Related Experiment Video
Updated: Dec 4, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
11.0K
Extracting useful signals from flawed sensor data: Developing hybrid data-driven approaches with physical factors
Cheng Yang1, Glen T Daigger1, Evangelia Belia2
1Civil and Environmental Engineering, University of Michigan, 2350 Hayward St, G.G. Brown Building, Ann Arbor, MI 48109, USA.
Water Research
|October 22, 2020
Summary
Hybrid approaches extract valuable data from flawed water quality sensor signals. Incorporating physical factors with machine learning improves signal processing for better water and wastewater management.
Area of Science:
- Environmental science
- Water quality monitoring
- Data science
Background:
- Water quality sensors are increasingly available but prone to errors from complex influents and instability.
- Extracting reliable data from flawed sensor measurements is a significant challenge in water and wastewater systems.
- Traditional methods focus on sensor maintenance, but data-driven approaches offer alternative solutions.
Purpose of the Study:
- To demonstrate a hybrid approach for processing flawed sensor signals by integrating physical factors with data-driven tools.
- To provide a case study using five-day biochemical oxygen demand (BOD5) sensor data to showcase signal processing improvements.
- To propose an Improved Standard Signal Processing Architecture (ISSPA) for enhanced sensor data utilization.
Main Methods:
- Utilized a hybrid methodology combining physical factors (process knowledge, constraints, observations) with data-driven tools.
- Applied machine learning algorithms to process flawed five-day biochemical oxygen demand (BOD5) sensor measurements.
- Incorporated physical insights throughout the signal processing pipeline, from raw data to validated output.
Main Results:
- Successfully extracted and validated useful signals from flawed BOD5 sensor measurements.
- Demonstrated that integrating physical factors significantly improved the performance of machine learning tools in signal processing.
- The proposed Improved Standard Signal Processing Architecture (ISSPA) offers a structured framework for hybrid signal processing.
Conclusions:
- Hybrid approaches are effective for extracting valuable information from imperfect sensor data in water systems.
- Physical factors are crucial for enhancing the accuracy and reliability of data-driven signal processing techniques.
- The developed methodology and ISSPA provide a pathway for more robust water quality monitoring and management.
