基于卷积神经网络的智能制造过程质量异常诊断  

Quality Abnormal Recognition Model Based on Convolutional Neural Network

作　　者：王宁[1] 李盼盼 赵哲耘 杨剑锋[1] WANG Ning;LI Pan-pan;ZHAO Zhe-yun;YANG Jian-feng(Business School,Zhengzhou University,Zhengzhou 450001,China;School of Marxism,Zhengzhou University,Zhengzhou 450001,China;Department of Development and Planning Off,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学商学院,河南郑州450001 [2]郑州大学马克思主义学院,河南郑州450001 [3]郑州大学发展规划处,河南郑州450001

出　　处：《运筹与管理》2022年第6期220-225,共6页Operations Research and Management Science

基　　金：国家社科基金资助项目(20BTJ059);国家自然科学基金资助项目(U1904211);河南省高等学校青年骨干教师培养项目(2021GGJS006)。

摘　　要：针对现有方法在智能制造过程中诊断能力有限和识别精度不高的问题,提出了一种与智能制造过程相适应的基于卷积神经网络的质量异常诊断模型。首先建立基于实时数据的过程质量图谱,以精准表达制造过程运行状态。其次,构建用于识别质量图谱的卷积神经网络诊断模型。最后,利用滑动窗口取值的方式对当前过程运行状态进行动态诊断,并通过某球磨过程验证了所提方法的有效性与实用性。结果表明,所提方法优于传统浅层模型,能够有效的对过程异常状态进行识别与诊断。Aiming at the problems of limited diagnostic ability and low recognition accuracy of existing methods in the intelligent manufacturing process,a quality anomaly diagnosis model based on convolutional neural network is proposed to adapt to intelligent manufacturing process.Firstly,the process quality spectra based on real-time data are established to accurately express the operating status of the manufacturing process.Secondly,a convolutional neural network diagnosis model is constructed to identify quality spectra.Finally,the dynamic diagnosis of the current process running state is carried out by using the sliding window value method,and the effectiveness and practicability of the proposed method are verified by a ball milling process.The results show that the proposed method is superior to the traditional shallow model and can effectively identify and diagnose abnormal process states.

关 键 词：制造过程 卷积神经网络 质量图谱 

分 类 号：F273[经济管理—企业管理]