a lower density during the entry was the underlying cause. The reconstructed density profile that Spirit experienced is shown in Fig. 2, which is plotted as a fraction of the pre-entry baseline prediction that was used for all the entry, descent, and landing (EDL) design analyses. The reconstructed density is observed to be less dense throughout the descent reaching a maximum reduction of 15% at 21 km. This lower density corresponded to approximately a 1- low profile relative to the dispersions predicted. Nearly all the deceleration during the entry occurs within 10- 50 km. As such, prediction of density within this altitude band is most critical for entry flight dynamics analyses and design (e.g., aerodynamic and aerothermodynamic predictions, landing location, etc.).

Desai, Prasun N.↗

Multi-Terrain Earth Landing Systems Applicable for Manned Space Capsules

A key element of the President's Vision for Space Exploration is the development of a new space transportation system to replace Shuttle that will enable manned exploration of the moon, Mars, and beyond. NASA has tasked the Constellation Program with the development of this architecture, which includes the Ares launch vehicle and Orion manned spacecraft. The Orion spacecraft must carry six astronauts and its primary structure should be reusable, if practical. These requirements led the Constellation Program to consider a baseline land landing on return to earth. To assess the landing system options for Orion, a review of current operational parachute landing systems such as those used for the F-111 escape module and the Soyuz is performed. In particular, landing systems with airbags and retrorockets that would enable reusability of the Orion capsule are investigated. In addition, Apollo tests and analyses conducted in the 1960's for both water and land landings are reviewed. Finally, tests and dynamic finite element simulations to understand land landings for the Orion spacecraft are also presented.

Fasanella, Edwin L.↗

LC-Opt: Benchmarking Reinforcement Learning and Agentic AI for End-to-End Liquid Cooling Optimization in Data Centers

Liquid cooling is critical for thermal management in high-density data centers with the rising AI workloads. However, machine learning-based controllers are essential to unlock greater energy efficiency and reliability, promoting sustainability. We present LC-Opt, a Sustainable Liquid Cooling (LC) benchmark environment, for reinforcement learning (RL) control strategies in energy-efficient liquid cooling of high-performance computing (HPC) systems. Built on the baseline of a high-fidelity digital twin of Oak Ridge National Lab's Frontier Supercomputer cooling system, LC-Opt provides detailed Modelica-based end-to-end models spanning site-level cooling towers to data center cabinets and server blade groups. RL agents optimize critical thermal controls like liquid supply temperature, flow rate, and granular valve actuation at the IT cabinet level, as well as cooling tower (CT) setpoints through a Gymnasium interface, with dynamic changes in workloads. This environment creates a multi-objective real-time optimization challenge balancing local thermal regulation and global energy efficiency, and also supports additional components like a heat recovery unit (HRU). We benchmark centralized and decentralized multi-agent RL approaches, demonstrate policy distillation into decision and regression trees for interpretable control, and explore LLM-based methods that explain control actions in natural language through an agentic mesh architecture designed to foster user trust and simplify system management. LC-Opt democratizes access to detailed, customizable liquid cooling models, enabling the ML community, operators, and vendors to develop sustainable data center liquid cooling control solutions.

Naug, Avisek [Hewlett Packard Enterprise]↗