多组分稀土萃取分离过程优化控制关键技术及应用

This project is aimed at difficulties in the extraction process optimization and control of rare earth with multiple component contents, such as lack of effective means in process reconstruction and technology optimization, shortness of effective methods in rapid detection of multiple component contents, and devoid of reliable models for control of the whole process. This project presents a new intelligent modeling approach and a new simulation method of the whole process for extraction process of rare earth with multiple component contents based on the operational principle, information and process data, and develops a self-allocation method and a technology optimization method with consideration of multiple technical and economic target. Moreover, a rapid detection method based on multiple detection modes is addressed, and a new control approach based on the extraction process model and data drive method is designed. Such a research will result in new intelligent optimal control theory and technology of rare earth’s extraction process which has self-owned intellectual property. Its application will change the old mode in rare earth industry which is off-line analysis, empirical control and hand regulation, improve technical and economic level, and reach industry restructuring as well as sustainable development. It develops new modeling and control theory and techniques for complex industrial processes, which also improves related disciplines.

项目针对多组分稀土萃取分离全流程运行优化面临的流程重组与工艺优化缺乏有效手段、多组分含量快速检测缺乏有效方法、全流程优化控制缺乏可靠模型支撑的难题，研究提出融合萃取机理、知识与运行数据的多组分稀土萃取过程智能建模与全流程模拟方法，基于综合技术经济指标的多组分稀土萃取流程动态自主配置及全流程工艺优化方法，融合多检测模式的稀土萃取过程多组分含量快速检测方法，基于模型与数据驱动的多组分稀土萃取全流程优化控制方法；形成具有自主知识产权的稀土萃取过程智能优化控制理论和关键技术。成果的工业应用将突破稀土行业长期依赖“离线分析、经验控制、手动调整”的人工控制模式，显著提升稀土行业技术经济指标，促进行业转型升级和绿色可持续发展。项目研究同时可为复杂工业过程建模与优化控制提供新理论与方法，对促进相关学科的发展具有十分重要的科学意义。

稀土被誉为“工业维生素”，是我国不可再生的重要战略资源。项目团队首先针对稀土萃取机理的流程仿真模型没有考虑萃取槽的萃取效率，引入能效分离系数概念，建立符合实际萃取工业的各萃取槽稀土元素组分含量模型，构建求解能效分离系数的优化目标函数。基于稀土萃取工业大数据，首次采用多分支深层神经网络构建稀土萃取流程模拟模型。其次，针对复杂工业生产现场光照复杂、光衰等光照条件变化导致颜色特征失色问题，提出一种基于Grey Edge参数主动变化的光照补偿算法和基于改进即时学习算法的元素组分含量检测方法。最后，基于工业数据和专家经验，提出一种案例推理静态设定和模糊推理动态补偿相结合的药剂量优化控制方法，提出一种基于事件触发的多组分稀土萃取过程分布式预测控制方法以减少通信频次和控制器计算负担。此外，项目团队针对稀土工业数据提出了一系列降维、特征选择、聚类等原创性机器学习方法。所形成的多组分稀土萃取全分离智能优化控制关键技术及系统，对实现两化深度融合、稀土工业绿色可持续发展具有重要现实意义和广阔的应用前景。. 项目执行期间，申请美国、中国、日本发明专利26项（已获授权15项），获软件著作权8项；在国际、国内重要期刊和学术会议发表学术论文50余篇（SCI/EI收录30余篇）；组织承办国际、国内学术会议5次；培养博士、硕士研究生20余人；超额完成预期目标。