Supervised Contrastive Learning based Dual-Mixer Model for Remaining Useful Life Prediction

TL;DR

This work addresses RUL prediction for multivariate time series by introducing a spatial-temporal homogeneous feature extractor, the Dual-Mixer, to enable flexible, layer-wise feature fusion. It couples this with Feature Space Global Relationship Invariance (FSGRI), a supervised contrastive training scheme that preserves degradation-related relationships in feature space via DW-InfoNCE and Gaussian threshold sampling. Empirical results on the C-MAPSS dataset show that FSGRI yields average improvements of $7.00\%$ in RMSE and $2.41\%$ in MAPE across baselines, while the Dual-Mixer achieves superior performance on most metrics. The combination offers a lightweight, transferable framework for robust RUL prediction with practical impact for maintenance scheduling and PHM systems, and the authors provide public code for replication.

Abstract

The problem of the Remaining Useful Life (RUL) prediction, aiming at providing an accurate estimate of the remaining time from the current predicting moment to the complete failure of the device, has gained significant attention from researchers in recent years. In this paper, to overcome the shortcomings of rigid combination for temporal and spatial features in most existing RUL prediction approaches, a spatial-temporal homogeneous feature extractor, named Dual-Mixer model, is firstly proposed. Flexible layer-wise progressive feature fusion is employed to ensure the homogeneity of spatial-temporal features and enhance the prediction accuracy. Secondly, the Feature Space Global Relationship Invariance (FSGRI) training method is introduced based on supervised contrastive learning. This method maintains the consistency of relationships among sample features with their degradation patterns during model training, simplifying the subsequently regression task in the output layer and improving the model's performance in RUL prediction. Finally, the effectiveness of the proposed method is validated through comparisons with other latest research works on the C-MAPSS dataset. The Dual-Mixer model demonstrates superiority across most metrics, while the FSGRI training method shows an average improvement of 7.00% and 2.41% in RMSE and MAPE, respectively, for all baseline models. Our experiments and model code are publicly available at https://github.com/fuen1590/PhmDeepLearningProjects.

Figures (10)

• Figure 1: Illustration of progression relationship between samples for equipment degradation process.
• Figure 2: Basic MLP Block and Gate Block in the proposed Dual-Mixer model.
• Figure 3: Proposed Dual-Mixer Layer structure.
• Figure 4: Dual-path Mixer Model structure.
• Figure 5: The illustration of proposed Gaussian Threshold sampling method.