基于随机森林算法的酒精浓度在线测量系统

On-Line Alcohol Concentration Measurement System Based on Random Forest Algorithm

酒精度是酒类生产过程中重要的技术指标,目前对于酒精度的测量方法过程繁琐,耗时较长,容易污染浪费,本文提出一种基于微波测量和随机森林算法的在线酒精浓度测量系统。首先基于传输反射法和有限元法(FEM)设计双端口测量装置,优化腔体结构,使S11和S21与酒精浓度呈单调变化;然后基于FEM算法计算腔体内不同介电常数的S11和S21,得到4081组训练数据;基于随机森林算法,训练反演网络;最后引入非线性回归校准算法,依据浓度与介电实部的关系获得待测样品的酒精浓度,测试准确度高达99%。本系统结构简单,可以用于酒精浓度的非接触式在线测量,测量精度高,避免酒精的污染浪费,有效地提高了生产效率。

Alcohol content is an important technical index in wine production process.Present measurement methods of alcohol content are cumbersome,time-consuming and easy to pollution and waste.An on-line alcohol concentration measurement system based on microwave measurement and random forest algorithm is proposed.Firstly,a two-port measuring device is designed based on transmission reflection method and finite element method(FEM),and the cavity structure is optimized to make S 11 and S 21 parameters changing monotonically with alcohol concentration.Then the S 11 and S 21 with different permittivity in the cavity are calculated based on FEM algorithm.4081 groups of training data are obtained,and the inversion network is trained based on random forest algorithm.And finally,a nonlinear regression calibration algorithm is introduced to obtain the alcohol concentration of the tested sample according to the relation between concentration and dielectric real part,which shows a testing accuracy up to 99%.The system has a simple structure and can be used for non-contact and on-line measurement of alcohol concentration with high measurement accuracy,avoiding alcohol pollution and waste,and effectively improving production efficiency.