基于卷积神经网络CNN的示功图智能分析方法  

Intelligent Analysis Method of Indicator Diagram Based on Convolutional Neural Network CNN

作　　者：王必改[1] WANG Bi-gai(Fujian Chuanzheng Communications College,Fuzhou Fujian 350007,China)

机构地区：[1]福建船政交通职业学院航海学院,福建福州350007

出　　处：《广州航海学院学报》2024年第2期39-44,共6页Journal of Guangzhou Maritime University

摘　　要：针对目前对船舶柴油机示功图的分析评估大多通过人工进行,其结果依赖于轮机管理人员的业务能力与经验,存在不确定性的问题,构建了一种基于卷积神经网络的示功图智能分析监测模型.该模型包括5层卷积神经网络和2层全连接神经网络,采用多种优化算法并使用GPU进行硬件加速.训练与实验结果表明,示功图训练精度为99.5%,识别精度为95.9%,平均识别时间为0.032 s,验证了该模型的可靠性和准确性,可以对示功图进行智能分析,满足柴油机燃烧工况智能监测的精度要求,为实现船舶动力系统智能化管理提供支持.A convolutional neural network-based intelligent analysis and monitoring model for marine diesel engine indicator diagrams is constructed to address the issue of uncertainty in the current analysis and evaluation of indicator diagrams,which mostly rely on manual work and rely on the business capabilities and experience of engine management personnel.The model includes 5 layers of convolutional neural networks and 2 layers of fully connected neural networks,and adopts various optimization algorithms and hardware acceleration using GPU.Through training and experiments,the results show that the training accuracy of the indicator diagram is 99.5%,the recognition accuracy is 95.9%,and the average recognition time is 0.032 seconds,which verifies the reliability and accuracy of the model.It can achieve intelligent analysis of the indicator diagram,meet the accuracy requirements of intelligent monitoring of diesel engine combustion conditions,and provide support for the intelligent management of ship power systems.

关 键 词：船舶柴油机 示功图 卷积神经网络 智能分析 燃烧工况 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]