摘要
基于深度学习的故障诊断是当前工业安全智能化管理的重要研究方向。工业实际生产中故障时常发生,导致生产效率下降,严重时会造成停产甚至人员伤亡。由于生产环境复杂多变,导致故障特征难以提取和识别,且工业现场需要实时监测和快速诊断,传统故障诊断方法通常依赖专家经验进行特征提取和模式识别,难以适应复杂动态的工业环境。针对上述问题,提出了一种基于深度学习的多模态工业故障诊断方法。采用极端梯度提升(XGBoost)筛选与工业故障相关的工艺参数,以此作为模型输入的多模态数据。通过深度极限学习机(DELM)提取生产工艺参数的非线性和高维特征,识别出异常状态的工业设备,并利用霜冰优化算法(RIME)优化DELM的关键参数,使模型达到最佳性能。RIME-DELM输出正常状态的设备样本,异常设备样本则继续输入至深度置信网络(DBN)和最小二乘支持向量机(LSSVM),对异常样本进行故障类型的具体判别。将所提出的方法应用于铝电解生产过程,验证了模型的有效性。经铝电解生产现场实验结果表明,该模型的异常状态检测的准确率为97.96%,F1-score为0.9753,故障类型诊断的准确率为96.75%,Macro-F1分数为0.9447,通过消融实验、与常见深度学习模型对比,本文构建的DMIFD模型诊断精度更高,为实际工业生产的故障诊断提高了更准确、可靠的支持。
Fault diagnosis based on deep learning is an important research direction in the current intelligent management of industrial safety.Failures often occur in the actual production of industry,resulting in a decline in production efficiency,which can cause serious production stoppages and even casualties.Due to the complexity and variability of the production environment,which makes it difficult to extract and identify fault features,and the industrial site requires real-time monitoring and rapid diagnosis,the traditional fault diagnosis methods usually rely on the experience of experts to carry out feature extraction and pattern recognition,which is difficult to adapt to the complex and dynamic industrial environment.To address the above problems,a deep learning based Multimodal Industrial Fault Diagnosis DMIFD(Deep learning based Multimodal Industrial Fault Diagnosis)method is proposed.Extreme gradient boosting(XGBoost)is used to filter the process parameters related to industrial faults,which is used as the multimodal data for model input.The nonlinear and high-dimensional features of the production process parameters are extracted by Deep Extreme Learning Machine(DELM)to identify industrial equipment in abnormal states,and the key parameters of DELM are optimised using the Frost and Ice Optimisation Algorithm(RIME)to achieve the best performance of the model.RIME-DELM outputs samples of equipment in normal states,and samples of abnormal equipment are continued to be inputted into the Deep Belief Network(DBN)and the Minimal Squares Support Vector Machine(LSSVM)for specific discrimination of fault types for abnormal samples.The effectiveness of the model is verified by applying the proposed method to the aluminium electrolysis production process.The results of the aluminium electrolysis production field experiments show that the model has an abnormal state detection accuracy of 97.96%,an F1-score of 0.9753,a fault type diagnostic accuracy of 96.75%,and a Macro-F1 score of 0.9447.Through ablation experiments,and comparisons with the common deep learning models,the DMIFD model constructed in this paper has a higher diagnostic accuracy,and it can be used for practical industrial production fault diagnosis improves more accurate and reliable support.Fault diagnosis based on deep learning is an important research direction in the intelligent management of industrial safety.Failures frequently occur in the actual industry production,leading to reduced production efficiency,and in severe cases,production stoppages or even casualties.Due to the complex and variable production environment,fault features are difficult to extract and recognize,andreal-time monitoring and rapid diagnosis are required at industrial sites.Traditional fault diagnosis methods typically rely on expert experience for feature extraction and pattern recognition,which makes them difficult to adapt to the complex and dynamic industrial environments.To address these issues,a deep learning based multimodal industrial fault diagnosis(DMIFD)method is proposed.Extreme gradient boosting(XGBoost)is employed to select process parameters related to industrial faults,which are then used as the multimodal input data for the model.The deep extreme learning machine(DELM)is used to extract nonlinear and high-dimensional features from the production process parameters to identify equipment in abnormal states.The key parameters of DELM are optimised using the frost and ice optimisation algorithm(FIOA)to achieve optimal model performance.The RIME-DELM module outputs equipment samples in normal states,while the samples of abnormal equipment are further input into a deep belief network(DBN)and a least minimal squares support vector machine(LSSVM)to perform specific fault type classification.The proposed method is applied to the aluminium electrolysis production process to validate its effectiveness.Field conducted in an aluminum electrolysis plant show that the model achieves an abnormal state detection accuracy of 97.96%,an F1-score of 0.9753,a fault type diagnostic accuracy of 96.75%,and a Macro-F1 score of 0.9447.Compared with common deep learning models and through ablation experiments,the DMIFD model demonstrates higher diagnostic accuracy and provides more accurate and reliable support for fault diagnosis in practical industrial production.
作者
尹刚
朱淼
颜玥涵
王怀江
江茂华
刘期烈
Yin Gang;Zhu Miao;Yan Yuehan;Wang Huaijiang;Jiang Maohua;Liu Qilie(State Key Laboratory of Coal Mine Disaster Dynamics and Control,Chongqing University,Chongqing 400044,China;College of Resource and Safety Engineering,Chongqing University,Chongqing 400044,China;School of Electrical Engineering,Guizhou University,Guiyang 550025,China;Chinalco Zhengzhou Research Institute,Zhengzhou 450041,China;Chongqing Gaofeng New Materials Technology Co.,Ltd.,Chongqing 404160,China;Chongqing Post and Communications University,Chongqing 400065,China)
出处
《仪器仪表学报》
北大核心
2025年第6期215-227,共13页
Chinese Journal of Scientific Instrument
基金
重庆市自然科学基金创新发展联合基金项目(CSTB2024NSCQ-LZX0166)
重庆英才·创新创业示范团队项目(cstc2024ycjh-bgzxm0131)
科技转化重大项目(H20201555)资助。
关键词
深度学习
故障诊断
多模态融合
神经网络
过程控制
deep learning
fault diagnosis
multimodal fusion
neural networks
process control
