A Multi-output Spatio-Temporal Synergistic Network-Based Soft Sensor Model for Engine Oil Indicators in Construction Machinery

Engine oil indicators, such as water content, di-electric constant, and viscosity, are crucial for maintaining the stable operation of construction machinery. This paper proposes a Multi-ouput Spatio- Temporal Synergistic Network (MSTSN) soft sensor model to enhance the prediction accuracy of engine oil indicators. First, the engine operating data is augmented by reordering and stacking to facilitate better extraction of its spatial correlations. Then, the CNN extracts spatial features from the augmented data, which are combined with the original data and fed into the LSTM to capture temporal information, producing spatio-temporal features. Finally, a multi-channel structure, com-bined with self-attention mechanisms, further enhances prediction accuracy for multiple quality indicators. Validation on real-world engine oil datasets from various construction machinery demonstrates the superior performance and practical application potential of the model.