AI & ML
Discontinuity-Aware Solvers for Vapor-Compression Cycles
arXiv cs.NA0 views
arXiv:2611.20177v2 Announce Type: replace-cross Abstract: Physics-based models of vapor-compression cycles are often formulated as large, stiff DAE systems with derivative discontinuities introduced at the points of phase change. We use an implementation of a discontinuity-aware ODE and DAE solver scheme that uses domain-specific, user-declared information to compute appropriate time steps for more efficient integration. This scheme is made open-source within Julia's DifferentialEquations package. We demonstrate how discontinuity information can be specified symbolically in a declarative modeling language such as Dyad and passed to the solver to address the numerical challenges posed. We apply this discontinuity-aware technique to two applications: first, a scalar ODE with two discontinuities, and then a model of a full vapor-compression cycle. The discontinuity-aware solver yields significantly fewer step rejections and improved solver times, particularly with the Rodas5P solver at loose tolerances, as compared to standard error control, suggesting that exposing physical discontinuity information from equation-oriented models to adaptive solvers can improve practical simulation performance without requiring reassembly of the model.
Source:https://arxiv.org/abs/2611.20177PreviousDirect Numerical Simulation of Thermoacoustically Unstable Flames Via Non-Drifting Acoustic Delay Characteristic Boundary ConditionsNextOfficePrism: Benchmarking Agents on Single-Domain Operations Across Enterprise Business Functions
Related posts
How OpenAI scales PostgreSQL to 800 million users
0dev.to Database
Spec-Driven PBT: Deriving Semantic Invariants from Code Specifications for Agentic Property-based Testing
0arXiv cs.SE
A Grey Literature Review of Graduate Students' Research Practices and Skills in the LLM Era
0arXiv cs.SE
ReBoT-MiCS: Reliability-Aware Bounding Time for Enhancing Timing Behavior of Embedded Mixed-Criticality Systems
0arXiv cs.SE
VoxelSBFL: Voxel-Level Spectrum-Based Failure Prediction for Vision-Language-Action Models
0arXiv cs.SE
Memorization capacity of deep ReLU neural networks characterized by width and depth
0arXiv cs.NA