Calculation model of alcohol content based on LSTM and SAPSO-BP

Inhalt:
To solve the problem of inaccurate alcohol level detection in the current segmented wine picking process relying on manual "watch hop picking", a method for rapid alcohol level detection is proposed. The study collected experimental data on tuning forks in different modes and concentrations of alcohol solutions, and realized tuning fork frequency filtering and dynamic compensation based on the second-order filtering algorithm and long-short-term memory network (LSTM). The alcohol content calculation model is established based on the simulated annealing particle swarm optimization BP neural network (SAPSO-BP). The average calculation error of the alcohol content of the model is 0.405, which is superior to the alcohol content calculation model established by particle swarm optimization BP neural network and BP neural network. The alcohol content calculation method has high applicability and improves the accuracy of the alcohol content detection during the flow of wine.