Abstract: Mining the implied value of industrial big data is an important research direction of intelligent manufacturing,and carrying out outlier detection is a prerequisite for realizing data analysis.The main problems addressed by industrial big data anomaly detection were introduced,and the relevant definitions in this paper were proposed.Based on the Multi-Objective Particle Swarm Optimization algorithm(MOPSO),an improved DBSCAN model for industrial big data outlier detection was proposed.The algorithm design idea and algorithm steps of the model were introduced,the pseudo-code of the algorithm was completed,and the calculation the time complexity of the algorithm was proposed.The model simulation and experiments were carried out by using the manufacturing data of an electric core factory,and it was verified that the proposed model could improve the accuracy of industrial big data outlier detection.This paper provided a reference for the application of data mining in industrial outlier detection.

Key words: industrial big data, outlier detection, multi-objective particle swarm optimization algorithm, DBSCAN model

摘要： 挖掘工业大数据的隐含价值是智能制造的一个重要研究方向,针对工业大数据特点开展异常点检测是实现数据分析的前提。首先,介绍了工业大数据异常点检测解决的主要问题,提出相关定义。其次,基于多目标粒子群算法(MOPSO),提出一种工业大数据异常点检测的改进DBSCAN模型,介绍了模型的算法设计思想、算法步骤,完成了算法伪代码的编写,并提出了算法时间复杂度的计算方法。最后,通过某电芯工厂制造数据,进行了模型仿真与实验,经实验验证,所提模型提高了工业大数据异常点检测的准确率,为数据挖掘在工业异常点检测中的应用提供了参考。

关键词: 工业大数据, 异常点检测, 多目标粒子群算法, DBSCAN模型