Data‐driven modeling for crystal size distribution parameters in cane sugar crystallization process

• DOI: 10.1111/jfpe.13648
• Publication Date: 2021-01-19T01:17:26Z
• Authors: Yanmei Meng; Tao Yao; Shuangshuang Yu; Johnny Qin; Jinlai Zhang; Jianfan Wu
• Keywords: Crystal (programming language)

Abstract Crystal size distribution (CSD) is important in evaluating crystal quality cane sugar crystallization process. Due to the complex non‐linearity, time‐delay and strong coupling process, it difficult directly modeling mechanism of process obtain CSD parameters. In order two main parameters so that achieve better control production sugar, this article constructs a data‐driven model based on least squares support vector regression (LSSVR) particle swarm optimization (PSO). Based LSSVR, takes easy measureable variables (massecuite brix, massecuite level, temperature, steam pressure, feeding volume, vacuum degree) as input variables, outputs (crystal average size, coefficient variation size). PSO algorithm used optimize key primary get performance. Compared with other methods such back propagation, extreme learning machine, radial basis function, SVR, constructed PSO‐LSSVR has obvious advantages over models speed predictive effect, generalization ability. This potential be applied system product sugar. Practical applications The uses (which not measured online directly), can control. According relation between parameters, useful set up operation conditions, form environment, improve crystal. So, method will benefit efficiency also for study continuous