Masters Thesis

Download as pdf or txt
Download as pdf or txt
You are on page 1of 98

Modeling and Validation of an Automobile Evaporative

Emission Carbon Canister

A Thesis

Presented in Partial Fulfillment of the Requirements for the Degree


Master of Science in the Graduate School of The Ohio State
University

By

Anoop Sunil, B. Tech

Graduate Program in Mechanical Engineering

The Ohio State University

2019

Master’s Examination Committee:


Dr. Giorgio Rizzoni, Advisor
Dr. Yann Guezennec
© Copyright by

Anoop Sunil

2019
Abstract

The automotive carbon canister consists of activated carbon particles designed to

adsorb and desorb Volatile Organic Compounds (VOCs) that evaporate from gasoline

present in the fuel tank. This is done to both increase fuel economy and decrease fuel

emissions. Experimental tests were done on an activated carbon bed measuring the

temperature and mass change during loading and purging of the carbon canister.

Experiments were conducted using n-butane as the volatile organic compound

during the loading phase and dry air during the purging phase. A 1D model was

developed in MATLAB to predict the canister behavior during these experiments.

The model predicts the mass change in the canister which in turn estimates the

amount of fuel available in the canister.

Since adsorption and desorption are exothermic and endothermic reactions respec-

tively, the model also predicts the temperature change along the bed of the canister

to study the effect of temperature on the adsorption and desorption process. The

experiments were used to validate the model using the activated carbon present in

an automotive canister.

The model developed in this thesis can be used to predict the loading and purging

event that occurs in an ideal carbon canister. This model can be used as the foun-

dation for a more complex model in the future and provide inputs to create a more

efficient automotive canister.

ii
This is dedicated to my faculty advisor, friends and collegues who have helped me

complete this project.

iii
Acknowledgments

I would like to thank everyone who has helped me complete my project to model

automotive carbon canister which resulted in my thesis. I would like to give a special

thanks to my faculty advisor, Dr. Giorgio Rizzoni, and project teammate, Ruochen

Yang. Your guidance and support have been of tremendous help and encouragement

to me. I would also like to thank everyone from the Ohio state university who have

provided valuable feedback during my project.

Thank you Luca Romagnuolo and Brian Rahman who supported the project and

for giving me feedback on my thesis. It has been a pleasure working with everyone

at CAR and all the friends I made along the way.

I would also like to send a special thanks to my family and friends who supported

me during the course of my graduate school and helped me complete my degree.

iv
Vita

August 11, 1993 . . . . . . . . . . . . . . . . . . . . . . . . . . . . Born - Kerala, India

2011-2015 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.Tech. Mechanical Engineering,


National Institute of Technology
Karnataka, Surathkal
2015-2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Assistant Manager, Tata Motors

2017-present . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.S. Mechanical Engineering,


The Ohio State University.

Fields of Study

Major Field: Mechanical Engineering

v
Table of Contents

Page

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Problem introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1


1.2 Purpose and Objectives . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1 Effect of fuel vapor on environment . . . . . . . . . . . . . . . . . . 6


2.2 Brief overview of EVAP-system and carbon canister . . . . . . . . . 7
2.2.1 Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 Purging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Activated carbon . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

vi
3. Theory and Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.1 Adsorption fundamentals . . . . . . . . . . . . . . . . . . . . . . . 16


3.1.1 Effect of porosity . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Thermal properties of adsorption . . . . . . . . . . . . . . . . . . . 17
3.3 Adsorption Isotherms . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3.1 The Langmuir Isotherm Model . . . . . . . . . . . . . . . . 19
3.3.2 Linear Driving Force . . . . . . . . . . . . . . . . . . . . . . 21
3.3.3 Adsorption Isotherm Model . . . . . . . . . . . . . . . . . . 22
3.4 Breakthrough and Bleed emissions . . . . . . . . . . . . . . . . . . 25
3.4.1 Canister Breakthrough . . . . . . . . . . . . . . . . . . . . . 25
3.4.2 Bleed emissions . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5 Model development . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5.1 Mass transport equation . . . . . . . . . . . . . . . . . . . . 27
3.5.2 Thermal Transport equation . . . . . . . . . . . . . . . . . . 30

4. Experiment Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . 34


4.1.1 Canister geometry . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.2 Thermocouple connections . . . . . . . . . . . . . . . . . . . 36
4.1.3 Setup Description . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 Testing Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.1 Canister loading . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.2 Canister Purging . . . . . . . . . . . . . . . . . . . . . . . . 43

5. Simulation Method and Parameters . . . . . . . . . . . . . . . . . . . . . 46

5.1 Method of solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 46


5.2 Initial and boundary conditions . . . . . . . . . . . . . . . . . . . . 47
5.3 Loading and purging simulations . . . . . . . . . . . . . . . . . . . 48
5.4 Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . 49

6. Result and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6.1 Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . 52


6.1.1 Canister Loading Mass and Temperature Variation . . . . . 52
6.1.2 Canister Purging Mass and Temperature Variation . . . . . 56
6.1.3 Measurement errors and inaccuracies . . . . . . . . . . . . . 60
6.2 Model Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.3 Simulation and Experimental Comparison . . . . . . . . . . . . . . 66
6.3.1 Canister Loading . . . . . . . . . . . . . . . . . . . . . . . . 66

vii
6.3.2 Canister Purging . . . . . . . . . . . . . . . . . . . . . . . . 70

7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Appendices 76

A. Canister Dimensions for Model . . . . . . . . . . . . . . . . . . . . . . . 76

B. Thermocouple Positioning on Canister . . . . . . . . . . . . . . . . . . . 78

C. MATLAB code to calibrate model . . . . . . . . . . . . . . . . . . . . . . 79

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

viii
List of Tables

Table Page

3.1 Adsorption Isotherm curve-fit parameters . . . . . . . . . . . . . . . . 25

3.2 Antoine equation coefficients for n-Butane . . . . . . . . . . . . . . . 28

4.1 n-Butane and Nitrogen Mixture Gas cylinder Composition . . . . . . 41

5.1 Physical Properties of carbon bed and other Model Parameters . . . . 49

5.2 Canister Geometric Properties . . . . . . . . . . . . . . . . . . . . . . 50

B.1 Thermocouple Position along Carbon bed for both Canisters . . . . . 78

ix
List of Figures

Figure Page

2.1 A simplified schematic representation of the Fuel EVAP system . . . 8

2.2 Schematic diagram of a commonly used activated carbon canister . . 8

2.3 Approximate flowchart of vapor flow during loading process . . . . . . 10

2.4 Approximate flowchart of vapor flow during purging process . . . . . 11

3.1 Adsorption Isotherm for n-Butane on BAX 950 Activated Carbon using
f and v . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.1 Schematic Diagram of Canister 1 . . . . . . . . . . . . . . . . . . . . 35

4.2 Schematic Diagram of Canister 2 . . . . . . . . . . . . . . . . . . . . 35

4.3 Thermocouple positions on canister 1 . . . . . . . . . . . . . . . . . . 37

4.4 Thermocouple positions on canister 2 . . . . . . . . . . . . . . . . . . 37

4.5 Rotameter used to Measure Vapor volumetric flow rate . . . . . . . . 38

4.6 Schematic Diagram of the Canister loading experiment setup . . . . . 39

4.7 Schematic Diagram of the Canister Purging experiment setup . . . . 40

4.8 Experimental setup for loading experiments . . . . . . . . . . . . . . 43

4.9 Experimental setup for Purging experiments . . . . . . . . . . . . . . 44

6.1 Canister #1 Mass Profile during Loading Experiment . . . . . . . . . 53

x
6.2 Canister #1 Temperature Profile during Loading Experiment . . . . . 54

6.3 Canister #2 Mass Profile during Loading Experiment . . . . . . . . . 55

6.4 Canister #2 Mass Profile during Loading Experiment . . . . . . . . . 56

6.5 Canister #1 Mass Profile during Purging Experiment . . . . . . . . . 57

6.6 Canister #1 Temperature Profile during Purging Experiment . . . . . 58

6.7 Canister #2 Mass Profile during Purging Experiment . . . . . . . . . 59

6.8 Canister #2 Temperature Profile during Purging Experiment . . . . . 60

6.9 Canister mass profile with small disturbance during experimental testing 61

6.10 Cooldown Experiment Results - Carbon Bed and Skin Temperatures 62

6.11 Cooldown Experiment - Calibrating dTp /dTC . . . . . . . . . . . . . . 63

6.12 Cooldown Experiment - Calibrating Uloss . . . . . . . . . . . . . . . . 64

6.13 Cooldown Experiment - Validating Calibrated values on Canister #1 65

6.14 Cooldown Experiment - Validating Calibrated values on Canister #1 65

6.15 Comparison between Simulation and experimental results - Mass pro-


file Canister #1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6.16 Comparison between Simulation and experimental results - Mass pro-


file Canister #2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6.17 Comparison between Simulation and experimental results - Tempera-


ture profile Canister #1 . . . . . . . . . . . . . . . . . . . . . . . . . 68

6.18 Comparison between Simulation and experimental results - Tempera-


ture profile Canister #2 . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.19 Comparison between Simulation and experimental results - Mass pro-


file Canister #1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

xi
6.20 Comparison between Simulation and experimental results - Mass pro-
file Canister #2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.21 Comparison between Simulation and experimental results - Tempera-


ture profile Canister #1 . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.22 Comparison between Simulation and experimental results - Tempera-


ture profile Canister #1 . . . . . . . . . . . . . . . . . . . . . . . . . 73

A.1 Canister #2 geometry and dimensions approximation for model . . . 77

A.2 Canister #1 geometry and dimensions approximation for model . . . 77

xii
Chapter 1: Introduction

Automobiles have been expressed as one of the sources of air pollution since the

1960s. Emissions can cause harmful effects on the environment and public health.

Automotive emissions control was introduced to decrease the harmful effect of these

emissions on the environment. Automotive emissions control includes the study of re-

ducing vehicle emissions particularly vehicles with internal combustion engines. Most

of the automotive emissions control focus on tailpipe emissions such as CO and N Ox

emissions. This work is going to focus on the evaporative emissions which are caused

due to evaporation of the fuel present in the fuel tank [20].

1.1 Problem introduction

A major contributor to air pollution are hydrocarbons emissions (HC). Long ex-

posure to hydrocarbons can lead to asthma, lung disease and cancer. A subsection

of hydrocarbons that directly affects public health is the Volatile organic compounds

(VOCs). Volatile organic compounds are organic compounds whose boiling point

is below 250 ◦ C. The fuel present in internal combustion engines especially gasoline

consists of different types of hydrocarbons. The lighter hydrocarbons present in the

1
gasoline constitute the volatile organic compounds since they are likely to evapo-

rate into gaseous form. The compounds within this group contribute to the smog

formation and environmental pollution [4].

The volatile organic compounds can be exposed to high pressures while inside the

fuel tank. This increase in pressure can be caused due to two reasons: Refueling and

temperature increase. In a closed fuel tank system, while refueling the volume of liquid

gasoline present in the tank will increase compressing the gaseous gasoline present in

the tank. Another way the pressure of the gas present in the tank can increase is due

to the increase in ambient temperature as the pressure of a gas is directly proportional

to its temperature provided the volume remains a constant according to the ideal gas

law [2].

Ideally, a fuel tank should not be pressurized due to safety concerns and to prevent

back pressure inside the tank. To avoid this pressure build-up, the fuel tank is vented

to the atmosphere to get rid of the gaseous vapors and maintain atmospheric pressure

inside the fuel tank. As mentioned earlier these volatile organic compounds are major

contributors to the air pollution and the smog formation. To prevent releasing the

volatile organic compounds to the atmosphere automotive makers have introduced

an active carbon canister in between the fuel tank and the atmosphere. [16]

The activated carbon canister is used to adsorb the volatile organic compounds

on the surface of the activated carbon and expel only clean air into the atmosphere.

However due to its limited capacity, once the canister reaches saturation it will stop

adsorbing VOCs and will be released into the atmosphere. To be able to use the

canister again, the adsorbed gasoline vapors have to be purged back into the intake

manifold. The canister has to be purged regularly to maintain optimum emission

2
control. Purging involves passing air through the canister which desorbs the VOCs

and forcing them into the engine manifold. This process allows for better fuel economy

and reduces evaporative emissions.

The carbon canisters have existed for a long time but due to the new stringent

emission norms being introduced every year automotive companies are forced to im-

plement better carbon canisters with larger capacity and better-activated carbon

types. The need for efficient carbon canisters has led to the demand for canister

models which can accurately predict the amount of volatile organic compounds ad-

sorbed so that it can be used to improve the fuel economy and maintain the emission

standards.

1.2 Purpose and Objectives

The purpose of this thesis is to create a carbon canister model capable of pre-

dicting the amount of volatile organic compounds adsorbed and desorbed in a given

activated carbon canister. This model can be used by the automotive industry to run

simulations of the EVAP system and create new and improved carbon canisters to

comply with the advanced emission norms.

The goal of this project is to create a carbon canister model that can accurately

predict the amount of VOCs adsorbed in an ideal flow situation (diurnal setting) and

the amount of VOCs desorbed while purging the canister at a constant flow rate.

The model will be validated using experimental tests conducted on a canister filled

with activated carbon. Once the model is validated it can be used by an automotive

company to analyze the carbon canister inside their vehicles. The model can also

3
be used as a base to create new canister designs and help reduce the evaporative

emissions further.

1.3 Method

The project consists of two main parts to predict the behavior of an activated

carbon canister: modeling and experimental analysis. The modeling part includes

predicting the adsorption and desorption behavior of activated carbon using the car-

bon and canister properties in chemical equations. The model will predict the amount

of VOCs adsorbed or desorbed and the temperature changes in the canister during

this process. The experimental part consists of building a test rig of an activated

carbon canister with a weighing scale and thermocouples to studying the mass and

temperature changes of the canister as it is being loaded with VOCs.

The model is implemented in MATLAB to simulate the mass and temperature

change of the canister during the adsorption and desorption phases. The MATLAB

model will simultaneously solve the mass and temperature partial differential equa-

tions to model the carbon adsorption properties.

1.4 Thesis structure

This thesis will first give a background on the EVAP system and carbon canister.

The thesis then explains the theory behind the canister model and the equations

used to model the carbon canister. In chapter 4 the thesis describes the experimental

setup and analysis done to compare the model. In chapter 6 the thesis will provide

the results and discussions based on the findings of the experimental setup and the

4
accuracy of the model. Finally, the thesis will conclude with the use of the model

and any future work remaining.

5
Chapter 2: Background

This chapter will provide a background on the problem considered in this thesis.

First, it will explain the effect of fuel vapor on the environment and then provide

a brief overview of the EVAP system and carbon canister. Finally, it will discuss

various methods used by other authors to model the activated carbon canister.

2.1 Effect of fuel vapor on environment

Volatile organic compounds (VOCs) are a group of gaseous compounds containing

carbon that exist in the atmosphere, not including carbon monoxide, carbon dioxide

or carbonic acid. VOCs as pollutants have a major impact on the environment and

human health. VOCs at the ground level react with oxides of nitrogen in the presence

of sunlight in a photochemical reaction to create ozone in the ground levels. Ozone at

the ground level is the main ingredient in smog formation. Ozone can also be toxic to

humans and cause various health issues. VOCs are also responsible for its destruction

of the ozone layer at the stratosphere level and the role it plays in the greenhouse

effect [6]. Given the harmful effects of VOCs in the atmosphere, its release should be

minimized as much as possible.

As mentioned in the previous section, evaporative emissions from fuel present in

automobiles is a major source of VOCs. Increasing ambient temperature or heat from

6
the engine can cause the volatile compounds present in the fuel to evaporate inside

the fuel tank. In the absence of a carbon canister, the vapor build-up inside the fuel

tank is released directly into the atmosphere. To avoid releasing these harmful vapors

in the atmosphere, a carbon canister is introduced into the fuel system in between the

fuel tank and the vent to the atmosphere. Even with the carbon canister installed,

the built-up vapor in the fuel tank can leak into the atmosphere when the canister is

saturated. Therefore an accurate model of the carbon canister is required so that the

canister can be purged before it reaches saturation.

2.2 Brief overview of EVAP-system and carbon canister

The evaporative emission control system or EVAP for short is used to control

the amount of evaporative gasoline vapor escaping into the atmosphere from the fuel

system. The OBD II EVAP monitor on vehicles runs a self-diagnostics check on the

fuel system to detect any fuel vapor leaks. If the system finds a leak, it will turn

on the check engine light indicating a leak in the EVAP system to the owner of the

vehicle.

Figure 2.1 shows a simplified sketch of the main components of the EVAP system

which include the fuel tank, gas cap and EVAP canister. The fuel tank usually

is equipped with some expansion volume at the top to accommodate for the fuel

expansion during a hot day without overflowing or causing the EVAP system to leak.

The gas cap helps to seal the system from the atmosphere and allows the filling of

the fuel tank.

7
Figure 2.1: A simplified schematic representation of the Fuel EVAP system

Figure 2.2: Schematic diagram of a commonly used activated carbon canister

The main component of the EVAP system is the carbon canister. The location

of the activated carbon canister varies from vehicle to vehicle. Figure 2.2 shows

a schematic diagram of a commonly used activated carbon canister. The carbon

8
canister consists of a carbon bed placed inside a plastic housing with some filler and

protection to keep the carbon bed secured inside the housing. The canister has 3

main valves on the outside named the load valve, purge valve and atmosphere valve.

These valves allow the flow of air and gasoline vapors to and from the canister. The

load valve serves as a connection between the fuel tank and the carbon bed inside

the canister. The purge valve connects the carbon bed to the vehicle intake manifold

and finally, the atmosphere valve connects the carbon bed to the atmosphere.

2.2.1 Loading

The event where hydrocarbons are adsorbed on the carbon bed is known as load-

ing. For the loading event, the load valve and atmosphere valve remain open while

the purge valve remains closed. During the event, gasoline vapors flow through the

carbon bed in the canister through the load valve and out the atmosphere valve. The

loading event occurs due to two main reasons: an increase in ambient temperature or

a sudden increase in fuel vapors inside the fuel tank.

The first reason for loading is the increase in ambient temperature around the fuel

tank. This increase in temperature will cause the volatile compounds present in the

fuel to evaporate into gases. As a result of this evaporation, the air inside the tank is

concentrated with hydrocarbons present in the fuel vapors. To prevent damage to the

fuel tank by the increase in pressure and temperature the fuel vapors must be vented

away from the tank. As mentioned in the previous section we cannot vent this fuel

vapor into the atmosphere to avoid pollution. The canister is placed in between the

tank and the atmosphere so that all vented air from the fuel tank must pass through

the canister before it can pass to the atmosphere. As the mixture of air and fuel pass

9
Figure 2.3: Approximate flowchart of vapor flow during loading process

through the canister from the loading port, the fuel vapors adsorb on the carbon bed

of the canister and clean air is released into the atmosphere through the atmospheric

port.

Another reason canister loading may occur is during a refueling event. During

the refueling process, there is a sudden increase in the liquid fuel level inside the

fuel tank. This sudden liquid volume increase causes the fuel vapor present inside

the fuel tank to be pushed into the canister and the carbon bed is loaded with fuel

vapors during this process. During the refueling event, a lot of vapor is loaded into the

canister quickly it is known as a fast loading event and the loading due to temperature

changes is a slow loading event since the gasoline vapors are loaded on the carbon

bed at a much slower rate.

Figure 2.3 shows an approximate flow diagram of the gasoline vapors during the

adsorption or loading process. The gasoline vapors from the fuel tank flow through

the carbon canister and clean air is released into the atmosphere.

10
Some markets, like in the US, require the canister to adsorb all the vapor released

during the refueling event whereas in other markets such as Europe the fuel pump

will take in the majority of the fuel vapors released during the refueling event. This

is also a major factor involved in the canister design stage. US market canisters have

larger canisters with more activated carbon volume as compared to their European

counterparts [14].

2.2.2 Purging

The event where hydrocarbon vapors are desorbed from the carbon bed by passing

air inside the canister is called purging. During the purging event, the atmosphere

and purge valve remain open while the load valve remains closed. This event occurs

when the engine is running. Figure 2.4 shows an approximate flow diagram of the

airflow through the EVAP system during the purging process.

Figure 2.4: Approximate flowchart of vapor flow during purging process

11
Air is drawn into the canister with the help of the low pressure created during the

engine intake stroke. As air rushes into the canister through the atmosphere valve

the hydrocarbons adsorbed on the carbon bed get desorbed and are pushed into the

engine intake manifold through the purge valve and used for engine combustion. The

pressure created by the engine suction varies depends on the driving cycle and this

can affect the mass flow of air through the canister. The mass flow rate of air decides

the amount of fuel that will be desorbed from the carbon canister. The purge valve is

also a one-way check valve which is used to control the flow and avoid any backflow

into the canister. [22]

2.3 Activated carbon

Activated carbon is a form of carbon which has been processed to increase its

surface area to make it more susceptible to adsorption. The carbon bed inside the

canister consists of activated carbon. Activated carbon used in automotive appli-

cations is usually produced in very small sizes of granules or pellets. The carbon

is obtained from biomass and coal. The carbon is mixed with a binder and passes

through several processing techniques such as pelletizing, heat treatment and condi-

tioning. During this process, the carbon becomes porous and will lead to a larger

surface area allowing higher adsorption. This process is known as activating the

carbon particles [11].

2.4 Literature review

Studies have been conducted by various authors to understand canister perfor-

mance during the adsorption and desorption process. Various authors have used

12
experimental and modeling methods to find factors that affect canister loading and

purging events. A model is presented by Lavoie et al. in 1996 who used experimental

results to predict the behavior of the activated carbon canister. The authors col-

lected data from loading a wood-based activated carbon and predicted the 1D mass

and thermal transient behavior of the carbon canister. The model is written in FOR-

TRAN and was created to run on an IBM-compatible PC. The model still predicts

the canister behavior pretty well and is being used as a base for the model for this

project [12].

Fiani et al. created a model to describe experiments using both heat and mass

transfer. This paper does not handle canister purging but provides good insight

and discussion about conductive and convective heat transfer around a small carbon

sample [8].

Bai et al. developed a 2D model to simulate the adsorption and desorption of n-

butane over an activated carbon bed. They described that the Linear Driving Force

model accurately predicted the mass loading of butane on carbon. Since they created

a 2D model they were able to predict the temperature variation of carbon bed along

the sides of the plastic wall and found that the heat transfer to ambient created a

low-temperature region near the canister walls and in turn lead to better adsorption.

This shows that radial effects do play a part in the adsorption of gasoline vapors on

the canister. The only drawback of this study is the fact that it only used results

and parameters from other studies hence a strong validation could not be done by

the authors [1].

Sato et al. created a 3D CFD model which took twice the amount of time as a

real-time experiment to run the model even with lower resolution. The model was

13
later simplified using a Linear Driving Force model which significantly improved the

model prediction speed and still predicted the results quite accurately. This paper

also inspects the effect of time step which will be useful while modeling in MATLAB

[17].

Smith et al. used a 1D model calibrated using enthalpy of formation and Langmuir

model adsorption and desorption reaction rate ratio. This model was then later used

to simulate a purge valve [19].

Williams et al. studied the effect of canister geometry in bleed emissions and

breakthrough. These authors did a canister design study which helped to understand

how the shape of the canister affects the gas flow and various efforts taken to minimize

the bleed emissions from canisters [21].

Bergaoui et al. studied how different alkanes adsorb on the surface of BAX 1100

carbon and created a model to predict the isotherm for the same. They started

with data collected from C1 - C7 n-alkanes adsorbed on the surface of a Westvaco

BAX-1100 activated carbon. Then they used a statistical physics model to predict

the adsorption behavior of these alkanes on the activated carbon. This isotherm

prediction model can be used to predict the isotherm for n-butane over a wide range

of temperatures [3].

Yoo et al. explored the control-oriented modeling of the canister purging sys-

tem and on-line estimation of fuel loading on the canister. This model created an

analytical expression for the evaporative fuel using only the sensors available in the

production vehicle. This paper allows modeling the canister loading to be versatile

and used in any vehicle just by using the fuel trim value. The model was helpful to

14
understand the need for modeling the canister to be used in different types of vehicle

[22].

This section provided various models created by various authors to predict the

adsorption and desorption on activated carbon in one-, two- and three- dimensional

models. However, none of these models have been created in MATLAB and involve a

large simulation time. This provides the motivation for carrying out this project and

create a fully functional carbon canister model.

15
Chapter 3: Theory and Modeling

This chapter will discuss the fundamental theory behind the modeling of the

carbon canister. It will also discuss the adsorption process and the physical behavior

of the canister at the time of loading and purging.

3.1 Adsorption fundamentals

Adsorption is a surface phenomenon where the molecules of a gas or liquid adhere

on the surface of an insoluble solid. The gas molecules being adsorbed on the surface

is called adsorbate and the surface itself is called adsorbent. For this project, the

adsorbate is the gaseous vapor of fuel and the adsorbent is the activated carbon bed

inside the canister. Since the adsorbent and adsorbate are present in different states

(solid and gas respectively) this process is an example of heterogeneous adsorption.

[2]

The loading event in the carbon canister is an example of physical adsorption or

physisorption where the gas vapor molecules are held in the vacant spots in the acti-

vated carbon by Van Der Waal forces. Physisorption is a relatively weak interaction

as the Van Der Waal forces have weak interaction energy. This allows the adsorbed

particles to be removed quite easily in the case of the carbon canister. [2]

16
3.1.1 Effect of porosity

As mentioned in the previous chapter, the carbon bed is activated by various

processes that increase the carbon porosity and surface area. In the case of adsorption,

high porosity and large surface area of adsorbent are favorable for efficient loading of

adsorbate particles on the adsorbent [18].

An activated carbon adsorbent has different sizes of pores after the activation

process and each pore size plays a different role in the adsorption and desorption

process. Pore sizes larger than 500 Å called macro-pores are inefficient for adsorption

but they transport the hydrocarbon molecules into the smaller pores. Mesopores,

pores ranging from 20-500 Å in size, are the most important pores on the activated

carbon because they effectively adsorb the hydrocarbon vapor molecules and release

them quite easily during the purging process. The smallest pores, micropores, of sizes

less than 20 Å are highly effective for adsorption but are not very effective during

desorption. Due to the high efficiency of the mesopores activated carbon used in

automobile applications contain a large number of mesopores [15].

3.2 Thermal properties of adsorption

Physical adsorption, or physisorption, occurs when the intermolecular forces be-

tween the adsorbate molecules are lesser than that between the adsorbent and adsor-

bate. This makes physisorption an exothermic process as heat is released when the

adsorbate adsorbs on the adsorbent similar to condensation. The enthalpy change of

a reaction is given by:

∆H = ∆G + T ∆S (3.1)

17
Since the adsorbate molecules are more ordered in the adsorbed state the entropy

change, ∆S, of the system will be negative. Also since the adsorption process is

spontaneous the change in Gibbs Free Energy, ∆G, is also negative. Using this in

equation (3.1) shows that the enthalpy change is always negative (exothermic). This

shows that the temperature of the carbon bed will increase during adsorption and

decrease during desorption. Higher temperatures cause unfavorable conditions for the

adsorption process meaning the exothermic adsorption process will act against more

adsorption into the carbon bed [7].

As the adsorption process progresses the thermal properties of the carbon bed will

keep changing throughout the duration of the adsorption. This model will assume a

fixed average of these thermal properties to reduce the complexity of the mathematical

model and faster processing.

3.3 Adsorption Isotherms

Adsorption equilibrium describes the ratio of the amount of adsorbate adsorbed

to the amount remaining in the solution. The adsorption equilibrium is dependent

on the temperature and partial pressure of the adsorbate. The equilibrium will not

provide any information on the rate of the adsorption.

For gaseous adsorbates, the adsorption process is usually described using adsorp-

tion isotherms. An isotherm is a curve which indicates the amount of adsorbate

adsorbed on an adsorbent (adsorption equilibrium) as a function of the partial pres-

sure of the gaseous adsorbate at a constant temperature. The amount of adsorbate

adsorbed is normalized per unit mass of adsorbent so that it can be compared to

different adsorbate-adsorbent combinations [9].

18
Loading is expressed in an isotherm as, θ, a fraction of the maximum amount of

adsorbate that can possibly be loaded on the adsorbent.

Amount of adsorbate loaded


θ= (3.2)
Maximum number of vacant sites available

The loading fraction is a unitless variable and can take any value between 0

and 1. θ = 0 indicates that the system is empty and θ = 1 indicates that the

system is saturated and the adsorbent has no more vacant sites available for loading.

Each isotherm is unique for an adsorbate-adsorbent system but most isotherms follow

that lower temperature and higher partial pressure lead to a higher concentration of

adsorbate loading.

3.3.1 The Langmuir Isotherm Model

A large number of models have been developed to mathematically describe adsorp-

tion isotherms. The models help explain the interactions between the components in

the adsorption system. They are used to design and scale up the adsorption process.

Using the kinetic energy of gases and few assumptions Langmuir was able to

predict a type I isotherm (mono-layer adsorption). The assumptions for the Langmuir

model are:

• The adsorption was limited to a single layer of adsorbate

• The surface of the adsorbent was considered flat

• Each vacant site on the adsorbent surface has an equal probability of adsorbing

an adsorbate molecule.

19
• The adsorbate molecules do not interact with each other or any other molecule

on the site

Using these assumptions, the rate of adsorption and desorption for a molecule of

adsorbate X adsorbing on a vacant site on adsorbent S and forming the adsorbed

state SX can be written as follows:

ads r
S+X (−
−−− SX
+ (3.3)
rdes

The forward reaction is the adsorption reaction occurring at a rate of rads given

by

rads = kads pi (1 − θ) (3.4)

where θ is the fraction of adsorbate molecules that have occupied from the max-

imum available vacant sites, pi is the partial pressure of the adsorbate gas and kads

is a constant rate multiplier given by the expression kads = Navo / 2πMw RT . Navo

is the Avogadro’s number, Mw is the adsorbate molecular weight, R is the universal

gas constant and T is the temperature in Kelvin.

The reverse reaction is the desorption occurring at a rate of rdes given by

rdes = kdes e(−EA /(RT )) θ (3.5)

where kdes is constant rate multiplier given by the expression kdes = nm v1 . nm is

the number of adsorbate molecules to complete a monolayer and v1 is the vibration

frequency of the adsorbate particles. The term e(−EA /(RT )) represents the probability

20
of the adsorbate particle to overcome potential energy at the surface of the adsorbent

where EA represents the activation energy [7].

As mentioned in the earlier section, adsorption isotherms are expressed during

adsorption equilibrium where the rate of adsorption is equal to the rate of desorption.

Using this and solving for the amount of adsorbate loaded, θ,

rads = rdes

kads pi (1 − θ) = kdes e(−EA /(RT )) θ (3.6)


Kpi
θ=
1 + Kpi

where the K is a constant and is defined,

kads (EA /(RT ))


K= e (3.7)
kdes

Equations (3.6) and (3.7) show the simplicity and ease of use of the Langmuir

model. The model has only two unknown variables, the activation energy EA and the
kads
ratio of the reaction rate multipliers kdes
. This model shows that as the temperature

increases the loading will decrease. Higher temperature leads to a lower K value and

therefore a lower θ [7].

3.3.2 Linear Driving Force

Isotherms are used to determine the amount of adsorbate present on adsorbent

at equilibrium but they lack the information regarding the speed of the reaction.

Consider the adsorption reaction S + X *


) SX, where a molecule of adsorbate S

adsorbs on the vacant site in adsorbent X to form adsorbed state SX. This reaction is

21
a reversible reaction where the molecule S can be desorbed from the adsorbed state

resulting in a vacant spot and adsorbate molecule.

This reaction can be written in two ways - one with two counter-acting reactions

with their own reaction rates or as one reaction with one reversible reaction rate.

Equations (3.4) and (3.5) are examples of counteracting reactions. The same reaction

can be rewritten with only one reaction rate as

r
S+X ←
→ SX (3.8)

This is achieved by having the reaction rate as the driving force between the

current state and the equilibrium state of the reaction. The most simple reaction

rate is given by the linear driving force (LDF) which is written as:

r = k(θeqm − θ) (3.9)

where r is the reaction rate for (3.8), k is a time constant, θeqm is the loading level

at equilibrium and θ is the current local loading level of the adsorbate. Equation (3.9)

will be positive for an adsorption reaction and negative for a desorption reaction. The

LDF model can be used to display any isotherm with a reaction rate expression [5].

3.3.3 Adsorption Isotherm Model

For a particular family of adsorbates, the adsorption isotherms can be defined

using the relative volume of adsorbate adsorbed and the differential free energy of

adsorption normalized to a unit volume of liquid.

22
The relative volume of adsorbate adsorbed, v (unitless), is defined as the ratio

of the volume of adsorbate adsorbed per unit mass of carbon, V (m3 /kgC), to the

saturation volume of the carbon, V0 (m3 /kgC),

V
v= (3.10)
V0
The differential free energy of adsorption per unit volume, f (J/m3 ), can be written

as:

ρL RT psat
f= ln (3.11)
W pi
where ρL (kg/m3 ) is the liquid density of the adsorbate , R = 8.314 (JK −1 mol−1 )

is the universal gas constant, W (kg/mol) is the molecular mass of adsorbate, T (K)

is the absolute temperature, pi (P a) is the partial pressure of adsorbate and psat (P a)

is the saturation pressure of adsorbate at T [12].

Figure 3.1 shows an isotherm for n-butane (adsorbate) adsorbed on activated

carbon BAX 950 (adsorbent) plotted in terms of parameters v and f. This curve

has been plotted by curve fitting data collected of n-butane adsorbed on BAX950

activated carbon by Westvaco Company. The curve fit was identified by Lavoie [12]

using equations (3.12) and (3.13). For computational convenience, the data is divided

into two segments by breakpoints v ∗ and f ∗ . The breakpoint is the point where the

curve is continuous on both sides v ∗ = 0.9 and f ∗ = 4 J/ml. The curve fit parameters

are described in table 3.1.

Below the breakpoint, v ∗

  ∗ 1/n
A
f (v) = f1 ln v ≤ v∗ (3.12)
v

23
Figure 3.1: Adsorption Isotherm for n-Butane on BAX 950 Activated Carbon using
f and v

Above the breakpoint, v ∗

f0
f (v) = v v > v∗ (3.13)
B∗
−1

This plot only shows one hydrocarbon adsorbate but studies have shown that

hydrocarbons ranging from methane to hexane can be plotted on the same curve.

This allows adsorption curves to be created for hydrocarbons in the same family as

long as one curve is available [10].

24
Curve parameter Value Unit
6
f1 85 ∗ 10 J/m3
f0 0.5 ∗ 106 J/m3
n 1.2 -
v∗ 0.9 -

A 0.9233 -

B 0.8 -
− 3
V0 8.3 ∗ 10 4 m /kgC

Table 3.1: Adsorption Isotherm curve-fit parameters

3.4 Breakthrough and Bleed emissions

The carbon canister has a limited capacity based on the type of activated carbon

present inside the canister. The size of the canister and the type of activated carbon

used must meet the required minimum emission standards.

3.4.1 Canister Breakthrough

The carbon bed inside the canister has a limited number of vacant sites avail-

able for adsorption. While the canister is being loaded with hydrocarbons it will

eventually run out of vacant spots. This leads to canister saturation when all the

vacant spots are filled and further loading of the canister will lead to the hydrocar-

bon molecules escaping into the atmosphere. This phenomenon is known as canister

breakthrough. Lavoie (1996) shows data collected from canister loading to follow

a non-linear behavior with effective adsorption at low vapor pressures. This non-

linearity of the activated carbon leads to complete saturation of the carbon canister

before appreciable breakthrough occurs to the atmosphere [12].

25
An ideal canister would have a unit step function for canister breakthrough where

the concentration of hydrocarbons leaving the canister would be equal to the hydro-

carbons entering the canister. A non-ideal canister would have non-isothermal effects

and non-ideal geometry leading to a slightly earlier breakthrough as compared to an

ideal canister. For emission purposes, a canister is considered saturated when 2.0g of

hydrocarbon is released from the canister atmosphere port [21].

3.4.2 Bleed emissions

The carbon canister is loaded from one end to another. This tends to cause a

concentration gradient in the carbon bed after loading. This concentration gradient

can lead to diffusion of adsorbate particles within the canister even if there is no

convective flow inside the canister. As mentioned in the earlier section, non-ideal

canisters tend to have breakthrough slightly before the carbon bed is saturated. The

diffusion phenomenon can lead to breakthrough if the canister is left for a while, this

event is called bleed emissions.

Diffusion bleed emissions can be reduced efficiently by making the carbon beds

longer, having air pockets in between carbon beds or by using an auxiliary chamber

with another carbon bed since it will lead to lower pressure drops [21].

3.5 Model development

The adsorption and desorption process inside the canister can be simplified as a

flow inside a tube. To accurately model the adsorption and desorption the carbon

canister model is developed by discretizing the carbon bed inside the canister in

one axis and calculating the mass loading along this axis. As mentioned in the

earlier sections adsorption and desorption are exothermic and endothermic processes,

26
therefore, the corresponding temperature change accompanied by the mass loading

at each section also needs to be modeled for better accuracy.

3.5.1 Mass transport equation

To model the carbon canister, we consider a 1-D bed of activated carbon (ad-

sorbent) along a coordinate axis, x, of length L. The mass of the carbon bed inside

the canister is given by mcarbon and the canister has a uniform cross-sectional area of

Ax . The carbon bed is assumed to have an initial arbitrary loading of butane vapors

given by mini . The mass of butane loaded on the carbon bed will be defined by m1 .

A molar flow along the axis of the canister containing a mixture of butane and air is

introduced to change the distribution of the vapor loading inside the canister. The

mass transport equation will determine the mass loading of butane on the carbon bed

with time along the length of the bed with time t, m1 (x, t), given an initial loading

of m1 (x, 0) = mini .

Using equation (3.10) we can calculate the relative volume of adsorption for each

point along the length of the carbon bed. Once relative volume of adsorption is

calculated equations (3.12) or (3.13) can be used to calculate the differential free

energy of adsorption. At equilibrium the partial pressure of butane vapor is a function

of the loading of butane on the carbon bed and the temperature, T. Rearranging

equation (3.11) we can calculate the partial pressure of butane in the carbon bed.

fW
p1 = psat e ρL RT (3.14)

Saturation pressure for butane for a given temperature T can be calculated using

the Antoine equation:

27
B
log10 (psat ) = A − (3.15)
T −C

where A, B, C are empirical coefficients and T is the absolute temperature. An-

toine coefficients for n-butane at relevant temperatures are presented in table 3.2.

The parameters are based on pressure in bar and absolute temperature [13].

Temperature Range (K) A B C


195.11-272.81 3.85002 909.65 -36.146
272.66-425.00 4.35576 1175.581 -2.071

Table 3.2: Antoine equation coefficients for n-Butane

Using the partial pressure of the butane vapor in the canister and the total pressure

the local mole fraction can be calculated and similarly the mole fraction of air in the

system:

p1
y1 =
pt (3.16)
ya = 1 − y1

Convective term

The mixture of butane and airflow can be assumed to be uniform and equal to

the inlet along the length of the canister bed. The flow of butane due to convection

can be written as:

y1
Q1 conv = Qt y1 = Qa (3.17)
ya

28
where Qt is the total convective flow inside the canister and Qa is the convective

flow of air.

Diffusive term

The vapors that have loaded on the carbon bed can also diffuse within the canister

over time. This diffusion primarily occurs in the gas phase and is given by:

∂n1
Q1 dif f = −Ax βD1 (3.18)
∂x
where D1 is the diffusion coefficient and n1 is the molar concentration of butane

vapor in the carbon bed while β is a dimensionless factor to take into account changes

in cross-sectional area and longer flow length due to void space. For the simplified

model we can assume β = 1.

The molar concentration of butane can be written as the product of total molar

concentration and mole fraction of butane, n1 = nt .y1 . Using ideal gas law, nt =

pt /RT .

Simplifying equation (3.18) we get,

pt ∂y1
Q1 dif f = −Ax βD1 (3.19)
RT ∂x

Overall mass transport equation

Considering a small element, ∆x, of carbon bed with a cross-sectional area, Ax ,

the rate of change of adsorbed butane can be written by using continuity of moles:

∂N ∆[Ax n1 ] ∆(Q1 + Q1
conv dif f )
≡ =− (3.20)
∂t ∆t ∆x
Substituting equations (3.17) and (3.19)

29
 
∂n1 Qa ∂(y1 /ya ) βD1 pt ∂ ∂y1
=− + Ax (3.21)
∂t Ax ∂x Ax RT ∂x ∂x
Using ρapp to denote the apparent density of the carbon bed, D1 in terms of
p
diffusion coefficient of air D1 = Da WA /W1 , replacing Qa with ma /Wa and rewriting

the equation in terms of mass loading we get:

r  
∂m1 ṁa W1 ∂(y1 /ya ) ρa βDa W1 ∂ ∂y1
=− + Ax (3.22)
∂t ρapp Ax Wa ∂x ρa pp Ax Wa ∂x ∂x
where m1 is the specific mass loading of butane per unit mass of carbon, ρa is the

density of air, Wa is molecular weight of air, W1 is molecular weight of butane.

3.5.2 Thermal Transport equation

During the loading of canister the local carbon bed temperatures have seen to

go as high as 90◦ C and during purging the temperatures have dropped to as low as

−10◦ C. These high transients bring the need to model the thermal effects of the

carbon bed during these processes. The thermal effects of the carbon bed can be

modeled using the conservation of energy equation [12].

Energy conservation equation

Similar to the mass transport equation we consider a small element of the canister,

∆X along the flow direction. Conservation of energy states that:

∂E ∂QEconv ∂QEcond dQloss


=− − − (3.23)
∂t ∂x ∂x dt
where E is the total energy in the system, QEconv is the energy flux due to con-

vection in the x-direction, QEcond is the conductive energy flux and dQloss /dt is the

heat loss to and from the canister to the ambient air through the plastic wall.

30
Expanding on equation (3.23) we have:

∂E ∂
= [ρapp Ax {ec + mp ep + ma ea + m1 e1 + Es (m1 , T )}] (3.24)
∂t ∂t

where ρapp is the apparent density of the carbon bed (prior to adsorption), Ax is

the cross-sectional area of the canister, ec , ep , ea and e1 are the mass-specific internal

energy of carbon, plastic wall, air and butane respectively. mp , ma , m1v , m1 are the

mass quantities relative to the mass of carbon for plastic, air, butane vapor and

adsorbed butane respectively. The energy of the adsorbed material, Es is a function

of the total volume adsorbed.

Convection

The energy flux due to convection of butane vapor along the carbon bed is given

by:

∂QEconv ∂
= [Qt (h1 W1 y1 + ha Wa ya )] (3.25)
∂x ∂x

where Qt is the total molar flow given in (3.17), h is the mass-specific enthalpies,

W is the molecular weight of and y is the mole fraction.

Conduction

The energy flux due to conduction is given by:

 
∂QEcond ∂ ∂T
=− Ax kapp (3.26)
∂x ∂x ∂x

where kapp is the apparent effective thermal conductivity of the carbon bed. The

thermal transport due to the diffusion of butane vapor is expected to be small as

31
compared to the conductive term and can therefore be ignored in the overall thermal

transport equation [12].

Heat Loss

The energy loss through the plastic container of the canister is given by,

dQloss
= Pext (x)Uext (T − Tamb ) (3.27)
dt

where Pext (x) is the perimeter of the canister at point x, Uext is the overall external

heat transfer coefficient between the ambient air and carbon bed and Tamb is the

absolute ambient temperature.

The overall external heat loss transfer coefficient can be expressed in terms of

heat transfer conductance through the plastic wall to ambient air, Uloss , apparent

conductivity of the activated carbon, kapp and cross-sectional area as:

1
Uext = √ (3.28)
1/Uloss + Ax /3kapp

Energy properties

Assuming constant specific heats, the mass-specific internal energy can be written

as:

ec = Cvc T

ep = Cvp Tp
(3.29)
ea = Cva T

e1 = Cv1 T
The energy of the adsorbed phase can be written in the form [12]:

32
Zm1 
f (v)
Es (m1 , T ) = Cp1 T − ∆hvap1 (T ) − dm1
ρL
0 (3.30)
d f (v) dm1
Es (m1 , T ) = Cp1 T − ∆hvap1 (T ) −
dt ρL dt

Overall Thermal Energy Release Equation

Using equations from (3.24) to (3.30) to substitute into equation (3.23) and noting

that the gaseous energy terms ma ea andm1 e1 are smaller compared to the carbon

energy term we get,

  
dTp ∂∆hvap ∂T
CV C + mp CP P + m1 CP − =
dTC ∂T ∂t
 
f (v) ∂m1
− CP T − ∆hvap −
ρL ∂t

1 ∂ ṁa T Cp W1 y1 + CPa Wa ya kapp ∂T
 (3.31)
− − Ax
Ax ∂x ρapp Wa ya ρapp ∂x
Pext 1
− √ (T − Tamb )
ρapp Ax 1 + Ax
Uloss 3 kapp

The mass and temperature terms have been separated to obtain the temperature

equation.
dTp
The first two terms on the LHS of the equation CV C and mp CP P dT C
represent the
 
thermal inertia of the carbon and the canister wall. The next term, m1 CP − ∂∆h ∂T
vap
,

represents the thermal inertia of the adsorbed mass of butane vapor.


h i
The first term on the RHS, CP T − ∆hvap − fρ(v) L
∂m1
∂t
, represents the energy

added to the system by adsorption of butane on the carbon bed. The next two
h i
1 ∂ ṁa T Cp W1 y1 +CPa Wa ya kapp ∂T
terms, Ax ∂x ρapp W a ya
− Ax ρapp ∂x , represent the energy added to the

system through convection and conduction along the bed. The last term, − Pext
ρapp
(x)
,

represents the heat loss from the system to ambient through the internal boundary.

33
Chapter 4: Experiment Analysis

This chapter describes the experimental setup to collect data required to calibrate

and validate the canister model described in chapter 3.

4.1 Experimental Setup

In order to collect reliable data, an automotive canister from two different vehicles

were used. The canisters were fitted with multiple thermocouples along the length to

measure the temperature changes on the activated carbon bed.

4.1.1 Canister geometry

Two canisters of different sizes and geometry were used to get a wide range of

model validation. The first canister, taken from a mid-size pickup truck, consists

of BAX-1100 activated carbon bed arranged in a straight line. Figure 4.1 shows a

sketch of the canister. The canister contains 1270 g of activated carbon encompassing

a volume of 3.5 L. The load valve and purge valve are on the one end of the canister

and the atmosphere valve is placed on the opposite end of the canister. During the

loading phase, the purge valve will be sealed off with sealant caps provided with the

canister and air will flow from the load valve to the atmosphere valve. The cross-

sectional area of the airflow path can be approximated to around 81.5cm2 .

34
Figure 4.1: Schematic Diagram of Canister 1

The second canister, taken from an SUV, consists of BAX-1100 activated carbon

bed arranged in a U-shape inside a rectangular casing. Figure 4.2 shows a sketch of

the canister. This canister contains 1000 g of carbon encompassing a volume of 2.7L

inside the canister. The cross-sectional area of the airflow path can be approximated

to around 67cm2 .

Figure 4.2: Schematic Diagram of Canister 2

35
The reason for using two different types of canisters is to validate the model as

long as the geometry of the canister is provided.

4.1.2 Thermocouple connections

The aim of the experiment is to observe the changes in carbon bed temperature

as the activated carbon is adsorbing butane vapors and while it is being purged with

clean air. Since activated carbon has a non-linear behavior when it comes to vapor

loading, the carbon bed will have different concentrations of butane vapor during the

course of the experiment. Different rates of loading along the length of the bed lead

to different temperatures along the length of the carbon bed. To account for this

thermocouples have to be placed at different sections of the carbon bed to observe an

accurate temperature variation at different points along the length of the bed.

To place a thermocouple first a hole the size of the thermocouple was drilled

on the plastic casing of the canister. After drilling, the canister was prepared by

inserting K-type thermocouples into the carbon bed to ensure that the thermocouple

will be able to accurately measure the bed temperature. Along with the K-type

thermocouples, T-type thermocouples were also used on the body of the canister to

measure the temperature variation on the walls of the canister during loading and

purging experiments. Another T-type thermocouple was used to measure the ambient

air during the experimental process.

The first canister was prepared with 5 K-type thermocouples placed along the

length of the carbon bed and 1 T-type thermocouple on the canister wall. The second

canister was prepared with 4 K-type thermocouples with 2 placed on one edge of the

U-shape and the 2 more placed on the other edge of the U-shape. The second canister

36
Figure 4.3: Thermocouple positions on canister 1

was also fitted with 1 T-type thermocouple to measure the temperature changes to

the wall of the canister. Figures 4.3 and 4.4 show the sketches of the thermocouple

position on each canister.

Figure 4.4: Thermocouple positions on canister 2

37
4.1.3 Setup Description

During the experimental testing, the canister under testing will be placed on top

of an Adam Equipment GFK 165ah weighing scale with a capacity of 75 kg and

readability of 1g. This weighing scale is used to measure the mass change of the

canister during the loading and purging process. The thermocouples attached on the

canister are connected to an ETAS ES650.1 A/D and Thermo module to observe the

temperature changes in the carbon bed and plastic wall. Data is acquired from the

ETAS module with the help of a software named INCA which is published by ETAS

for data measurement and calibration. The data collected from INCA is converted

to an m-file readable in MATLAB and the data is plotted.

To allow for the flow of gas the canister is connected to other components in the

experimental setup with the help of rubber tubes and brass connectors.

Figure 4.5: Rotameter used to Measure Vapor volumetric flow rate

38
During the loading phase, the butane gas is provided from a cylinder that contains

a mixture of n-butane and nitrogen mixed in a 50/50 volume mix. The butane

nitrogen mixture is usually used in testing to avoid errors from using different gasoline

types with different composition of volatile compounds. The cylinder is controlled

using a regulator which measures the cylinder pressure and allows the user to control

the exit pressure of the mixture from the cylinder. This regulator is connected to a

Praxair PRS3002-FL3-CGA rotameter which allows the user to control the volumetric

flow rate of the gas mixture coming out the cylinder. Figure 4.5 shows a picture of

the flowmeter used to measure the flow rate of gas into the carbon canister.

Figure 4.6: Schematic Diagram of the Canister loading experiment setup

In real-life conditions, the flow rate of gasoline vapors into the carbon canister

is very slow. The flow-meter was selected such that it can measure a flow range of

39
94.8 - 948 sccm of air. The flow-meter is connected to the inlet port (load valve) of

the canister. The exit port (atmosphere valve) is connected to another spare canister

on a weighing scale to measure 2g breakthrough from the test canister. The setup

is kept near a shop ventilation pipe to avoid any butane leak into the ambient air.

Figure 4.6 shows a schematic diagram of the canister loading setup for experimental

analysis.

Figure 4.7: Schematic Diagram of the Canister Purging experiment setup

During the purging phase, air to purge the canister is provided by an air com-

pressor. The air compressor is connected to an Omega Engineering FMA1845A mass

flow meter which displays the flow rate of air in LPM using an LCD display. The flow

is manually controlled by increasing or decreasing the pressure of air exiting the air

compressor. The flow-meter is connected to the atmosphere valve port of the canister.

40
The reverse airflow through the canister desorbs the adsorbed n-butane and purges

it out the canister purge valve. Figure 4.7 shows a schematic diagram of the canister

purging setup.

4.2 Testing Procedures

A number of tests with varying conditions have been conducted to obtain data

required to validate the carbon canister model.

4.2.1 Canister loading

In order to maintain similar initial conditions for all experiments, before each

loading experiment the canister is purged with air at a constant rate of 50 L/min till

the canister displayed a weight reduction of 0.5g in 10 minutes which is the equivalent

to 7x10−4 exit hydrocarbon concentration.

n-Butane vapor is loaded into the canister using a cylinder and flow-meter. The

cylinder contains a mixture of butane and nitrogen in 50/50 ratio by volume. The

composition of the cylinder is shown in table 4.1. To simulate real-time loading

conditions as in a vehicle, n-butane is loaded into the canister at a rate of 50 g/hr.

Gas Pressure Volume Mass Moles


(Psig) (m3 ) (g) (mol)
Butane 4.92 0.059 145.79 2.5085
Nitrogen 25.00 0.059 70.26 2.5085

Totals @70 F 0.119 216.05 5.0169

Table 4.1: n-Butane and Nitrogen Mixture Gas cylinder Composition

41
Since the flow-meter only measures the volumetric flow of air we need to apply a

correction factor and calculate the required flow of the n-butane and nitrogen mixture

to obtain 40 g/hr of n-butane. To calculate the correction factor, we need to calculate

the specific gravity of the mixture:

molecular weight of gas mixture


SG =
molecular weight of air
(% of N2 ∗ mol. wt of N2 + % of C4 H10 ∗ mol. wt. of C4 H10 )
SG =
mol. wt. of air (4.1)
0.5 ∗ 27 + 0.5 ∗ 58.12
SG =
28.97
SG = 1.486

The correction factor of a rotameter flowmeter is given by:

r
1
CF =
SG (4.2)
CF = 0.8202
Using the correction factor the new flowmeter range is calculated for a mixture of

n-butane and nitrogen in a 50/50 ratio by volume. Range of flow meter for n-butane

and nitrogen mixture = Range of flowmeter for air * correction factor = 94.8*0.8202

- 948 *0.8202 = 77.755 - 777.550 cc/min of n-butane and nitrogen mixture at 50/50

by volume.

Table 4.1 is used to calculate the required flow of the mixture required to obtain

40 g/hr of n-butane in the following steps.

• Weight fraction of n-butane in the mixture = mass of n-butane in mixture


Total mass of mixture

145.79 g
= 216.05 g
= 0.6748

• 1 g of mixture contains 0.6748 g of n-butane

42
• To get 50 g/hr of n-butane we will need 50
0.6748
= 74.10 g/hr of mixture

• 0.119 m3 volume contains 216.05 g of n-butane and nitrogen mixture.

0.119m3
• Volume containing 74.1 g/hr of mixture = 216.05g
∗ 74.1 g/hr

= 0.04081 m3 /hr = 680.2 cc/min

Therefore the required flow rate of the mixture to obtain 50 g/hr of n-butane is

544.19 cc/min. Figure 4.8 shows a picture of the setup used to load the canister.

Figure 4.8: Experimental setup for loading experiments

4.2.2 Canister Purging

Once the carbon canister is completely loaded with 2g breakthrough the canister

is said to be saturated. This canister is now ready to be purged. Once the canister

is saturated it is allowed to cool down to room temperature. To purge the canister,

43
Figure 4.9: Experimental setup for Purging experiments

the air is passed through the canister from the atmosphere vent and adsorbed butane

vapors are desorbed from the pores present in the activated carbon bed. Figure 4.9

shows the picture of the experimental setup for purging experiments.

Before the experiment begins all connections are sealed using vacuum tape to

prevent any gas leaks and the flowmeter is switched on. Since the flowmeter measures

flow by introducing a heat flux into the gas flow and measuring the amount of heat

carried by the gas, the flowmeter takes around 15 minutes to warm up after initial

powering up. Once a flowmeter is warmed up the experiment is ready to collect data.

The data acquisition system is powered on and INCA starts recording data collected

by the ETAS equipment.

44
To start the airflow, the air compressor is switched on and the pressure is varied

till the Omega Engineering flowmeter reading reads 50 LPM. Once the air is flowing

through the canister the ETAS equipment will measure the changes in the carbon bed

temperature and the weighing scale will measure the canister mass changes during

the desorption process. To ensure that inflammable butane vapor is not released into

the lab, the pipe from the purge valve is placed near the lab exhaust pipe.

45
Chapter 5: Simulation Method and Parameters

5.1 Method of solution

The partial differential equations of mass and temperature can be solved by dis-

cretization of the PDE into sections. For an initial condition ordinary differential

equation, the value of the first differential is given. Using this initial condition the

subsequent states of the equation are determined using a finite difference scheme.

Similarly, for a boundary value problem with a second-order ordinary differential

equation the boundary conditions reduce the degree of freedom of the system by the

number of boundary conditions available. The idea for a partial differential equation

is similar to the ordinary differential equation.

The equations (3.22) and (3.31) show that the equations are a 1st order PDE in

time and a 2nd order PDE in space. The system is discretized on a two-dimensional

grid for space, length of the carbon bed x, and time, t, using the notation mi,j ≡

m(i∆x, j∆t). Here ∆x refers to the size of nodes or discretization length of the carbon

bed and ∆t refers to the time step of the simulation. The domain for the PDE along

x is bounded by the boundary conditions and the domain for t is semi-infinite with

an initial condition and run till the simulation is stopped.

46
To perform the integration in t, the PDE is discretized by approximating ∂m/∂t

with the forward difference method and ∂m/∂x with the central difference method

as shown in equations (5.1) and (5.2).

 
∂m mi,j+1 − mi,j
≈ (5.1)
∂t i,j ∆t

 
∂m mi+1,j − mi−1,j
≈ (5.2)
∂x i,j 2∆x

The second partial derivatives in the PDE are approximated by the 2nd order

central difference method as shown in equation (5.3).

∂ 2m
 
mi−1,j − 2mi,j + mi+1,j
≈ (5.3)
∂x2 i,j (∆x)2

5.2 Initial and boundary conditions

The state variables in the equations are the mass of adsorbate adsorbed per mass

of carbon and the temperature of the carbon bed. The partial differential equations

(3.22) and (3.31) are single derivatives in time and their solution in the time domain

require an initial condition given by:

m1 (x, 0) = m1 init (x) (5.4)

and

T (x, 0) = Tinit (x) (5.5)

47
The carbon bed is assumed to have an initial adsorbed mass loading of 0.35 g/gC

and the initial temperature of the system is taken as average room temperature of

21 ◦ C. Using the initial mass condition, m1 init , the mole fraction of the butane

vapor, y1 init , at each point on the carbon bed is calculated to get the initial loading

condition for equation (3.22).

The partial differential equations are second derivatives in space and therefore

need two boundary conditions for the solution. The boundary conditions at the inlet

(at x=0) for positive flow are given by:

m1 (0, t) = m10 (t)


(5.6)
T (0, t) = T0 (t)
At the outlet of the canister (x=L) we can neglect all heat and mass transport

except convection thus giving a constant gradient:

∂m1
(L, t) = 0
∂x (5.7)
∂T
(L, t) = 0
∂x
For purging condition, the flow is reversed and the boundary conditions are also

reversed. At x=L the boundary conditions are set by the flow rate of gas and at x=0

the gradients are set to 0. The purging experiment is conducted with a completed

saturated carbon canister, therefore, the initial conditions of the canister will include

saturated canister.

5.3 Loading and purging simulations

The simulations are set up to replicate the environmental and physical conditions

of the experimental setup to accurately predict the behavior of the carbon canister.

48
Solving the partial differential equations occurs in two steps. In the first step, using

the initial and boundary conditions the equation for the mass transport (equation

3.22) is used to advance the mass loading on butane along the carbon bed in time.

In the second step, the updated mass loading of butane, gas concentrations along

the bed and energy flux terms are calculated using the first step and the temperature

along the bed is updated using equation (3.31). Due to the coupling between the mass

and temperature equations some instabilities may arise during this two-step solution

process.

The physical properties of the carbon bed, plastic canister and other model pa-

rameters are mentioned in table 5.1 and 5.2.

Property Symbol Value Unit


Apparent Density of Activated Carbon ρapp 345 kg/m3
Ratio of plastic mass to carbon mass mp 0.5 -
Specific Heat of Carbon CvC 700 J/(kg K)
Specific Heat of Butane Cp1 1600 J/(kg K)
Specific Heat of Plastic Cpp 1900 J/(kg K)
Apparent Thermal Conductivity of Carbon bed kapp 0.13 W/(m K)
Self Diffusion Coefficient of air Da 1.5 ∗ 103 m2 /s
Axial Diffusivity Factor in Carbon Bed β 1 -

Table 5.1: Physical Properties of carbon bed and other Model Parameters

5.4 Parameter Estimation

In equations (3.22) and (3.31) all parameters are properties of the carbon bed,

plastic canister or the gas flowing inside the canister except for the ratio of change of

49
Property Unit Canister 1 Canister 2
Length cm 53 46
2
Cross-Sectional Area cm 81 70
Canister Volume L 3.5 2.7
Carbon Mass g 1270 1000
Activated Carbon Type - BAX 1100 BAX 1100
Discretization Length mm 5 5

Table 5.2: Canister Geometric Properties

plastic wall temperature with respect to carbon bed (dTp /dTC ) and the heat loss con-

ductance through the wall (Uloss ). These parameters are calibrated from experimental

data before they can be used in the model.

A canister cool down experiment is conducted to estimate these parameters. This

experiment is done by loading the carbon canister till the part of the canister under

consideration reaches its maximum temperature and then the butane flow into the

canister is shut off. Since there is no flow and no new butane being added to the

canister the temperature of the canister will begin to cool down to room temperature.

The stagnant air will prevent heat loss through the ends of the canister and heat loss

will only occur through the plastic walls of the canister [12].

To calibrate the two parameters canister 1 is loaded with butane till thermocouple

1 (figure 4.3) reaches its maximum temperature and starts to cool down. This indi-

cated that this area of the carbon bed is saturated. Once saturated the temperature

of the carbon bed and the skin thermocouple is collected. Since there is only heat loss

across the walls of the canister and very little heat loss across the ends, the equation

(3.31) can be modified to get:

50
" Pext √1
#
∂T ρapp A (1/Uloss + A/(3 kapp ))
= (T − Tamb ) (5.8)
∂t CvC + mp Cpp dTdT
p
+ m 1 (C p1 − ∂∆hvap
∂T
)
The value of dTp /dTC can be determined by studying the decrease in temperature

of the plastic wall with respect to the decrease in temperature of the carbon bed.

Using this value, equation (5.8) is modeled in MATLAB and the values for Uloss is

varied until the equation fit matches the data for the cooldown experiment.

Once the values are calibrated the cool down test is done once more to validate

the calibrated values. The values are also validated by modeling the cool down test

for canister 2 and predicting the heat loss in the canister bed after saturation.

51
Chapter 6: Result and Discussions

6.1 Experimental results

This section will present the experimental results of canister testing. First, the

experimental results for loading the two canisters are presented and then the results

for purging the canister is presented. The end of the section will discuss a few mea-

surement errors and inaccuracies present in the data collection for this project.

6.1.1 Canister Loading Mass and Temperature Variation

As described in section 4.2.1 to study the effect of vapor loading on canister mass

change and carbon bed temperature, a constant flow of n-butane is provided to the

canister through the load valve. Figures 4.3 and 4.4 show the positions of the ther-

mocouples in two canister under testing. These thermocouples are positioned in the

radial center of the carbon bed to obtain a good measure of the bed temperature.

A problem that can arise with this method is that the thermocouples can only pro-

vide temperature information at discrete points on the carbon bed, thus limiting the

information available.

Figure 6.1 and 6.2 mass and temperature profiles as the canister is loaded with

n-butane at a constant flow. The experiment lasted 6 hours and 20 minutes before

52
Figure 6.1: Canister #1 Mass Profile during Loading Experiment

the 2g breakthrough was observed in the spare canister. The mass of the canister

increased by 321g showing that the canister adsorbed 321 grams of n-butane. Since

the n-butane was flowing at a constant flow the mass of the canister also increased

linearly at the same constant rate.

Figure 6.2 shows the temperature profile as the adsorption moves across the can-

ister bed. As explained in 3 adsorption is an exothermic process and the temperature

of the carbon bed increases as the carbon adsorbs n-butane. Once a portion of the

carbon bed is saturated no more butane can be adsorbed and the temperature of

the carbon bed decreases to room temperature. At the end of 6 hours and 20 min-

utes, the concentration wave of butane has moved completely across the bed. Before

thermocouple 4, the area of canister #1 changes which account for the difference in

53
Figure 6.2: Canister #1 Temperature Profile during Loading Experiment

temperature rise of the canister at this position. The cross-sectional area of the can-

ister decreases before thermocouple 4 and leads to an uneven temperature increase.

The temperature at the center of the carbon bed, where the thermocouples are fitted,

is higher than the temperature at the ends of the canister as seen by the difference

in temperature increase between the normal thermocouple and skin thermocouples.

Figure 6.3 shows the mass increase as canister #2 is loaded with n-butane at a

constant flow rate. Canister #2 has a smaller cross-section and shorter carbon bed

length as compared to canister #1. This leads to canister #2 reaching 2g break-

through earlier than canister #1 as seen in the figure. The experiment lasted 4 hours

and 50 minutes and the mass of the canister increased by 253g. Similar to the pre-

vious experiment, the mass of the canister increased in the same linear rate as the

butane gas was loaded into the canister.

54
Figure 6.3: Canister #2 Mass Profile during Loading Experiment

Figure 6.4 shows the temperature profile of canister #2 as it is loaded with butane

gas at a constant rate. The concentration wave moves along the length of the carbon

bed and the temperature in each portion of the bed increases until it is saturated.

Once a section of the bed is saturated no more butane vapor can be adsorbed on the

carbon bed and the concentration wave of butane will move ahead. As mentioned

earlier, the cross-sectional area of canister #2 is smaller than canister #1 and this

causes a larger increase in temperature at each section in the canister as compared

to canister #1. This is evidenced in equation (3.31) where a larger area leads to a

smaller increase in carbon bed temperature due to better convection and heat loss

to the environment. The heat generated by the adsorption is dissipated easily when

the cross-sectional area of the carbon bed is larger. Due to this factor, the carbon

bed temperature of canister #2 increases higher than canister #1. This factor is

55
Figure 6.4: Canister #2 Mass Profile during Loading Experiment

also the reason why the temperature measured by thermocouple 4 rises to a higher

temperature as compared to the temperature at the previous thermocouples.

6.1.2 Canister Purging Mass and Temperature Variation

To study the effect of purging on a saturated canister, the mass change and tem-

perature change along the carbon bed of the canister is recorded as the canister is

purged with 50 LPM of air through the atmosphere vent. Once a fully saturated

canister has cooled down to close to room temperature the canister is purged with

air from an air compressor. The flow is controlled by changing the output pressure

out of the compressor and the flowmeter provides a reading of the airflow rate.

Figure 6.5 shows the mass loss in canister #1 as it is being purged with 50 LPM of

air through the vent port. The first few minutes are very critical to canister purging

56
Figure 6.5: Canister #1 Mass Profile during Purging Experiment

as the canister releases a lot of butane during this time period. The figure shows a

steep decrease in the canister mass which then reduces to a gradual decrease in mass.

The mass flow rate of air is much higher than the rate at which the canister is being

loaded which leads to a much faster decrease in canister mass as compared to the

increase in mass during loading experiment.

Figure 6.6 shows the temperature profile of canister #1 as it is being purged with

50 LPM of air. Since purging through the atmosphere port, the thermocouple closest

to the inlet air is thermocouple 5. As mentioned earlier, since purging is a much faster

process than loading, desorption occurs in all parts of the canister simultaneously.

This leads to the sudden decrease in temperature at all parts of the canister as

evidenced in the figure. Once the canister has been completely purged of butane

vapors, the temperature of the carbon bed returns to ambient temperature. The

57
Figure 6.6: Canister #1 Temperature Profile during Purging Experiment

end of the carbon bed closest to the atmosphere vent reaches ambient temperature

first due to its proximity to the incoming air and faster heat loss due to convection.

Positions on carbon bed further down will reach ambient temperature slower since

the butane vapor from the atmosphere vent end flow along the carbon bed to the

purge vent and may get adsorbed along the way. The plastic wall of the canister

also experiences a decrease in temperature similar to the carbon bed as seen by the

reading from the skin thermocouple.

Figure 6.7 shows the mass profile of canister #2 during the purging experiment.

Similar to canister #1, most of the butane is removed from canister 2 in the first 10

minutes of the experiment. The experiment lasted 75 minutes after which the mass

of the canister decreased by less than 0.5g every 10 minutes.

58
Figure 6.7: Canister #2 Mass Profile during Purging Experiment

Figure 6.8 shows the temperature profile of canister #2 as it is being purged with

50 LPM of air. The n-butane starts to desorb from the carbon bed quickly once the

inlet air is introduced into the canister and this causes a decrease in the carbon bed

temperature. In canister 2, thermocouple 4 was placed too close to the atmosphere

port which did not allow the thermocouple to get an accurate reading of the carbon

bed temperature and instead was only able to read the incoming air temperature.

The remaining three thermocouples displayed an accurate reading of the carbon bed

as it was being purged. Thermocouples 2 and 3 started at a higher temperature than

thermocouple 1. Similar to canister 1 the last thermocouple along the carbon bed

(thermocouple 1) took the longest to return to ambient temperature as butane vapor

was desorbed from this section of the canister last.

59
Figure 6.8: Canister #2 Temperature Profile during Purging Experiment

6.1.3 Measurement errors and inaccuracies

The experiments in this project were conducted in non-ideal conditions. The flow

of butane into the canister is controlled by using the manually operated rotameter

which is mainly used to indicate the flow rate of gas. This can cause a small dis-

crepancy in the flow meter readings which adjusting the flow. The flowmeter used to

measure the flow of air for purging experiments has an accuracy of ±10 LPM. This

can cause a variation in the airflow while purging the canister. To verify this flow an

anemometer was used to make sure that the airflow was constant.

The weighing scales used to measure the weight of the canister as it is being loaded

or purged are very sensitive. Any small changes to the experimental setup caused a

sudden increase or decrease to the weighing scale and rendering the data collected

60
for that experiment unusable. The pipes that were used to flow gas in and out of the

canisters must be kept stationary and fixed in order to avoid them influencing the

weighing scale. Figure 6.9 shows an experiment where the test setup was disturbed

during purging canister 2. Any changes to the setup can cause a sudden change

in the weighing scale and it may not return to the previous reading corrupting the

experimental data.

Figure 6.9: Canister mass profile with small disturbance during experimental testing

The temperature sensors in the carbon bed are placed in the center of the carbon

bed assuming a perfectly cylindrical carbon shape with equal amounts of activated

carbon on each side of the thermocouple. This assumption may not always be true

as the carbon inside the canister may be shaped differently and the thermocouples

may not always get the accurate average temperature reading of the carbon bed.

61
One way to avoid this inaccuracy is to include more thermocouples to get a better

understanding of the activated carbon bed temperature.

6.2 Model Calibration

Observer design and parameter estimation are used to identify the parameters of

the model which are not directly measurable. A cooldown test is used to estimate

these parameters and calibrate the carbon canister model as explained in chapter 5.

Figure 6.10: Cooldown Experiment Results - Carbon Bed and Skin Temperatures

Figure 6.10 shows the results of a cooldown test conducted on canister #1. The

canister is loaded with butane vapor and once a section of the canister has been

saturated the butane flow into the canister is cut off and the temperature of the

carbon bed and plastic wall is recorded as they cool to ambient temperature.

62
Figure 6.11: Cooldown Experiment - Calibrating dTp /dTC

To determine the rate of change of plastic wall temperature with respect to the

carbon bed, the temperature of the plastic wall is plotted vs the temperature of the

carbon bed as seen on figure 6.11 and a linear approximation of the curve is calculated.

The slope of this curve will give the rate of change of plastic wall temperature with

respect to the carbon bed. From the graph the value of dTp /dTC is estimated to be

0.2078.

Once the value of dTp /dTC is estimated this value can be used in equation (5.8) to

determine the value of heat loss conductance through the walls of the plastic canister.

Equation (5.8) is modeled in MATLAB and different values of Uloss are considered

and the cooldown of the carbon bed temperature is calculated. The value for Uloss

that predicts the carbon bed temperature most accurately is selected as the parameter

63
Figure 6.12: Cooldown Experiment - Calibrating Uloss

value as seen in figure 6.12. From the figure, the best value for Uloss to predict the

carbon bed temperature is 0.05 W/(m2 K).

Once both the parameters have been estimated these values are used in equa-

tion (5.8) to predict the carbon bed temperature for a different cooldown experiment

on canister 1 and 2 (figures 6.13 and 6.14). These figures show that the parame-

ter estimation has been successful and the model is able to predict the carbon bed

temperature with very good accuracy.

64
Figure 6.13: Cooldown Experiment - Validating Calibrated values on Canister #1

Figure 6.14: Cooldown Experiment - Validating Calibrated values on Canister #1

65
6.3 Simulation and Experimental Comparison

Once the parameters of the model are estimated the model is used to predict the

mass of butane vapor that has been adsorbed and desorbed by the activated carbon

bed during the loading and purging experiments. Once the simulations are run the

prediction is compared against experimental values to determine the accuracy of the

model. Since the model is aimed to be used in an automotive ECU to predict the

amount of gasoline vapor inside the carbon canister, an assumption of a uniform

cross-sectional area across the length of the canister is made on the model to reduce

its complexity. The main priority of the model is to predict the mass loading and

purging of the canister and only secondary importance is given to the temperature

of the carbon bed. For the comparison, the simulation and experimental results are

plotted on the same graph and compared.

The experimental results are presented in the solid lines and the simulation results

are presented in the same graph with dotted lines. Since the experimental results

include temperature plots from different sections on the canister the temperature

measurement and predictions have been color-coded for ease of understanding.

6.3.1 Canister Loading

The loading the canister was modeled using equations (3.22) and (3.31). The

canister loading is simulated for both canisters at 50 g/hr loading of n-butane to

match the loading rate of butane in the experimental results.

66
Figure 6.15: Comparison between Simulation and experimental results - Mass profile
Canister #1

Figure 6.16: Comparison between Simulation and experimental results - Mass profile
Canister #2

67
Figures 6.15 and 6.16 shows the comparison between the model prediction of

the canister mass increase as the canister is being loaded with butane vapor. The

figure shows that the model is able to predict the amount of butane adsorbed on the

activated carbon inside the canister quite accurately. The larger canister #1 is able

to store more butane vapors due to the larger volume and greater amount of activated

carbon inside the canister. Since the adsorption process prefers a cooler environment

the larger canister provides better heat transfer away from the carbon loading sites

and allow for better adsorption inside the carbon bed.

Figure 6.17: Comparison between Simulation and experimental results - Temperature


profile Canister #1

68
Figure 6.18: Comparison between Simulation and experimental results - Temperature
profile Canister #2

Figures 6.17 and 6.18 show the comparison between the simulation and experi-

mental results temperature profile of each canister during the loading phase.

The model predicts the homogeneous temperature at each section of the canister.

In the experiments, the thermocouple measures the temperature of the carbon bed

at the axially central position. Due to this, there is a small discrepancy between the

simulation temperature measurement and the experimental measurement. Also, an-

other factor contributing to the different temperature prediction and actual readings

is due to the assumption made in the simulation where the cross-section area of the

canister is assumed to be constant along the length of the carbon bed. This is not

true in the case of both canisters as the cross-sectional area of canister 1 decreases

as we reach the position of thermocouple 4 and 5 and similarly, the cross-sectional

area of canister 2 is smaller at the locations of thermocouples 4 and 5. The small

69
increase in temperature at thermocouple 4 of canister #2 is due to the shape of the

canister. As seen in figure 4.4 once the position of thermocouple 1 is loaded heat is

released and this heat is dissipated in all directions. Since thermocouple 4 is very

close by the carbon bed temperature at thermocouple 4 also increases due to cross

bed conduction. This conduction term has not been included in the simulation model

hence the model does not predict this small increase in carbon bed temperature.

A comparison between the mass and temperature profiles show that the change in

cross-sectional area of the canisters does not have a significant enough change in the

loading rate or the amount of butane adsorbed on the activated carbon. Therefore

the assumption of a uniform cross-sectional area across the carbon bed is justified

and will make the model less complex and easier to use.

6.3.2 Canister Purging

Canister purging was simulated using the equations (3.22) and (3.31) similar to the

loading simulation but for the purging simulation the initial and boundary conditions

of the simulation have been reversed.

Figures 6.19 and 6.20 present the comparison between the simulation prediction

and experimental results of purging the canister at 50 LPM. Similar to canister loading

the model is able to predict the mass loss in both the canisters. The larger canister

#1 has stored a larger amount of butane vapors due to the larger volume of activated

carbon. The larger canister cross-sectional area also allows for faster desorption of

butane vapor from the activated carbon bed as seen in the figures. Once the majority

of the carbon bed has been desorbed the temperature of the carbon bed decreases

70
Figure 6.19: Comparison between Simulation and experimental results - Mass profile
Canister #1

significantly which creates an unfavorable condition for desorption and therefore the

amount of butane desorbed decreases.

Figures 6.21 and 6.22 present the comparison between the temperature profiles of

the canister as they are being purged of n-butane vapors adsorbed on the activated

carbon bed. The simulation model is able to predict the temperature drop in the

carbon beds fairly accurately.

Similar to the loading phase the simulation model predicts the homogeneous tem-

perature of the carbon bed at each position in the carbon bed and the experiment

results present the carbon bed temperature at the axial center of the carbon bed.

This leads to small variations between the simulation and experimental results. Due

to the assumption of a uniform cross-sectional area the model does not predict the

71
Figure 6.20: Comparison between Simulation and experimental results - Mass profile
Canister #2

larger drops in carbon bed temperature as well. The thermocouples 2, 3 and 4 in

canister #2 started at a higher temperature than thermocouple 1 and the model has

been adjusted to this starting temperature to get an accurate representation of the

experimental results.

72
Figure 6.21: Comparison between Simulation and experimental results - Temperature
profile Canister #1

Figure 6.22: Comparison between Simulation and experimental results - Temperature


profile Canister #1

73
Chapter 7: Conclusions

This project investigated and developed a simplified one-dimensional physics-

based model of the automotive evaporative emission activated carbon canister. The

model was simulated in MATLAB and was validated using experimental results on

two activated carbon canisters. The objective of this thesis was to create a model

for the automotive industry that can accurately predict the amount of gasoline va-

por adsorbed and desorbed by the carbon canister during normal stopped vehicle

conditions.

The model predicts the amount of vapor adsorbed on the activated carbon bed

during slow-loading conditions such as a diurnal stopped vehicle. The model is devel-

oped by discretizing the carbon bed into small areas and assuming a local equilibrium

of adsorption in these areas. The model also assumes a local heat source at the car-

bon bed to include the effects of the heat of adsorption released during the loading

of vapor.

The model predicts the amount of vapor desorbed from the activated carbon bed

during normal purging conditions such as engine ignition. Similar to the loading case

the model is discretized and a local heat sink is used to include the effects of the

endothermic desorption process.

74
Experimental results were conducted on two different shaped EVAP carbon can-

isters to validate the model prediction of the amount of vapor adsorbed and desorbed

by the canister during the loading and purging phases. By comparing the experimen-

tal results with the simulation results the model successfully predicted the amount of

vapor adsorbed and desorbed from the activated carbon canister during loading and

purging phases respectively. The main goal of the project was to create a model that

can predict the vapor loading in the carbon canister with a small number of inputs

so that the model can be used for different vehicles by the automotive manufacturer

in the vehicle modeling phase.

Future work in this project can include predicting the vapor loading in the carbon

canister at higher load rates such as Onboard Refueling Vapor Recovery (ORVR)

and validate the model using further experimental results. The model can be further

refined by including two-dimensional thermal effects on the carbon bed. As seen in

the experimental results the temperature of the carbon bed near the plastic wall is

lesser than the temperature of the carbon bed at the center of the canister.

A more accurate model of the carbon canister can be modeled for future work

where loading of gasoline vapor due to diffusion effects and cross-bed heat conduction

are also considered as part of the 2D or 3D canister model.

75
Appendix A: Canister Dimensions for Model

Figures A.2 and A.1 shows the approximation made to the geometry of the canister

#2 and canister #1 respectively while modeling to create a uniform cross-sectional

area across the carbon bed.

76
Figure A.1: Canister #2 geometry and dimensions approximation for model

Figure A.2: Canister #1 geometry and dimensions approximation for model

77
Appendix B: Thermocouple Positioning on Canister

Tables B.1 presents the position of the thermocouples along the length of the

canister bed. The end of the load valve is considered as x=0 and measurements are

taken from this point towards the atmosphere valve of the canister. In the case of

canister #2 the whole bed of the canister is assumed to be in a straight line and

measurements are made accordingly.

Canister # Thermocouple Number Position along Carbon bed (cm)


TC 1 6
TC 2 14
1 TC 3 20
TC 4 30
TC 5 40
TC 1 5
TC 2 18
2
TC 3 23
TC 4 41

Table B.1: Thermocouple Position along Carbon bed for both Canisters

78
Appendix C: MATLAB code to calibrate model

1
2 clc;
3 clear;
4 close all;
5
6 % ETAS output showing the number of samples collected and the ...
data points
7 % corresponding to thermocouple outputs.
8 % sampleCnt = 24985 ;
9 % recordFormat = 'f8 f8 f8 f8 f8 f8 f8 f8 '
10 % time = 1 ;
11 % TC 1qiES650qNqrqNADqrThermoqj1 = 2 ;
12 % TC 2qiES650qNqrqNADqrThermoqj1 = 3 ;
13 % TC 3qiES650qNqrqNADqrThermoqj1 = 4 ;
14 % TC 4qiES650qNqrqNADqrThermoqj1 = 5 ;
15 % TC 5qiES650qNqrqNADqrThermoqj1 = 6 ;
16 % SkinTCqiES650qNqrqNADqrThermoqj1 = 7 ;
17 % AmbTqiES650qNqrqNADqrThermoqj1 = 8 ;
18
19 %% data points
20

21 data=data points; % function to load values from data collected ...


by ETAS DAQ
22
23 %% temperature plots
24
25 time2=data(:,1)/60;
26 TC 1=data(:,2);
27 TC 2=data(:,3);
28 TC 3=data(:,4);
29 TC 4=data(:,5);
30 TC 5=data(:,6);
31 skinTC=data(:,7);
32 ambT=data(:,8);
33
34 % plot temperature variation for loading experiment

79
35 figure
36 plot(time2,TC 1,'linewidth',2);
37 hold all
38 grid on
39 plot(time2,TC 2,'linewidth',2);
40 plot(time2,TC 3,'linewidth',2);
41 plot(time2,TC 4,'linewidth',2);
42 plot(time2,TC 5,'linewidth',2);
43 plot(time2,skinTC,'linewidth',2);
44 plot(time2,ambT);
45 xlabel('Time [min]')
46 ylabel('Temperature [ˆoC]');
47 title({'Canister 1';' Temperature Variation during Loading ...
Experiment'});
48 legend('TC1','TC2','TC3','TC4','TC5','skin TC','Ambient');
49 set(gca,'fontsize',15)
50
51 %% weight data points and plots
52
53

54 can m start=2869; % canister start weight


55 can m end=3190; % canister end weight
56
57 % collect mass data
58 data1=xlsread("Experiment 5 weight calculations 061619.xlsx");
59

60 can m=data1(:,2)-can m start; % to measure change in mass


61 time1=data1(:,1);
62
63 % Plot mass profile of canister during loading
64 figure
65 plot(time1,can m,'linewidth',2);
66 grid on;
67 hold all;
68 xlabel('Time [min]');
69 ylabel('Canister Mass [g]');
70 title({'Canister 1';' Mass Variation during Loading Experiment'});
71 set(gca,'fontsize',14)
72
73
74
75
76 %% cooldown plots
77

78 % clear;
79 % clc;
80 % close all;
81
82 % to find start point and 1 hour after start
83 temp1=find(TC 1>40,1,'last');
84 temp2=find(time2<time2(temp1)*60+3600,1,'last');

80
85
86 % cooldown plot from experiment
87 figure
88 plot((time2(temp1:temp2)*60)-time2(temp1)*60,TC 1(temp1:temp2),...
89 'linewidth',2);
90 hold all;
91 grid on;
92 plot((time2(temp1:temp2)*60)-time2(temp1)*60,skinTC(temp1:temp2),...
93 'linewidth',2);
94 xlabel('Time [s]');
95 ylabel('Temperature [ˆoC]');
96 grid on;
97 set(gca,'fontsize',14)
98 title('Cooldown Experiment Plots - For Calibration');
99 xlim([0 3600]);
100 ylim([20 50]);
101 legend('TC 1','Skin TC');
102
103 % % to find rate of change of plastic body
104 p=polyfit(TC 1(temp1:temp2),skinTC(temp1:temp2),1);
105 x1=linspace(TC 1(temp1),TC 1(temp2));
106 y1=polyval(p,x1);
107 dT p=p(1);
108
109 figure
110 plot(TC 1(temp1:temp2),skinTC(temp1:temp2));
111 hold all;
112 grid on;
113 plot(x1,y1);
114 xlabel('Carbon Temperature [ˆoC]');
115 ylabel('Plastic Temperature [ˆoC]');
116 set(gca,'fontsize',14)
117 title({'Cooldown Experiment Plots - For Calibration';'Determine ...
dT p/dT'});
118 legend('Experimental Data','Linear Approximation');
119
120 % canister properties
121 A=99; % Cross sectional area [cmˆ2]
122 % r=sqrt(A/pi); % assuming cylinderical canister, radius [cm]
123 P ext=40; % exterior perimeter at x
124 rho app=0.345; % apparent density of the carbon canister [g/ml]
125 m p=0.5; % mass of plastic to mass of carbon ...
(m plastic/m carbon)
126 C vc=0.7; % specific heat of carbon [kJ/kg/C]
127 C p1=1.6; % specific heat of species 1 [kJ/kg/C]
128 C pp=1.9; % specific heat of plastic wall [kJ/kg/C]
129 k app=0.13; % Apparent thermal conductivity of carbon bed ...
[W/m/C]
130 U loss=13; % Wall-air heat conductance [W/mˆ2/C]
131 m s1=0.35; % adsorbed mass of species 1 [g/gC)
132 dh vap=-0.2; % heat of vaporization [J/g]

81
133 T amb=mean(ambT); % Ambient temperature [C]
134 T0=TC 1(temp1); % Initial ccarbon ...
temperature [C]
135 t=100; % time for simulation
136 % dT p=0.2078;
137 sim('Temperatureparameters model');
138
139 figure
140 plot(time*100,T,'linewidth',2);
141 hold on;
142 plot((time2(temp1:temp2)*60)-time2(temp1)*60,TC 1(temp1:temp2),'--r',...
143 'linewidth',2);
144 xlabel('Time [s]');
145 ylabel('Temperature [ˆoC]');
146 grid on;
147 set(gca,'fontsize',14)
148
149 title('Plot of experimental cooldown data - Lavoie values');
150 % set(gca,'fontsize',15);
151 xlim([0 3600]);
152 ylim([20 50]);
153 legend('Model','Experiment');
154
155 % changing uloss
156 figure
157 plot((time2(temp1:temp2)*60)-time2(temp1)*60,TC 1(temp1:temp2)...
158 ,'--r','linewidth',2,'DisplayName','Exp data');
159 hold all;
160 for i=1:10
161 U loss=i*0.01;
162 sim('Temperatureparameters model');
163

164 % figure
165 plot(time*100,T,'linewidth',1,'DisplayName',num2str(U loss));
166 hold on;
167
168 end
169 xlabel('Time [s]');
170 ylabel('Temperature [ˆoC]');
171 grid on;
172 legend('show','fontsize',10,'Orientation','horizontal','NumColumns',2)
173 title({'Cooldown Experiment Plots- For Calibration';'Determine ...
U {loss}'});
174 set(gca,'fontsize',14);
175 xlim([0 3600]);
176 ylim([20 50]);
177
178 % Final plot with correct uloss and dTp
179 U loss=0.045;
180 dT p=0.2078;
181 sim('Temperatureparameters model');

82
182
183 figure
184 plot(time*100,T,'linewidth',2);
185 hold on;
186 plot((time2(temp1:temp2)*60)-time2(temp1)*60,TC 1(temp1:temp2),'--r',...
187 'linewidth',2);
188 xlabel('Time [s]');
189 ylabel('Temperature [ˆoC]');
190 grid on;
191 set(gca,'fontsize',14)
192
193 title({'Cooldown Experiment Plots';'Experiment5- Validation'});
194 set(gca,'fontsize',15);
195 xlim([0 3600]);
196 ylim([20 50]);
197 legend('Model','Experiment');

83
Bibliography

[1] Xingyan Bai, K M Isaac, R Banerjee, D Klein, W Breig, and L Oliver. Modeling
and simulation of n-butane adsorption/desorption in a carbon canister. SAE
Technical Papers, 03 2004.

[2] Roop Chand Bansal and Meenakshi Goyal. Activated carbon adsorption. Boca
Raton : Dekker/CRC Press, 2005. Includes bibliographical references and in-
dexes.

[3] Manel Bergaoui, Asma Nakhli, Shaheen Al-Muhtaseb, and Mohamed Khalfaoui.
Adsorption process of n-alkanes onto bax-1100 activated carbon: Theoretical es-
timation of isosteric heat of adsorption and energy distribution of heterogeneous
surfaces. Journal of Molecular Liquids, 252:399 – 407, 2018.

[4] Richard G. Derwent. Sources, distributions, and fates of vocs in the atmosphere.
In R. E. Hester and R. M. Harrison, editors, Volatile Organic Compounds in the
Atmosphere, volume 4, pages 1–16. The Royal Society of Chemistry, 1995.

[5] D.D. Do. Chapter 16. dynamics of adsorption in heterogeneous solids. In


W. Rudziski, W.A. Steele, and G. Zgrablich, editors, Equilibria and Dynam-
ics of Gas Adsorption on Heterogeneous Solid Surfaces, volume 104 of Studies in
Surface Science and Catalysis, pages 777 – 835. Elsevier, 1997.

[6] Ground-level ozone basics, Oct 2018.

[7] Can Erkey. Chapter 4 - thermodynamics and dynamics of adsorption of metal


complexes on surfaces from supercritical solutions. In Can Erkey, editor, Super-
critical Fluids and Organometallic Compounds, volume 1 of Supercritical Fluid
Science and Technology, pages 41 – 77. Elsevier, 2011.

[8] E. Fiani, L. Perier-Cambry, and G. Thomas. Non-isothermal modelling of hydro-


carbon adsorption on a granulated active carbon. Journal of Thermal Analysis
and Calorimetry, 60(2):557–570, May 2000.

[9] K.Y. Foo and B.H. Hameed. Insights into the modeling of adsorption isotherm
systems. Chemical Engineering Journal, 156(1):2 – 10, 2010.

84
[10] R. J. Grant, Milton Manes, and S. B. Smith. Adsorption of normal paraffins and
sulfur compounds on activated carbon. AIChE Journal, 8(3):403–406, 1962.

[11] Setsuda D J Johnson, P J and R S Williams. Activated carbon for automotive


applications, Jul 1999.

[12] G.A. Lavoie, P.J. Johnson, and J.F. Hood. Carbon canister modeling for evap-
orative emissions: Adsorption and thermal effects. SAE Technical Papers, 01
1996.

[13] Mallard W. G. National Institute of Standards Linstrom, Peter J. and Technology


(U.S.). Nist chemistry webbook., 1997. ID: 37586124.

[14] Huan Liu, Hanyang Man, Michael Tschantz, Ye Wu, Kebin He, and Jiming Hao.
Voc from vehicular evaporation emissions: Status and control strategy. Environ-
mental Science & Technology, 49(24):14424–14431, 2015. PMID: 26599318.

[15] Harry Marsh and Francisco Rodrguez-Reinoso. Chapter 2 - activated carbon


(origins). In Harry Marsh and Francisco Rodrguez-Reinoso, editors, Activated
Carbon, pages 13 – 86. Elsevier Science Ltd, Oxford, 2006.

[16] Kazunari Sato and Noriyuki Kobayashi. Adsorption and desorption simulation
of carbon canister using ¡i¿n¡/i¿-butane as model compound of gasoline. Journal
of the Japan Petroleum Institute, 54(3):136–145, 2011.

[17] Kazunari Sato, Noriyuki Kobayashi, and Masanobu Hasatani. Calculation


method of adsorption and desorption performance of butane gas in carbon can-
ister. SAE Int. J. Engines, 1:409–416, 04 2008.

[18] J.D. Seader, E.J. Henley, and D.K. Roper. Separation Process Principles with
Applications Using Process Simulators, 4th Edition. Wiley, 2016.

[19] Hussain A. Pautasso E. Servetto E. Graziano E. Brown J. Smith, L. Evap system


fluid-dynamics and chemistry modelling for ems purge control development and
optimization. SIA Powertrain Conference, Versailles, 2015.

[20] Ann Y. Watson. Automotive emissions, Jan 1988.

[21] Roger S. Williams and C. Reid Clontz. Impact and control of canister bleed
emissions. SAE Transactions, 110:579–587, 2001.

[22] In Kwang Yoo, Devesh Upadhyay, and Giorgio Rizzoni. A control-oriented car-
bon canister model. SAE Transactions, 108:2133–2142, 1999.

85

You might also like