.png){kind=icon}
.png){kind=icon}
Return
Robust ensemble learning framework through symbolic regression for soft sensor modeling
.png){kind=icon}
OA
AI
DOI：10.1016/j.jocs.2026.102886.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The selection of an optimal mathematical representation for soft sensor modeling in industrial applications is challenging, as existing methods struggle to balance accuracy, interpretability, and computational efficiency. Although Symbolic Regression (SR) models provide interpretability, they often lack a systematic selection framework that allows decision-makers to balance accuracy and complexity effectively. This study addresses this gap by introducing the Tuned Evolutionary Selection for Symbolic Regression (TESSR), an ensemble SR model that integrates diverse state-of-the-art SR models with multiple attribute decision-making algorithms, offering a structured framework for model selection that can be updated over time. TESSR consistently outperformed standalone models, achieving an R2 (Coefficient of Determination) of 0.997 and an RMSE (Root Mean Squared Error) of 1210 with a 15-operation equation for modeling a heavy-duty engine’s turbo-speed soft sensor. This represents a 10.7% reduction in error and an 11.8% decrease in hardware usage compared to one of the current state-of-the-art SR models, demonstrating an ideal balance between accuracy and complexity. TESSR offers a robust solution for industrial applications, combining interpretability, accuracy, and computational efficiency, making it a valuable tool for decision-makers in complex system modeling for industrial applications.
Keywords:
Combustion engines
Ensemble learning
Time-series
Soft sensors
Symbolic regression
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
J
IF:
3.7
Papers:
195
Citations:
0