May 21, 2016 - Even with the more stringent emission standards (Bharat IV through ..... population density may or may not be considered durng this pilot stage.
COMMODITY SENSING FOR AIR QUALITY MONITORING A Graduate Project Report submitted to Manipal University in partial fulfillment of the requirements for the award of the degree of
BACHELOR OF TECHNOLOGY in
Mechatronics Engineering Submitted by
Agni Biswas(120929006)
Under the guidance of Vibha Damodara K Asst. Professor Department of Mechatronics Engineering MANIPAL INSTITUTE OF TECHNOLOGY
Sriram Reddy Founder Sensors Without Borders
DEPARTMENT OF MECHATRONICS ENGINEERING
MANIPAL INSTITUTE OF TECHNOLOGY (A constituent Institute of MANIPAL UNIVERSITY)
MANIPAL - 576 104, KARNATAKA, INDIA May 2016
DEPARTMENT OF MECHATRONICS ENGINEERING
MANIPAL INSTITUTE OF TECHNOLOGY (A constituent Institute of MANIPAL UNIVERSITY)
MANIPAL - 576 104, KARNATAKA, INDIA Manipal 21/05/2016
CERTIFICATE This is to certify that the project titled Commodity Sensing for Air Quality Monitoring is a record of the bonafide work done by Agni Biswas (120929006) submitted in partial fulfillment of the requirements for the award of the degree of BACHELOR OF TECHNOLOGY in MECHATRONICS ENGINEERING of Manipal Institute of Technology, Manipal, Karnataka (A constituent college of Manipal University) during the year 2015-2016.
Vibha Damodara K. Asst. Professor, Mechatronics
Head of the Department
ii
Several years ago, I personally hand carried this device into India from the United States, after it was delivered to me by the Shipper from their office in North Carolina, to begin our R & D work here in India. After the recent failure of the device’s pump (which you should find shipped alongside the repaired and returned device), I also hand-carried the faulty device back to the United States on my recent trip there in January 2016, and shipped it locally to RTI for replacement purposes. Dated: 19-5-2016 From, I am
happy to show you proof of travel by providing you relevant copies of my passport pages along with immigration stamps.
Sriram Guddireddigari, KarmaCorps, Chennai, India.
As the understanding with the Shipper was that the device should be replaced with a new one rather than repaired and returned, given this background, we are not [Subject: of Completion] willing to accept this shipment as Certificate is and kindly request that it should be returned to the shipper as an RTO with only the minimum of charges levied.
To Whomsoever It May Concern,
Please & Biswas oblige.has successfully completed a four months of internship as our This is to do statethe thatneedful Mr. Agni Hardware Projects Developer at Karma Corps, as stated in the Letter of Appointment between Mr. Agni
Thanking YouCorps on 14-12-2015. Biswas and Karma Authorized Signatory
Yours Faithfully
Sriram Guddireddigari, Managing Trustee, Karma Corps.
Flat 1, Ground Floor, La Serena Ceebros, 12 Sterling Road, Second Cross St, Nungambakkam Chennai 600034, Tamil Nadu, India. www.karmacorps.in
iii
Acknowledgements
I wish to express my sincere thanks to Sriram Reddy, Adithya Pasupuleti, Vinayak N, Ram N. and all associates, for providing me with all the necessary facilities for the research. I place on record, my sincere thanks to Dr. Chandrasekhar Bhatt, H.O.D, Dept. of Mechatronics, for the continuous encouragement. I am also grateful to Mrs. Vibha Damodara K, Asst. Professor, Department of Mechatronics. I am extremely thankful and indebted to her for sharing expertise, the valuable guidance and encouragement extended to me despite her daily engagements. I take this opportunity to express gratitude to all of the Department faculty members for their help and support. I also thank my parents for the unceasing encouragement, support and attention. I also place on record, my sense of gratitude to one and all, who directly or indirectly, have lent their hand in this venture.
iv
Table of Contents CERTIFICATE ............................................................................................................................... ii Acknowledgements ........................................................................................................................ iv Table of Contents ............................................................................................................................ v ABSTRACT ................................................................................................................................. viii LIST OF FIGURES ....................................................................................................................... ix LIST OF TABLES .......................................................................................................................... x LIST OF NOTATIONS AND ABBREVIATIONS ...................................................................... xi Chapter 1.
Introduction ............................................................................................................... 1
Chapter 2.
Literature Review...................................................................................................... 2
2.1
Background ...................................................................................................................... 2
2.2
Screening and Assessing of Pollution Contributed by Standing Traffic .......................... 3
2.3
Traffic Management ......................................................................................................... 4
2.4
PM And Health................................................................................................................. 5
2.5
Traffic Light as as a special case of Traffic Congestion .................................................. 5
2.6
Health Impact of Transport on Populations living near Roadways : ............................... 6
2.7
Further Explanations ...................................................................................................... 11
2.7.1
Community-Oriented Monitoring ........................................................................... 11
2.7.2
Emissions Zone of Influence .................................................................................. 11
2.7.3
Community Monitoring Zone (CMZ) ..................................................................... 12
2.8
Site Types ....................................................................................................................... 13
2.9
PM10 .............................................................................................................................. 13
2.10 Middle ............................................................................................................................ 14 2.11 Neighborhood ................................................................................................................. 14 2.12 PM2.5 ............................................................................................................................. 15 v
2.12.1
Microscale ............................................................................................................... 15
2.12.2
Middle ..................................................................................................................... 15
2.12.3
Neighborhood ......................................................................................................... 16
2.12.4
Urban....................................................................................................................... 16
2.12.5
Regional .................................................................................................................. 16
Chapter 3.
Project Methodology............................................................................................... 18
Chapter 4.
Sensoric Analysis and Development ...................................................................... 21
4.1 Trident ................................................................................................................................. 21 4.1.1
General layout ......................................................................................................... 22
4.1.2
Enclosure................................................................................................................. 23
4.1.3
Hardware ................................................................................................................. 24
4.1.4
MQTT ..................................................................................................................... 25
4.1.5
Post Data Acquisition Calibration .......................................................................... 26
4.1.6
Calibrations ............................................................................................................. 30
4.1.7
Assumptions............................................................................................................ 32
4.2
Jellyfish .......................................................................................................................... 32
4.2.1
Introduction ............................................................................................................. 33
4.2.2
Design Constraints .................................................................................................. 34
4.2.3
Testing..................................................................................................................... 34
4.2.4
CFD analysis ........................................................................................................... 35
4.3
Iteration 2 ....................................................................................................................... 36
Chapter 5.
Results ..................................................................................................................... 38
Chapter 6.
Conclusion .............................................................................................................. 40
References ..................................................................................................................................... 41 Appendix ....................................................................................................................................... 43 vi
Acrylic Bending Calculations ................................................................................................... 43 List of competing devices. ........................................................................................................ 45
vii
ABSTRACT Right to decent and clean environment is the fundamental right of the people and the Environmental Protection Act, 1986 and Air (Prevention and Control of Pollution) Act, 1981 impose a definite obligation and duty upon the State to ensure availability of air quality conforming to the prescribed standards. This is perhaps most important among the young of our Nation such as school going students. The overall objective of this research is to understand emissions, exposures and health risks that arise to near-road communities, in particular schools, from traffic-related air pollutant in the case of traffic congestion due to the presence of traffic infrastructure in the traffic light. In this study, we will deploy a sensor network grid these CAQSDs to estimate the exposure levels of PM2.5 and PM10 levels to nearby school at a major traffic junction in Bangalore The Study aims at understanding spaces, the environments and the devices that pertain to the field of air quality monitoring and create/ adapt devices to create a mixed set of monitoring and modelling data to suit the healthcare segments while restraining itself to global accuracy norms.
viii
LIST OF FIGURES Figure 4.1. Trident : concept ......................................................................................................... 21 Figure 4.2 Trident Working .......................................................................................................... 22 Figure 4.3 Schema : SWB ............................................................................................................. 22 Figure 4.4 Hollow View: Trident.................................................................................................. 23 Figure 4.5 Holding Plate: Trident ................................................................................................. 23 Figure 4.6 Assembly as Seen from below Trident ........................................................................ 23 Figure 4.7 Flow: Real time correction ........................................................................................ 27 Figure 4.8 Jellyfish Alpha ............................................................................................................ 32 Figure 4.9 Rear and Front Shots of the Dylos .............................................................................. 33 Figure 4.10 CFD analysis Results ................................................................................................ 35 Figure 4.11 Jellyfish Version2.0 ................................................................................................... 36 Figure 4.12 External Flow under no fan condition ....................................................................... 37 Figure 4.13 Internal Flow with forced input ................................................................................. 37 Figure 5.5.1 JF Guided Flow Simulations .................................................................................... 38 Figure 5.2 Flow analysis JF. ......................................................................................................... 38 Figure A.1 Showing the appropriate dimensions of T.C. ............................................................ 43 Figure A.2 Description of unfold T.C. in to a planar surface ....................................................... 44
ix
LIST OF TABLES Table 4.1: True mean data compression ....................................................................................... 28 Table 4.2: Dynamic Mean Point Compression ............................................................................. 29 Table 4.3 Dynamic Median Point Compression ........................................................................... 30 Table 4.4 : Wind testing Jellyfish Alpha. ..................................................................................... 34 Table 5.1 : Nova reference test ..................................................................................................... 39 Table A.1Competing Devices ....................................................................................................... 45
x
LIST OF NOTATIONS AND ABBREVIATIONS EPA:
The EPA is the Environmental Protection Authority which is the main regulatory agency for environmental quality, including within the air and water domains, whose guidelines regulate pollution standards in USA but typically followed by other regulatory agencies around the world including those of the CPCB (Central Pollution Control Board), India.
PEM :
Personal exposure Monitor
JF :
Jellyfish; an adaptive enclosure design being developed at Sensors Without Borders
CAQSDs :
Commodity Air Quality Sensing Devices
NAAQS:
National Ambient Air Quality Sensors
CMZs:
Community Monitoring Zones
Trident:
A replicate Sensor Platform under development at sensors without borders.
Nova:
Nova-fitness SDS011 a commodity sensor from China
Shinyei:
A Korean brand of air quality monitors that allow for commodity sensing as a base unit mostly as a building block for many DIY sensor Kits as well as a number commodity AQMs inclusive of the Air Beam and Air Quality Egg.
xi
Chapter 1.
Introduction
This air quality initiative aims to gather, analyze and interpret a high quality dataset of PM2.5 and PM10 levels, using a praxis-orientation that ties into our larger mission at SWB, corresponding to a spatial grid microscale pattern around a major traffic intersection in Bengaluru, Karnataka, India of commodity air quality instruments - in two variations of the Trident, the Dylos 1700 and the MicroPEM. These transformed datasets based on NAAQS standards of daily PM2.5 and PM10 levels are then correlated with varying ‘transportation fluidity categories’ (self-termed) - standing traffic / freely moving traffic - captured via a vision capture system running on a IoT-specific hardware platform to determine the contributions of this major PM sources to ambient levels on a microscale which also lends itself to developing understandings on short-term timescales.
We believe this practical understanding of by how much and how often daily PM levels around major transportation hubs exceed NAAQS levels and their short-term, microscale health impacts on commuters and residents who live and work in-and-around these highly polluted transport hubs will assist with planning for new transportation grids in India’s smart cities and adding to transportation grids in existing Indian metros and Tier 2 cities.
1
Chapter 2. 2.1
Literature Review
Background:
India’s policymakers recently announced the development of 15 Smart Cities [1]. The development of effective transport infrastructure, that coalesce around transport hubs, that minimizes the health burden to local and distant populations living and working around these hubs should and is a key driver and outcome of design planning. Heavy trucks in India which are a major enabler of intra- and inter-city trade still run on diesel and less efficient forms of fuel. This coupled with less stringent regulatory norms and enforcement [2] is considered a major source of PM which has been conclusively linked to a variety of pulmonary and cardiovascular issues in humans. Standing traffic, or vehicles that are stationary, is considered a particularly large source of PM given the inefficiency of the combustion process, and it is our objective to understand the variation of levels of PM as a function of the volume and frequency of standing traffic at select major traffic junctions. This study will be conducted across various metro cities beginning with Bangalore, Chennai, Mumbai and Delhi, and will attempt to ascertain the hourly and daily PM averages that will allow us to determine the short-term, microscale impacts of PM to passengers at these junctions using existing research in this area. Finally, we will conduct ground-level community surveys data of retail shop owners located around these junctions that are typically completely exposed to PM on a daily basis. There is also the possibility of linking this research up with the recent odd-even plan in Delhi to determine how certain legislative changes may have an impact on PM levels on a practical level with attendant durational impacts.
2
2.2
Screening and Assessing of Pollution Contributed by Standing Traffic
New Jersey has linked high levels of air [3] to idling construction vehicles and equipment and says it "will be taking action" to protect air quality. It is illegal to idle [4] for more than three minutes if a vehicle isn't moving or equipment is not in use, according to a Department of Environmental Protection compliance advisory [5] Previous research has highlighted the inadequacy of regulatory monitors to characterize the microscale impacts (given that these monitors are typically representative of larger scale community-level, neighborhood-level and even regional-level ambient levels. Concentrations of air pollutants from vehicles are elevated along roadways, indicating that human exposure in transportation microenvironments may not be adequately characterized by centrally located monitors. We report results from ∼180 h of real-time measurements of fine particle and black carbon mass concentration (PM2.5, BC) and ultrafine particle number concentration (PN) inside a common vehicle, the auto-rickshaw, in New Delhi, India. Measured exposure concentrations are much higher in this study (geometric mean for ∼60 trip-averaged concentrations: 190 μg m−3 PM2.5, 42 μg m−3 BC, 280 × 103 particles cm−3; GSD ∼1.3 for all three pollutants) than reported for transportation microenvironments in other megacities. In-vehicle concentrations exceeded simultaneously measured ambient levels by 1.5× for PM2.5, 3.6× for BC, and 8.4× for PN. Shortduration peak concentrations (averaging time: 10 s), attributable to exhaust plumes of nearby vehicles, were greater than 300 μg m−3 for PM2.5, 85 μg m−3 for BC, and 650 × 103 particles cm−3 for PN. The incremental increase of within-vehicle concentration above ambient levels— which we attribute to in- and near-roadway emission sources—accounted for 30%, 68% and 86% of time-averaged in-vehicle PM2.5, BC and PN concentrations, respectively. Based on these results, we estimate that one's exposure during a daily commute by auto-rickshaw in Delhi is as least as large as full-day exposures experienced by urban residents of many high-income countries. This study illuminates an environmental health concern that may be common in many populous, low-income cities (Joshua)[6].
3
2.3
Traffic Management
In order to be able to reactively combat high PM levels, cities must be able to identify hot-spots and temporary peaks in exposure. For this, both a high temporal resolution as well as the timeliness of readings are important: A possible application case could for instance be, that city governments exercise concentration-related control of traffic by temporarily prohibiting vehicular access to certain hot-spot areas. A low latency is crucial for such a reactive systems [7]
The Chennai dialogue forum discussed and debated on issues related to air quality benefits of first generation reforms, emerging air quality and health challenges, issues related to sub-utilization of public transport systems including the tramways in the city and poor facilities for pedestrians, cyclists and public transport users. The expert panel recommended that air pollution is major concern and parking norms should be tightened across the city and the city must shift focus from car centric development towards providing mobility choices to majority of people who either walks, cycle, or take public transport [8].
For this, both a high temporal resolution as well as the timeliness of readings are important: A possible application case could for instance be, that city governments exercise concentrationrelated control of traffic by temporarily prohibiting vehicular access to certain hot-spot areas. A low latency is crucial for such a reactive system.
Similarly at source, monitoring of industrial point sources should leverage advances in Continuous Emissions Monitoring Systems (CEMS) technology to produce complete and accurate records of air pollution from every chimney of significant enough size. It is simply not possible to produce a complete record of air pollution sources through intermittent manual samplings taken once or twice in a year.
4
Summarizing the overall use case for low cost optical sensors, [9] D. M. Holstius had to say," In community monitoring or near-roadway contexts, a dense network or gradient with de- liberate oversampling could provide high-quality estimates of spatiotemporally resolved concentrations. More flexible saturation monitoring, based on less expensive and more 20 portable instruments, could also respond more readily to changing land use, enable more timely empirical verifications of emission-reduction policies, facilitate rapid responses to natural or accidental releases of observed aerosols, and support more efficient screening campaigns for urban “hot spots”, with follow-up measurements made by “reference techniques”.
2.4
PM And Health
PM is considered to be an indicator of overall pollution and the most dangerous of all. The health impacts of PM particularly for children and our older population can be quite detrimental. There is well documented research in this regard. In India, the morbidity and mortality burden of outdoor air pollution, is particularly costly in terms of workdays lost, lost productivity, and loss in terms of gross domestic product, approximately USD 23.4 billion and 1.7% of national GDP in 2009.[10] . Since, most of the pollution related deaths occur within a year or two of exposure, reducing PM pollution from sources like transport and power plants has immediate benefits for health and national economy. While other sources of PM can be, among others, cement plants, construction sites, garbage burning, industrial emissions, some estimates put the contribution of traffic to PM at around 70% of ambient levels.
2.5
Traffic Light as as a special case of Traffic Congestion
This bottom-up air quality data collection-oriented (as opposed to simulation-based studies) aims to gather, analyse and interpret a high quality dataset of PM2.5 and PM10 levels, using a praxisorientation that ties into our larger mission at SWB, corresponding to a spatial grid microscale pattern around a major traffic intersection in Bengaluru, Karnataka, India of commodity air quality instruments - in two variations of the Trident, the Dylos 1700 and the MicroPEM. These transformed datasets based on NAAQS standards of daily PM2.5 and PM10 levels are then 5
correlated with varying ‘transportation fluidity categories’ (self-termed) - standing traffic / freely moving traffic - captured via a vision capture system running on a IoT-specific hardware platform to determine the contributions of this major PM sources to ambient levels on a microscale which also lends itself to developing understandings on short-term timescales
We believe this practical understanding of by how much and how often daily PM levels around major transportation hubs exceed NAAQS levels and their short-term, microscale health impacts on commuters and residents who live and work in-and-around these highly polluted transport hubs will assist with planning for new transportation grids in India’s smart cities and adding to transportation grids in existing Indian metros and Tier 2 cities.
Transport is another sector that contributes to air pollution, especially in urban areas. A good action plan must be developed for traffic management in urban areas as the levels of pollution in the ambient air due to traffic is increasing at an alarming rate [11] 2.6
Health Impact of Transport on Populations living near Roadways :
Congestion can be caused by physical bottlenecks (40% of cases in the U.S.), traffic incidents (25%), work zones (10%), weather (15%), traffic control devices (5%), special events (5%), and fluctuations in normal traffic [12]. In addition to degrading urban air quality, consequences of congestion include travel delays, wasted fuel, decreased economic competitiveness, and decreased quality of life (Downs, 2004). Congestion in 438 U.S. urban areas in 2007 was estimated to cause approximately 4.2 billion hours of travel delay and waste 2.8 billion gallons of fuel, at a total cost of $87.2 billion (Schrank and Lomax, 2009). Congestion can be divided into recurring congestion and incident congestion, the latter caused by an accident or disabled vehicle
A few studies have examined congestion-related impacts on exposure and health risks. predicted that congestion charging zone an area that drivers must pay to enter) in London, UK would extend 183 years-of-life per 100,000 population in the congestion charging zone, and would provide a total of 1,888 additional years of life in the greater London area. Suggest that the congestion pricing 6
zone in Stockholm, Sweden would avoid 25-30 deaths annually due to traffic air pollution, in a region with approximately 1.4 million inhabitants. These two studies were conducted in Europe and focused on congestion charging zones’ impacts. No study is known that has investigated the impact of rush hour congestion on health in the U.S. population.
The derived linkage may be seen as being useful to the design of traffic control measures such as the allowing of U-turn / No U-turn at traffic intersections, traffic light optimization including local as opposed to global optimization of traffic lights, and other health-related measures such as air filter masks for traffic policemen, among other such benefits.
It is estimated that traffic contributes to as much as x% of PM-related pollution in Bengaluru . Add diagram from report here. The contribution of transport to PM10 levels in Bangalore is 42%. Even with the more stringent emission standards (Bharat IV through to Bharat VI) and cleaner fuel standards (Euro IV), the significant increase in transport figures every year in Bangalore - close to 1200 vehicles- means that PM contribution from transport is projected to increase significantly year-on-year. Therefore, means by which to mitigate this issue is both important and imperative. Standing traffic, or vehicles that are stationary, is considered a particularly large source of PM given the inefficiency of the combustion process, and it is our objective to understand the extent and variation of levels of PM as a function of the volume and of standing traffic at select major traffic junctions in Bengaluru city, India.
New Jersey, for instance, has linked high levels of air pollution to idling construction vehicles and equipment and says it "will be taking action" to protect air quality. It is illegal to idle for more than three minutes if a vehicle isn't moving or equipment is not in use, according to a Department of Environmental Protection compliance advisory.
7
Concentrations of air pollutants from vehicles are elevated along roadways, indicating that human exposure in transportation microenvironments may not be adequately characterized by centrally located monitors.
The relationship between congestion and vehicle emissions is complex (TRB, 2002). Emission rates are associated with the distributions of speed and acceleration, which depend on road type, traffic flow and other factors [14]. Congestion alters driving patterns, specifically causing frequent acceleration and deceleration in stop-and- go traffic, which increases emissions[14].
Acceleration increases the load placed on engines, and thus engines are operated in a fuel-rich and high emission mode that can overload catalytic converters [14]. CO and VOC emissions are most affected [14] . Effects on NOx emissions are limited because fuel-lean modes cause higher NOx emissions [14]. Deceleration contributes particularly to PM and VOC emissions because unburned fuel can be emitted under fuel enleanment conditions [15]
However, congestion does not always increase emissions since vehicle emissions may be reduced at low speeds [14]. Information regarding emissions that pertain to congestion is very limited. A few experimental studies have explored the relationship between congestion and emissions.
Anderson et al. (1996) found that congestion increased CO, HC and NOx emissions by 71%, 53% and 4% respectively, compared to free flow conditions. Sjodin et al. (1998) showed a 10-fold increase in CO and HC emissions with congestion (average speed, 20 km/h) compared to uncongested conditions (average speed, 60-70 km/h). De Vlieger et al. (2000) indicated that CO and HC emissions in rush hour increased by 60% and 10%, respectively, compared to smooth conditions, but NOx emissions were unchanged. Frey et al. (2001) used on-board measurements for CO, NO and HC and found that emissions for all three pollutants increased by 50% in congestion. 8
Other research has taken a different approach ascertaining exposure to people within transport such as autorickshaw. For instance, Apte collected ∼180 h of real-time measurements of fine particle and black carbon mass concentration (PM2.5, BC) and ultrafine particle number concentration (PN) inside a common vehicle, the auto-rickshaw, in New Delhi, India.
Measured exposure concentrations are much higher in this study (geometric mean for ∼60 tripaveraged concentrations: 190 μg m−3 PM2.5, 42 μg m−3 BC, 280 × 103 particles cm−3; GSD ∼1.3 for all three pollutants) than reported for transportation microenvironments in other megacities. In-vehicle concentrations exceeded simultaneously measured ambient levels by 1.5× for PM2.5, 3.6× for BC, and 8.4× for PN. Short-duration peak concentrations (averaging time: 10 s), attributable to exhaust plumes of nearby vehicles, were greater than 300 μg m−3 for PM2.5, 85 μg m−3 for BC, and 650 × 103 particles cm−3 for PN.
The incremental increase of within-vehicle concentration above ambient levels—which we attribute to in- and near-roadway emission sources—accounted for 30%, 68% and 86% of timeaveraged in-vehicle PM2.5, BC and PN concentrations, respectively.
Traffic congestion occurs when vehicle volume exceeds road capacity, which slows vehicle speeds, sometimes to a crawl or stop. The primary causes of congestion include physical bottlenecks (40% of cases in the U.S.), traffic incidents (25%), work zones (10%), weather (15%), traffic control devices (5%), special events (5%), and fluctuations in normal traffic (CAMSYS and TTI, 2005).
9
The Chennai dialogue forum discussed and debated on issues related to air quality benefits of first generation reforms, emerging air quality and health challenges, issues related to sub-utilisation of public transport systems including the tramways in the city and poor facilities for pedestrians, cyclists and public transport users. The expert panel recommended that air pollution is major concern and parking norms should be tightened across the city and the city must shift focus from car centric development towards providing mobility choices to majority of people who either walks, cycle, or take public transport (http://cseindia.org/node/5072).
Standing traffic, or vehicles that are stationary, is considered a particularly large source of PM given the inefficiency of the combustion process, and it is our objective to understand the variation of levels of PM as a function of the volume and frequency of standing traffic at select major traffic junctions. This study will be conducted across various metro cities beginning with Bangalore, Chennai, Mumbai and Delhi, and will attempt to ascertain the hourly and daily PM averages that will allow us to determine the short-term, microscale impacts of PM to passengers at these junctions using existing research in this area. Finally, we will conduct ground-level community surveys data of retail shop owners located around these junctions that are typically completely exposed to PM on a daily basis.
There is also the possibility of linking this research up with the recent odd-even plan in Delhi to determine how certain legislative changes may have an impact on PM levels on a practical level with attendant durational impacts.
New Jersey has linked high levels of air pollution to idling construction vehicles and equipment and says it "will be taking action" to protect air quality.
It is illegal to idle for more than three minutes if a vehicle isn't moving or equipment is not in use, according to a Department of Environmental Protection compliance advisory. 10
2.7
Further Explanations:
2.7.1 Community-Oriented Monitoring Community-oriented (core) monitoring sites are beyond the zone of influence of a single source, and should have neighborhood- to urban- scale zones of representation. The principal purpose of community-oriented monitoring sites is to approximate the short-term and long-term exposures of large numbers of people where they live, work, and play. A monitor placed at the fence line of an emissions source would not be considered to represent community exposures, even though there might be residences abutting that fence line. A monitor placed in the middle of an area adjacent to a source would, however, be deemed a community exposure monitor for that neighborhood provided that the location represented a zone of at least 0.5 km in diameter. The fence line monitor might still be operated because it provides information on how much the nearby source contributes to the community-oriented site. The data from the fence line monitor would not be used to determine annual NAAQS compliance, though it might be used to make comparisons to the 24hour standard or to design control strategies to bring the area into compliance with the annual NAAQS
2.7.2 Emissions Zone of Influence The zone of influence of a source is the distance at which PM from that specific source contributes no more than 10% of the measured PM concentration. The zone of influence refers to a specific emitter, rather than to a source category. For example, though suspended road dust may contribute 50% of PM10 over a wide region, the majority of emissions from a specific road influence concentrations over a few tens of meters from the emissions point. The actual size of a zone of influence varies with meteorology, being larger downwind than upwind, and the nature of the source (point, elevated, area, line, etc.). Zones of influence are, therefore, expressed as orders of magnitude rather than as exact distances. The concept is useful for locating community exposure sites that are intended to represent concentrations for sources with large rather than small zones of influence. Actual zones of influence must be determined empirically, by spatially dense monitoring networks, or theoretically by applying air quality and meteorological models
11
2.7.3 Community Monitoring Zone (CMZ): When spatial averaging is utilized for making comparisons to the annual PM2.5 NAAQS, Community Monitoring Zones must be defined in the monitoring network description. Otherwise, they may be used as a more informal manner, as a means to describe the communities surrounding one or more core monitoring sites. CMZs have dimensions of 4 to 50 km with boundaries defined by existing political demarcations (e.g., aggregates of zip codes, census tracts) with population attributes. They could be smaller in densely populated areas with large pollutant gradients. Each CMZ would ideally equal the collective zone of representation of one or more community-oriented monitors within that zone. The CMZ, applicable only to PM2.5, is intended to represent the spatial uniformity of PM2.5 concentrations. In practice, more than one monitor may be needed within each CMZ to evaluate the spatial uniformity of PM2.5 concentrations and to accurately calculate the spatial average for comparison with the annual PM2.5 NAAQS. When spatial averaging is used, each MPA would be completely covered by one or more contiguous CMZs. A separate appendix for each MPA should present the following detailed information: 2.7.3.1 Introduction: Describe the physical setting of the MPA, population characteristics, climate and weather, dominant economic activities, and emissions sources. Much of this information can be concisely and efficiently summarized on maps such as USGS (1:50000 or 1:2400), aerial photographic, or commercial maps available on CD-ROM (Appendix A identifies several options). 2.7.3.2 Community Monitoring Areas or Zones: Show maps of the selected community monitoring areas or community monitoring zones, if appropriate, and justify them based on the procedures in Section 4. Document modeling and data analysis activities that were conducted to determine these zones. These maps should include, where available and appropriate: 1) populated entity boundaries (e.g. census tracts or blocks), 2) relevant jurisdictional boundaries; 3) commercial, residential, agricultural, and industrial land uses; 4) suspected area source emissions hot spots (e.g. wood burning communities, diesel emissions hot spots identified from bus transfer locations, railyards, marine terminals, waste recycling, land preparation, unpaved roads, etc.). Include tables of spatial uniformity measures (spatial averages, spatial coefficients of variation, 98th percentile concentrations, and spatial correlations) using existing data to determine the zone of representation of existing monitors. Use wind roses and 12
trajectories to identify potential transport pathways and terrain maps to identify potential barriers to transport.
3. Sampling Site Descriptions: Provide site descriptions, including maps showing surrounding sources as well as verbal descriptions of activities surrounding the site. Define the variables measured at each site in terms of observables measured (e.g., PM10 mass, PM2.5 mass, chemical composition), sample duration, frequency, and measurement method. State and Local Air Monitoring Stations (SLAMS): SLAMS are designed and operated by local air pollution control districts to determine: 1) the highest concentrations expected to occur in each MPA; 2) representative concentrations in areas of high population density; 3) the impact on ambient pollution levels of significant sources or source categories; 2.8
Site Types
Several types of sampling sites, not all of which are designated for determining compliance with NAAQS, will be part of the PM2.5 measurement networks. Community-Oriented (Core) Sites: Community-oriented sites are located where people live, work, and play rather than at the expected maximum impact point for specific source emissions. These sites are not located within the microscale or middle-scale zone of influence of a specific, nearby particle emitter. Community- oriented sites may be located in industrial areas as well as and in residential, commercial, recreational, and other areas where a substantial number of people may spend a significant fraction of their day. Daily Compliance Sites: Daily compliance sites are used to determine NAAQS compliance for the 24-hour (daily) PM2.5 standard, but not for the annual standard. Because a daily compliance site does not necessarily represent community-oriented monitoring, it may be located near an emitter with a microscale or middle-scale zone of influence. 2.9
PM10
This scale would typify areas such as downtown street canyons, traffic corridors, and fence line stationary source monitoring locations where the general public could be exposed to maximum PM10 concentrations. Microscale particulate matter sites should be located near inhabited 13
buildings or locations where the general public can be expected to be exposed to the concentration measured. Emissions from stationary sources such as primary and secondary smelters, power plants, and other large industrial processes may, under certain plume conditions, likewise result in high ground level concentrations at the microscale. In the latter case, the microscale would represent an area impacted by the plume with dimensions extending up to approximately 100 meters. Data collected at microscale sites provide information for evaluating and developing hot spot control measures. 2.10 Middle Much of the short-term public exposure to coarse fraction particles (PM10) is on this scale and on the neighborhood scale. People moving through downtown areas or living near major roadways or stationary sources, may encounter particulate pollution that would be adequately characterized by measurements of this spatial scale. Middle scale PM10 measurements can be appropriate for the evaluation of possible short-term exposure public health effects. In many situations, monitoring sites that are representative of micro-scale or middle-scale impacts are not unique and are representative of many similar situations. This can occur along traffic corridors or other locations in a residential district. In this case, one location is representative of a neighborhood of small scale sites and is appropriate for evaluation of long-term or chronic effects. This scale also includes the characteristic concentrations for other areas with dimensions of a few hundred meters such as the parking lot and feeder streets associated with shopping centers, stadia, and office buildings. In the case of PM10, unpaved or seldomly swept parking lots associated with these sources could be an important source.
2.11 Neighborhood Measurements in this category represent conditions throughout some reasonably homogeneous urban sub region with dimensions of a few kilometers and of generally more regular shape than the middle scale. Homogeneity refers to the particulate matter concentrations, as well as the land use and land surface characteristics. In some cases, a location carefully chosen to provide neighborhood scale data would represent not only the immediate neighborhood but also neighborhoods of the same type in other parts of the city. Neighborhood scale PM10 sites provide information about trends and compliance with standards because they often represent conditions 14
in areas where people commonly live and work for extended periods. Neighborhood scale data could provide valuable information for developing, testing, and revising models that describe the larger scale concentration patterns, especially those models relying on spatially smoothed emission fields for inputs. The neighborhood scale measurements could also be used for neighborhood comparisons within or between cities.
2.12 PM2.5
2.12.1 Microscale Areas such as downtown street canyons and traffic corridors where the general public would be exposed to maximum concentrations from mobile sources. In some circumstances, the microscale is appropriate for particulate sites; community-oriented SLAMS sites measured at the microscale level should, however, be limited to urban sites that are representative of long-term human exposure and of many such microenvironments in the area. In general, microscale particulate matter sites should be located near inhabited buildings or locations where the general public can be expected to be exposed to the concentration measured. Emissions from stationary sources such as primary and secondary smelters, power plants, and other large industrial processes may, under certain plume conditions, likewise result in high ground level concentrations at the microscale. In the latter case, the microscale would represent an area impacted by the plume with dimensions extending up to approximately 100 meters. Data collected at microscale sites provide information for evaluating and developing hot spot control measures.
2.12.2 Middle People moving through downtown areas, or living near major roadways, encounter particle concentrations that would be adequately characterized by this spatial scale. Thus, measurements of this type would be appropriate for the evaluation of possible short-term exposure public health effects of particulate matter pollution. In many situations, monitoring sites that are representative of microscale or middle-scale impacts are not unique and are representative of many similar situations. This can occur along traffic corridors or other locations in a residential district. In this 15
case, one location is representative of a number of small scale sites and is appropriate for evaluation of long-term or chronic effects. This scale also includes the characteristic concentrations for other areas with dimensions of a few hundred meters such as the parking lot and feeder streets associated with shopping centers, stadia, and office buildings.
2.12.3 Neighborhood Measurements in this category would represent conditions throughout some reasonably homogeneous urban sub-region with dimensions of a few kilometers and of generally more regular shape than the middle scale. Homogeneity refers to the particulate matter concentrations, as well as the land use and land surface characteristics. Much of the PM2.5 exposures are expected to be associated with this scale of measurement. In some cases, a location carefully chosen to provide neighborhood scale data would represent the immediate neighborhood as well as neighborhoods of the same type in other parts of the city. PM2.5 sites of this kind provide good information about trends and compliance with standards because they often represent conditions in areas where people commonly live and work for periods comparable to those specified in the NAAQS. In general, most PM2.5 monitoring in urban areas should have this scale.
2.12.4 Urban This class of measurement would be used to characterize the particulate matter concentration over an entire metropolitan or rural area ranging in size from 4 to 50 kilometers. Such measurements would be useful for assessing trends in area-wide air quality, and hence, the effectiveness of large scale air pollution control strategies. Community-oriented PM2.5 sites may have this scale.
2.12.5 Regional These measurements would characterize conditions over areas with dimensions of as much as hundreds of kilometers. As noted earlier, using representative conditions for an area implies some degree of homogeneity in that area. For this reason, regional scale measurements would be most applicable to sparsely populated areas. Data characteristics of this scale would provide information about larger scale processes of particulate matter emissions, losses and transport. PM2.5 transport 16
contributes to elevated particulate concentrations and may affect multiple urban and State entities with large populations such as in the eastern United States. Development of effective pollution control strategies requires an understanding at regional geographical scales of the emission sources and atmospheric processes that are responsible for elevated PM2.5 levels and may also be associated with elevated O3 and regional haze.
17
Chapter 3.
Project Methodology
Co-locate SWB commodity sensor (Trident, Dylos 1700, MicroPEM) against 2 CEMS and 2 Gravimetric stations as identified by the KSPCB. Determine appropriate calibration curve using Gravimetric / CEMS measurements against MicroPEM gravimetric, MicroPEM gravimetric Vs MicroPEM Optical , and MicroPEM Optical compared to SWB Sensor Use calibrated SWB sensors along with open source visual system developed on RPi to determine correlates between daily PM2.5 and PM10 levels, their accordance with NAAQS levels, and the potential health impact. Quantitative data is cross-validated with qualitative data obtained through community surveys regarding health impacts of permanent work staff of exposed retail stores and commercial enterprises at ground level. Our approach will be guided by the approach below which, along with many of the diagrams and illustrations that follow, have been obtained from the report “ Air quality assessment, emission inventory and source apportionment study for Bangalore city ”. Aspects of landuse patterns and population density may or may not be considered durng this pilot stage.
18
This bottom-up air quality data collection-oriented (as opposed to simulation-based studies) aims to gather, analyse and interpret a high quality dataset of PM2.5 and PM10 levels, using a praxisorientation that ties into our larger mission at SWB, corresponding to a spatial grid microscale pattern around a major traffic intersection in Bengaluru, Karnataka, India of commodity air quality instruments - in two variations of the Trident, the Dylos 1700 and the MicroPEM. These transformed datasets based on NAAQS standards of daily PM2.5 and PM10 levels are then correlated with varying ‘transportation fluidity categories’ (self-termed) - standing traffic / freely moving traffic - captured via a vision capture system running on a IoT-specific hardware platform to determine the contributions of this major PM sources to ambient levels on a microscale which also lends itself to developing understandings on short-term timescales . We believe this practical understanding of by how much and how often daily PM levels around major transportation hubs exceed NAAQS levels and their short-term, microscale health impacts on commuters and residents who live and work in-and-around these highly polluted transport hubs will assist with planning for new transportation grids in India’s smart cities and adding to transportation grids in existing Indian metros and Tier 2 cities.
Statistical Method:
Application of a multi-lag regression approach to determine on-road PM10 and PM2.5 emission rates
Use knowledge of VKT to determine total exposure levels
Concentration - Dosage approach to determine health impact
Simulation-based Approach:
Using CFD modelling (described in the relevant chapter below)
19
Qualitative method (Potential): Interviews of bike commuters / shop owners near traffic junctions are they are likely to be experiencing the greatest levels of pollution
These CAQSDs are capable of capturing data on 1 minute timescales which are necessary for emissions screening applications; We use video recordings placed at major traffic intersections to understand the densities of various ‘transportation fluidity categories’ (self-termed) along with the extent of standing traffic / freely moving traffic; The PM2.5 and PM10 levels front these CAQSDs are then correlated with these traffic density parameters to obtain an understanding of the interrelationships; We believe this practical understanding of by how much and how often daily PM levels around major transportation hubs exceed NAAQS levels and their short-term, microscale health impacts on commuters and residents who live and work in-and-around these highly polluted transport hubs will assist with planning for new transportation grids in India’s smart cities and adding to transportation grids in existing Indian metros and Tier 2 cities. We will also be using a co-location approach where we place our CAQSD next to the CPCB regulatory monitor at Alandur to validate its accuracy; We will be using the “live traffic” feature provided by Google Maps to understand the current and historical intensity of traffic levels in the coverage area thus enabling us to focus our site survey.
20
Chapter 4.
Sensoric Analysis and Development
4.1 Trident Replicate particle air monitor
Figure 4.1. Trident : concept
The explosion of DIY, AQ devices like our Trident that incorporate multiple sensors on board that can capture emissions in a 360 degree view, and which are highly portable, real-time, can be seeded densely in a local environment (which is particularly prone to the such fugitive emissions), and coupled with smart machine learning algorithms with a sound mathematical basis in plume dispersion models offers the possibility for the first time to tracking such fugitive emissions.
21
4.1.1 General layout
Enclosure design
MQTT
Hardware
Trident
gammification platform
post daq callibration
Figure 4.2 Trident Working
Figure 4.3: Schema : SWB
22
4.1.2 Enclosure
Figure 4.4 Hollow View: Trident
Figure 4.5 Holding Plate: Trident
Figure 4.6 Assembly as Seen from below Trident
23
The trident has 3 external holes at an offset 120 degrees from each other, these holes serve as the basis for the air inlet. Entering air is sucked up in a swirl by a fan at the center via 3 holes each offset from its nearest external hole by 60 degrees. Given that between every two external holes there is one internal hole, we assume that the swirl generated will be sufficiently strong to mix pm and negate differential airflow .
The sensor base plate above tight fits the suction end and forces air to pass through the sensors placed on the plate. This flow is what is primarily sensed by the ppd 60/45 in the plate. once the plate is tight fitted to the suction end, the sensor plate also acts as a support to distribute pressure from top of the sensor to the base By placing sensors in this fashion we are able to obtain a 360 degree representative view of the sensor’s immediate environment which means that a representative PM reading can be obtained no matter the direction of the wind, which in case of a single-sided vent entry would not be representative of the actual PM levels emitted as it may neglect to account for rapidly changing PM levels (which will be important for emissions monitoring given that localisation of the emission is going to be a key success parameter) 4.1.3 Hardware We suggest a replicate sensor platform that enables us to reduce noise and random error the requirements are:
Nova /Shinyei/ Dylos
1 Arduino boot loaded atmega328
1* esp8266 /wt 8266
1* DHT22 / DHT11
1* fan driver circuits
1* battery
1* batter ports
4* solar panel
24
The atmega 328 has sufficient ram to allow for data to get burnt onto the Arduino while the esp runs and also enough gpios for considering further device expansion so in the event where more sensors are added, the line can be utilized to add more devices or control circuitry. The fan driver circuits is a simple transistor based switch with a 12ohm 12 Watt current limiting resistor The Esp 8266 has an rf balun which allows it to connect to any IEEE 802.11.4 b/g/n device 4.1.4 MQTT MQTT stands for MQ Telemetry Transport. It is a publish /subscribe, extremely simple and lightweight messaging protocol, designed for constrained devices and low-bandwidth, highlatency or unreliable networks. The design principles are to minimize network bandwidth and device resource requirements whilst also attempting to ensure reliability and some degree of assurance of delivery. These principles also turn out to make the protocol ideal of the emerging “machine-to-machine” (M2M) or “Internet of Things” world of connected devices, and for mobile applications where bandwidth and battery power are at a premium. MQ Telemetry Transport (MQTT) is a lightweight broker-based publish/subscribe messaging protocol designed to be open, simple, lightweight and easy to implement. These characteristics make it ideal for use in constrained environments, for example, but not limited to:
Where the network is expensive, has low bandwidth or is unreliable
When run on an embedded device with limited processor or memory resources
Features of the protocol include: The publish/subscribe message pattern to provide one-to-many message distribution and decoupling of applications A messaging transport that is agnostic to the content of the payload The use of TCP/IP to provide basic network connectivity Three qualities of service for message delivery:
25
"At most once", where messages are delivered according to the best efforts of the underlying TCP/IP network. Message loss or duplication can occur. This level could be used, for example, with ambient sensor data where it does not matter if an individual reading is lost as the next one will be published soon after. "At least once", where messages are assured to arrive but duplicates may occur.
"Exactly once", where message are assured to arrive exactly once. This level could be used, for example, with billing systems where duplicate or lost messages could lead to incorrect charges being applied.
A small transport overhead (the fixed-length header is just 2 bytes), and protocol exchanges minimized to reduce network traffic. A mechanism to notify interested parties to an abnormal disconnection of a client using the Last Will and Testament feature
4.1.5 Post Data Acquisition Calibration Currently we are running 3 types of data calibration methodologies 4.1.5.1 Local averaging Every sensors data is averaged locally for 10 iterations (30 analog read cycles ) each read being 40 ms apart the data is then stored and converted to microgram per meter cube by as 12 degree polynomial. This data for each sensor input is stored with a unique identifier and transmitted until a successful transmission occurs the data is time stamped on the computer assuming no significant lag. 4.1.5.2 Spike removal As the name indicated the spike removal algorithm has two phases :The algorithm provides for a warm up time that allows the device to reach a stable value before the collected data is stored for analysis. The algorithm seeks and eliminated data sets among data clusters where the standard deviation is way too high for a stable reading. This subsequently leads to a drop in all nearby replicate sensor data sets if they are found to be within a limit for standard deviation. In the event all sensors are found to spike the data is noted and the sampling frequency doubled to estimate if it’s a FEE. 26
4.1.5.3
Further Calibration work
Over the past few tests, we have come across a few methods for the three sensor Shinyei platform what we call the trident. Shinyei ppd60s have been actively used as sensors for real time data acquisition for a number of commodity (pm2.5) sensing devices. But even on colocation, during testing in India, we found that on most occasions the three devices rarely come to a conclusive value directly. We propose a system that combines these values into one value based on the behavior of each Shinyei during the sensing time window. The data for a small window is then compressed into a point which represents the overall sensor behavior in that particular window. The newly obtained data can be then sent to the calibration equation which takes sensor values closer to the actual readings of the reference sensor. 4.1.5.4 Real-time collection The procedure is a 3 step process:
DAQ
Point compression
Real time Calibration
Figure 4.7 Flow: Real time correction
This assumes that testing environments have been charted and the trident responses have been used to gather calibration equations pertaining to those environments. 4.1.5.4.1 Point compression: Each point acquired by default is an un-weighed average of 10 samples. Then a small window is then subjected to one of the three described techniques to condense into a single point. The selection of the technique is solely based on the data’s intensity and standard deviation characteristics. 4.1.5.4.2 True mean point compression: This method assumes that all data points in the window are equally important to the window so it directly applies a mean filter on the given data set. Such sets were sometimes found to be less volatile while in clean/ low pollution conditions.
27
The sample below shows that there is more stability in the true mean data-set (pre-callibration) which was still found to be about 9 times higher than the deviations sensed by the micro PEM, which was found to be 0.492153.
Table 4.1: True mean data compression
time median filtered data mean filteredtrue datamean 1/21/2016 12:36 42.33053178 42.19681405 32.295 1/21/2016 12:36 36.15441384 37.29890299 26.68917 1/21/2016 12:36 28.50115467 29.00374393 22.95917 1/21/2016 12:36 28.64633364 28.33483379 23.16833 1/21/2016 12:37 26.52004529 26.98575542 21.38583 1/21/2016 12:37 26.04679724 26.00939086 21.86417 1/21/2016 12:37 28.41876224 28.33446434 22.89583 1/21/2016 12:37 28.98666457 28.64935729 19.405 1/21/2016 12:38 32.49380685 32.23111437 23.26167 1/21/2016 12:38 31.5124507 31.65827841 21.41833 1/21/2016 12:38 36.57369542 36.33385593 26.82417 1/21/2016 12:38 44.3487494 44.46373513 34.00417 1/21/2016 12:38 39.67915648 39.56181512 28.32083 Standard dev 5.897633747 5.910775145 4.237909
4.1.5.5 Dynamic Weighed mean In relatively more stable environments we found that often the true mean method failed to reduce the standard deviation of the data to a more stable value. So we concluded that a weight had to be factored in corresponding to the sensor standard deviation in that very time window so we came up with this equation: avg(S1) avg(S2) avg(S3) + + std(S1) std(S2) std(S3) 1 1 1 + + std(S1) std(S1) std(S1)
28
Where: avg() is the average of Sensor data in the particular sensing window std() is the standard deviation of Sensor data in the particular sensing window S1,S2,S3 correspond to the three sensors in the trident This was found to be particularly useful under low pm counts, with high sampling rates. The example
below has 5 second sampling. Table 4.2: Dynamic Mean Point Compression
Sensor 1
Sensor 2
Mean
dynamic mean
18.33
13.73
16.03
14.91267574
18.23
14.48
16.03
15.40845192
17.95
14.07
16.01
15.83299803
18.23
13.73
15.98
15.78094673
17.21
14.07
15.64
15.50796744
16.85
13.61
15.23
15.11074771
16.88
13.73
15.305
15.19378146
16.46
13.74
15.1
15.01110811
15.85
14.34
15.095
15.05159746
16.46
13.97
15.215
15.15358108
0.841644224 0.276949454 0.393414857 0.305613814
4.1.5.5.1 Dynamic median Similar to the dynamic mean the method here utilizes the formula: med(S1) Med(S2) Med(S3) + + std(S1) std(S2) std(S3) 1 1 1 + + std(S1) std(S1) std(S1) 29
Where med(x) is the median filter applied to the dataset, while the remaining are the same as above. In highly noisy environments the median filter often gives a selected point which is most isolated from the noise. The samples below were from the rooftop of a 5 floored building next to a highly active traffic junction. Table 4.3 Dynamic Median Point Compression
median filtered data
mean filtered data
58.16704374
48.66
70.22074168
60.416795
71.9713449
81.975
83.91648332
73.541
86.39311571
86.341
92.48246705
102.4824671
97.0830403
94.0256
83.00912626
92.00989
85.25117487
76.2518723
82.96670926
83.93341
Std
dev
- 15.66344713
10.83013951
4.1.6 Calibrations Now while developing calibration equations we observed that the equations with 𝑅 2 < 60% were highly environment specific. So to actually have a common equation that goes well with all the use cases, for instance hold correct for both the calibration equations obtained in a clean environment as well as a more polluted environment, for instance, a traffic junction during the peak hours, we believe that we need a system that can classify data based on the environment and 30
use the pre-stored calibration equation which pertains to that environment. This would work well when the environment under mapping would be increasingly similar to the environments under which the calibration equation was obtained but would begin to fail when the mapped environment would slowly start to move to an environmental condition that belongs to neither of the tested environments, requiring a method to transform the obtained calibration equations into each other. For instance a huge room with a person smoking in it would neither be as clean as a lab, nor be as polluted as a traffic junction during the peak traffic hours and would hence require a new calibration equation altogether. We assume that a data set has pm over time and no other parameter as an output, with inputs being the Voltage output, time, relative humidity(relationship not ascertained yet ) and temp (relationship not ascertained yet ). This can transform the pm v/s time plot into an intensity versus standard deviation form, which can be utilized to ascertain the type of test environment the dataset bears a resemblance to and to what degree. We propose the usage of a reference sensor, the micro PEM as a supervisor classify the raw test environment readings of the “Trident” into two or more categories and classify the real-time data so as to better “fit” the correct calibration slope, while using the proximity to the core of the classified regions (the training environment) as a factor to modify the new dataset to match a theoretical extrapolation of the reference’s (micro PEM’s) behavior at that point. One possible way of doing it can be to extrapolate the intensity- Standard deviation plots on closest training environments to a ratio of distances, where the inverse of the distance to the testing region would be multiplied as a weight to the output of the corresponding calibration equations and then normalized. 𝑛𝑡ℎ 𝑠𝑒𝑡
𝐶𝑎𝑙𝑖𝑏𝑟𝑎𝑡𝑒(𝑥) = −𝑎0 + ∑ (
𝑊𝑖 × 𝐹𝑖 (𝑥𝑑𝑎𝑡𝑎 ) ) 𝑊𝑛
𝑖=1
Where: 𝐶𝑎𝑙𝑖𝑏𝑟𝑎𝑡𝑒(𝑥) is the new equation 𝑊𝑖 : the weight is the inverse of the distance from the training set from the calibration equation 31
𝐹𝑖 (𝑥𝑑𝑎𝑡𝑎 ): is the corresponding calibration equation for that training set applied onto the current data set 𝑎0 : the zero offset of the trident. 4.1.7 Assumptions In a small sensing window, the noise lies at the extremities of the data set.
In a small sensing window, accuracy is proportional to the stability.
The sensors have not been subjected to poisoning in anyway.
The micro PEM, which is factory calibrated, is accurate.
The calibration equations achieved is not overfit.
4.2
Jellyfish
Test conditions show that the dylos DC1700 and most air quality sensors demonstrate sensitivity to airflows the jelly fish is an attempt to create an enclosure that can limit wind peak influence within the sensor mainframe. The first prototype was made with a funnel like shape to slow down air flow to compliment the dylos’s internal air flow.
Figure 4.8 Jellyfish Alpha
32
4.2.1 Introduction
Design Goal : to provide drawings for a suitable housing arrangement for our Air Quality Sensor Device (AQSD) The sensor device being housed, called the Dylos1, whose specs are available at the URL, uses optical technology to measure particulate matter (or PM) of average sizes of approx 2.5 and 10 microns respectively. The basic principles uses the detection of light scattering intensity to size and enumerate particles that are entrained across a focussed light source (either laser beam or LED).
There is a passive air inlet coupled to an active air exhaust (see Fig 1 below) to ensure that the same (rather, similar) volume of airflow is provided across the light source within each sampling window. It is very important that the same volume of air be sampled as it will inaccurately reflect the actual PM count which is represented as no of particles / unit volume.
Figure 4.9Rear and Front Shots of the Dylos
1
http://www.dylosproducts.com/dc1700.html
33
4.2.2 Design Constraints To limit the impact of strong wind on the Dylos’ intake which, as has been explained below, can lead to distortions in the volume of air being sampled leading to inaccuracies;
To prevent rain ingress into the housing and potentially into the AQSD which will not only facilitate corrosion but also lead to measurement errors as humidity has been shown to be a strong influence on an optical particle counter - on this note, arrangement that limit humidity within the housing, and by extension the AQSD, will be considered a plus;
The housing should be designed to prevent ambient light from entering the AQSD to prevent stray light from interfering with the AQSD’s detection sensitivity;
To prevent the ingress of insects such as mosquitoes, flies, and spiders.
To map the behavior of another device, assuming a similar footprint but with a lower flow rate would behave in the so created setting
4.2.3 Testing The setup used had a jellyfish derived from locally available material and housed a dylos DC1700 within the device, as well as the same device without the jellyfish enclosure. Table 4.4 : Wind testing Jellyfish Alpha.
Dylos Jf exposed error Converted JF converted free 27.6 30.88 -3.28 16.47651978 18.4346 -1.95808 Green indicates sensor Agreeent 27.54 30.14 -2.6 16.44070126 17.99284 -1.55214 red Indicates Disagreement 27.09 30.4 -3.31 16.17206235 18.14805 -1.97599 (red) NOT due to Wind peaks 26.03 29.72 -3.69 15.53926848 17.74211 -2.20284 Yellow indicates possible Wind peaks 42.48 34.06 8.42 25.35951306 20.33298 5.026532 41.13 33.49 7.64 24.55359633 19.9927 4.560892 41.39 32.93 8.46 24.70880992 19.6584 5.050411 27.33 30.54 -3.21 16.31533644 18.23163 -1.91629 28.32 30.38 -2.06 16.90634204 18.13611 -1.22977 18.86 29.46 -10.6 11.25895519 17.58689 -6.32794 17.03 30.6 -13.57 10.16649029 18.26745 -8.10096 16.76 31.74 -14.98 10.00530694 18.948 -8.94269 17.22 30.99 -13.77 10.2799156 18.50027 -8.22035 17.71 31.58 -13.87 10.57243353 18.85248 -8.28005 17.76 29.6 -11.84 10.6022823 17.67047 -7.06819 18.3 29.44 -11.14 10.92464899 17.57495 -6.65031 47.73 44.55 3.18 28.49363367 26.59525 1.898382 49.56 45.9 3.66 29.58609857 27.40117 2.18493 40.35534 35.96605 4.3893 24.0911434 21.47085 2.620298 Result (green): Safe Sensor Agreement
34
In the practical scenario the sensors were found to be functional over a span of 24 hours, there was a sensor agreement based on the 80-20 rule established by the EPA . However the sensors inside the Jellyfish was found to be relatively stable 4.2.4 CFD analysis Followed by simulation in Ansys fluent, the CFD boundary conditions are the two fans and ambient wind speed 3m/s and atmospheric pressure 101325 pa.
Figure 4.10 CFD analysis Results
35
4.3
Iteration 2
Iteration 2 of the Jf was found to be needed due to the high cost of the manufacturing processes involved, specially keeping in mind that the most important constraint was a commodity sensor enclosure, which would imply that the enclosure has to be a fraction of the cost of the sensors under consideration, the version 2 was implemented.
Figure 4.11 Jellyfish Version2.0
The prototype was found to possess the same characteristics of its cone counterpart and was more functional from the manufacturing design point of view. The CFD analysis below proves the efficiency of the design.
36
Figure 4.12 External Flow under no fan condition
Figure 4.13 Internal Flow with forced input
With the fans guiding the flow only a flow redundancy of 0.38CFM remains which is then guided out by the two fans. 37
Chapter 5.
Results
Figure 5.5.1 JF Guided Flow Simulations
Figure 5.2 Flow analysis JF.
38
As seen by the fluent flow simulation above, even a gale force wind, which is at 13m/s has no effect on the internal flow which in turn maintains itself the trident was tested for high dust conditions and was found to work with regards to the overall average. Table 5.1 : Nova reference test
Timestamp Nova1 Nova2 Nova3 Nova3+offsetHumidity Micropem 5/17/2016 18:51 48.1 47.8 41.1 48.1 35 5/17/2016 18:51 48.2 48.1 40.9 47.9 34 5/17/2016 18:51 48.5 48.1 41.2 48.2 34 5/17/2016 18:51 48.9 48.2 41.2 48.2 34 5/17/2016 18:51 48.9 48.2 41.4 48.4 35 5/17/2016 18:51 49.3 48.5 41.4 48.4 35 5/17/2016 18:51 49.3 48.4 41.4 48.4 35 5/17/2016 18:51 49.4 48.6 41.3 48.3 35 5/17/2016 18:51 49.4 48.7 41.4 48.4 36 5/17/2016 18:51 49.4 48.7 41.4 48.4 36 52 Results 48.94 48.33 48.27 34.9 52
The novas are independently found to agree with the micropem. With the true mean filter , The agreement gives and an 𝑅 2 of about 0.9329 which is 0.1329 higher than the EPA recommended value of the total 𝑅 2 . The True mean, (Table 1), Dynamic mean (Table 2) and Dynamic Median (Table 3) were Found to work with respect to the Shinyei PPD60, allowing a calibration to at most an 𝑅 2 of 69 and was hence removed as the core unit of the however the NOVA never made it beyond the threshold of 5 µG/m^3 as a standard deviation. The ability to sense trends instantaneously, still remains with the shinyei ppd60 and was noted as a result of sensoric profiling.
39
Chapter 6.
Conclusion
Although the devices in the air quality segment are mostly uncharted, and to a vast degree require, a sound study in terms of profiling and accuracy studies. The project successfully covered profiling of 6 Air Quality sensors : Air Quality Egg, Air beam, Dylos DC1700, DC1100, Micropem, IndiaSpends AQT and the Foobot as exposed to external environments in India and successfully built a common indoor to outdoor converter housing to stabilize the sensors and help them adapt. In terms of calibration, the multilevel classifier worked on dummy data sets, and is soon to be used for real time pilot studies to better understand its capabilities to create an adaptive code for sensor calibration given a set of parameters, currently in use with, but not restricted to only RH and PM 2.5 levels. The trident as a device was found to be as stable as our reference quality monitor and is under further validation.
40
References [ 1 ]. [ 2 ].
http://aqicn.org/city/india/bangalore/bwssb/ SPECIFICATIONS AND GUIDELINES
FOR
CONTINUOUS
EMISSIONS
MONITORING SYSTEMS (CEMS) FOR PM MEASUREMENT WITH SPECIAL REFERENCE TO EMISSION TRADING PROGRAMS [ 3 ].
(http://blogs.app.com/enviroguy/2015/06/08/high-air-pollution-from-idling-constructionvehicles/)
[ 4 ].
(http://www.stopthesoot.org/sts-idle.htm)
[ 5 ].
(http://www.nj.gov/dep/enforcement/advisories/2015-08.pdf).
[ 6 ].
Joshua :http://eetd.lbl.gov/node/55616
[ 7 ].
Enabling Low-Cost Particulate Matter Measurement for Participatory Sensing Scenarios by M Budde, 2013.
[ 8 ].
http://cseindia.org/node/5072.
[ 9 ].
D. M. Holstius et al., 2014, Field calibrations of a low-cost aerosol sensor, AMTD, 7, 605– 632
[ 10 ].
World Bank report , 2012
[ 11 ].
Air Pollution Database for Tamil Nadu - 2014.
[ 12 ].
The Keys to Estimating Mobility in Urban Areas, CAMSYS and TTI,2005
[ 13 ].
http://www.cpcb.gov.in/CAAQM/frmStationdetails.aspx?cityID=136
[ 14 ].
Traffic Regulation Board
[ 15 ].
https://en.wikipedia.org/wiki/Traffic_simulation Real-time Traffic Information System (RTIS)
[ 16 ].
http://www.newindianexpress.com/cities/chennai/IIT-M-Comes-Up-With-Real-TimeInfo-System-on-South-Chennai-Traffic/2014/08/20/article2387711.ece
[ 17 ].
Mahmoud Abu-Allaban*, John A.Gillies, Alan W.Gertler, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, US,A, 13
41
[ 18 ].
February 2003, Application of a multi-lag regression approach to determine on-road PM10 and PM2.5 emission rates, Atmospheric
[ 19 ].
Environment 37 (2003) 5157–5164
[ 20 ].
Air Pollution Database in Tamil Nadu
[ 21 ].
NATIONAL AMBIENT AIR QUALITY STATUS 2009
[ 22 ].
NATIONAL AMBIENT AIR QUALITY STATUS & TRENDS - 2012
[ 23 ].
Cars not to blame for Delhi pollution
[ 24 ].
Forget Shanghai, Bengaluru is worse off than even China’s 13th largest city
[ 25 ].
MODELLING URBAN TRAFFIC AIR POLLUTION DISPERSION
[ 26 ].
Estimation of the contribution of road traffic emissions to particulate matter concentrations from field measurements: A review
[ 27 ].
B'luru air 200 pc more polluted than national quality standard
[ 28 ].
Real-time estimation of pollution emissions and dispersion from highway traffic
[ 29 ].
Urban transportation emissions mitigation: Coupling high-resolution vehicular emissions and traffic models for traffic signal optimization
[ 30 ].
Air quality assessment, emission inventory and source apportionment study for Chennai city
[ 31 ].
Air Quality Monitoring Project-Indian Clean Air Programme (ICAP)
[ 32 ].
CONCEPTUAL GUIDELINES AND COMMON METHODOLOGY FOR AIR QUALITY MONITORING, EMISSION INVENTORY & SOURCE
[ 33 ].
APPORTIONMENT STUDIES FOR INDIAN CITIES
[ 34 ].
Emission Factor development for Indian Vehicles as a part of Ambient Air Quality Monitoring and Emission Source Apportionment
[ 35 ].
Studies
[ 36 ].
CRITICAL ANALYSIS OF STRATEGIES FOR PM REDUCTION IN URBAN AREAS
[ 37 ].
EXPOSURES AND HEALTH RISKS DUE TO TRAFFIC CONGESTION, Kai Zhang, A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Environmental Health Sciences) in The University of Michigan, 2010
42
Appendix Acrylic Bending Calculations For using Acrylic bending the following calculations were made:Upper circle diameter = a = 0.186mts. Lower circle diameter = b = 0.381mts. Height of T.C. = H = 0.268mts The calculations are as follows: Circumference of upper circle =Π x a = 0.584mts. Circumference of lower circle = Π x b = 1.197mts.
Apothem (h) 0.286mts. (H)
0.584mts.
0.0975mts
(a)
(c)
Figure A.1 Showing the appropriate dimensions of T.C.
Apothem of T.C. (slant height) = h = √𝑐 2 + 𝐻 2 = 0.30212mts. The apothem can be approximated as 0.3 mts.
43
Figure A.2 Description of unfold T.C. in to a planar surface
From the figure 3 – we can observe two arcs with same center and center angle with different radii. The larger arc represents the lower circle of T.C. and the smaller arc represent the upper circle of the T.C. So The circumference of upper circle of T.C. = the circumference of smaller arc 0.584 = radius of smaller circle (r1) x center angle (θ) Similarly The circumference of lower circle of T.C. = the circumference of larger arc 1.197 = radius of larger circle (r2) x center angle (θ) Difference in radii = apothem of truncated cone
44
List of competing devices. Competitive Matrix Names Status Features Clad Equipment Pre-Beta PM2.5, UV, H, T ChemiSense Pre-Beta VOC, NO2, SO2, CO, CO2 Carnegie Speck Beta PM2.5 Sensaris SensBlocks Beta UV, "Air Pollution", H, T Plume Beta PM2.5, PM10 Berkeley Pats+ Beta PM2.5 Pimi Airbox Beta PM2.5 AirCasting AirBeam Beta PM2.5 Alima Beta VOC, CO, CO2, A, L, PM2.5, H, T Airnut Beta CO2, PM2.5, H, T Kanarci DIY PM2.5 SensorDrone Available Light, IR, CO, H, T Lapka Environmental Available Monitor Nitrates, Radiation, EMF, H Air Quality Egg Available NO2, CO, H, T Smart Citizen Available N, NO2, CO, H, T Dylos Available PM2.5, PM10 Netatmo Weather Available Station N, B, W, "Air Quality", CO2, H, T Breezometer Available Mobile App, AQI Yu Cheng Tech Available PM2.5 Asian company.. Available PM2.5 Haier Air Box Available PM2.5, VOC Danaus Available VOC iKair Available PM2.5, more options Blit Available Formaldehyde Xnose Sensor Available Formaldehyde SunSprite Available UV, L AirAir Failure? PM2.5 Community Sense Failure? CO, Nox, Ozone, H, T SilverCanary ??
Cost TBA $150 $200
$200
$150 $250 $200 $175 $300 $180 $0 ¥338 ¥780 ¥399 ¥399 ¥369 ¥388 ¥400 $105 $70
T=Temperature H=Humidity B=Barometric N=Noise W=Weather A=Accelerometer L=Light
Table A.1Competing Devices
45
46
