Answers are grounded in this paper and cite their sources.
The accurate assessment of lubricant condition is critical for predictive maintenance in heavy-duty industrial and military vehicles, where oil degradation directly affects engine and transmission performance. Traditional laboratory methods are precise but unsuitable for in-field or tactical scenarios due to cost, time, and infrastructure requirements. This study proposes an intelligent diagnostic system using computer vision and deep learning to classify used lubricants as compliant or non-compliant. High-resolution images, acquired under two optical setups—white-background crucible and spectroscopic cuvette—were analyzed using a ResNet-34 convolutional neural network. Two lubricant types were studied: synthetic transmission oil (O-1178) and mineral engine oil (O-1236), sourced from Spanish Army heavy vehicles. Both partial and full fine-tuning strategies were tested. Results showed high classification accuracy, especially under cuvette imaging. Grad-CAM analysis revealed that predictions were based on interpretable features like chromatic gradients and opacity. The system is a low-cost, autonomous alternative to lab diagnostics, suitable for real-time deployment in defense maintenance workflows. This dual-use solution supports military readiness and predictive logistics, while also aligning with Industry 4.0 paradigms.
No takes yet. Share an insight, caveat, or question.
López-Pingarrón et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: