Deep MLP-Based Prediction of Process Parameters for Roller Kilns

DOI: https://doi.org/10.62517/jiem.202503404

Author(s)

Xiaoyu Zhang1, Mingming Gong2, Jiahao Zhang1

Affiliation(s)

1School of Artificial Intelligence and Big Data, Henan University of Technology, Zhengzhou, Henan, China 2iFlytek Co., Ltd., Hefei, Anhui, China

Abstract

Under high-load aviation conditions, internal non-uniform sintering within battery cathode materials readily induces thermal runaway, presenting a critical bottleneck constraining the safe development of low-altitude aviation. Addressing this, this paper proposes a temperature prediction model based on Deep Multi-Layer Perceptron (Deep MLP) to tackle the non-linear and strongly coupled characteristics of temperature field distribution during roller kiln sintering. The model incorporates critical process parameters such as heating rod setpoint temperature and airflow rate as inputs. It employs a layer-wise contraction network architecture, utilising ReLU activation functions and Dropout mechanisms to enhance nonlinear expressive capability and generalisation performance. Standardised processing via StandardScaler, MSE loss function, and Adam optimisation algorithm ensure efficient and stable training. Experimental results demonstrate the model's capability to accurately capture complex mapping relationships between input parameters and temperatures at multiple measurement points, exhibiting outstanding performance in metrics such as RMSE and MAPE. This research offers a data-driven approach for thermal field modelling and energy efficiency optimisation in cathode material sintering processes. It establishes a technical foundation for achieving closed-loop "perception-prediction-regulation" control in roller kilns, holding significant implications for advancing intelligent manufacturing in the lithium battery industry and ensuring the safe, high-quality development of the low-altitude economy.

Keywords

Roller Kiln; Process Indicators; Temperature Prediction; Deep Learning; Multi-Layer Perceptron; Low-Altitude Economy

References

[1] Wang Pengbo, Zheng Junchao. Current Status and Prospects of Lithium-ion Batteries. Nature Magazine, 2017, 39 (04): 283-289. [2] Zhan Linghao, Wang Jinkun, Sun Weiming, Xie Leqiong, Wang Li, Yu Lanjian, He Xiangming. Development History and Outlook of Sintering Kilns for Lithium-ion Battery Cathode Materials. Battery Industry, 2024, 28 (04): 199-205. [3] Liu Rui, Chen Ning, Guo Yuqian, Chen Jiayao, Yang Chunhua, Gui Weihua. Multi-objective Optimisation of Process Temperature for High-Quality Energy-saving Sintering of Ternary Cathode Materials Proceedings of the 34th China Conference on Process Control. Process Control Professional Committee of the Chinese Association for Automation, Chinese Association for Automation, Process Control Professional Committee of the Chinese Association for Automation, 2023: 1. [4] Liu Hualing, Ma Jun, Zhang Guoxiang. A Review of Content Recommendation Algorithms Based on Deep Learning. Computer Engineering, 2021, 47 (07): 4-5 [5] Zhu Junwen, Yang Haidong, Xu Kangkang, Song Cairong, Bao Hao. Temperature Prediction for Ceramic Roller Kilns Based on KPCA-CNN-GRU. Sensors and Microsystems, 2025, 44(05): 12-15+19. [6] Xu Peng, Su Wensheng, Xie Lifei. Structural and Economic Analysis of Lithium Battery Roller Kilns. Industrial Furnaces, 2016, 38 (04): 31-34+39. [7] Liu Yongshi. Research on Robust Sparse Optimisation Methods for Multi-Layer Perceptron Soft Measurement Models. Qilu University of Technology, 2023. [8] Cui Bingqian, Gao Haiyang, Han Fang. Multi-Layer Perceptron-Based Model Correction Method for Rotating Machinery. Acta Mechanica Sinica, 2025, 57(10): 2446-2456. [9] Huo Guannan, Liu Jie. A Method for Diagnosing Minor Faults in Industrial Processes Based on an Improved Multi-Layer Perceptron. Chemical Industry Automation and Instrumentation, 2024, 51(05): 892-899. [10] Luo Bowei. Data Visualisation Based on Python. Information Recording Materials, 2019, 20 (12): 72-74.