You are on page 1of 29

PREM CHANDRAN

GAS LEAK / AIR QUALITY DETECTOR


DETECTING GAS (LPG) LEAKS/ AIR QUALITY

USING

INTERNET OF THINGS (IoT)


Agenda
● Motivations behind product concept
● Research on current products
● Design considerations
● Prototyping Product
● Testing product
● Analysis of product performance
● Challenges & Suggested improvements
Motivations behind product concept

https://www.channelnewsasia.com/news/singapore/gas-leak-causes-fire-disrupts-supply-to-1-305-jurong-west-8822726

https://www.straitstimes.com/singapore/cameras-to-be-deployed-across-island-to-detect-illegal-smoking

https://www.mindef.gov.sg/oms/imindef/mindef_websites/topics/elifestyle/articles/smoking/deadly_form_of_air.html
Flour dust
What harm can it do ?

https://www.wired.com/2008/03/the-explosive-t/

http://dustcontrol.com/post/213-dust-hazard-of-asthma-in-bakeries.html
Research on Current Products
Design Brief

To design and make a portable air quality detector that can inform user(s) when
air quality in location has reached a certain threshold posing a health risk
(e.g hazardous).

The air quality IoT system will be able to detect

● Smoke (cigarette smoke or soldering fumes causing breathing difficulty)


● Carbon monoxide (car exhaust gas leak, poor car park ventilation)
● Gas leak (gas leak in kitchen/lab leading to fires)-LPG(liquid petroleum) gas
● Dust ( factory conditions with fine particles, eg flour mill )

The IoT system will also operate actuators as a feedback to control the air
quality to reduce the effect of the above components
Design considerations

● How to place / expose the sensors effectively to measure


the environment variables of air pollutants

● Calibration of sensors for optimum responsiveness


(hardware and coding issue)

● What feedback systems to put in place to alert user of the threshold being
reached (will require immediate and remote)
Brainstorming
Prototyping Product
Testing Environmental Sensors
Dust Sensor Temperature and
Humidity Sensors
Prototyping Product

Testing Sensors
MQ2 Gas Sensors
Gas Sensor Characteristics

Name of Gas Cost Accuracy and Operating Environmental Reliability


Sensor Repeatability Range Range

Analog Gas Expensive Precision without Same Same Low


Sensor accuracy
SKU:SEN0127

Gas Sensor Affordable Precision without Same Same High


FC-22 (MQ2) accuracy

Grove - Gas Moderate Precision and Same Same Medium


Sensor(MQ2) accuracy
Prototyping Product
Testing Product ( 4 scenarios)
Gas Leak (in kitchen)- detect LPG gas Detect Smoke combined with Gas (LPG) leak
Concept Diagram of Gas Leak Detector
Raspberry Pi As Router
Link to Gas Leak Detector Sketch – Source Codes & Relevant Libraries
Process Flow Diagram of Gas Leak Detector with IoT
Testing of Gas Leak Detector with LPG
Testing of Gas Leak Detector with Smoke
Testing of Gas Leak Detector with CO from car exhaust
DASHBOARD DESIGN
Considerations

- What variables to measure?


(air quality, smoke, LPG , temp & humidity)
- What type of feedback indicators to use ?
(gauge or level)

- What actuators to use ?


- to warn user / environment victims (light, buzzer)
- to control the environment (fan)
Machine Learning (Classification problem)

GAS ON
GAS OFF
FLAME ON What happens if the
flame goes off and there
is gas leak?
Did I leave
the stove on ? How do i know if
a fire started ?

GAS ON GAS ON
FLAME OFF
+SMOKE
FLAME OFF
Sensor Variables
1st Experiment (4 variables)
https://docs.google.com/spreadsheets/d/1LVjZ6NA0DFSXrt6Iw2ujYWPGVuy_qiumn3azo4WQzFc/edit#gid=0

- Dust ( pcs/0.01cf)
- CO concentration-ppm (incomplete combustion and smoke)
- LPG level-ppm (liquid petroleum gas for gas leak)
- Smoke level -ppm (indicator of fire)

2nd Experiment (put in units) (6 variables)


https://docs.google.com/spreadsheets/d/1LVjZ6NA0DFSXrt6Iw2ujYWPGVuy_qiumn3azo4WQzFc/edit#gid=0
+ Temperature
- Dust ( pcs/0.01cf)
Humidity level
- CO concentration-ppm (incomplete combustion and smoke)
- LPG level-ppm (liquid petroleum gas for gas leak)
- Smoke level -ppm (indicator of fire)
Machine Learning applied to detecting Gas Leak / Smoke
detection

● 4 features ( 1st experiment)

● 6 features ( 2nd experiment)

Objective
To identify based on combination of sensor readings whether there is
Potential hazardous situation in the form of a
GAS LEAK or SMOKE (indicator of possible fire,
Or incomplete burning that could lead to unhealthy environment)
Machine Learning applied to detecting Gas Leak / Smoke detection
Machine Learning Algorithms
Machine Learning Results
75 samples, 4 features, 4 categories

29 samples, 6 features, 4 categories


Challenges in Data Analysis
● Getting sufficient data points in Hazardous zones (due to process danger)
● Possible by interpolating data (but will lead to overfitting of data ?)
● Controlled environment where ambient conditions are consistent so that only
measured features make a difference.
Suggested Improvements / Possible Applications
● Measuring more related variables ( eg temp & humidity) for stronger
correlation of sensor data to determine hazardous zones during time intervals
● More sampling of different use scenarios with varying ambient conditions (eg
wind) to correlate relevant sensor data (features) to categories
● Incorporate GPS module to include Coordinate when sending alarm

You might also like