摘 要：
本研究针对水利水电工程施工质量异常检测问题，提出了基于机器学习的检测框架。通过整合多模态数据，解决了数据孤岛问题，为模型提供了全面数据支撑。研究设计了基于XGBoost的集成学习模型，结合LightGBM特性，优化了检测性能。实验表明，模型在准确率、召回率上显著优于传统方法。此外，研究还通过多算法融合与数据增强技术提升了模型的鲁棒性。为了推动行业智能化转型，研究提出了政策制定与行业规范建议，包括出台数据共享激励政策、推动数据标准化建设、加强AI人才储备以及加强行业交流与合作。这些建议旨在促进数据共享与利用，提高数据质量和可用性，满足机器学习模型对数据的需求，并推动水利水电工程施工质量异常检测领域的快速发展。
关键词：水利水电工程；机器学习；多模态数据融合
Abstract:
This study proposes a detection framework based on machine learning for the problem of abnormal detection of construction quality in water conservancy and hydropower projects. By integrating multimodal data, the problem of data silos has been solved, providing comprehensive data support for the model. The research designed an ensemble learning model based on XGBoost, combined with the characteristics of LightGBM, to optimize the detection performance. Experiments show that the model is significantly superior to traditional methods in terms of accuracy and recall rate. In addition, the research has also enhanced the robustness of the model through multi-algorithm fusion and data augmentation techniques. To promote the intelligent transformation of the industry, the research has put forward suggestions for policy formulation and industry norms, including introducing incentive policies for data sharing, promoting the construction of data standardization, strengthening the reserve of AI talents, and enhancing industry exchanges and cooperation. These suggestions aim to promote data sharing and utilization, enhance data quality and availability, meet the data requirements of machine learning models, and drive the rapid development of the field of abnormal detection of construction quality in water conservancy and hydropower projects.
Keywords: Water resources and hydropower engineering; Machine learning; Multimodal data fusion
--
