# Toward accurate RUL and SoH estimation using reinforced graph-based physics-informed neural networks enhanced with dynamic weights

Source: [Toward accurate RUL and SoH estimation using reinforced graph-based physics-informed neural networks enhanced with dynamic weights](<https://www.sota2.com/research/paper/68fa79f81e209131cea8054c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-02T05:00:51.557Z

Authors: Mohamadreza Akbari Pour, Ali Ghasemzadeh, Mohamad Ali Bijarchi, Mohammad Behshad Shafii

Published: 2025-07-13T19:49:12.000Z

## Abstract

Accurate estimation of Remaining Useful Life (RUL) and State of Health (SoH) is essential for reliable Prognostics and Health Management (PHM), supporting timely maintenance and dependable industrial operation. However, hybrid models that combine data-driven learning with physics-based regularization often rely on fixed loss weights and therefore lose accuracy when transferred across assets with different degradation behaviors. This study introduces Reinforced Graph-based Physics-informed Networks with Dynamic Weighting (RGPD), a unified framework for spatio-temporal degradation modeling and adaptive physics-guided regularization. Graph-based representation learning captures inter-sensor degradation structure, a Soft Actor-Critic (SAC) module refines latent features under noisy conditions, and a lightweight Q-learning policy adaptively balances monotonicity, smoothness, and latent-dynamics residual losses during training. The framework is evaluated on the C-MAPSS, PHM2012, and XJTU datasets, which represent engine, bearing, and battery degradation processes. Relative to the strongest compared baselines reported in the corresponding benchmark tables, RGPD improves average RMSE by up to 12 percent on PHM2012 and C-MAPSS, and reduces average MAPE by 20 percent on XJTU compared with the second-best reported model. Performance on these heterogeneous benchmarks further suggests the model's generalizability across degradation systems. The physics-informed component is implemented through degradation-consistent priors together with a Deep Hidden Physics Model-style residual, which improves physical plausibility without requiring a full first-principles model for each asset type.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [RUL Estimation on C-MAPSS FD001 (RMSE, Score)](<https://www.sota2.com/research/sota/rul-estimation-on-c-mapss-fd001-rmse-score.md>): RUL Estimation; C-MAPSS FD001 1.0; RMSE: 10.77.

- [RUL Estimation on C-MAPSS FD002 1.0 (RMSE, Score)](<https://www.sota2.com/research/sota/rul-estimation-on-c-mapss-fd002-1-0-rmse-score.md>): RUL Estimation; C-MAPSS FD002 1.0; RMSE: 11.58.

- [RUL Estimation on C-MAPSS FD003 1.0 (RMSE, Score)](<https://www.sota2.com/research/sota/rul-estimation-on-c-mapss-fd003-1-0-rmse-score.md>): RUL Estimation; C-MAPSS FD003 1.0; RMSE: 10.78.

- [RUL Estimation on C-MAPSS FD004 1.0 (RMSE, Score)](<https://www.sota2.com/research/sota/rul-estimation-on-c-mapss-fd004-1-0-rmse-score.md>): RUL Estimation; C-MAPSS FD004 1.0; RMSE: 12.38.

- [RUL Estimation on C-MAPSS Average 1.0 (RMSE, Score)](<https://www.sota2.com/research/sota/rul-estimation-on-c-mapss-average-1-0-rmse-score.md>): RUL Estimation; C-MAPSS Average 1.0; RMSE: 11.38.

- [SoH estimation on XJTU Batch 6](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-6.md>): SOH estimation; XJTU Batch 6; MAPE: 0.0055.

- [Remaining Useful Life Prediction on PHM 2012 (test)](<https://www.sota2.com/research/sota/remaining-useful-life-prediction-on-phm-2012-test.md>): Remaining Useful Life prediction; PHM 2012 (test); RMSE: 0.0778.

- [SoH estimation on XJTU Batch 1](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-1.md>): SOH estimation; XJTU Batch 1; MAPE: 0.46.

- [SoH estimation on XJTU Batch 2](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-2.md>): SOH estimation; XJTU Batch 2; MAPE: 0.86.

- [SoH estimation on XJTU Batch 3](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-3.md>): SOH estimation; XJTU Batch 3; MAPE: 0.8.

- [SoH estimation on XJTU Batch 4](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-4.md>): SOH estimation; XJTU Batch 4; MAPE: 1.11.

- [SoH estimation on XJTU Batch 5](<https://www.sota2.com/research/sota/soh-estimation-on-xjtu-batch-5.md>): SOH estimation; XJTU Batch 5; MAPE: 1.37.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2507.09766>)

[PDF](<https://arxiv.org/pdf/2507.09766.pdf>)
