Thinking Like an Expert: Aligning LLM Thought Processes for Automated Safety Modeling of High-Altitude Solar Drones

Highlights: This section summarizes the core contributions and practical value of the research on the Knowledge-Enhanced Graph of Thoughts (K-EGoT) framework for automated safety modeling of high-altitude solar drones, focusing on key experimental results and their significance for both academic research and industrial applications. What are the main findings? The proposed K-EGoT framework, when applied to a 7B-parameter model (Qwen2-7B-Instruct), achieves a Safety Extension Score (SES) of 92.7 in high-altitude solar drone safety modeling, significantly outperforming standard Graph of Thoughts (GoT) prompting (84.7) and other fine-tuning baselines (e.g., SFT + DPO-Behavioral with 89.4). The “Safety Rationale”—a verifiable link between LLM-generated model extensions and expert-curated safety principles—is the core driver of K-EGoT’s performance; removing it leads to a 7.6-point drop in total SES and a 14.2-point drop in Rationale Quality (Srat) in ablation tests. What are the implications of the main findings? For specialized safety-critical domains like high-altitude solar drones, aligning LLM reasoning with expert logic (via rationale-centric optimization) is more impactful than general-purpose prompting or behavioral-only fine-tuning, enabling smaller models to outperform larger generic models in domain-specific tasks. K-EGoT provides an auditable and reliable solution for early-stage automated safety modeling, addressing the inefficiency of manual expert analysis and the opaqueness of traditional LLM applications, which can reduce late-stage design modification costs and accelerate drone development cycles. As the application of high-altitude solar drones expands, ensuring their safety is paramount. Traditional safety modeling, which relies on manual expert analysis, struggles to keep pace with rapid development cycles. While Large Language Models (LLMs) offer a path to automation, state-of-the-art reasoning frameworks like Graph of Thoughts (GoT) are too generic, lacking the domain-specific knowledge required for effective application. To address this gap, we introduce K-EGoT, a framework that grounds LLM reasoning in a verifiable, domain-specific knowledge base. Our method introduces a “Safety Rationale”—a mandatory, auditable link between LLM-generated model extensions and expert-curated safety principles. We then train a specialized model using a novel “thought process alignment” strategy, applying Direct Preference Optimization (DPO) to the quality of these rationales to ensure the model’sreasoning aligns with expert logic. On a high-fidelity dataset for the flight control–energy coupling problem, our 7B K-EGoT model achieved a Safety Extension Score (SES) of 92.7, significantly outperforming the 84.7 score from standard GoT prompting. Our work delivers a reliable and auditable solution for automated safety modeling for this critical class of drones.