Abstract:

Chemical biomanufacturing process has the advantages of green, low-carbon, environmentally friendly and sustainable, and its role in the chemical industry is becoming increasingly important. However, the biosynthesis monitoring, control and optimization as well as the separation of products during the biomanufacturing processes always have real-time dynamic complexity due to the influences of metabolicregulation and external environmental factors. The complex nonlinear relationships in data from these dynamic processes can be obtained by machine learning methodology without the need for explicit understanding of process mechanisms, which could then be helpful to reveal the bioprocess laws, and thus improve the optimization and prediction of biomanufacturing processes. In this paper, recent advances regarding the machine learning methodology, key algorithms and typical applications in the optimization, monitoring and control of biosynthesis processes, the development of bioseparation processes, and the production of biofuel chemicals, were summarized and analyzed. The challenges to be addressed in further applications of machine learning in the manufacturing process of chemicals were also discussed.

Key words: machine learning, biomanufacturing, bioprocess engineering, bioseparation, algorithm

摘要：

化学品生物制造过程具有绿色低碳、环境友好和可持续性优势，在化学工业中的作用日益重要。然而，化学品生物制造过程受代谢调控和外源环境等诸多因素影响，其生物合成的监测、控制、优化及产物分离，都具有实时动态复杂性。机器学习在无须明确机理条件下即可捕捉动态过程数据中的复杂非线性关系，有助于掌握化学品生物制造过程的复杂规律，进行过程优化和预测。本文对机器学习范式、主流算法及其在生物合成过程优化、监测与控制、生物分离过程开发和生物燃料化学品生产中的研究进展进行了综述分析，探讨了未来机器学习用于化学品生物制造过程中的问题。

关键词: 机器学习, 生物制造, 生物过程工程, 生物分离, 算法