基于改进深度信念网络的锻锤磨损状态预测  

Prediction of Wear Status of Forging Hammers Based on Improved Deep Belief Network

作　　者：田萌[1] Tian Meng(Department of Mechanical and Electrical Engineering,Henan Vocational College of Industry and Trade,Zhengzhou Henan 451191,China)

机构地区：[1]河南工业贸易职业学院机电工程系,河南郑州451191

出　　处：《山西冶金》2023年第9期83-84,92,共3页Shanxi Metallurgy

基　　金：河南省高等学校重点科研项目(22A470005)。

摘　　要：锻床是冶金领域最常见的设备之一,其运行效率影响着冶金成形的质量。建立了一种锻床快速在位检测系统,可以高效辨别锻锤图像,并以锤面磨损数据建立预测仿真模型。研究结果表明:加入Dropout后,有助于提升模型学习性能,促进网络预测性能提升,还能够更快完成特征匹配收敛过程。通过优化DBN模型,预测结果更加稳定与准确。该研究具有很好的理论支撑价值,可有效应用于实际的机加工过程。Forging bed is one of the most common equipment in the field of metallurgy,and its operation efficiency affects the quality of metallurgical forming.A fast in-place detection system was established for NC forging bed,which can realize the high efficiency discrimination of forging hammer image,and then the prediction simulation model was established based on the tool surface wear data.The results show that adding Dropout can improve the model learning performance and improve the network prediction performance.It can also complete the feature matching convergence process faster,and the optimized DBN model shows more stable and accurate prediction results.The research can be effectively applied in the actual machining process and has a good theoretical support value.

关 键 词：视觉检测 全生命周期 深度学习 DROPOUT 锻锤磨损预测 

分 类 号：TH16[机械工程—机械制造及自动化]