From Physics to Machine Learning and Back: Part II - Learning and Observational Bias in PHM

2025-09-25

AI Summary

In real-world industrial settings, Prognostics and Health Management (PHM) faces challenges including sensor noise/missing data, scarce labeled degradation samples, highly nonlinear system degradation dynamics, and complex cross-device dependencies. To address these, this paper proposes a dual-perspective modeling framework—“learning bias” and “observation bias”—that synergistically integrates physics-informed priors with data-driven learning. Specifically, we design a physics-constrained loss function to ensure model consistency; introduce virtual sensing and multi-source heterogeneous data fusion to alleviate label scarcity; and unify Physics-Informed Neural Networks (PINNs), few-shot meta-learning, and reinforcement-based decision-making for closed-loop optimization of prediction, diagnosis, and maintenance. The method enables rapid zero-shot or few-shot transfer from single-device models to fleet-level deployment. Evaluated on multiple industrial datasets, it significantly improves generalizability, robustness, and interpretability, establishing a novel paradigm for reliability modeling and maintenance decision-making in complex systems.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Prognostics and Health Management ensures the reliability, safety, and efficiency of complex engineered systems by enabling fault detection, anticipating equipment failures, and optimizing maintenance activities throughout an asset lifecycle. However, real-world PHM presents persistent challenges: sensor data is often noisy or incomplete, available labels are limited, and degradation behaviors and system interdependencies can be highly complex and nonlinear. Physics-informed machine learning has emerged as a promising approach to address these limitations by embedding physical knowledge into data-driven models. This review examines how incorporating learning and observational biases through physics-informed modeling and data strategies can guide models toward physically consistent and reliable predictions. Learning biases embed physical constraints into model training through physics-informed loss functions and governing equations, or by incorporating properties like monotonicity. Observational biases influence data selection and synthesis to ensure models capture realistic system behavior through virtual sensing for estimating unmeasured states, physics-based simulation for data augmentation, and multi-sensor fusion strategies. The review then examines how these approaches enable the transition from passive prediction to active decision-making through reinforcement learning, which allows agents to learn maintenance policies that respect physical constraints while optimizing operational objectives. This closes the loop between model-based predictions, simulation, and actual system operation, empowering adaptive decision-making. Finally, the review addresses the critical challenge of scaling PHM solutions from individual assets to fleet-wide deployment. Fast adaptation methods including meta-learning and few-shot learning are reviewed alongside domain generalization techniques ...

Problem

• Addressing noisy sensor data and limited labels in Prognostics and Health Management
• Modeling complex nonlinear degradation behaviors and system interdependencies
• Scaling PHM solutions from individual assets to fleet-wide deployment

Innovation

• Physics-informed loss functions embed constraints
• Virtual sensing and simulation augment data
• Reinforcement learning optimizes maintenance policies