PaperlayerView exact revision on arXiv

Source-linked AI summary

Sensing Data Fusion for Enhanced Indoor Air Quality Monitoring

Q. P. Ha, S. Metia, M. D. Phung

arXiv:2001.01976v1eess.SYcs.SY

TL;DR

The paper addresses the need to fuse heterogeneous indoor measurements for air-quality assessment while accounting for humidity. It combines IAQI and humidex into EIAQI and applies EFKF to sensor data; its case study reports pollutant-profile findings, while noting sensor-sensitivity variation as a measurement constraint.

  • Problem

    Indoor air quality depends on many pollutants and environmental factors, making heterogeneous sensor fusion important for calculating IAQI and monitoring building pollutants.

    Source: §I · p. 1

  • Method

    The paper combines IAQI and humidex into EIAQI and uses an Extended Fractional Kalman Filter with Matérn covariance to model pollutant profiles and address missing or inaccurate sensor data.

    Source: §I · pp. 1–2

  • Results

    Over 20% oxygen was assessed as moderate; EFKF estimated CO above Waspmote measurements and unhealthy for sensitive people, while toluene rose over 0.7 ppm and was unhealthy for a sensitive person.

    Source: C. Indoor Air Pollutant Profiles with… · p. 7

  • Takeaways & Limitations

    The authors emphasize integrating accurate indoor air-quality monitoring and prediction into building management to better maintain inhabitants’ wellbeing.

    Source: §VI · p. 11

    Waspmote sensitivity may vary widely between units, and calibration may be needed for the operating range and application conditions.

    Source: §II · p. 2

Abstract

from arXiv · show

Multisensor fusion of air pollutant data in smart buildings remains an important input to address the well-being and comfort perceived by their inhabitants. An integrated sensing system is part of a smart building where real-time indoor air quality data are monitored round the clock using sensors and operating in the Internet-of-Things (IoT) environment. In this work, we propose an air quality management system merging indoor air quality index (IAQI) and humidex into an enhanced indoor air quality index (EIAQI) by using sensor data on a real-time basis. Here, indoor air pollutant levels are measured by a network of waspmote sensors while IAQI and humidex data are fused together using an extended fractional-order Kalman filter (EFKF). According to the obtained EIAQI, overall air quality alerts are provided in a timely fashion for accurate prediction with enhanced performance against measurement noise and nonlinearity. The estimation scheme is implemented by using the fractional-order modeling and control (FOMCON) toolbox. A case study is analysed to prove the effectiveness and validity of the proposed approach.

I. INTRODUCTION

Indoor air quality assessment must account for many pollutant sources and environmental factors, making heterogeneous sensor data fusion important. The paper proposes an EIAQI that combines IAQI with humidex and describes a Waspmote-based building monitoring system.

I. INTRODUCTION

  • Indoor air quality reflects diverse pollutant sources and occupant activities, motivating heterogeneous sensor fusion to calculate IAQI and monitor building pollutants.The paper also links indoor air quality to human health and work efficiency.Source: §I · p. 1
  • Prior work addresses indoor air quality through seasonal models, ventilation control, pollutant monitoring, and Kalman-filter data fusion.Examples include subway, residential, commercial-building, and aircraft-cabin settings.Source: §I · p. 1
  • The proposed EIAQI combines humidex with IAQI, while an Extended Fractional Kalman Filter using Matérn covariance addresses nonlinear, uncertain profiles and missing sensor data.The authors identify humidity inclusion and recovery of missing sensor data as contributions.Source: §I · pp. 1–2

II. SYSTEM DESCRIPTION AND MOTIVATION

  • The building network uses Waspmote sensors to measure eight gases, temperature, and humidity; 16 sensors were installed on one floor and over 100 elsewhere in the building.A Meshlium gateway sends sensor data to the cloud.Source: §II · p. 2
  • Waspmote sensor sensitivity can vary widely between units, so calibration may be needed for the operating range and application conditions.The paper notes that calibration under controlled temperature and humidity can improve monitoring accuracy.Source: §II · p. 2
  • A student’s fainting episode on August 24, 2016, was initially assessed as low oxygen but was attributed after investigation to high indoor pollutant levels.The incident motivated development of an enhanced index intended to forecast risks to occupants.Source: §II · p. 3

III. IAQI DATA FUSION FRAMEWORK

The framework uses IAQI and humidity-related information to calculate an enhanced indoor air quality index. Figures 5 and 6 show pollutant levels measured by waspmotes.

III. IAQI DATA FUSION FRAMEWORK

  • The framework describes data fusion for calculating IAQI and an enhanced index that incorporates humidity.The supplied figure captions identify CO2, O2, ethanol, and ammonia levels measured by waspmotes.Source: §III · p. 3, Figure 5 · p. 3, Figure 6 · p. 3
  • Figure 5 presents CO2 (ppm) and O2 (%) levels measured by waspmotes.Source: Figure 5 · p. 3
  • Figure 6 presents ethanol (ppm) and ammonia (ppm) levels measured by waspmotes.Source: Figure 6 · p. 3

A. Indoor Air Quality Index (IAQI)

The section extends AQI-based assessment to indoor air, where additional pollutants and oxygen are considered. IAQI values are calculated by linear interpolation between concentration breakpoints and corresponding index values.

A. Indoor Air Quality Index (IAQI)

  • AQI is described as a real-time scale of air-pollution levels used to inform people about outdoor activity adjustments to avoid health risks.The text notes that AQI is usually applied outdoors, although indoor environments also affect human health.Source: A. Indoor Air Quality Index (IAQI) · pp. 3–4
  • The paper extends AQI-based assessment to indoor air, where additional pollutants including CO2, VOCs, radon, and formaldehyde are identified as relevant.The text also identifies hydrogen sulfide, toluene, ethanol, ammonia, hydrogen, and oxygen as pollutants with health-related breakpoint information.Source: A. Indoor Air Quality Index (IAQI) · p. 4
  • IAQI is calculated by linear interpolation between a pollutant concentration's lower and upper breakpoints and their corresponding index values.For oxygen, the calculation uses oxygen concentration in percentage and its corresponding breakpoints and index values.Source: A. Indoor Air Quality Index (IAQI) · p. 4
  • For the example oxygen reading of 19.7347%, the calculated IAQI is 69.9475, classified as moderate.Source: A. Indoor Air Quality Index (IAQI) · p. 5
  • Eight pollutant profiles from the waspmote sensor are used to calculate IAQI, with humidity and temperature considered for the proposed EIAQI.Source: A. Indoor Air Quality Index (IAQI) · p. 5

B. Humidex

The section presents humidex as a measure combining heat and humidity to describe perceived hot or cold conditions. Its ratings are summarized in Table III.

B. Humidex

  • Humidex combines heat and humidity to describe an average person's hot or cold feelings across seasons.The text describes it as a dimensionless quantity based on dew point theory.Source: B. Humidex · p. 5
  • The humidex calculation uses air temperature in °C and relative humidity in %.Source: B. Humidex · p. 5
  • Table III summarizes humidex ratings.Source: Table III · p. 5, B. Humidex · p. 5

C. Enhanced Indoor Air Quality Index (EIAQI)

The proposed procedure first estimates an EFKF transfer function using a Matérn covariance function, then calculates EIAQI. EIAQI combines IAQI and humidex weightings into an overall room-condition weightage.

C. Enhanced Indoor Air Quality Index (EIAQI)

IV. EXTENDED FRACTIONAL KALMAN FILTER

The section uses extended fractional Kalman filtering to estimate indoor pollutant profiles, with fractional-order system identification from Waspmote data. It combines Matérn covariance modeling with a FOMCON-based identification process.

IV. EXTENDED FRACTIONAL KALMAN FILTER

  • EFKF is used to estimate pollutant profiles and predict indoor air quality that installed Waspmote sensors may overlook.The authors describe it as suitable for nonlinear state estimation affected by uncertainty, noise, disturbance, and missing measurements.Source: §IV · p. 5, A. EFKF Estimation Scheme · pp. 5–6

A. EFKF Estimation Scheme

  • Matérn covariance and a fractional-order system are incorporated into EFKF to model spatial distributions and nonlinear, uncertain indoor-air data.The Gaussian-process correlation length l informs the model parameter λ.Source: A. EFKF Estimation Scheme · p. 6

B. Fractional Order Identification

  • Fractional-order identification uses FOMCON in MATLAB to infer a dynamic system model from experimentally collected Waspmote data.The model relates system inputs and outputs under external stimuli; the section describes a single-input, single-output case.Source: B. Fractional Order Identification · p. 6
  • The identification example uses 78 CO2 samples spanning August 21–23, with three hours added before and after the three-day interval.A rational transfer function is used as the conventional system-identification starting point.Source: B. Fractional Order Identification · pp. 6–7
  • For seven contaminant gases, FOMCON-estimated fractional-order models had mean squared errors ranging from 0.3 to 0.9 across 78 samples.The pollutants were CO2, CO, H2, NH3, C2H6O, H2S, and C7H8.Source: B. Fractional Order Identification · p. 7

C. Indoor Air Pollutant Profiles with EFKF

The case study compares pollutant and oxygen profiles from August 23–26 using Waspmote measurements, EKF, and EFKF. On August 24, the filtered profiles indicate unhealthy gaseous pollutants alongside a substantial humidity rise, rather than low oxygen, as the context for the fainting incident.

C. Indoor Air Pollutant Profiles with EFKF

  • The analysis calculates IAQI, humidex, and EIAQI during August 23–25, 2016, after identifying the transfer function during August 21–23.Source: C. Indoor Air Pollutant Profiles with… · p. 7
  • On August 24, EFKF profiles indicate unhealthy CO and toluene levels with a substantial humidity rise, while oxygen remained above 20%; the authors attribute the fainting incident to poor air quality, not low oxygen.The toluene rise exceeded 0.7 ppm; carbon monoxide was higher than Waspmote measurements.Source: C. Indoor Air Pollutant Profiles with… · p. 7
  • Hydrogen, ammonia, ethanol, and hydrogen sulfide peaked on August 24 in moderate ranges that may be unhealthy for highly sensitive people.Source: C. Indoor Air Pollutant Profiles with… · p. 7

D. Statistical Analysis

The statistical analysis compares EKF and EFKF prediction data and reports improved precise estimation with EFKF. It also notes higher CO2 RMSE than for other IAQ levels, attributing this to human respiration and CO2’s broad permissible range.

D. Statistical Analysis

  • The analysis uses MAPE and RMSE to quantify forecast errors and R2 to assess the relationship between estimates and accurate observations.Source: D. Statistical Analysis · pp. 7–8
  • EFKF improves precise indoor-air-quality estimation over EKF, with the authors attributing the gain to Matérn covariance enabling better correlation at a suitable length scale.Table V presents descriptive statistics for EKF and EFKF prediction data on August 23 and 24, 2016.Source: D. Statistical Analysis · p. 9, Table V · p. 9
  • CO2 has higher RMSE than other IAQ levels, which the authors explain by human respiration and CO2’s wide permissible range.Source: D. Statistical Analysis · p. 9

V. INDOOR AIR QUALITY ASSESSMENT

The authors argue that indoor-air-quality monitoring should be accurate, comprehensive, and continuous, account for human comfort, and integrate prediction into building management.

V. INDOOR AIR QUALITY ASSESSMENT

  • The authors recommend integrating continuous IAQ monitoring and prediction into building management, including possible ventilation control to help prevent incidents.Source: §V · p. 9

A. Time Series Plot of IAQI and Humidex Using Real Data and Estimated Data

The section compares measured and EFKF-processed IAQI data, reporting very unhealthy IAQI on 24 August 2016. It also describes pollutant interpolation and plots the overall indoor-air index.

A. Time Series Plot of IAQI and Humidex Using Real Data and Estimated Data

B. Time Series Plot of EIAQI Using Real Data and Estimated Data

The EFKF-based EIAQI is reported as more accurate during the episode day, although it is generally similar to real data. The conclusion presents the method as addressing nonlinear, stochastic pollutant profiles and missing or inaccurate sensor data.

B. Time Series Plot of EIAQI Using Real Data and Estimated Data

VI. CONCLUSION

  • The conclusion presents fractional-order EKF estimation as improving pollutant-profile prediction while addressing nonlinear, stochastic conditions and missing or inaccurate sensor data.Source: §VI · p. 10
  • A slight-fainting incident is used as a case study, and the proposed indoor-air index combines IAQI with humidex to account for humidity.Source: §VI · p. 11
Loading 2001.01976v1…