制造过程数据分布式协调优化方法研究

Nonferrous smelting and petrochemical industries are typical long manufacturing processes. This kind of process is characterized by a series of multiple stages, serious noise pollution of process data, large data temporal and spatial span, and strong coupling and uncertainties of each stage, resulting in low data accuracy. With the development of big data and artificial intelligence, there is an urgent need to improve data quality. To this end, a data distributed reconciliation optimization method is proposed for manufacturing process. At first, the temporal and spatial characteristics of the process data are studied, and the multi-delay interval of material staying in different equipment is estimated. Using the multi-delay intervals, a fast method for solving integer time sequence is designed and a time registration method is investigated based on time correlation matrix. Then, in order to realize data type partition and data redundancy analysis, the measurement network structures of multiple sub-stages are constructed based on graph theory. Meanwhile, the robust estimation function is designed and the robust data reconciliation modeling method with uncertainty information is studied. After that, a mode identification method based on the Gaussian mixture model is proposed. Moreover, process correlation knowledge and data characteristics are mined according to the bayesian network, distributed data reconciliation optimization method is researched for long process with multiple stages. This project is of great significant to further promote the effectiveness of process data, ensure a high-quality, high-efficient operation, and low-consuming for manufacturing process.

有色冶炼、石油化工等属于典型长流程制造生产过程，这类过程存在多工序级联、数据噪声污染严重、数据时空跨度大、各工序关联耦合且不确定性强等特点，造成数据准确性低。随着大数据、人工智能的发展，提高数据质量具有迫切需求。为此，本项目提出制造过程数据分布式协调优化方法。首先，研究过程数据时空特性，预估物料停留不同设备的多重时滞区间；利用多重时滞区间，设计整数型时基序列快速求解方法，研究基于时效关联矩阵的数据时间配准方法。其次，基于图论知识构造子工序的测量网络结构图，以此实现数据类型划分及数据冗余性分析；设计鲁棒估计函数，研究具有不确定信息的鲁棒数据协调建模方法。最后，基于高斯混合模型提出一种不同生产模态辨识方法，根据贝叶斯网络挖掘工序间关联知识和数据特征，研究长流程多工序多层级分布式数据协调优化方法。本项目研究对进一步提高制造过程数据的有效性，保证过程优质高效运行和节能降耗具有重要意义。

本项目从高数据质量需求角度研究制造过程数据分布式协调优化方法，以此解决长流程制造生产过程中多工序级联、数据噪声污染严重、数据时空跨度大、各工序关联耦合且不确定性强等特点造成数据准确性低等问题。研究了过程数据时空特性，预估了物料在不同设备中的多重时滞区间，并设计了整数型时基序列快速求解方法，提出了基于时效关联矩阵的数据时间配准方法。针对不确定因素干扰下的数据协调建模问题，基于图论知识和鲁棒估计函数研究了具有不确定信息的鲁棒数据协调建模方法。考虑复杂流程结构以及多变工况特征，基于高斯混合模型提出了一种不同生产模态辨识方法，研究了贝叶斯网络下关联知识挖掘的多工序多层级分布式数据协调优化策略。本项目的研究成果有效地推动了过程数据协调方法的理论发展，对进一步提高长流程制造过程数据可靠性，确保过程优质高效运行和节能降耗具有重要意义。在本项目实施过程中，共发表SCI论文6篇，申请发明专利2项，按计划完成项目研究。本项目成果也应用于有色冶炼、石油化工等领域，所涉及的长流程制造过程数据协调优化方法具有一定的推广前景。