Related Experiment Video
Updated: Jul 25, 2026
Measuring Sub-23 Nanometer Real Driving Particle Number Emissions Using the Portable DownToTen Sampling System
Published on: May 22, 2020
A robust localized soft sensor for particulate matter modeling in Seoul metro systems
1Jiangsu Provincial Key Lab of Pulp and Paper Science and Technology, College of Light Industry Science and Engineering, Nanjing Forestry University, Nanjing 210037, China; Department of Environmental Science and Engineering, College of Engineering, Kyung Hee University, Yongin 446701, South Korea.
This study introduces a new robust soft sensor, Just-in-Time Least Squares Support Vector Regression (JIT-LSSVR), for monitoring indoor air quality (IAQ) in subway systems. The JIT-LSSVR method significantly improves PM2.5 prediction accuracy compared to traditional approaches.
Area of Science:
- Environmental Science
- Sensor Technology
- Data Science
Background:
- Indoor air quality (IAQ) monitoring in underground metro systems is critical for public health.
- Traditional soft sensor models like Partial Least Squares (PLS) struggle with dynamic and nonlinear processes.
- Hazardous pollutants, such as PM2.5, require accurate real-time prediction and monitoring.
Purpose of the Study:
- To develop a robust soft sensor for accurate prediction and monitoring of PM2.5 in subway stations.
- To enhance soft sensor modeling by leveraging the local learning capabilities of Just-in-Time (JIT) learning.
- To improve the prediction performance of PM2.5 soft sensors in dynamic and potentially noisy environments.
Main Methods:
- A novel Just-in-Time Least Squares Support Vector Regression (JIT-LSSVR) soft sensor was developed.
- The JIT learning technique was integrated with LSSVR to effectively track process variations.
- An outlier detection step was incorporated into the JIT-LSSVR model to mitigate the impact of erroneous data.
Main Results:
- The proposed JIT-LSSVR soft sensor demonstrated capability in modeling nonlinear and dynamic subway environments.
- The JIT-LSSVR model achieved a 55% improvement in root mean square error (RMSE) compared to the standard LSSVR model.
- The integrated outlier detection enhanced the robustness of the soft sensor.
Conclusions:
- The JIT-LSSVR soft sensor offers a superior approach for real-time IAQ monitoring in underground metro systems.
- This method provides more accurate predictions of PM2.5 concentrations than conventional techniques.
- The robust JIT-LSSVR is a promising tool for ensuring safer indoor air quality in public transport infrastructure.
