Using Machine Learning for the Calibration of Airborne Particulate Sensors

Wijeratne, Lakitha O. H.; Kiv, Daniel; Aker, Adam; Talebi, Shawhin; Lary, David J.

doi:10.3390/s20010099

Cited by 34 publications

(19 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to literature evaluation studies, AlphaSense shows a good correlation with reference devices even though the PM readings were found to be underestimated [ 57 , 58 , 59 ]. On the other hand, a machine learning algorithm can effectively improve the calibration and performance of low-cost optical sensors that could be used in the future [ 60 ].…”

Section: Resultsmentioning

confidence: 99%

Comparison of Low-Cost Particulate Matter Sensors for Indoor Air Monitoring during COVID-19 Lockdown

Kaliszewski

Włodarski

Młyńczak

et al. 2020

Sensors

View full text Add to dashboard Cite

This study shows the results of air monitoring in high- and low-occupancy rooms using two combinations of sensors, AeroTrak8220(TSI)/OPC-N3 (AlphaSense, Great Notley, UK) and OPC-N3/PMS5003 (Plantower, Beijing, China), respectively. The tests were conducted in a flat in Warsaw during the restrictions imposed due to the COVID-19 lockdown. The results showed that OPC-N3 underestimates the PN (particle number concentration) by about 2–3 times compared to the AeroTrak8220. Subsequently, the OPC-N3 was compared with another low-cost sensor, the PMS5003. Both devices showed similar efficiency in PN estimation, whereas PM (particulate matter) concentration estimation differed significantly. Moreover, the relationship among the PM1–PM2.5–PM10 readings obtained with the PMS5003 appeared improbably linear regarding the natural indoor conditions. The correlation of PM concentrations obtained with the PMS5003 suggests an oversimplified calculation method of PM. The studies also demonstrated that PM1, PM2.5, and PM10 concentrations in the high- to low-occupancy rooms were about 3, 2, and 1.5 times, respectively. On the other hand, the use of an air purifier considerably reduced the PM concentrations to similar levels in both rooms. All the sensors showed that frying and toast-making were the major sources of particulate matter, about 10 times higher compared to average levels. Considerably lower particle levels were measured in the low-occupancy room.

show abstract

Section: Resultsmentioning

confidence: 99%

Comparison of Low-Cost Particulate Matter Sensors for Indoor Air Monitoring during COVID-19 Lockdown

Kaliszewski

Włodarski

Młyńczak

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…This study also provided an elaborate example of statistical correction methods for devices evaluated in laboratory conditions. However, one should note that other valid approaches are available, for example based on machine-learning principles [ 28 , 29 , 30 , 31 ]. More investigation is required to determine the influence of different statistical methods and which method is most suitable for the evaluation and correction of occupational PM monitors readings.…”

Section: Discussionmentioning

confidence: 99%

Exploring Evaluation Variables for Low-Cost Particulate Matter Monitors to Assess Occupational Exposure

Ruiter

Kuijpers

Saunders

et al. 2020

IJERPH

View full text Add to dashboard Cite

(1) Background: Small, lightweight, low-cost optical particulate matter (PM) monitors are becoming popular in the field of occupational exposure monitoring, because these devices allow for real-time static measurements to be collected at multiple locations throughout a work site as well as being used as wearables providing personal exposure estimates. Prior to deployment, devices should be evaluated to optimize and quantify measurement accuracy. However, this can turn out to be difficult, as no standardized methods are yet available and different deployments may require different evaluation procedures. To gain insight in the relevance of different variables that may affect the monitor readings, six PM monitors were selected based on current availability and evaluated in the laboratory; (2) Methods: Existing strategies that were judged appropriate for the evaluation of PM monitors were reviewed and seven evaluation variables were selected, namely the type of dust, within- and between-device variations, nature of the power supply, temperature, relative humidity, and exposure pattern (peak and constant). Each variable was tested and analyzed individually and, if found to affect the readings significantly, included in a final correction model specific to each monitor. Finally, the accuracy for each monitor after correction was calculated; (3) Results: The reference materials and exposure patterns were found to be main factors needing correction for most monitors. One PM monitor was found to be sufficiently accurate at concentrations up to 2000 µg/m3 PM2.5, with other monitors appropriate at lower concentrations. The average accuracy increased by up to three-fold compared to when the correction model did not include evaluation variables; (4) Conclusions: Laboratory evaluation and readings correction can greatly increase the accuracy of PM monitors and set boundaries for appropriate use. However, this requires identifying the relevant evaluation variables, which are heavily reliant on how the monitors are used in the workplace. This, together with the lack of current consensus on standardized procedures, shows the need for harmonized PM monitor evaluation methods for occupational exposure monitoring.

show abstract

“…Thus, recent research on data-driven sensor self-calibration have combined and adopted machine learning approaches to help to build a data driven model. Examples of this type of research are artificial neural networks for laser scanner calibration [7], decision tree based learning for the calibration of airborne particulate sensors [13], and support vector machine based pressure sensor calibration model [14]. In this paper, we follow this idea of combing the concept of data-driven modeling and approaches of machine learning.…”

Section: Introductionmentioning

confidence: 99%

Hidden Markov Model Based Data-driven Calibration of Non-dispersive Infrared Gas Sensor

You

Oechtering

2021

2020 28th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

Non-dispersive infrared gas sensing is one of the best gas measurement method for air quality monitoring. However, sensors drift over time due to sensor aging and environmental factors, which makes calibration necessary. In this paper, we propose a hidden Markov model approach for sensor selfcalibration, which builds on the physical model of gas sensors based on the Beer-Lambert law. We focus on the statistical dependency between a calibration coefficient and the temperature change. Supervised and unsupervised learning algorithms to learn the stochastic paramentes of the hidden Markov model are derived and numerically tested. The true calibration coefficient at each time instant is estimated using the Viterbi algorithm. The numerical experiments using CO2 sensor data show excellent initial results which confirms that data-driven calibration of non-dispersive infrared gas sensors is possible. Meanwhile, the challenge in the practical design is to find an appropriate quantization scheme to keep the computation burden reasonable while achieving a good performance. Index Terms-Non-dispersive infrared gas sensor, drift, self calibration, data-driven modeling, hidden Markov model, statistical inference. This research project has received funding from the European Unions Horizon 2020 research and innovation programme under grant agreement No 825272 (ULISSES). We would also like to acknowlegde Henrik Rödjegård and Hans Martin from SenseAir for the helpful technical discussions and for providing us the sensor data used in this work.

show abstract

Using Machine Learning for the Calibration of Airborne Particulate Sensors

Cited by 34 publications

References 55 publications

Comparison of Low-Cost Particulate Matter Sensors for Indoor Air Monitoring during COVID-19 Lockdown

Comparison of Low-Cost Particulate Matter Sensors for Indoor Air Monitoring during COVID-19 Lockdown

Exploring Evaluation Variables for Low-Cost Particulate Matter Monitors to Assess Occupational Exposure

Hidden Markov Model Based Data-driven Calibration of Non-dispersive Infrared Gas Sensor

Contact Info

Product

Resources

About