Abstract:

The batch distillation process is an important separation and purification process, and its operating condition prediction plays an important role in ensuring the smooth operation of the batch distillation process, optimizing the production quality and yield of batch distillation. This article conducted in-depth research on the model establishment, operating condition prediction algorithm, and simulation software design of the fine chemical D1 batch distillation process. Firstly, a data-driven model for the D1 batch distillation process was established using operation data of historical production, combined with the characteristics of long short term memory (LSTM) and back propagation (BP) neural networks, to predict the rising vapor temperature, distillate temperature, distillation endpoint time, and final product purity. Then, the above work was combined through Matlab's graphical user interface (GUI) to develop a simulation GUI for the D1 batch distillation process, which achieved the prediction of operating parameters and control simulation from data processing to final results. The simulation test results showed that the prediction of batch distillation conditions was fast and accurate, and had important reference value for guiding actual process operations.

Key words: batch distillation process, model, prediction, neural networks

摘要：

间歇蒸馏过程是一种重要的分离提纯工艺过程，其工况预测对保障间歇蒸馏过程平稳运行、优化间歇蒸馏生产质量和产量具有重要作用。本文对精细化学品D1间歇蒸馏过程的模型建立、工况预测算法以及仿真软件设计进行了深入研究。首先，利用历史生产运行数据，结合长短期记忆神经网络（long short term memory, LSTM）和反向传播（back propagation，BP）神经网络特点建立D1间歇蒸馏过程的数据驱动模型，实现了上升气温度、转馏分温度、蒸馏终点时间以及最终产品纯度的预测；然后，通过Matlab图形用户界面（graphical user interface，GUI）将上述工作结合起来，开发了D1间歇蒸馏过程仿真GUI，实现了从数据处理到最终工况参数预测以及控制仿真。仿真测试结果表明，间歇蒸馏工况预测快速、准确，对指导实际工艺操作具有重要参考价值。

关键词: 间歇蒸馏过程, 模型, 预测, 神经网络