Article: Optimisation of PID control for tobacco feeding rail based on convolutional neural network Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.4 pp.1 - 15 Abstract: The tobacco wire guide system is a key component in cigarette production equipment. This paper proposes a tobacco wire guide PID control optimisation model based on convolutional neural network (CNN-PID). The tobacco wire guide speed, guide position, guide temperature, ambient temperature, ambient humidity, and PID parameters at the previous moment are selected as model dependent variables. After normalisation, they are input into the lightweight convolutional neural network. After model parameter adjustment, the predicted proportional gain, integral gain and derivative gain are finally output. After training, the R2-score values of CNN-PID on proportional gain, integral gain and derivative gain are 0.992, 0.984, and 0.982, respectively. In addition, the R2-score values of the CNN-PID model are better than those of traditional PID control, BP neural network PID, and PSO optimised PID. Inderscience Publishers - linking academia, business and industry through research

Title: Optimisation of PID control for tobacco feeding rail based on convolutional neural network

Authors: Guoqing Xu; Bin Peng; Yufei Li

Addresses: Liuzhou Cigarette Factory, China Tobacco Guangxi Industrial Co., Ltd., Liuzhou, 545005, China ' Qingdao Vic Powder Metallurgy Co., Ltd., Qingdao, 266000, China ' China Tobacco Guangxi Industrial Co., Ltd., Nanning, 530001, China

Abstract: The tobacco wire guide system is a key component in cigarette production equipment. This paper proposes a tobacco wire guide PID control optimisation model based on convolutional neural network (CNN-PID). The tobacco wire guide speed, guide position, guide temperature, ambient temperature, ambient humidity, and PID parameters at the previous moment are selected as model dependent variables. After normalisation, they are input into the lightweight convolutional neural network. After model parameter adjustment, the predicted proportional gain, integral gain and derivative gain are finally output. After training, the R²-score values of CNN-PID on proportional gain, integral gain and derivative gain are 0.992, 0.984, and 0.982, respectively. In addition, the R²-score values of the CNN-PID model are better than those of traditional PID control, BP neural network PID, and PSO optimised PID.

Keywords: PID control optimisation; convolutional neural network; CNN; intelligent control algorithm.

DOI: 10.1504/IJICT.2026.151528

International Journal of Information and Communication Technology, 2026 Vol.27 No.4, pp.1 - 15

Received: 02 Aug 2025
Accepted: 23 Nov 2025
Published online: 04 Feb 2026 *

Title: Optimisation of PID control for tobacco feeding rail based on convolutional neural network

Keep up-to-date