TY - JOUR

T1 - MultiPINN: A Modular Framework for Physics-Informed Neural Networks

AU - Tsgoev, Chermen

AU - Kalinin, Vladislav

AU - Bratenkov, Miron

AU - Seredkin, Alexander

AU - Sakharov, Danil

AU - Travnikov, Vlad

AU - Kasyanov, Danil

AU - Matushkin, Mikhail

AU - Nekriach, Alexander

AU - Fomichev, Dmitriy

AU - Mullyadzhanov, Rustam

PY - 2026

Y1 - 2026

N2 - Physics-informed neural networks (PINNs) approximate solutions of differential equations by embedding governing equations and boundary conditions into the training objective. Controlled PINN comparisons are difficult because geometry representation, collocation sampling, loss construction, neural architecture, and optimization settings can each affect the result. This paper presents MultiPINN, a modular PyTorch framework for component-level PINN research. MultiPINN decomposes the PINN workflow into composable geometries, boundary entities, physical conditions, point generators, neural models, loss terms, and training services. It supports analytical and mesh-based geometries, constructive operations, boundary-shell abstractions with normal-vector handling, condition-specific non-adaptive and residual-adaptive sampling, loss regularization, mixed-precision training, gradient accumulation, distributed multi-GPU execution, and callback-based monitoring. We evaluate the framework on representative benchmarks covering geometry handling, sampling strategies, neural architectures, three-dimensional nonlinear simulations, and computational performance. The results show that MultiPINN enables controlled studies of individual PINN components while preserving a unified implementation workflow. We also compare MultiPINN with existing PINN frameworks and position it as a compact research environment for experiments with complex geometries and configurable training pipelines.

AB - Physics-informed neural networks (PINNs) approximate solutions of differential equations by embedding governing equations and boundary conditions into the training objective. Controlled PINN comparisons are difficult because geometry representation, collocation sampling, loss construction, neural architecture, and optimization settings can each affect the result. This paper presents MultiPINN, a modular PyTorch framework for component-level PINN research. MultiPINN decomposes the PINN workflow into composable geometries, boundary entities, physical conditions, point generators, neural models, loss terms, and training services. It supports analytical and mesh-based geometries, constructive operations, boundary-shell abstractions with normal-vector handling, condition-specific non-adaptive and residual-adaptive sampling, loss regularization, mixed-precision training, gradient accumulation, distributed multi-GPU execution, and callback-based monitoring. We evaluate the framework on representative benchmarks covering geometry handling, sampling strategies, neural architectures, three-dimensional nonlinear simulations, and computational performance. The results show that MultiPINN enables controlled studies of individual PINN components while preserving a unified implementation workflow. We also compare MultiPINN with existing PINN frameworks and position it as a compact research environment for experiments with complex geometries and configurable training pipelines.

UR - https://www.scopus.com/pages/publications/105054875666

UR - https://www.mendeley.com/catalogue/0fc02489-c824-3572-ab3c-21626e839f20/

U2 - 10.3390/make8100340

DO - 10.3390/make8100340

M3 - Article

VL - 8

SP - 340

JO - Machine Learning and Knowledge Extraction

JF - Machine Learning and Knowledge Extraction

SN - 2504-4990

IS - 10

ER -