Research on alcohol prediction model based on LSTM and IGA-BP

Abstract

Objective: In order to solve the problem of inaccurate detection of alcohol accuracy due to manual "flower picking" in the segmented liquor picking process. Methods: Designed and built a segmented liquor taking system based on alcohol accuracy modeling. The research collected tuning fork frequency values, tuning fork built-in temperature values, alcohol solution temperature values and pump speed values under dynamic conditions in different modes of alcohol solutions with different concentrations, implemented adaptive tuning fork frequency filtering and dynamic compensation based on least mean square algorithm （LMS） and long short-term memory network （LSTM）, and built a liquor accuracy prediction model based on improved genetic algorithm optimized BP neural network （IGA-BP）. Results: The model outperformed the traditional genetic algorithm optimized BP neural network and BP neural network in terms of the number of iterations and prediction accuracy, and the average prediction error of the alcoholic beverages was 0.381. Conclusion: which verifies the reasonableness of the model. In order to solve the limitations of the current manual "flower picking", a method is proposed to improve the accuracy of alcohol detection in the segmented liquor taking process.

Keywords

• liquor taking
• alcohol content
• long short-term memory network
• genetic algorithm
• bp neural network