With industrial embodied intelligent equipment, such as intelligent robots, being rapidly deployed on the shop floor, high-quality multimodal operational data has become a critical bottleneck for policy learning, robust planning, and online optimization. Existing public datasets are largely collected in laboratory or simulated environments, leaving systematic gaps to industrial settings in terms of process constraints, hardware heterogeneity, operational safety, and data security/compliance. To address this, this paper proposes a structured modeling framework for industrial embodied-intelligence operational data that spans data types–acquisition architecture–processing pipeline–sharing mechanism, and summarizes key engineering steps including multi-source data organization and key processing steps. Furthermore, aligned with the Trusted Data Space Development Action Plan (2024–2028) and its three core capabilities of trusted governance and control of data, resource interaction, and value co-creation, we develop a reference architecture based on the trusted data spaces system. By abstracting workshop-level embodied-intelligence units as data nodes and leveraging edge-side pre-processing, connectors, data catalogs, and digital contracts with usage control, the proposed solution enables controllable authorization, real-time evidence preservation, and traceable usage across organizations, thereby supporting trusted sharing and reuse of operational data. Based on literature review and engineering cases, we outline a practical implementation path, providing references for industrial data-solution design and standardization toward trusted data spaces.