Selected work

Representative Papers

Overview of Publicly Available Degradation Data Sets for Tasks within Prognostics and Health Management

Mar 20, 2024arXiv.org

The PHM (Prognostics and Health Management) community has long suffered from a lack of systematically evaluated, freely accessible degradation datasets. Method: This work establishes the first multi-dimensional unified evaluation framework for PHM datasets, incorporating critical dimensions—data provenance, equipment types, sensor configurations, failure modes, and annotation completeness—integrated with structured metadata analysis, cross-dataset comparative assessment, task-specific PHM mapping, and physics-of-failure-informed semantic annotation. Contribution/Results: We systematically curate and analyze 32 high-quality public datasets, identifying 11 recurrent deficiencies. Based on this analysis, we provide task-oriented data selection guidelines and benchmarking recommendations. This study fills a critical gap in systematic surveys of PHM public data resources, explicitly delineates the applicability boundaries and modeling limitations of existing datasets, and has been widely cited and adopted within the PHM research community.

1 citationsRead paper

Physics-Informed Machine Learning in Prognostics and Health Management: A Systematic Literature Review

Aug 10, 2026

This study addresses the limitations of purely data-driven machine learning in prognostics and health management (PHM)—notably poor generalization, lack of causal reasoning, and limited interpretability—by proposing a systematic literature review that establishes a four-category taxonomy of physics-informed machine learning (PIML): observational bias, inductive bias, learning bias, and hybrid approaches. The framework is evaluated through categorization by PHM tasks across 212 studies. Findings indicate that PIML consistently outperforms conventional methods in applications such as lithium-ion batteries and bearings; however, its adoption remains constrained by narrow application scope, absence of a universal design paradigm, and insufficient empirical validation for certain claimed advantages. This work provides a structured perspective and a research roadmap to advance the systematic development of PIML in PHM.

0 citationsRead paper

Explainable AI for the EU Right to Explanation: A Systematic Review of the Law-XAI Translation Gap

Aug 03, 2026

This study addresses the significant gap between legal requirements and technical capabilities in explainable artificial intelligence (XAI), particularly concerning the EU’s legal right to explanation under Article 15(1)(h) of the General Data Protection Regulation and Article 86 of the AI Act. Through a systematic literature review—screening 2,643 papers and identifying only 19 that meaningfully integrate legal and technical perspectives—combined with legal text analysis and XAI evaluation, the work proposes a “recipient/purpose framework” to clarify the form and content of algorithmic explanations and outlines a four-stage implementation blueprint. It identifies three recurring problem patterns and distills six key open challenges, offering the first systematic account of the structural misalignment between legal mandates and current XAI approaches, thereby providing theoretical and practical guidance for achieving regulatory compliance.

0 citationsRead paper

Constraint-Data-Value-Maximization: Utilizing Data Attribution for Effective Data Pruning in Low-Data Environments

May 11, 2026

Existing Shapley value–based data pruning methods struggle to effectively retain critical samples in low-data regimes. This work proposes Constrained Data Value Maximization (CDVM), which introduces constrained optimization into data value–driven pruning for the first time. By formulating the problem as a constrained optimization task, CDVM simultaneously maximizes the overall influence of the retained dataset while limiting the excessive contribution of any individual test sample, thereby achieving a balance between global impact and local fidelity. Experiments on the OpenDataVal benchmark demonstrate that CDVM significantly outperforms existing approaches when retaining only a small fraction of data, offering both superior performance and computational efficiency.

0 citationsRead paper

Foundation-Model-Based Agents in Industrial Automation: Purposes, Capabilities, and Open Challenges

May 04, 2026

This study addresses the lack of systematic understanding regarding the objectives, capability boundaries, and challenges of foundation model–based agents in industrial automation. Following PRISMA 2020 guidelines, the authors screened 2,341 publications to identify 88 core studies, which were analyzed through a systematic literature review, structured coding, and a Technology Readiness Level (TRL) assessment framework. The work proposes the first operational definition of industrial agents that integrates classical agent theory, automation engineering standards, and foundation model paradigms. Quantitative analysis reveals notable shifts in agent capabilities—human–machine interaction (+37%) and uncertainty handling (+35%) have improved, whereas negotiation capacity declined by 39%. Findings indicate that 75% of systems remain at prototype or early validation stages, with only 9.1% demonstrating deployment evidence, primarily serving assistive, monitoring, and optimization roles, constrained by limited generalization, hallucination risks, data scarcity, and inference latency.

0 citationsRead paper

Recent publications

Latest Papers

Physics-Informed Machine Learning in Prognostics and Health Management: A Systematic Literature Review

Aug 10, 2026

This study addresses the limitations of purely data-driven machine learning in prognostics and health management (PHM)—notably poor generalization, lack of causal reasoning, and limited interpretability—by proposing a systematic literature review that establishes a four-category taxonomy of physics-informed machine learning (PIML): observational bias, inductive bias, learning bias, and hybrid approaches. The framework is evaluated through categorization by PHM tasks across 212 studies. Findings indicate that PIML consistently outperforms conventional methods in applications such as lithium-ion batteries and bearings; however, its adoption remains constrained by narrow application scope, absence of a universal design paradigm, and insufficient empirical validation for certain claimed advantages. This work provides a structured perspective and a research roadmap to advance the systematic development of PIML in PHM.

0 citationsRead paper

Explainable AI for the EU Right to Explanation: A Systematic Review of the Law-XAI Translation Gap

Aug 03, 2026

This study addresses the significant gap between legal requirements and technical capabilities in explainable artificial intelligence (XAI), particularly concerning the EU’s legal right to explanation under Article 15(1)(h) of the General Data Protection Regulation and Article 86 of the AI Act. Through a systematic literature review—screening 2,643 papers and identifying only 19 that meaningfully integrate legal and technical perspectives—combined with legal text analysis and XAI evaluation, the work proposes a “recipient/purpose framework” to clarify the form and content of algorithmic explanations and outlines a four-stage implementation blueprint. It identifies three recurring problem patterns and distills six key open challenges, offering the first systematic account of the structural misalignment between legal mandates and current XAI approaches, thereby providing theoretical and practical guidance for achieving regulatory compliance.

0 citationsRead paper

Constraint-Data-Value-Maximization: Utilizing Data Attribution for Effective Data Pruning in Low-Data Environments

May 11, 2026

Existing Shapley value–based data pruning methods struggle to effectively retain critical samples in low-data regimes. This work proposes Constrained Data Value Maximization (CDVM), which introduces constrained optimization into data value–driven pruning for the first time. By formulating the problem as a constrained optimization task, CDVM simultaneously maximizes the overall influence of the retained dataset while limiting the excessive contribution of any individual test sample, thereby achieving a balance between global impact and local fidelity. Experiments on the OpenDataVal benchmark demonstrate that CDVM significantly outperforms existing approaches when retaining only a small fraction of data, offering both superior performance and computational efficiency.

0 citationsRead paper

Foundation-Model-Based Agents in Industrial Automation: Purposes, Capabilities, and Open Challenges

May 04, 2026

This study addresses the lack of systematic understanding regarding the objectives, capability boundaries, and challenges of foundation model–based agents in industrial automation. Following PRISMA 2020 guidelines, the authors screened 2,341 publications to identify 88 core studies, which were analyzed through a systematic literature review, structured coding, and a Technology Readiness Level (TRL) assessment framework. The work proposes the first operational definition of industrial agents that integrates classical agent theory, automation engineering standards, and foundation model paradigms. Quantitative analysis reveals notable shifts in agent capabilities—human–machine interaction (+37%) and uncertainty handling (+35%) have improved, whereas negotiation capacity declined by 39%. Findings indicate that 75% of systems remain at prototype or early validation stages, with only 9.1% demonstrating deployment evidence, primarily serving assistive, monitoring, and optimization roles, constrained by limited generalization, hallucination risks, data scarcity, and inference latency.

0 citationsRead paper

Actor-Critic Pretraining for Proximal Policy Optimization

Feb 27, 2026

This work addresses the high sample cost that limits reinforcement learning in robotic applications by proposing a novel dual-network pretraining framework for Actor-Critic algorithms such as PPO. Unlike existing approaches that pretrain only the policy network (Actor), this method simultaneously initializes both the Actor and the value network (Critic) using expert demonstration data, thereby significantly improving sample efficiency. By integrating behavioral cloning with reward estimation, the approach achieves a 86.1% improvement in sample efficiency over no pretraining and a 30.9% gain compared to Actor-only pretraining across 15 simulated robotic tasks. These results demonstrate the effectiveness and novelty of jointly pretraining both networks in an Actor-Critic architecture.

0 citationsRead paper