Work

Wear prediction of internal combustion engine cylinder liners based on adaptive generative adversarial networks

Year: 2025

Type: article

Abstract: The cylinder liner is an important component of internal combustion engine, and its wear directly affects the performance and service life of the engines. Smaller cylinder liner wear data reduces the accuracy of wear evaluation, making it difficult to effectively monitor its operational condition. To address this issue, a prediction model for wear of cylinder liners is proposed based on adaptive generative adversarial network (DA-GAN), and a small-sample data enhancement method is introduced into the model. To strengthen temporal feature extraction, a DA integrated with standard deviation interpolation is introduced to the generator, enabling real-time updates of wear volume standard deviations through adaptive noise mechanisms. A cosine-annealing learning rate scheduling strategy is further adopted to mitigate overfitting and underfitting risks by dynamically adjusting optimization trajectories during training. The results demonstrate that the DA-GAN model outperforms conventional long short-term memory neural network and synthetic minority over-sampling technique-based methods, and the determination coefficient (R²), mean square error (MSE), mean absolute error of cylinder liner wear obtained by the proposed model are 0.9682, 0.0368, and 0.1653, respectively. The framework effectively addresses the limitations of traditional approaches when handling complex environmental variables and sparse datasets, establishing a novel technical pathway for health management and intelligent maintenance of internal combustion engines.

Source: Results in Engineering

DOI: 10.1016/j.rineng.2025.105740

Authors: Jianxiong Kang, Chang Shu, Haoran Yang, Yaofeng Liu, Changfeng Yan, Feifei Zhang

Institution: Lanzhou University of Technology

Language: English

Topic: Machine Fault Diagnosis Techniques

Subfield: Control and Systems Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $1,940