Prediction of Pressure Drop based on IFA-LSTM for Filter Rod

Pressure drop directly affects the quality of filter rod, and the measurement to which is often influenced by the environmental temperature, humidity and other aspects. To obtain high accuracy of pressure drop prediction, this paper improves the firefly algorithm (FA) to carry out hyper-parameter optimization on the long short-term memory (LSTM) neural network to establish the pressure drop computation model. Firstly, the search efficiency and optimization accuracy of firefly algorithm are boosted with tent chaotic mapping and adaptive t-distribution ways. Then, the effectiveness and superiority of the enhanced strategy are verified by the benchmark test functions. Finally, the pressure drop is calculated with IFA-LSTM which combines optimization method with computation model. The experimental results show that the proposed mechanism can capture the fluctuations and variation information from the input impact factors to pressure drop prediction and make the accurate estimation to the filter rod pressure drop possible.