Recent advancements in Artificial Intelligence (AI) and Machine Learning (ML) have achieved remarkable success in natural language processing, computer vision, and reinforcement learning. However, when applied to human behavior understanding and motion prediction, most deep learning architectures remain black-box models—lacking interpretability in how perceptual cues and motor actions interact to form human decision-making patterns. To address this limitation, this study proposes an Explainable Multimodal Event Knowledge Graph (E²-KG) framework that integrates eye-tracking data and OpenPose-based joint angles to construct a semantic representation of cognitive–motor coupling in dynamic tasks. Instead of treating perception and motion as separate streams, E²-KG embeds them within a unified perception–action manifold, where local energy gradients automatically detect event transitions (fixation, preparation, release, follow-through). These events are organized into a knowledge graph structure, enabling interpretable reasoning about causal links between visual attention, body movement, and task outcomes. A perception–action manifold model capturing the joint evolution of gaze and motion through local energy variations. A multimodal event extraction pipeline for synchronized eye–pose data. An event knowledge graph that supports cause–effect reasoning. Validation through a throwing-distance prediction task, demonstrating higher interpretability and transparency than data-driven baselines. E²-KG thus bridges manifold learning and knowledge-graph reasoning, offering a transparent, cognitively grounded approach for explainable human performance modeling—applicable to domains such as human factors, adaptive training, and human–robot collaboration.

This study presents an explainable knowledge graph (KG) framework that transforms continuous spindle monitoring time-series data into transparent, reasoning-ready diagnostic structures. Existing data-driven approaches, while accurate, often lack the interpretability required for high-stakes industrial decision-making and are sensitive to operating condition drifts. To address these limitations, we propose a two-level temporal segmentation method combining label transition detection and statistical drift analysis to identify meaningful state boundaries. Furthermore, a percentile-based discretization mechanism converts statistical features into interpretable semantic tags. A Neo4j-based state–event–feature schema captures lifecycle evolution and evidence relations, enabling attribution path reasoning that links failure events to salient precursor features. Experiments on real industrial spindle data demonstrate a fault detection accuracy of 84.97% and a false alarm rate of 3.43%, effectively capturing stable baselines and intermittent abnormal bursts. The proposed framework provides a distinct novelty in bridging the gap between numerical time-series and symbolic reasoning, offering a practical pathway for deploying explainable and maintainable spindle health analytics.