【作者】 陈曦；

【导师】 李言； 杨明顺；

【作者基本信息】 西安理工大学 ， 机械制造及其自动化， 2021， 硕士

【摘要】 对加工设备故障准确预测,合理规划运维活动是制造企业保障生产秩序、预防安全事故、节约维护资源、提高经济效益的有效途径。传统制造过程中,由于缺少对加工设备运行状态的有效监控,难以通过实时获取运行参数来对加工设备故障进行预判,设备维护通常只能采用事后或定期方式进行维护,容易因突发故障造成生产停机和维护浪费。物联网、大数据、人工智能等技术在制造业中的应用,促进了传统制造向智能制造的转变,为传统制造业转型升级提供了有效的解决途径。论文通过大数据、深度学习等技术,对加工设备的故障进行预测、维护决策以及系统实现进行研究,具有重要的意义。论文分析了基于物联网和大数据的加工设备故障预测和维护决策的方法和流程,建立了加工设备故障预测和维护决策整体架构,讨论了实现加工设备故障预测和基于预测信息进行维护决策的关键技术。基于估计威布尔分布参数的多任务故障预测框架,构建了一种门控循环单元与时域卷积网络相结合的“威布尔分布-时域卷积-门控循环神经网络”。结合故障预测模型,通过预测故障概率及其分布规律参数,得到设备故障的变化趋势,算例结果表明,与同类算法相比,该方法在预测精度、召回率等指标上有更好的优越性。建立一种多样本均值聚类与分组搜索近邻算法相结合的故障预测模型,有效解决了因加工设备状态参数采集缺失或参数异常导致预测不准确的问题,算例验证了该方法是参数缺失或异常情况下故障预测的理想方法。针对设备运维问题,综合考虑故障的预测信息、维护资源、维护累积效果衰退效应等因素,构建了以维护调度总费用率最小为目标的设备维护决策模型,采用禁忌搜索算法对模型求解,算例验证了模型的正确有效性。在上述研究的基础上,采用Hadoop、Spark等技术建立了加工设备故障预测与维护决策原型系统,利用大数据分布式文件系统实现对海量数据的存储并优化系统的计算能力,并实现了故障预测、维护决策、信息管理等多个功能,有效辅助管理维护人员的决策。更多还原

【Abstract】 Accurate prediction of processing equipment failures and reasonable planning of operation and maintenance activities are effective ways for manufacturing companies to ensure production order,prevent safety accidents,save maintenance resources,and improve economic benefits.In the traditional manufacturing process,due to the lack of effective monitoring of the operating status of processing equipment,it is difficult to predict processing equipment failures by obtaininng operating parameters in real time.Equipment maintenance usually can only be done after the fact or on a regular basis,and production downtime and maintenance waste is easily caused by sudden failures.The application of technologies such as the Internet of Things,big data,and artificial intelligence in the manufacturing industry has promoted the transformation from traditional manufacturing to intelligent manufacturing,and has provided an effective solution for the transformation and upgrading of traditional manufacturing.The paper uses big data,deep learning and other technologies to study the fault prediction of processing equipment,maintenance decision-making and system realization,which is of great significance.Analyze the methods and processes of processing equipment failure prediction and maintenance decision-making based on the Internet of Things and big data,establishes an overall framework for processing equipment failure prediction and maintenance decision-making,and discusses the key to realizing processing equipment failure prediction and maintenance decisionmaking based on predictive information technology.Based on the multi-task fault prediction framework for estimating Weibull distribution parameters,a "Weibull distribution-time domain convolution-gated recurrent neural network"combining gated recurrent units and time domain convolutional networks is constructed.Combined with the failure prediction model,the change trend of equipment failure is obtained by predicting the failure probability and its distribution law parameters.The results of calculation examples show that compared with similar algorithms,this method has better advantages in prediction accuracy,recall rate and other indicators.A fault prediction model combining multi-sample mean clustering and group search nearest neighbor algorithm is established,which effectively solves the problem of inaccurate prediction due to missing or abnormal parameters of processing equipment status.The calculation example verifies that the method is an ideal method for failure prediction under missing parameters or abnormal conditions.Aiming at the equipment operation and maintenance problem,comprehensively considering factors such as failure prediction information,maintenance resources,and the cumulative effect of maintenance,the equipment maintenance decision-making model is constructed with the goal of minimizing the total cost of maintenance and dispatching.The tabu search algorithm is used to solve the model.The example verifies the correctness and validity of the model.On the basis of the above research,a prototype system for processing equipment fault prediction and maintenance decision-making was established using Hadoop,Spark and other technologies,and a large data distributed file system was used to store massive amounts of data and optimize the system’s computing power,and to achieve fault prediction,maintenance decision-making,information management and other functions,effectively assisting management and maintenance personnel in decision-making.更多还原