Deep learning-based on-line image analysis for continuous industrial crystallization processes--颗粒学报PARTICUOLOGY

Zong, S., Zhou, G., Li, M., & Wang, X. (2023). Deep learning-based on-line image analysis for continuous industrial crystallization processes. Particuology, 74, 173-183. https://doi.org/10.1016/j.partic.2022.07.002

Deep learning-based on-line image analysis for continuous industrial crystallization processes

Shiliang Zong, Guangzheng Zhou ^*, Meng Li, Xuezhong Wang ^*

Beijing City Key Laboratory of Enze Biomass and Fine Chemicals, College of New Materials and Chemical Engineering, Beijing Institute of Petrochemical Technology, Beijing, 102617, China

10.1016/j.partic.2022.07.002

Volume 74, March 2023, Pages 173-183

Received 24 January 2022, Revised 27 June 2022, Accepted 2 July 2022, Available online 16 July 2022, Version of Record 3 August 2022.

E-mail: zhouguangzheng@bipt.edu.cn; wangxuezhong@bipt.edu.cn

Highlights

• Deep learning is investigated for crystal segmentation of high solid concentrations.

• Large increase of crystal number with detailed label points enhances model performance.

• Mask R-CNN well recognizes overlapped crystals at high concentrations.

• Mask R-CNN performs much better than multi-scale method in precision and recall.

• Accurate statistics of geometrical features facilitate crystallization understanding.

Abstract

In situ microscopic imaging is a useful tool in monitoring crystallization processes, including crystal nucleation, growth, aggregation and breakage, as well as possible polymorphic transition. To convert the qualitative information to be quantitative for the purpose of process optimization and control, accurate analysis of crystal images is essential. However, the accuracy of image segmentation with traditional methods is largely affected by many factors, including solid concentration and image quality. In this study, the deep learning technique using mask region-based convolutional neural network (Mask R-CNN) is investigated for the analysis of on-line images from an industrial crystallizer of 10 m³ operated in continuous mode with high solid concentration and overlapped particles. With detailed label points for each crystal and transfer learning technique, two models trained with 70,908 and 7,709 crystals respectively are compared for the effect of training data amount. The former model effectively segments the aggregated and overlapped crystals even at high solid concentrations. Moreover, it performs much better than the latter one and traditional multi-scale method both in terms of precision and recall, revealing the importance of large number of crystals in deep learning. Some geometrical characteristics of segmented crystals are also analyzed, involving equivalent diameter, circularity, and aspect ratio.

Graphical abstract

Keywords

Continuous crystallization; Crystal shapeImage analysis; Deep learningInstance segmentation; Process analytical technology

文章导读