基于复数卷积神经网络烧结台车定位跟踪算法

Positioning and tracking algorithm for sintering trolleys based on convolutional neural network of complex number

烧结台车定位跟踪是非稳态烧结高质量过程控制的基础,同时也是台车状态监控的关键。针对现有烧结生产过程中的物料跟踪和过程控制的问题,提出了一种全新的台车定位跟踪算法。该算法针对烧结台车定位场景设计了复数卷积神经网络结构,运用复数网络的俯角解决了相位从360°到0°跳变时定位系统误差大的问题;同时采用GradCAM技术解析网络的工作原理,以确保模型的泛化能力;通过在台车上加装定位二维码来二次确认定位信息的准确性。经过长达一年半的论证,证明系统具有高度的可靠性和稳定性,并能够有效地应用于烧结生产过程中的物料跟踪和过程控制。

The positioning and tracking of sintering trolleys is the basis for high-quality process control of unsteady sintering,and it is also the key to the condition monitoring of trolleys.In view of the problem of material tracking and process control in the existing sintering production process,a new trolley positioning and tracking algorithm is proposed.The algorithm designs a convolutional neural network structure of complex number for the positioning scenario of sintering trolleys,and uses the depression angle of the complex number network to solve the problem of large positioning system error when the phase jumps from 360°to 0°.At the same time,GradCAM technology is used to analyze the working principle of the network to ensure the generalization ability of the model.In addition,the accuracy of the positioning information is confirmed twice by installing a positioning QR code on trolleys.After a year and a half of demonstration,it has been proved that the system has a high degree of reliability and stability,and can be effectively applied to material tracking and process control in the sintering production process.