Sponsored by:

Implementing Embedded AI to Measure Indoor Air Quality

Implementing Embedded AI to Measure Indoor Air Quality

Learn More
.
The Challenge
THE PERSONAL IMPACT
Studies show that in today’s lifestyle 90% of a person’s time is spent indoors—indoor air quality matters!
Comfort Impact

Poor indoor air quality can cause fatigue and drowsiness in the short term to dehydration and muscle cramps in the long term. Exposure to high levels of temperature and humidity can even cause heat exhaustion and fainting.

 

Health Impact

Poor indoor air quality can produce eye and throat irritation in the short term to respiratory complications in the long term. Exposure to high levels of some pollutants can result in cancer and even death.

 

 

Economic Impact

Poor indoor air quality can have multiple levels of economic impact from the cost of HVAC equipment to regulate temperature and humidity to the costs incurred to alleviate the health and comfort impacts mentioned above through products and medications.

Gases...Gases Everywhere
You may not know it, but you encounter these potentially harmful gases in your daily life.
Home Concerns

Everything from outgassing of furniture and carpets to the chemicals used in detergents and perfumes, as well as cooking odors, automobile fluids and exhausts, and pesticides housed inside can provide indoor air quality challenges.

School Concerns
Everything from outgassing of furniture and carpets to the chemicals used for cleaning classrooms and bathrooms as well as the mixture of perfumes, deodorants, and clothes washing detergents brought in by students and teachers can provide indoor air quality challenges.
Office Concerns
Combine the concerns of the home and school and you have the office environment, with the addition of the chemicals used in printers and copiers—as well as other office equipment—and your indoor air quality challenges can increase quickly.
What We’re Talking About
What We’re Talking About
TVOC  (Total Volatile Organic Compounds)
0 - 1 mg/m3
1 - 3 mg/m3
3 – 10 mg/m3
TVOCs come from everywhere: burning fuels, tobacco, personal care products, cleaning agents, paints and lacquers, and printers and copiers.
CO2  (Carbon Dioxide)
400 - 1000 PPM
1000 - 2000 PPM
2000 - 5000 PPM
Some of the highest levels of CO2 come from human breath in enclosed spaces. Other sources include commonly used items such as heaters, dryers, and gas appliances.
Humidity
0% - 30%
30% - 60%
60% - 100%
Once again, increased humidity levels can come from showers and bathrooms, as well as kitchen dishwashers and clothes dryers. Other sources of humidity include rising dampness, leaks, and porous walls.

Household Locations and Sources of VOCs/Ratings

Use cases for indoor air quality sensing include every room in your house, as well as your garage and close outside areas, such as porches, where you might cook on an outside grill.
Living Room
  • Furniture
  • Upholstery
  • Wall Paint
  • General TVOC
Kitchen
  • Cooking
  • Detergents
  • General TVOC
Bathrooms
  • Detergents
  • Perfumes
  • General TVOC
Bedroom
  • Furniture
  • Upholstery
  • Carpets
  • General TVOC
Garage
  • Automotive Fluids
  • Exhaust Products
  • Pesticides
  • General TVOC

Other Sources of Indoor Air Quality Pollutants

Renesas IAQ Rating

≤ 1.99

 

2.00 - 2.99
 
3.00 - 3.99
 

3.00 - 3.99

 

≥ 5.00

 

Reference Level*

Level 1

 

Level 2
 
Level 3
 

Level 4

 

Level 5

 

Air Information
Clean Hygenic Air
(Target value)
Good Air Quality
(If no threshold is exceeded)
Noticeable Comfort Issues
(Not recommended for exposure > 12 months)
Significant Comfort Issues
(Not recommended for exposure > 1 month)

Unacceptable Conditions
(Not recommended)

TVOC (mg/m3)

< 0.3

 

0.3 - 1.0
 

3.0 - 10.0

 

3.0 - 10.0
 

> 10.0

 

Air Quality

Very Good

 

Good
 

Medium

 

Poor

 

Bad

 

*Based on a study by the German Environment Agency (UBA)
The Sensor
Key attributes of an indoor air quality gas sensor should include the ability to sense harmful levels of VOCs and / or CO2. Sensors should have low detection limits, require no maintenance, and provide long life, all for a reasonably low price.
SENSOR COMPARISONS
Chemiresistor Sensors

Detects VOC and e-CO2


Low detection limits (ppb to ppm)

OFFERS

  • Long life
  • No maintenance
  • Low price
Electrochemical Sensors

Detects VOC, usually ppm levels

 

 

OFFERS

  • Less robust

  • Short life

  • Expensive

PID
Sensors

Detects VOC and methane

 

Low detection limits (ppb to ppm)

OFFERS

  • Short life

  • Higher maintenance

  • Expensive

NDIR
Sensors

Detects VOC and CO2

 

 

OFFERS

  • Good CO2 detection

  • Not for trace detection (ppb to low ppm) of VOCs in ambient air

  • Expensive

Mox Sensor Technology Principles

Chemiresistor Working Principle
Gas generates a free charge carrier in the MOx (Metal Oxide) sensor.
Chemiresistor Reaction on MOx Resistivity

Oxygen and gas molecules are adsorbed on the MOx surface. The reaction of these events causes the oxygen equilibrium on the surface to be disturbed, transferring a charge.

Influences to MOx Resistivity

The factors which influence the MOx resistivity include the surface geometry of the sensor, the oxygen concentration, what gases are present, and the temperature, all of which help the MOx to sense multiple gases.

Mox Sensor Technology Principles

1

Digital communications with I2C interface with up to 400kHz.

 

2

Accurate heater with temperature control of ±05K.

 

3

Tailored sequences enable sensitivity to target gases and selectivity.

4

New methods and algorithms via software upgrade to ASIC settings.

 

Artificial Intelligence
AI has become an extremely powerful tool. Through the use of neural networks, more complex problems can be solved, which enables companies to leverage AI in multiple ways.
UNDERSTANDING THE BASICS
Definition

Computer systems that are able to perform tasks that normally require human intelligence and interaction.

Machine Learning

“Field of study that gives computers the ability to learn without being explicitly programmed.”
— Arthur Samuel, 1959

 

Neural Networks

A type of AI that uses a multilayer neural network to integrate large number of known input data samples with corresponding outputs in order to train and optimize the neural network model enabling it to predict outputs given an unknown input.

Emedded AI

Offline Training
All training is completed prior to algorithm deployment.
Capture Results

All results are captured in the algorithm for use with the sensor.

Expected Use Cases
Comprehensive data sets are required for training and should encompass all expected use cases.
Training a Gas Sensor
Training and testing must be performed using a lab (artificial) as well as real world locations in order to collect an extensive amount of data so that the sensor can provide stability and sensitivity to the gasses of interest.

Algorithm Development and Refinement

1
Project Definition

To begin, the project must be fully defined so that the sensor is able to detect the target gas.

 

 

2
Data Collection & Analysis
Test sensors are incorporated to target the gases defined, collect relevant data, analyze the data, and optimize the sensor’s operating method.
3
Modeling & Deployment

Once the neural network model has been trained for the specific application, the algorithm can be finalized and deployed in firmware.

 

Odor Description Example

Algorithm Output — Classification of Odors by Categories
  1. Using single MOx operated with temperature modulation.
     
  2. Collect sensor response data with lab gas exposures.
     
  3. Target gases can be classified into groups using AI.
     
  4. Pre-trained models mean low computational complexity when deployed in firmware.

System Solutions for End Products

Accelerating time to market is a key focus for engineers.  Application-specific MCUs combined with indoor & outdoor air quality as well as humidity sensors that have pre-configured firmware that enable rapid time to market with exceptional performance.

 

Examples of end products that feature these solutions include vent fans, air purifiers, air conditioners, lighting and electrical fixtures, smart TV monitors, and conference phones.

Additional Resources

On-Demand Webinar

View Webinar

TVOC and Indoor Air Quality

View White Paper

ZMOD4410 Overview Video

View Overview Video