数据与知识协同的铝电解生产工况溯因识别方法

Aluminum electrolysis industry is facing the austere issues, such as overcapacity, high cost and environmental protection, in addition, aluminum electrolysis enterprises are often in a critical state of profit and loss. Therefore it is urgent to put forward the fine control and precise identification of the production process for aluminum electrolysis industry. However, traditional manual analysis and operation are unable to cope with the challenge. The precise and automatic identification is restricted by knowledge relationship hard to describe, inefficient method of data-knowledge fusion and lack of automatic identification algorithms. To this end, according to the ideology of knowledge automation, combining with the process knowledge solidification and extraction of characteristics of individualized conditions from data for reasoning, to build the relationship network model of knowledge with posterior relevance and layered cross-domain. Research on the collaborative method of characteristics of individualized and knowledge for heterogeneous data, to propose the root cause analysis and cell condition identification method for knowledge network model based on joint reasoning, in addition, it will get application validation of cell condition identification based on root cause analysis, resulting in getting the identification method and system of individualized cell conditions for the aluminum electrolysis process based on the synergy between data and knowledge in this subject. This subject provides a effective and scientific way to support the fine, personalized and intelligent control for the aluminum electrolysis process in the cell condition identification, and it is the enrichment of knowledge automation theory and methods. Moreover, the subject plays an important role in the improvement of production efficiency, promoting energy-saving emission reduction and the green production model changing of aluminum electrolytic industry.

铝电解行业面临资源、能源、环保和高素质人员短缺等方面的重大挑战，企业常常处于临界盈亏状态，对生产过程精细化控制和工况自动准确识别提出了迫切需求，而目前仍依赖操作人员进行电解槽工况识别方式。实现工况自动准确识别的难点主要在于铝电解过程领域知识复杂，知识关系描述困难，缺乏有效的数据-知识融合机制和智能识别方法。为此，本课题基于知识自动化的思想，将工艺知识库、数据-知识协同方法、推理识别方法相结合，构建具有后验关联性的分层跨域知识关系网络模型，研究异构关联数据的个性化特征提取与知识协同方法，提出基于联合推理的知识网络溯因分析与工况识别方法，实现个性化槽况智能识别验证，从而形成数据与知识协同的铝电解生产过程个性化工况溯因识别方法。本课题将在工况识别方面为实现铝电解生产过程精细化、个性化、智能化控制提供有力的理论方法支撑，丰富知识自动化的科学研究内容，对于实现铝电解工业精细化绿色化生产具有重要意义。

本课题基于知识自动化的思想，将工艺知识、数据-知识协同方法、推理识别方法相结合，提出了数据与知识协同的铝电解生产工况溯因识别方法。主要成果包括建立了2类分层跨域知识关系网络模型；提出了3种异构关联数据的个性化特征提取与知识协同方法；提出3种基于联合推理的知识网络溯因分析与工况识别方法；并开发了基于溯因分析的槽况自动识别系统作为铝电解工业数据分析与智能决策系统的重要功能模块，相关成果经过成套技术应用获2021年度中国有色金属工业科学技术一等奖。项目组在研究期限内发表相关学术期刊论文19篇，其中SCI收录15篇（含JCR1区论文4篇），EI收录4篇，授权国家发明专利12项，申请软件著作权2项，培养博士研究生5人，硕士研究生14人，完成了全部预期研究目标。本项目为铝电解生产过程精细化、个性化、智能化的工况识别和溯因分析提供了有力的理论方法支撑，对于工业人工智能和知识自动化领域研究具有重要的科学意义。

内容获取失败，请点击重试