Lie Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Lie Xu has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Daxiong Ji
Zhejiang University
5 shared publications - Birui Wu
School of Electronics and Information Engineering, Hangzhou Dianzi University, Hangzhou, Zhejiang 310018, China
4 shared publications - Marcelo H. Ang
Department of Mechanical Engineering, National University of Singapore, Singapore
3 shared publications Birui Wu
2 shared publications- Wenle Lan2 shared publications
- Yan Zhi Tan
National University of Singapore
2 shared publications
Latest Publications
- 18 November 2026
- Published by Elsevier in Journal of Energy Storage
- Vol. 183
Latent flow-matching regularization for lithium-ion battery state-of-health estimation under dynamic operating conditions
- 18 November 2026
- Published by Elsevier in Journal of Energy Storage
- Vol. 183
Latent flow-matching regularization for lithium-ion battery state-of-health estimation under dynamic operating conditions
- Open Access
- 13 June 2026
- Published by Elsevier in SSRN Electronic Journal
A Generative Latent Flow Matching Framework for Robust Battery Health Prognosis Under Dynamic Load Fluctuations
Accurate State of Health estimation is critical for the reliable operation and lifecycle management of lithium-ion batteries in modern energy storage systems. However, existing deterministic regression methods frequently struggle with statistical drift, unpredictable dynamic load fluctuations, and limited cross-domain generalization under highly heterogeneous energy dispatch conditions. To overcome these limitations, this paper proposes a novel Latent Flow Matching (LFM) framework that reformulates battery health prognosis as a physics-consistent generative transport problem. Acting as a continuous manifold learner governed by an ordinary differential equation, the LFM constraint maps complex operational profiles into a stable, continuous physical degradation manifold, effectively filtering out stochastic system-level disturbances. An enhanced local-aware encoder incorporating Group Normalization and channel-wise attention ensures operational feature robustness by isolating inter-sample interference during highly dynamic power demands. Additionally, a smart bias initialization strategy and a Huber loss objective resolve gradient optimization bottlenecks caused by highly skewed lifespan data distributions. Comprehensive evaluations across four diverse datasets—encompassing varying cathode materials, fast-charging protocols, and temperatures—demonstrate that LFM achieves state-of-the-art prognostic accuracy, significantly outperforming established sequence architectures. The framework exhibits exceptional data efficiency and reliability under constrained operational scenarios, maintaining high-fidelity estimations utilizing merely 30\% of historical calibration data. Crucially, entirely decoupling the computationally intensive vector field network during inference guarantees highly efficient constant temporal complexity, demonstrating immense practical scalability for real-time deployment in onboard Battery Management Systems and macro-level Energy Management Systems.
- 13 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Generative Latent Flow Matching Framework for Robust Battery Health Prognosis Under Dynamic Load Fluctuations
Accurate State of Health estimation is critical for the reliable operation and lifecycle management of lithium-ion batteries in modern energy storage systems. However, existing deterministic regression methods frequently struggle with statistical drift, unpredictable dynamic load fluctuations, and limited cross-domain generalization under highly heterogeneous energy dispatch conditions. To overcome these limitations, this paper proposes a novel Latent Flow Matching (LFM) framework that reformulates battery health prognosis as a physics-consistent generative transport problem. Acting as a continuous manifold learner governed by an ordinary differential equation, the LFM constraint maps complex operational profiles into a stable, continuous physical degradation manifold, effectively filtering out stochastic system-level disturbances. An enhanced local-aware encoder incorporating Group Normalization and channel-wise attention ensures operational feature robustness by isolating inter-sample interference during highly dynamic power demands. Additionally, a smart bias initialization strategy and a Huber loss objective resolve gradient optimization bottlenecks caused by highly skewed lifespan data distributions. Comprehensive evaluations across four diverse datasets—encompassing varying cathode materials, fast-charging protocols, and temperatures—demonstrate that LFM achieves state-of-the-art prognostic accuracy, significantly outperforming established sequence architectures. The framework exhibits exceptional data efficiency and reliability under constrained operational scenarios, maintaining high-fidelity estimations utilizing merely 30\% of historical calibration data. Crucially, entirely decoupling the computationally intensive vector field network during inference guarantees highly efficient constant temporal complexity, demonstrating immense practical scalability for real-time deployment in onboard Battery Management Systems and macro-level Energy Management Systems.
- 9 June 2026
- Published by Elsevier in Ocean Engineering
- Vol. 362
PH-FM: A Predictive Hard-Constrained Flow Matching framework for robust AUV path planning in complex dynamic environments
- 9 June 2026
- Published by Elsevier in Ocean Engineering
- Vol. 362
PH-FM: A Predictive Hard-Constrained Flow Matching framework for robust AUV path planning in complex dynamic environments
- Open Access
- 8 June 2026
- Published by Elsevier in SSRN Electronic Journal
Physics-Guided Causal Diffusion with Confidence-Rectified Ensemble for Generalized Zero-Shot Compound Fault Diagnosis
In complex mechanical systems, data-driven fault diagnosis faces significant challenges when encountering novel compound faults absent from the training set, a problem framed as Generalized Zero-Shot Learning (GZSL). Existing generative approaches often fail to capture the intrinsic physical coupling of fault signatures, leading to poor generalization due to the domain shift between synthetic and real data. To address this, we propose a novel framework termed Physics-Guided Causal Diffusion with Confidence-Rectified Ensemble (PGCD-CRE). First, we formulate a causal diffusion model that integrates multi-hot semantic conditioning with an adaptive physics envelope loss. This mechanism explicitly models transfer path attenuation and enforces the nonlinear superposition of fault impulses, enabling the high-fidelity synthesis of complex compound signals using only seen single-fault data. Second, to break the restrictive assumption of knowing the exact number of unseen fault types in advance and to mitigate the spectral confusion inherent in coupled faults, we design a specialist ensemble diagnosis strategy augmented by a density-based topological discovery (HDBSCAN) and a threshold-guided cross-confidence rectification mechanism. This module dynamically isolates environmental noise and aligns real unseen samples with their correct semantic categories without relying on predefined cluster counts. Extensive experiments on the XJTU and HUST datasets under complex industrial conditions demonstrate that PGCD-CRE significantly outperforms state-of-the-art methods, achieving robust accuracy in identifying unseen multi-component faults by bridging the semantic gap through physical causality and open-world discovery.
- 8 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPhysics-Guided Causal Diffusion with Confidence-Rectified Ensemble for Generalized Zero-Shot Compound Fault Diagnosis
In complex mechanical systems, data-driven fault diagnosis faces significant challenges when encountering novel compound faults absent from the training set, a problem framed as Generalized Zero-Shot Learning (GZSL). Existing generative approaches often fail to capture the intrinsic physical coupling of fault signatures, leading to poor generalization due to the domain shift between synthetic and real data. To address this, we propose a novel framework termed Physics-Guided Causal Diffusion with Confidence-Rectified Ensemble (PGCD-CRE). First, we formulate a causal diffusion model that integrates multi-hot semantic conditioning with an adaptive physics envelope loss. This mechanism explicitly models transfer path attenuation and enforces the nonlinear superposition of fault impulses, enabling the high-fidelity synthesis of complex compound signals using only seen single-fault data. Second, to break the restrictive assumption of knowing the exact number of unseen fault types in advance and to mitigate the spectral confusion inherent in coupled faults, we design a specialist ensemble diagnosis strategy augmented by a density-based topological discovery (HDBSCAN) and a threshold-guided cross-confidence rectification mechanism. This module dynamically isolates environmental noise and aligns real unseen samples with their correct semantic categories without relying on predefined cluster counts. Extensive experiments on the XJTU and HUST datasets under complex industrial conditions demonstrate that PGCD-CRE significantly outperforms state-of-the-art methods, achieving robust accuracy in identifying unseen multi-component faults by bridging the semantic gap through physical causality and open-world discovery.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cybernetics
- Vol. PP, 1-11
Saturated Precise Robust Singular Spectrum Analysis-Diffusion Model Predictive Control of AUVs
Accurate path following by autonomous underwater vehicles (AUVs) in wave-dominated environments is limited by two coupled engineering constraints: feedback-driven disturbance observers (DOBs) exhibit phase lag, whereas aggressive cancellation of high-frequency components can exceed the vehicle and actuator bandwidths. This article develops a closed-loop predictive compensation architecture, termed singular spectrum analysis-diffusion (S2D)-model predictive control (MPC) that integrates a S2D observer, a dynamics-informed adaptive filter (DIAF), and MPC. The observer combines instantaneous Kalman-filter (KF) estimates, singular spectrum analysis (SSA)-based trend-residual decomposition, and conditional skip-step diffusion to forecast the disturbance sequence over the MPC horizon. The DIAF then applies saturation and bandwidth-aware filtering so that only physically actionable components enter the feedforward channel. This perception-prediction-decision-action loop retains constraint handling while converting delayed disturbance estimates into phase-leading compensation. Regional input-to-state practical stability (RISPS) is established for the resulting closed loop. Under JONSWAP wave disturbances, the proposed method yields the lowest integrated absolute error (2.09 m $\cdot $ s), an average execution time of 33.25 ms for a 50-ms sampling interval, and integral absolute error (IAE) values of 2.054-2.136 across the tested sea states.- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cybernetics
- Vol. PP, 1-11
Saturated Precise Robust Singular Spectrum Analysis-Diffusion Model Predictive Control of AUVs
Accurate path following by autonomous underwater vehicles (AUVs) in wave-dominated environments is limited by two coupled engineering constraints: feedback-driven disturbance observers (DOBs) exhibit phase lag, whereas aggressive cancellation of high-frequency components can exceed the vehicle and actuator bandwidths. This article develops a closed-loop predictive compensation architecture, termed singular spectrum analysis-diffusion (S2D)-model predictive control (MPC) that integrates a S2D observer, a dynamics-informed adaptive filter (DIAF), and MPC. The observer combines instantaneous Kalman-filter (KF) estimates, singular spectrum analysis (SSA)-based trend-residual decomposition, and conditional skip-step diffusion to forecast the disturbance sequence over the MPC horizon. The DIAF then applies saturation and bandwidth-aware filtering so that only physically actionable components enter the feedforward channel. This perception-prediction-decision-action loop retains constraint handling while converting delayed disturbance estimates into phase-leading compensation. Regional input-to-state practical stability (RISPS) is established for the resulting closed loop. Under JONSWAP wave disturbances, the proposed method yields the lowest integrated absolute error (2.09 m $\cdot $ s), an average execution time of 33.25 ms for a 50-ms sampling interval, and integral absolute error (IAE) values of 2.054-2.136 across the tested sea states.- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 75, 3193-3207
Task-Factorized Riemannian Neural Operator for Bearing Fault Diagnosis Across Heterogeneous Operational Shifts
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 75, 3193-3207
Task-Factorized Riemannian Neural Operator for Bearing Fault Diagnosis Across Heterogeneous Operational Shifts
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
Physics-Guided GAN With Manhattan Attention: A Novel Approach for Imbalanced Bearing Vibration Fault Diagnosis
This study introduces a novel Physics-Guided Generative Adversarial Network (PGAN) tailored explicitly for diagnosing rolling bearing faults under severely imbalanced datasets. PGAN integrates domain-specific physical constraints into the generative process, thereby enhancing both the physical realism and interpretability of the generated data. Key innovations include a generator conditioned on physics-based features and random noise, the introduction of dedicated physics-guided loss functions, and the incorporation of Manhattan Attention to improve the extraction of essential vibration features. Extensive experiments conducted using the CWRU and HUST datasets demonstrate PGAN’s superior performance compared to several state-of-the-art methods. Results indicate that PGAN significantly mitigates the impact of dataset imbalance, achieving robust diagnostic accuracy even in extreme conditions. Further visualization and comparative analyses underscore the method’s capability to deliver highly discriminative and interpretable features.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
Physics-Guided GAN With Manhattan Attention: A Novel Approach for Imbalanced Bearing Vibration Fault Diagnosis
This study introduces a novel Physics-Guided Generative Adversarial Network (PGAN) tailored explicitly for diagnosing rolling bearing faults under severely imbalanced datasets. PGAN integrates domain-specific physical constraints into the generative process, thereby enhancing both the physical realism and interpretability of the generated data. Key innovations include a generator conditioned on physics-based features and random noise, the introduction of dedicated physics-guided loss functions, and the incorporation of Manhattan Attention to improve the extraction of essential vibration features. Extensive experiments conducted using the CWRU and HUST datasets demonstrate PGAN’s superior performance compared to several state-of-the-art methods. Results indicate that PGAN significantly mitigates the impact of dataset imbalance, achieving robust diagnostic accuracy even in extreme conditions. Further visualization and comparative analyses underscore the method’s capability to deliver highly discriminative and interpretable features.