32
© All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

© All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Embed Size (px)

Citation preview

Page 1: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

© All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Page 2: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Did you ever measure a smell? Can you tell whether one smell is just twice strong as another? Can you measure the difference between one kind of smell and another? It is very obvious that we have very many different kinds of smells, all the way from the odor of violets and roses to asafetida. But until you can measure their likeness and differences, you can have no science of odor.

If you are ambitious to find a new science, measure a smell.

Alexander Graham Bell (1914)

Page 3: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

The Department of

presents…

Page 4: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Emerging Interdisciplinary Challenges

Robi PolikarOctober 16, 2002

Page 5: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Outline

Introduction: emerging interdisciplinary challenges Motivation and background The mammalian olfactory system vs. the electronic nose Commercially available electronic nose systems Quartz crystal microbalances Experimental setup Identification of volatile organic compounds (VOCs)

An uncooperative database / sensitivity / selectivity issues Dealing with an uncooperative database

Automated Identification Neural Networks

Conclusions Questions, comments and suggestions

Page 6: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Introduction:Emerging Interdisciplinary Challenges

.

.

.

.

.

Olfactory Physiology

Organic Chemistry

Signal Processing

Pattern Recognition Computational Learning

ElectronicNose

Chemical Sensors /Analytical Chemistry

Page 7: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

IntroductionMotivation & Background

Food industries: detection of food quality / wholesomeness

Airport security: drug smuggling, detection of explosives Anti-personnel land-mine detection Detection of household chemicals Detection of hazardous gases

VX, CO, radon, etc Detection of volatile organic compounds Wastewater odor control

Many industries, institutions and organizations can benefit from a device capable of identifying odors:

Page 8: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Selectivity & Sensitivity Issues

• Humans can identify 10000 types of odors at varying sensitivity levels.

• 10000 odors are considered to be combination of a few basic types of odors: floral, musky, camphorous, pepperminty, ethereal, pungent (stinging), and putrid (rotten).

• Another group of researchers believe that this number is actually around 50.

• More recently, it has been suggested that there are actually over 1000 smell genes in the nose, each of which encodes a unique receptor protein.

• Sensitivity: 5.83 mg/L of ethyl ether,

3.30 mg/L of chloroform,

0.0000004 mg/L of methyl mercaptan (1/25 trillionth of a gram)

Page 9: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Mammalian Nose Vs. Electronic Nose

Mammalian Nose Electronic Nose

Receptor neuron Sensor / transducer

Odorant binding protein Coating

10000000 receptors 6-30 sensors (array)

Glomeruli Signal processing module

Brain Pattern recognition module

Sens. 1 part per trillion 1 part per million

Selec. 10000~20000 odors <50 odors

Page 10: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Electronic Nose Systems

Page 11: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Sensor Technologies

Metal Oxide Semiconductor sensors (MOS) Chemical Field Effect Transistors (ChemFET) Conducting Polymers (CP) Fiber Optical Sensors (FOS) Quartz Crystal Microbalances (QCM) Surface Acoustic Wave devices (SAW) Mass Spectrometry Gas Chromatography

Page 12: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Pattern Recognitiontechnologies

Statistical pattern recognition (SPR)Bayes classifiersDiscriminant analysis (DA)Maximum likelihood estimatePrincipal component analysis (PCA)

Non-parametric techniquesArtificial neural networks (ANN)Fuzzy logic (FL)Rule-based / expert systems

Page 13: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Com

merc

ially

Availa

ble

S

yst

em

s

Page 14: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Quartz Crystal Microbalances & Gas Sensing

Bare piezoelectric crystal

Central part of the crystal coated with first gold, and then polymer material

Electrode on front Electrode

on back

Crystal holder

A

WFF

26103.2

Page 15: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Coating Selection Considerations

For desired levels of selectivity and sensitivity • Thickness, softness / stiffness, reversibility, operation temperature

• Viscoelastic properties: thermal expansion, swelling due to sorption,

film resonance• Solubility parameters: coating – analyte interactions

Advantages Disadvantages

Thickness sensitivity resistance, phase lag, attenuation

Softness response time, reversibility

Attenuation

Stiffness Attenuation Reversibility

Temperature

Softness and hence response time

sorption and hence

sensitivity.

Page 16: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

VOCs and Coatings Used

O

Apiezon (grease, not a polymer)

APZ

Poly(isobutylene) PIB

Poly(diethyleneglycoladipate)

DEGA

Sol-gel SG

Poly(siloxane) OV275

Poly(diphenoxylphosphorazene)

PDPP

• 12 individual VOCs at 7 different concentrations (84 patters).• 24 Binary Mixtures of VOCs at 16 different concentrations (384 patterns)

Page 17: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Block Diagram of theExperimental Setup

Page 18: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Experimental Setup

Switching Box

Mass FlowController

NetworkAnalyzer

VOC inbubbler

NitrogenVOC

PC

SensorCell

Page 19: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

EXPERIMENTAL SETUP

Mass FlowController

Network Analyzer

Gas Bubbler

SensorCell

Mass FlowMeter

SwitchingBox

Post-Itnotes

Page 20: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

How Does Odor Signallook Like?

Page 21: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

• Existence of dominant VOCs

• Approach: Identify dominant VOC first, and identify secondary VOC based on the identification of the dominant VOC.

Problems With Problems With Identification Of MixturesIdentification Of Mixtures

APZ: Apiezon, PIB: Polyisobutelene, DEGA:Poly(diethyleneglycoladipate),

SG: Solgel, OV:Poly(siloxane), PDPP: Poly (diphenoxylphosphorazene)

Page 22: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Pattern Separability Issues

(a) Well separated patterns and (b) densely packed / overlapping patterns

Page 23: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Pattern (In)separabilityin Mixture VOC Problem

Sensor 1Sensor 2

Sen

sor

3

ETHANOL

TOLUENE

TCE

OCTANE

XYLENE

Page 24: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Identification of VOCs

Preprocessing

Increasing Pattern Separability

Neural Network Training

Neural Network Validation

VOC Identification

Raw Sensor Readings (6-D)

Filtering, Normalization, De-trending, etc.

Fuzzy nose (FNOSE), Feature range stretching, or

Nonlinear cluster transformation

Multilayer perceptron LEARN++ (for incremental learning)

Classification

.

.

.

.

.

Page 25: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Nonlinear Cluster Transformation

Outlier Removal

Cluster Translation

Nonlinear Cluster Transformation

C

jijii C 1

1mmMS

Generalized regression neural networks Similar to RBF networks Do not require iterative training Successful in multidimensional function approximation

Page 26: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

PRINCIPLE COMPONENT ANALYSISA Comparison

ETHANOL

TOLUENE

TCE

OCTANE

TOLUENE

OCTANE

XYLENE

ETHANOL

TCE

XYLENE

Page 27: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Artificial Neural Networks

SignalsOutput signal based on a

weighted average of input signals

.

.

.

.

.

Toluene

Xylene

From sensors(six)

Page 28: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

The Multilayer PerceptronNeural Network

d

iiJiJ xwnetf

1

……

....

……

...

Wji

Wkj

netj

i=1,2,…dj=1,2,…,Hk=1,2,…c

xd

x(d-1)

x2

x1

d inputnodes H hidden

layer nodes

c outputnodes

zc

z1

..

J

x1

xd

wJi

d

iijijj xwfnetfy

1

H

jjkjkk ywfnetfz

1

netk

netj

..zk netk

yj

.

.

.

.

.

Page 29: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

ResultsSingle VOC Identification

7 patterns obtained for each VOC, corresponding to seven different concentration values between 70 ppm and 700 ppm.

Thirty (30) of the total 12*7=84 patterns were used to train the neural network.

Remaining patterns were used to validate the performance of the network

All 54 validation patterns were identified correctly !

Page 30: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

ResultsBinary Mixture of VOCs

Dominant VOC Performance: 96%

Secondary VOC Performance: 96%

196 (50%) patterns used for training and remaining 196 used for testing.

Page 31: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Conclusions

QCM technology along with neural network identification can be used as an efficient tool for electronic nose applications

Challenges: Identification of components in mixtures Identification of gases at very low concentrations (ppb

levels ?) Adverse environmental conditions (temperature,

humidity, etc.) New sensor technologies for improved sensitivity and

selectivity Incremental learning of additional odorant (Algorithm:

Learn++)

Page 32: © All Rights Reserved, Robi Polikar, Electrical and Computer Eng. Rowan University, Glassboro, NJ 08028

Questions