<p>Process industries are characterized by high continuity, strong parameter coupling, and stringent safety requirements, where anomalies can easily propagate through material flows and energy transfers to cause interruptions or accidents. Root cause analysis in such systems requires not only precise identification of the root causes but also the construction of an interpretable causal chain linking “root cause-intermediate factors-anomaly features”. However, existing approaches suffer from an inherent data-knowledge disconnect: Data-based methods rely on massive sensor data but suffer from spurious causality and poor interpretability; knowledge-based methods depend on static prior knowledge but struggle to adapt to the dynamic changes of industrial processes. To address this, this paper proposes KDACL (Knowledge-Data Fusion for Anomaly Causal Chain Construction with LLMs), an explainable root cause analysis method for process industries that integrates Large Language Model (LLM), Chain of Thought (CoT), and Retrieval-Augmented Generation (RAG) technologies. It constructs a traceable reasoning framework with knowledge-data dual-modal synergistic enhancement. It employs RAG to selectively retrieve relevant mechanistic knowledge, providing physical anchors for LLM while constraining data causality mining to avoid spurious causality. Simultaneously, it uses real-time sensor data to perform temporal and consistency verification to LLM-generated initial causal hypotheses. This verification eliminates misaligned knowledge to dynamically correct biases, enabling the LLM to iteratively refine causal chains, while also preserving the reasoning decision bases to reduce LLM knowledge hallucinations. Finally, experimental validation using the Tennessee Eastman Process (TEP) and a real-world Slag Mill Production (SMP) system demonstrates that KDACL can accurately identify root causes and generate explainable causal chains with clear physical mechanisms. This method outperforms traditional approaches in anomaly inference accuracy and causal chain interpretability, providing effective technical support for the safe and reliable operation of process industries.</p>