Aircraft skin defects pose a threat to the safety and airworthiness of the aircraft. The front line of engineering has requirements of high precision and stable defect detection, which cannot be met by existing deep learning methods, due to conflicting information between multi-scale features. Herein, a Fine-Coordinated YOLO (FC-YOLO) algorithm is proposed to detect aircraft skin defects. Firstly, the ELAN-C module with Coordinate & Channel Attention mechanism is applied to the backbone network to enhance multi-scale detection precision. Secondly, the Adaptive-Path Aggregation Network structure is proposed to make features containing more information by adding a shortcut weighted by the Adaptively Spatial Feature Fusion (ASFF) module. The ASFF adaptively allocates the weights of features with different sizes to reduce the inconsistency of features between different levels during feature fusion to improve detection precision. Finally, the SCYLLA-IoU loss function is introduced to calculate the directional loss between the bounding box and the ground truth box to elevate the stability of the training. Experiments are executed with a self-constructed ASD-DET dataset and the public NEU-DET dataset. Results show that the mAP of FC-YOLO is improved by 3.1% and 2.7% compared to that of the original YOLOv7 on the ASD-DET dataset and the NEU-DET dataset. In addition, on the ASD-DET dataset and NEU-DET dataset, the mAP of FC-YOLO was higher than that of YOLOv8, RT-DETR by 1.4%, 1.6% and 2.2%, 3.8%, respectively. By which, it is shown that the proposed FC-YOLO algorithm is promising for the future automatic visual inspection of aircraft skin.

Reducing carbon emissions and improving revenue in the face of global warming and economic challenges is a growing concern for airlines. This paper addresses the inefficiencies and high costs associated with current aero-engine on-wing washing strategies. To tackle this issue, we propose a reinforcement learning framework consisting of a Similar Sequence Method and a Taylor DQN model. The Similar Sequence Method, comprising a sample library, DTW algorithm, and boundary adjustment, predicts washed aero-engine data for the Taylor DQN model. Leveraging the proposed Taylor neural networks, our model outputs Q-values to make informed washing decisions using data from the Similar Sequence Method. Through simulations, we demonstrate the effectiveness of our approach.

Currently, the research on academic early warning assessment and prediction for college students under the credit system in colleges and universities is mainly based on methods such as machine learning. However, the existing prediction models often have problems such as difficulty in network structure, parameter selection, and extraction of context time series information. In response to these issues, this study, based on students’ historical academic performance, proposes LTBoost, a novel framework of the XGBoost classification prediction model. The proposed model framework integrates the BiLSTM module to handle the time series information in the original data. It also integrates the proposed Enhanced Transformer module to perceive global information and obtain enhanced features. Through experiments, the LTBoost prediction model was compared with four other machine learning algorithms. The accuracy of the proposed LTBoost classification prediction model increased by 0.7% to 99.4%, demonstrating a good prediction effect on whether students are at risk of prolonging their studies. It provides a new paradigm and path for the construction of talent cultivation plans in credit-based universities.