The causal factors underlying anomalies during carrier-based aircraft launch and recovery are often rare and unpredictable. To address the hallucination issues commonly encountered by large language models (LLMs) when generating these anomaly causes, we propose a scenario generation for carrier aircraft driven by large-small model collaboration (SGCAD) method.

First, a knowledge base for carrier-based aircraft launch and recovery is constructed by integrating professional literature and retrieval-augmented generation (RAG) techniques, creating a dataset of normal operation descriptions. These normal descriptions serve as templates, which, combined with scenario-specific prompts, guide the large language model to generate potential anomaly causes. A smaller model is then employed to discriminate between reasonable and unreasonable anomaly causes. Finally, the large model undergoes fine-tuning using direct preference optimization (DPO) and is iteratively refined to progressively increase the proportion of reasonable anomaly scenarios.

Experimental results show that, after multiple iterations of the SGCAD method, the proportion of reasonable anomaly causes in the generated dataset reaches 94%, effectively mitigating hallucination issues and significantly improving the rationality and realism of the generated content. Expert evaluations further confirm that the generated anomaly causes encompass diverse and complex scenarios, utilize standardized terminology, and adhere to physical laws.

The proposed approach provides a valuable reference for analyzing anomalies during carrier-based aircraft launch and recovery operations.

The launch and recovery of carrier-based aircraft are characterized by rare incident occurrences, high unpredictability, and severe consequences. As a result, flight deck commanders cannot rely solely on past incident cases to accumulate experiential knowledge. To address this challenge, this study proposes a large-model-driven framework for comprehensive case-based simulation of carrier-based aircraft launch and recovery contingencies.

To analyze the evolution of contingency events under specific causative factors, a launch and recovery knowledge base was first constructed using carrier-based aircraft operation manuals and relevant literature. Additional domain-specific knowledge was incorporated to enhance the professionalism and relevance of content generated by the large model. Using prompt engineering, the large model was guided to generate latent situational trends, representing potential event evolutions under given causative factors. A decision-making model, augmented with the launch and recovery knowledge base, was then employed to evaluate these latent trends, filter out implausible evolutions, and update the contingency state iteratively.

Experimental results demonstrate that the proposed framework can, through multiple iterations, generate a complete set of contingency evolution processes corresponding to specific causative factors. By simulating diverse incident triggers and decision-making paths, the framework facilitates the creation of a comprehensive contingency case database for carrier-based aircraft operations.

The proposed approach effectively utilizes the extensive prior knowledge and strong logical reasoning capabilities of large models to address the scarcity of real-world data on carrier-based contingencies. The generated case simulations serve as valuable reference material and learning scenarios, enhancing the emergency decision-making capabilities of flight deck commanders.

The critical factor in the safe landing of carrier-based aircraft is the successful locking of the tailhook and arresting wires. However, in the existing research, there is relatively little work on using intelligent means to assist the landing signal officer (LSO) in identifying the arrested landing state.

This paper proposes a model for identifying the arrested landing state which integrates coordinate attention (CA) and a weighted bi-directional feature pyramid network (BiFPN). First, CA is used to enhance the network's feature extraction ability in both the spatial and channel dimensions. Next, BiFPN introduces learnable weights to learn the weights of different input features by repeatedly using top-down and bottom-up multi-scale feature fusion. A C2F lightweight model structure is adopted to reduce the parameters and computational complexity. Finally, the proposed model is compared with five baseline models through simulation experiments.

The results reveal that the proposed model outperforms the baseline model in detecting the tailhook and arresting wires of carrier-based aircraft.

The findings of this study can provide valuable references for improving the accuracy and robustness of the detection of the tailhook and arresting wires of carrier-based aircraft, and is of great significance for improving the efficiency of carrier-based aircraft landing operations and preventing potential personnel injuries and equipment losses.