组合模型对管道腐蚀速率预测的效能研究--基于注意力机制增强的CNN与LSTM模型  

Research on the effectiveness of a combined model for predicting the corrosion rate of pipelines-CNN and LSTM models based on attention mechanism enhancement

作　　者：骆正山[1] 杜丹 骆济豪 王小完[1] LUO Zhengshan;DU Dan;LUO Jihao;WANG Xiaowan(School of Management,Xi'an University of Architecture and Technology,Xi'an 710055,China;School of Information and Electronics,Beijing Institute of Technology,Beijing 102488,Beijing 102488,China)

机构地区：[1]西安建筑科技大学管理学院,西安710055 [2]北京理工大学信息与电子学院,北京102488

出　　处：《安全与环境学报》2024年第11期4263-4269,共7页Journal of Safety and Environment

基　　金：国家自然科学基金项目(41877527)。

摘　　要：为评估卷积神经网络(Convolutional Neural Network,CNN)、长短期记忆(Long Short-Term Memory,LSTM)网络及结合的CNN-LSTM模型在管道腐蚀速率预测中的性能表现,特别引入注意力机制,以期提高模型对关键特征的捕捉能力和预测的准确性。分析影响管道腐蚀速率的环境因素作为模型输入,并通过注意力机制优化特征表示。结果表明,结合注意力机制的CNN-LSTM模型在准确性和可靠性上超越了单独的CNN或LSTM模型。这一结果不仅展示了深度学习模型通过技术增强了处理复杂数据的能力,也为实际工业应用中的时间序列预测提供了新的视角,同时证实了利用深度学习技术对管道腐蚀速率进行精确预测的可行性和有效性。To evaluate the performance of Convolutional Neural Network(CNN),Long Short-Term Memory(LSTM),and their combined CNN-LSTM models in predicting pipeline corrosion rates,this study introduced an attention mechanism.This mechanism aims to enhance the model's capability to capture critical features,thereby improving prediction accuracy.By analyzing environmental factors influencing pipeline corrosion rates and incorporating them as inputs to the model,this study optimized feature representation using the introduced attention mechanism.The results demonstrate that the CNN-LSTM model,enhanced with the attention mechanism,outperformed standalone CNN or LSTM models in terms of accuracy and reliability.This finding underscores several critical points:Firstly,it highlights the capacity of deep learning models,augmented with advanced techniques such as the attention mechanism,to effectively manage complex data.Secondly,it introduces a fresh perspective on time series prediction within practical industrial applications.The enhanced CNN-LSTM model not only captures intricate patterns within the data but also prioritizes key features essential for accurate predictions.Furthermore,the study confirms the feasibility and effectiveness of using deep learning technologies for the precise prediction of pipeline corrosion rates.The utilization of the attention mechanism significantly enhances the predictive capabilities of the CNN-LSTM model.By emphasizing the most pertinent aspects of the input data,the model becomes more proficient in forecasting corrosion rates,crucial for safeguarding the integrity and safety of pipelines in industrial settings.In conclusion,the integration of the attention mechanism into the CNN-LSTM model signifies a substantial advancement in predicting corrosion rates.This approach not only harnesses the strengths of CNN and LSTM architectures but also enhances them through precise feature optimization.By focusing on relevant aspects of input data,the model achieves greater proficiency in forecasting corrosio

关 键 词：安全工程 管道腐蚀速率预测 卷积神经网络(CNN) 长短期记忆(LSTM) 注意力机制 时间序列分析 

分 类 号：X937[环境科学与工程—安全科学]