Drift-Aware Online Dynamic Learning for Nonstationary Multivariate Time Series: Application to Sintering Quality Prediction

2026-04-10

AI Summary

This work addresses the performance degradation in non-stationary multivariate time series forecasting during industrial sintering processes, caused by concept drift and label latency. To tackle this challenge, the authors propose a drift-aware multi-scale dynamic learning framework that integrates unsupervised Maximum Mean Discrepancy (MMD)-based drift detection, a multi-scale dual-branch convolutional network, a dynamic memory queue, and a prioritized experience replay mechanism. Furthermore, a drift-severity-guided hierarchical fine-tuning strategy is introduced to effectively balance model stability and plasticity under delayed labeling, thereby mitigating catastrophic forgetting. Experimental results on real-world sintering data and public benchmarks demonstrate that the proposed framework significantly outperforms existing methods, exhibiting superior cross-domain generalization and prediction robustness.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Reasoning under Uncertainty: Stochastic Optimization
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurate prediction of nonstationary multivariate time series remains a critical challenge in complex industrial systems such as iron ore sintering. In practice, pronounced concept drift compounded by significant label verification latency rapidly degrades the performance of offline-trained models. Existing methods based on static architectures or passive update strategies struggle to simultaneously extract multi-scale spatiotemporal features and overcome the stability-plasticity dilemma without immediate supervision. To address these limitations, a Drift-Aware Multi-Scale Dynamic Learning (DA-MSDL) framework is proposed to maintain robust multi-output predictive performance via online adaptive mechanisms on nonstationary data streams. The framework employs a multi-scale bi-branch convolutional network as its backbone to disentangle local fluctuations from long-term trends, thereby enhancing representational capacity for complex dynamic patterns. To circumvent the label latency bottleneck, DA-MSDL leverages Maximum Mean Discrepancy (MMD) for unsupervised drift detection. By quantifying online statistical deviations in feature distributions, DA-MSDL proactively triggers model adaptation prior to inference. Furthermore, a drift-severity-guided hierarchical fine-tuning strategy is developed. Supported by prioritized experience replay from a dynamic memory queue, this approach achieves rapid distribution alignment while effectively mitigating catastrophic forgetting. Long-horizon experiments on real-world industrial sintering data and a public benchmark dataset demonstrate that DA-MSDL consistently outperforms representative baselines under severe concept drift. Exhibiting strong cross-domain generalization and predictive stability, the proposed framework provides an effective online dynamic learning paradigm for quality monitoring in nonstationary environments.

Problem

• nonstationary time series
• concept drift
• label latency
• online prediction
• multivariate time series

Innovation

• concept drift
• online learning
• multi-scale architecture
• unsupervised drift detection
• catastrophic forgetting