Prediction of Filter Rod Pressure Drop Based on Improved KNN

Pressure drop is an important indicator of filter rods, and controlling the pressure drop is the key to quality evaluation. However, there are many factors affecting the pressure drop of filter rods and each factor contains a large number of nonlinear relationships with each other, so it is difficult to conduct in-depth research, and the current pressure drop is still based on experimental test data as the basis for evaluation. In order to better control the pressure drop of filter rods and reduce the judgment time of pressure drop in the current production process, this paper proposes a method based on the production of historical data and the use of K-Nearest Neighbor (KNN) machine learning algorithm to realize the rapid prediction of pressure drop, which extracts the important features from the original data and establishes the prediction model of pressure drop. And the model parameters were tuned using the ant colony algorithm (AOC) to obtain the best model prediction. The results show that: (1) The weight has the greatest influence on pressure drop. (2) The improved KNN model based on AOC algorithm achieves better pressure drop prediction. It is verified that the mean square error of the prediction model is 3.8249, and the error is controlled within $4 \%$. (3) The prediction model established based on historical data can realize rapid prediction of pressure drop with high accuracy, which provides reference for the key quality control of filter rod molding.