Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 



MODIFICATION OF AIR FLOWS WITHIN AN 

INDUSTRIAL CARCASS CHILLER USING DELTA 

WING VORTEX GENERATORS 

Royce Russell Heap, B. Tech (Hons) (Massey) 



ABSTRACT 

The chilling of carcasses after slaughter has a considerable bearing on process costs 

and quality of the meat. Uniform air distribution is essential for the optimal operation 

of carcass chillers , yet many existing chillers have highly variable and ineffective air 

flow. This directly affects the uniformity of evaporative weight loss and carcass 

cooling rates within the chiller. 

Delta plan aerofoils are known to convert a unidirectional air stream, progressively 

into a rotational and then turbulent multidirectional fragments of moving air. The aim 

of this research was to evaluate the use of delta wings to improve air flows inside a 

venison carcass chiller of typical design used in the New Zealand meat industry. 

Air flow patterns within the chiller were characterised by measuring mean air speeds 

with a hot-wire anemometer over 61 grid points at 4 levels in height. Air speeds were 

found to be highly time-variable so the mean, standard deviation and range of 60 one 

second air speed measurements were used to represent the air flow at each point on 

the grid. The measurement of air velocity (speed and direction) using three 

othoganall y mounted propeller anemometers had limited success, as air speeds within 

the chiller were often below the threshold of the anemometer. 

Measurements before the installation of the delta wings indicated that a poor air flow 

di stribution existed within the chiller as the majority of air was found to circulate 

around the wall s and floor, producing near stagnant conditions between the carcasses. 

Delta wings were constrncted in two sizes from thin aluminium sheets. Wings were 

installed into the chiller by suspending them from the ceiling within the evaporator fan 

delivery air stream. Two wing configurations were trialed: The first wing 

configuration utilised 3 large delta wings mounted in front of the evaporator fans 

followed by a row of 6 small wings then a row of 7 small wings (3,6,7). The second 

delta wing configuration utilised a row of 6 large wings closest to the evaporator fans 

followed by a row of 9 small wings then a row of 13 small wings (6,9,13). 

The second delta wing configuration showed superior performance over the first. In 

comparison to the unmodified chiller without wings, the mean air speeds in the critical 

region amongst the carcasses increased from 0.4 mis to 0.6 mis; the standard deviation 

of mean air speeds decreased from 0.33 mis to 0.22 mis and the percentage of mean 



ll 

air speeds between half and twice the mean increased from 84% to 95%. The second 

configuration of delta wings also produced a 14% increase in the mean air turbulence 

intensity (measure of the time-variability in air speed) and reduced the variability of 

evaporative weight loss within the chiller. 

Overall, the delta wings were found to be an economic way to improve the 

performance of a chiller by providing a more uniform and effective air distribution 

without increasing fan power. This can result in a reduction in chilling times and less 

potential for weight loss. Their use in both new and existing chillers is recommended. 



111 

ACKNOWLEDGEMENTS 

The author would like to thank the following persons for advice and assistance during the 
course of this project: 

Senior Lecturer Ross J Davies, Department of Process and Environmental 
Technology, Massey University, Chief Supervisor. 

Professor Donald J Cleland, Department of Process and Environmental 
Technology, Massey University, Supervisor. 

The author would li ke to thank the following persons for technical support during the 
experimenta l phase of this project: 

Chuck Ng, Department of Process and Environmental Technology, Massey 
University. 

Bruce R Collins, Department of Process and Environmental Technology, Massey 
University. 

John T Alger, Department of Process and Environmental Technology, Massey 
University. 

Don McLean, Department of Process and Environmental Technology, Massey 
University. 

Neil Jamieson, Scientist (Aerodynamics), Works Consultancy Services Limited. 

The author would like to thank the Foundation for Research, Science and Technology, 
Steelfort New Zealand Company Limited, Venison Packers Trading Limited and the 
Electricity Corporation of New Zealand Limited for funding this project. In particular the 
author would like to acknowledge the support of: 

John Jenkins, Steelfort New Zealand Company Limited. 

John Signal, Plant Manager, Venison Packers Trading Limited. 

Raymond Pratt, Plant Supervisor, Venison Packers Trading Limited. 

Rodger Kallu, Programme Manager -Refrigeration and Heat Pumps, Electricity 
Corporation of New Zealand Limited. 



CONTENTS 

Abstract 
Acknowledgments 
Contents 

CHAPTER 

1. INTRODUCTION 

2. LITERATURE REVIEiV 

2.1 IMPORTANCE OF AIR CIRCULATION IN CHILLING 
2. l. l Chilling Rates 
2. 1.2 Weight Loss 
2.1.3 Microbial Growth 
2.1.4 Uniformity of chilling 

2.2 FACTORS AFFECTING AIR FLOW 
2.2. I Flow Along a Fixed Swface 
2.2.2 Flow Over Baffles 
2.2.3 Flow Over Aerofoils 
2.2.4 Fans 

2.3 IMPROVING AIR DISTRIBUTION WITIN CHILLERS 
2.3. I Chiller Design 
2.3.2 Air Turbulence Within Chillers 

2.4 VISUALISING AIR FLOW PATTERNS 
2.4. / Scale Modeling 
2.4.2 Computational Fluid Dynamics 
2.4.3 Snwke or Bubble Technique 
2.4.4 Ribbon or Tape Technique 
2.4.5 Anemometry 

2.5 MEASURING AIR VELOCITIES IN CHILLERS 
2.5.l Measurement Duration 
2.5.2 Number of Measurements 

2.6 DELTA WING VORTEX GENERATORS 
2.6. I Vortex Generation 
2.6.2 Delta Wing Vortex Generators for Improved Air Mixing 

Ill 

JV 

1 

3 

3 
5 
6 
10 
11 

14 
14 
14 
15 
16 

17 
17 
20 

21 
21 
22 
23 
24 
24 

25 
26 
27 

27 
27 
28 

iv 



V 

2.7 RESEARCH OBJECTIVES 30 

3. MATERIALS AND METHODS 32 

3./ INTRODUCTION 32 

3.2 CHILLER SPECIFICATIONS 32 
3.2.J Chiller Construction 32 
3.2.2 Chiller Evaporator Spec~fzcations 33 
3.2.3 Chiller Operation 34 

3.3 AIR FLOW MEASUREMENTS 35 
3.3.J lvleasurement Methods 35 
3.3.2 Air Speed Measurements - Instrument Calibration and 36 

Testing 
3.3.2.1 Calibration 36 
3.3.2.2 Correlation of Voltage Signal With Displayed Air Speed 37 

Reading 
3.3.2.3 Transducer Housing Inte1jerence 38 
3.3.2.4 Measurement Time 40 
3.3.2.5 Estimation of Measurement Error Caused By Random 43 

Fluctuations in Air Flm,· 
3.3.2.6 Attainment of Steady State Conditions 44 
3.3.2.7 Estimation of the Air Speed Standard Deviation 46 
3.3.3 Directional Air Flow Measurements - Instrument Calibration 46 

and Testing 
3.3.3. I Calibration 46 
3.3.3.2 Response of a Single Anemometer at Various Orientation 47 

Angles to the Air Flow 
3.3.3.3 Response of Orthogonally Mounted Propeller 48 

Anemometers at Various Orientation Angles to the Air 
Flow 

3.4 DELTA WING CONSTRUCTION 51 

3.5 DELTA WING INSTALLATION 52 
3.5.l First Delta Wing Configuration 52 
3.5.2 Second Delta Wing Configuration 55 

3.6 CHILLER CHARACTERISATION 57 
3.6.l Chiller Measurement Sites 57 
3.6.2 Empty Chiller Air Flow Measurements 59 
3.6.3 Full Chiller Air Flow Measurements 59 



VI 

3.7 DATA ANALYSIS 61 
3. 7.1 Air Speed 61 
3.7.J.J Contour Plots 61 
3.7.1.2 Mean Air Speed Plots 61 
3.7. /.3 Difference in Mean Air Speed Plots 62 
3.7.2 Air Flow Direction 62 
3. 7.2. J Vector Plots 67 
3.7.3 Distribution of Air Speeds 63 
3. 7.3. l Histograms and Standard Deviations 63 
3. 7.3.2 MIR/NZ Indicator 64 
3.7.3.3 Weight Loss 64 
3.7.4 Turbulence 65 
3.7.4.J Turbulence Intensit1; 65 

3.8 DATA CONFIDENCE 65 
3.8./ Number of Measurements 65 
3.8.2 Data Reproducibility 66 
3.8.2.2 Air speed Measurements 66 
3.8.2.3 Directional Measurements 68 

4. RESULTS AND DISCUSSION 70 

4.1 INTRODUCTION 71 

4.2 EMPTY CHILLER CHARACTERISATION 71 

4.3 FULL CHILLER CHARACTERISATION 76 
4.3./ Air Speed 76 
4.3.2 Air Flow Distribution 81 
4.3.3 Weight Loss 85 
4.3.4 Air Turbulence 87 
4.3.5 Additional Tes ting 94 

4.4 DISCUSSION 87 
4.4.J Empty Chiller Characterisation 87 
4.4. I. I Without Delta Wings 87 
4.4. /.2 First Delta Wing Configuration 89 
4.4. 1.3 Second Delta Wing Configuration 90 
4.4.J.4 Summary of Results in Empty Chiller 91 

4.4.2 FULL CHILLER CHARACTERISATION 91 
4.4.2. l Air Speed 91 
4.4.2.2 Air Flow Distribution 94 
4.4.2.3 Weight loss 95 



vii 

4.4.2.4 Additional Testing 96 
4.4.2.5 Summary of Results in Full Chiller 96 

4.5 COMPARISONS TO PREVIOUS WORK 97 

4.6 IMPLICATIONS FOR INDUSTRY 98 

5. CONCLUSIONS 101 

6. RECOMMENDATIONS FOR FURTHER WORK 103 

REFERENCES 

APPENDIX A 

APPENDIXB 

APPENXIXD 

APPENDIX£ 

APPENDIXF 

105 



Chapter I 

INTRODUCTION 

The New Zealand meat industry continues to be a major contributor to New Zealand's 

economy. Since the I 880's it has been responsible for a large proportion of New 

Zealand's gross domestic product and foreign exchange earnings; in l 996 contributing 

16.3 percent to the total value of exports (Beattie, l 996). 

New Zealand produces around 1 million tonnes of meat per year, most of which is 

destined for overseas markets. Although, this makes up only 0.4 percent of the worlds 

total meat protein output, it equates to approximately 17 percent of all internationally 

traded meat. This can be broken down into a 7 percent share of world beef and veal 

exports and a 44 percent share of world sheep meat exports (Willis, 1992). 

New Zealand still faces strong tariff and non-tariff barriers to free trade of agricultural 

products. These have restricted the volume growth of exports and real increases in prices. 

Subsidised returns to farmers in the US and OECD countries encourages excess 

production. This surplus competes directly with New Zealand's exports. When this 

occurs, New Zealand farmers receive artificially low returns for their produce. The 

significant cost of long distance transportation to many of New Zealand's major markets 

and barriers to free trade means that New Zealand's agricultural products must be 

produced at a lower cost, more efficiently and be of superior quality to compete with 

those of its competitors. 

The refrigeration of meat makes up a significant proportion of the products final cost, as it 

is used extensively in the processing, transport and storage of meat. Capital investment, 

fan power, compressor energy and product weight loss all contribute to the cost of 

refrigeration. Refrigeration also has a significant impact on the final quality 

characteristics of the product. The microbial condition, tenderness and appearance are all 

affected by the environment, prior to and throughout the "cold-chain". 

Increased competition has led to a larger range of product choice for the overseas 

consumer. The market for frozen meat is declining as more consumers demand chilled 



2 Chapter I: Introduction 

"fresh meat" cuts . This trend has forced a change in the meat industry towards the further 

processing of meat before it reaches the point of sale. Higher prices due to increased 

demand for further processed chil led meat have driven this shift which has resulted in 

dramatic changes in the export of sheep and beef. 

In the year ended June 1992, 3 percent of New Zealand's sheep meat was exported in the 

unfrozen form . Since 1992 the quantity of chilled, fresh lamb exports has doubled, 

contributing to 7 percent of the total sheep meat exports in 1996. The volume of chilled 

lamb exports is expected to continue to grow and command premium prices (Lynch, 

1996). These shifts have also been evident in beef processing where the export of 

traditional frozen beef quarters have been replaced by chilled, boneless, plastic wrapped 

primal cuts. 

Such changes have been made possible by developments in the technology associated 

with the commerci al refrigeration of meat. In particular, developments to better control 

the environment in refrigerated shipping facilities has enabled the transport of chilled 

meat cuts to distant markets, replacing the need to transport whole frozen carcasses. The 

introduction of vacuum and modified atmosphere packaging has also contributed to this 

change by extending the products shelf life. 

Temperature, relative humidity, air velocity and air distribution make up the 

environmental conditions inside a refrigerated space. The effects of temperature and 

relative humidity on product characteristics are well understood and both can be 

control led relatively easily. In contrast, the effects of air velocity over the product and air 

distribution within the room are less well understood and tend to be more difficult to 

control. However, adequate air velocity and a uniform air distribution are crucial to 

effective chilling and the holding of product in a chilled state. The products microbial 

quality , chilling rate, weight loss and uniformity of chilling are all directly affected. 

The overall aim of this research is to consider meat carcass chiller design and operation 

from an air velocity and distribution perspective. In particular, the objective is to improve 

the uniformity of air flow across the product and increase chilling rates whilst minimising 

the rate of evaporative weight loss and fan energy requirements. 



3 

Chapter 2 

LITERATURE REVIEW 

This literature review is divided into seven sections. The first section considers the 

importance of air circulation to the effectiveness of carcass chilling. The second and third 

sections focus on the factors that influence air flow and ways that air circulation and 

distribution might be improved in chillers. The fourth and fifth sections cover the various 

methods used to visualise and measure air flow within chillers. The sixth section reviews 

the development and use of delta wing vortex generators and the effect these have on air 

movement. Lastly, the seventh section defines the specific objectives of this research. 

2.1 IMPORTANCE OF AIR CIRCULATION IN CHILLING 

The parameters which give rise to the physical environment in a chiller are of particular 

importance as they have a major effect on the product quality characteristics. Air 

velocity, humidity and temperature all effect ultimate meat quality. In particular, these 

parameters influence quality aspects such as the microbial status of the carcass, 

tenderness and appearance of the meat. They also affect processing aspects, such as the 

carcass weight loss, shrinkage, ease of cutting and the amount of condensation (produced 

on the chiller superstructure and on the carcass). Nottingham ( 1971) notes that as 

processing and quality requirements are often in conflict, optimal conditions for chilling 

must be a compromise. 

In general, a fast chilling rate results in less evaporative weight loss and lower microbial 

growth (Taylor, 1972). However, it also results in a higher process cost and an increased 

susceptibility to toughness through cold shortening (Bendall, 1972). A slow rate of 

chilling leads to a lower process cost and is more conducive to tenderness, but increases 

evaporative weight loss and microbial load. Therefore, a "window of compromise" exists 

were there is an acceptable balance of process cost, toughness, microbial growth, and 

weight loss (Figure 2.1 ). 



4 

Unacceptable 
process cos ts 

and toughness 

Chapter 2: Literature Review 

Chilling ture 

Figure 2.1: Compromise in the conditions for chilling 

Unacceptable 
weight loss 

and mlo growth 

In many instances the chilling regime cannot be controlled within this 'window' due to 

poor uniformity of air flow across the refrigerated space. Localised zones of high air flow 

enable the fast removal of heat from carcasses within these areas, but this product is more 

susceptible to toughness. Conversely, zones of low air flow produce areas where the 

chilling rate is slow yet carcasses in these locations are more susceptible microbial 

spoilage. 

In a batch situation, unloading of the chiller can only take place once the carcass with the 

slowest chilling rate has reached the required temperature. This situation produces 

conditions which are conducive to excessive weight loss. Carcasses which cool quickly, 

often must undergo a holding period in a location of high air flow as the required 

temperature is obtained much earlier than slowest cooling carcass. 

Although chilling and freezing of meat has been carried out in New Zealand for over one 

hundred years, considerable differences of opinion still exist as to the optimal 

environmental conditions in regard to air velocity, relative humidity and temperature. Of 

these three variables, air velocity (and air movement in general) is the most poorly 

understood and gives rise to the most controversy in chilling and freezing. It is well 

established that air velocity strongly affects the chilling rate (Frazerhurst, 1971; Longdill, 

1974; Daudin and Swain, 1990; Kondjoyan and Daudin, 1995). However, little is known 

about the distribution of the air flowing around the product (Daudin, 1992). In practice, 

air circulation is often only considered in a superficial manner in the design and operation 



Chapter 2: Literature Review 5 

of chillers despite the proven economic advantages obtained in systems where air flow is 

well managed. 

2. l. J CHILLING RATES 

There is obvious economic incentive in the rapid chilling of product. Rapid chilling 

achieves a quick turnover, better hygiene, reduced drip and in many cases a lower weight 

loss (Taylor, 1972: Gigiel et. al., 1989). However, the rapid chilling of carcasses can be 

detrimental to meat quality, for example, toughness induced by cold shortening of the 

meat (Bendall. 1972). This conflict between economic and quality aspects has given rise 

to two main processing specifications. The first is to slow chill, often with an initial 

delay stage at l 0°C or above to maintain product quality. The second is to rapid chill in 

two stages: an initial period of low air temperatures and high air velocities to reduce the 

potential for drip and weight loss: followed by a less severe second stage of chilling to 

allow for temperature equilibration (Gigiel et. al., 1989). 

The rate of chilling is determined by a combination of air velocity over the product, the 

difference between the temperature of the air and product surface as well as the nature 

and variability of the product itself (Garnett, 1974). Although the nature and variability 

of the product does have a large bearing on chilling rates, carcass variations such as 

surface area to mass ratio. side thickness, fat cover and initial surface moisture are largely 

beyond the control of the processor. Therefore. the velocity of air as well as the 

temperature of air are the two main variables that can be manipulated to influence chilling 

rates. 

Regulatory authorities require beef carcasses to be chilled to a deep bone temperature of 

10°C or lower within 16 hours, otherwise carcasses are considered to be susceptible to 

spoilage during subsequent storage (Hodgson, 1970). Hodgson (1970) stated that air 

velocities of at least 0.75 mis over the sides are required to reduce the deep bone 

temperature of a I 00 kg beef side to l 0°C within 16 hours in a room controlled at 0°C. 



6 Chapter 2: Literature Review 

The Meat Industry Research Institute of New Zealand (MIRINZ) specifies that air 

velocities of 0.5 to 1.0 mis with a minimum of 0.25 mis over any surface should be used 

for chilling and aging of beef (Garnett, 1974). Garnett (1974) carried out tests in a 

number of different chillers and confirmed that air velocities of l mis are required over 

the hind quarters of 135 kg prime beef sides if deep leg temperatures of 10°C are to be 

obtained over a 21 hour cycle in a chiller operating at 4°C. It was suggested that these 

velocities could be reduced to about 0.35 mis between the sides because of smaller 

carcass cross-sectional areas at this level. 

James and Bailey ( 1986) found that increasing air velocity during chilling produces a 

significant reduction in chilling times. Benefits were most pronounced at low air 

velocities. It was concluded that air velocities greater than l mis over the product are 

unlikely to be justified as the power required to move air within a refrigerated room 

increases with the cube of velocity. For example, a four-fold increase in air velocity from 

0.5 to 2 mis may result in a 4 to 7 hour reduction in chilling time for a 140 kg beef side, 

but requires a 64 times increase in fan power. 

2.1.2 WEIGHT LOSS 

Weight loss is critical in determining profitability in the meat industry. Currently, profit 

margins in the industry are small, so every gram of water retained within the meat before 

it is sold becomes an important transaction. 

Since evaporation is one of the mechanisms by which the carcass loses heat during 

chilling, some moisture loss is inevitable. Weight loss is effectively a slow low 

temperature drying process which is mass transfer limited (Daudin and Swain 1990). 

When air flows around a wet permeable body, mass transfer takes place both inside the 

body by diffusion and through the boundary layer which develops at the air-body 

interface. In the boundary layer, air flow is laminar and water vapour transfers to and 

from the surface by diffusion. This process depends on water vapour diffusivity in air 

and the thickness of the boundary layer. Thickness of the boundary layer is determined 

by shape, size and surface roughness of the body as well as air velocity and viscosity. 

Since the thickness of the boundary layer cannot be easily measured, its resistance to 



Chapter 2: literature Review 7 

mass transfer is usually represented by an average mass transfer coefficient which has to 

be determined experimentally. 

The mass transfer rate dM/dt at the air-body interface can be represented by : 

dM/dt = k A ( P wp aw,s - Pa> (2.1) 

(2.2) 

M Mass of product (kg) 

t Time (s) 

aw,s Water activity at the surface 

P
0 

Partial pressure of water vapour in bulk air (Pa) 

P wp Vapour pressure of water at product temperature TP (Pa) 

Pwa Vapour pressure of water at air temperature Ta (Pa) 

k Mass transfer coefficient (kg/m2sPa) 

A Transfer surface area (m2) 

H,. Relative Humidity(%) 

As equation 2. l shows, the rate of weight loss is a function of the mass transfer 

coefficient, the area available for mass transfer and the partial pressure driving force for 

mass transfer. 

The mass transfer coefficient is a measure of the "ease" with which water can break free 

from the product surface and move through the boundary layer into the bulk air (Cleland 

& Cleland, 1996). It depends on how readily moisture is transferred from the product 

surface to the air (interphase resistance), and from the air at the surface to the bulk air 

(transport resistance). Interphase resistance is a function of product characteristics. For 

example, a carcass with thick fat cover generally has a large interphase resistance, 

whereas lean meat has a low interphase resistance. Transport resistance is a function of 

air velocity over the product. Once the air velocity is increased above a critical level, the 

transport resistance becomes small compared to air resistance, so any further increases in 

ve locity will hardly change the mass transfer coefficient. 



8 Chapter 2: Literature Review 

Partial pressure driving force is dependent on the products surface temperature, as well as 

the bulk air temperature and its relative humidity (Cleland & Cleland, 1996). Relative 

humidity has very little effect on the rate of weight loss early in the chill cycle as the 

impact on the partial pressure diving force is small whi le the temperature difference is 

large. Initially, the partial pressure driving force varies directly with temperature 

difference between the product surface and air. 

As the products surface area is usually fixed and the mass transfer coefficient only varies 

slightly in a adequate air flow (as the interphase resistant is large in comparison to the 

transport resistance), the partial pressure driving force is the dominant variable in the 

chilling phase. As the partial pressure driving force varies directly with the temperature 

difference between the product surface and the air, the product is most susceptible to 

weight loss in the early stages of the chilling process because the product surface 

temperature is large , the partial pressure driving force will also be large. 

Higher air velocities reduce the length of chilling time by increasing the rate at which the 

product surface temperature drops. While product temperature is above the required 

chilled temperature, an increase in air velocity will reduce the total weight loss but 

increase the rate at which weight loss occurs (Cleland & Cleland, 1996). 

As the temperature of the product stabilises, the product can no longer be considered as 

being actively chilled. Under these conditions the partial pressure driving force is small 

and the 'drying power' of the air becomes dominant (Longdill , 1974). The drying power 

of the air is its ability to remove moisture from a wet surface at the same temperature. It 

is a function of air relative humidity and air velocity. Hence, once the carcass 

temperature is reduced, it is important to maintain a high relative humidity and a low air 

velocity in order to minimise weight loss . 

The basis, on which weight loss is measured by, is vital in determining whether a change 

in air velocity increases or decreases weight loss. Weight loss measurements can be taken 

over a fixed time interval or by chilling to a fixed final temperature (Lovett et. al, 1978). 



Chapter 2: Literature Review 9 

By conducting wind tunnel tests over a fixed time interval, Lovett et. al. (1978) 

concluded that in the initial stages of chilling, weight loss was not significantly affected 

by air velocity in the ranges from 0.56 to 3.7 mis. After 22 hours of chilling, however, 

there were significantly higher evaporative losses from the meat samples exposed to the 

higher air velocities. Alternative results were presented as a function of the temperature 

of the meat and air velocity, instead of a function of chilling time and air velocity. For 

the same meat temperature, a higher air velocity increased the rate of weight loss, but also 

gave a faster rate of cooling. Higher air velocities were found to reduce total evaporative 

losses at final meat temperatures between 4 and 13° C. 

Joseph and Smith (I 984) also used a a fixed chilling time to investigate the effects of air 

veloc ity on weight loss. It was found that when air speed over the surface of the cooling 

beef carcass was increased from 0.5 to 1.0 mis then evaporative weight loss decreased, 

but if the speed is increased further, weight loss increased again. 

To study the degree of weight loss caused by usmg conventional chilling or forced 

chilling techniques, van der Wal et. al. ( 1995) carried out experiments by chilling pigs to 

a fixed final temperature. Conventional chilling was carried out with an air temperature 

of 4°C and air veloc ity of 0.5 mis, whereas fo rced chilling was carried out usi ng an air 

temperature of -5°C fo r 120 minutes or -30°C for 30 minutes with air velocities of 1, 2 or 

4 mis. Results showed that losses in carcass weight were about 2% for conventional 

chilling and forced chilling at -5°C, but were reduced to 1.3% by 'ultra' rapid, forced 

chilling at -30° C with an air veloc ity of 4 mis. 

Cooper ( 1970) al so carried ou t studies on the weight loss of pig carcasses during quick 

and rapid chilling. Quick chilling of 60 kg dressed carcasses, using an air velocity of 0.25 

mis and temperature of 0.5° C , resulted in a weight loss of 1.9%. Rapid chilling, using an 

air velocity of 2 mis and temperature of -7° C, resulted in a lower weight loss of 1 .4% for 

the same degree of cooling (final temperature). 

James and Bailey ( 1986) compared weight loss of beef carcasses over a fixed interval of 

18 hours and found that changing the air velocity from 0.75 to 3 mis had only a small 

effect on weight loss. Similar results were obtained by Kerens ( 1976) where small 

increases in weight loss were detected when increasing the air velocity from 0.25 to 1.5 

mis. 



10 Chapter 2: Literature Review 

Longdill and Pham (1981) found that when chilling over a fixed time interval, weight 

losses increased proportionally with increasing air velocities for average air velocities 

under 0.6 mis. They concluded that the velocity of air flowing over the carcass was the 

most significant variable affecting weight loss in chilled lambs. 

Hodgson (1970) and Levy ( 1972) both recommended low air velocities for minimum 

weight loss when chilling over a fixed period of time. Hodgson ( 1970) stated that carcass 

chilling rooms should be controlled at the minimum value that is necessary to give the 

required product temperature at the terminal stage of the chilling process. 

In general , research suggests that increasing air velocities while chilling over a fixed time 

interval will increase weight loss. Whereas increasing air velocities while chilling to a 

fixed final temperature will reduce the overall weight loss, but a higher rate of weight loss 

will occur. Therefore, the most widely accepted process is to use a fast chilling rate 

initially to reduce the product surface temperature rapidly, followed by conditions of low 

air velocity and high relative humidity once the carcass is close to the required 

temperature. These conditions theoretically give the lowest overall weight loss (Longdill , 

1974). 

2.1. 3 MICROBIAL GROWTH 

Refrigeration of meat is principally concerned with the inhibition of microbial spoilage It 

is inevitable that bacteria will be transferred to the meat surface during slaughter. Many 

of these bacteria are psychotropic (ie. capable of growth at chill temperatures). Since 

microbial growth mainly occurs on the carcass surface, it is common to express the 

microbial condition of meat in terms of the number of viable organisms on a given 

surface area of meat (Shaw, 1972). 

The rate of microbial growth on the carcass surface is influenced by a number of factors 

such as the rate of fall in temperature, changes in the physiological factors (eg. decreasing 

pH within the tissue after slaughter) and the availability of water. The term used for 

phys iologically available water is 'water activity'. The water activity for chilled lean meat 

is approximately 0.993 (Scott, 1936). This availability of water offers ideal conditions for 

microorganisms to grow. 



Chapter 2: Literature Review 11 

Air velocity affects the available water on the carcass surface by contributing to surface 

drying and thus affects the rate of bacterial growth. Scott and Vickery (1939) found that 

provided the rate of air movement was sufficiently high (above 0.7 mis) bacterial growth 

could be prevented even if the air was relatively warm. If the air flow was halved 

bacterial numbers increased 26-fold in 72 hours. It was therefore concluded in this study 

that the rate of desiccation was likely to be more important than the rate of chilling in 

restricting microbial growth . 

Macfarlane ( 1973) took bacterial counts on the neck, loin and leg positions on beef sides 

during chilling in order to find the relationship between air velocity and bacterial growth. 

The data confirmed that where air flows tended to be low and consequently drying was 

minimal (eg. the neck region ), bacterial loads tended to be high. 

2. 1.4 UNIFORMITY OF CHILLING 

The heterogenei ty of air flow causes major problems in the control of batch chillers due to 

the resulting variation in chilling times , surface temperatures and unfavourable changes to 

the quality of the product. 

During chilling, the surface temperature of the product is largely determined by the 

surrounding air temperature and its velocity (Kondjoyan and Daudin , 1995). A long 

initial period in which the su rface temperature is above 10°C must be avoided to reduce 

the risk of bacterial growth, especially that of pathogenic bacteria. However, significant 

risks only exist when air velocities are low (below 0.5 mis ) and chilling times are long. 

The risk of surface freezing is much more common. The temperature at the surface must 

not drop below the freezing temperature of the product as this will give rise to thawing 

drip which is detrimental to the appearance of the product, increases overall weight loss 

and promotes bacterial growth. For lean meat the freezing temperature is about - I 0 C. 

However, Kondjoyan and Daudin ( 1995) suggest that in practice it is best to avoid surface 

temperatures below 0°C. 



12 Chapter 2: Literature Review 

For satisfactory operation, a carcass chilling room air distribution system should maintain 

a reasonably unifom1 air velocity distribution across the whole of the product. Localised 

recirculation patterns or free convection of warm air should be avoided, (as this can lead 

to inadequate cooling or condensation on the chiller superstructure) and an adequate flow 

of air across the thickest part of the carcass is needed to ensure a satisfactory chilling rate 

is achieved without increasing process costs unduly . 

Surveys on air flow characteristics of a number of industrial chillers have shown that the 

distribution of air velocities and air circulation are often far from an ideal in regards to 

uniform flow . (Macfarlane, 1974; Daudin et. al., 1992). The same situation has also been 

found in industrial cool stores and freezers (Falconer, 1993 ; Oliver, 1986; Edwards, 

1979). 

Studies of air flow characteristics in a chiller for pork offals by Daudin et. al. (1992) 

exemplified the large variation of air flow that exists in industrial plants . The air velocity 

was found to vary between 0.3 mis and 9 mis and had an average velocity of 2.3 mis with 

a standard deviation of 1.5 mis. 

Wootton ( 1986) investigated a number of carcass chilling rooms in Northern Ireland. So 

much vari ation in air velocity was found that the location of the carcass in the chill room 

had almost as much effect on its cooling rate as did carcass weight, which was expected 

to be the most important variable. 

Edwards ( 1979) suggested that if 95 % of the measured air velocity readings adjacent to 

the product were between ½ to 2 times the average air velocity, then there was an 

acceptably uniform air flow within the refrigerated room. Oliver (1986) carried out 

studies on three types of carton blast freezers and found that the best configuration 

studied only gave 88% of the air velocity readings between 1/2 and 2 times the mean 

velocity . 

Edwards (1979) showed that the basic multi -story vertical air flow carcass freezer was 

also unsatisfactory in terms of the uniformity of air flow . Only 91 % of the air velocity 

readings adjacent to the carcass were within the specified range. 



Chapter 2: Literature Review 13 

In conventional batch chillers with ceiling mounted evaporator fans, much of the air tends 

to bypass the carcass hanging area completely and recirculate back to the inlet of the 

evaporator. Test results on conventional batch chillers obtained by Murray ( 1972) 

indicated that air velocities near the walls and along the floor can be six times the 

velocities through the carcasses. 

Ammie ( 1993) carried out extensive air flow measurements m a pig pre-chiller. The 

room operated with a ceiling mounted evaporator unit which housed 3 continuous fans 

operating with a total volumetric flow rate of 11370 m3/h (3.16 m3/s). The room had a 

volume of 87.7 m3 and contained 5 rails mounted perpendicular to the air flow direction . 

The room could hold a maximum of 50 pigs. Air flow testing was undertaken in the 

empty room and again when the room was 2/ 3 full. Air velocity measurements were taken 

at the leg, loin and head positions of the carcasses (9 around each carcass). 

Measurements in the empty chiller were performed at intervals of I meter across 4 

different heights in the room. A total of 279 air velocity measurements were taken in the 

full chiller and 130 air velocity measurements were taken when the room was empty. In 

both cases results showed regions of high air movement located along the ceiling, walls 

and floor , whereas in the middle if the chiller, beneath the rails, the air movement was 

very s low. It was concluded that the area of high air movement can be considered 

turbulent and experiences only small velocity gradients. The area of low air movement 

could be considered to be non-turbulent with widely varying velocity gradients. 

At South Bank University the behaviour of thin delta plan aerofoils in air streams has 

been investigated. This was achieved by assessing the influence of strong leading edge 

vortices produced by the aerofoils on room air mixing (Missenden , 1987; Amraie, 1993). 

The influence of the vortices was examined experimentally using dummy carcasses to 

simulate the actual conditions inside a chill room. Air movement was investigated using 

the response of temperature sensors to rapid temperature changes of the inlet air stream. 

Visualisation of the air flow was achieved by using soap bubble generators. Experiments 

showed that these delta wings improved air movement between carcasses in conventional 

carcass meat chilling rooms and hence product quality was also improved (Missenden et. 

al., l 995). 



14 Chapter 2: Literature Review 

2.2 FACTORS AFFECTING AIR FLOW 

Air is a compressible fluid which when acted upon by a fan or other energy impelling 

means, flows from a region of higher absolute total pressure to one of lower absolute total 

pressure. The difference in pressures determine the characteristics of the flow. 

2.2. I FLOW ALONG A FIXED SURFACE 

When a fluid flows over a fixed surface, layers next to the surface are held back by the 

viscous forces and "stick" to the surface (John, 1969). This causes the velocity of the 

fluid at the fixed wall to be zero. Moving out from the wall, the velocity increases to its 

free stream value, and a velocity distribution is built up. The effects of viscosity are 

dominant in the region near the surface. Air viscosity is small, so the viscous effects are 

confined to a very thin layer at the surface called the 'boundary layer'. In the boundary 

layer, the velocity components are zero both normal and tangential to the wall surface. In 

many cases the boundary layer effects can be neglected in the analysis of air flows. 

However in cases such as the calculation of mass and heat transfer coefficients, the 

boundary layer effects are important. 

Air that flows along a fixed surface such as a wall or ceiling generates attraction forces 

between the surface and air. This phenomenon is known as the 'Coanda effect' (Amraie, 

1993). The Coanda effect contributes to the problem of air circulating around the walls of 

a chiller, giving rise to stagnant zones in the centre where good air movement is needed 

the most. 

2.2.2 FLOW OVER BAFFLES 

In chillers, baffles can be used to introduce an extra pressure drop into air pathways 

which would otherwise have less resistance to air flow. They disrupt the air flow and 

create turbulence, thereby enhancing air flow through pathways that the air does not 

preferentially move. In this way, baffles can be used to reduce the short circuiting of the 

air within a chiller (Cleland & Cleland, 1996). 



Chapter 2: Literature Review 15 

Figure 2.2: Flow over a fixed plate 

Laminar flow over an fixed object such as a baffle causes a three dimensional separated 

turbulent flow pattern as shown in Figure 2.2. Flow separation occurs at the sharp edges. 

The adverse static pressure gradient slows the air to a standstill and then reverses its 

direction. The separation of the flow causes a stream of detached vortices known as a 

"vortex sheet". These vortices are known to expand and decay under viscous action until 

they dissipate into random turbulence (Missenden, 1987). 

When the air flow is turbulent and there are many objects present, as in a loaded chiller, 

the detailed flow patterns that arise from air flowing over a baffle are difficult to describe 

(Bradshaw. l 97 l ). 

2.2.3 FLOW OVER AEROFOILS 

If a flat plate is inclined at an angle to a moving stream of air there will be a net force 

exerted on the plate by the fluid due to a region of low pressure developing above the 

plate and a region of high pressure forming below the plate. The vertical component of 

the force exerted on the plate is called the lift whereas, the horizontal component is 

known as the drag. 

The pressure difference between the surfaces causes an outward expansion over the lower 

surface which is balanced by an inward compression over the upper surface. This forces a 

lateral and upward circulation to occur at the tips where the upper and lower boundary 

layers meet (Figure 2.3). A trailing smface of discontinuity is formed as the extremities 

are rolled up. The resulting air flow pattern is called a Lanchester's vortex pair 



16 Chapter 2: Literature Review 

(Missenden, 1987). The vorticies are known to create high turbulence and a strong air 

mixing effect (Kuethe & Schetzer, 1961 ). 

Figure 2 .3: Flow over a inclined plate 

2.2.4 FANS 

In a chiller, air movement 1s required to transport heat from the product to the 

evaporators. For chillers and freezers, in particular, the amount of air movement is highly 

influential on the rate of heat removal from the product. Higher air movement allows 

chillers and freezers to be more compact and capital costs to be decreased, although at the 

expense of higher operating costs. However, there exists a limit where increasing the air 

velocity further leads to little benefit. Therefore, the amount of fan power used represents 

a compromise between capital and operating costs. As fans can typically provide over 30 

% of the total heat load in chillers and freezers, optimisation of air flowrate is crucial 

(Cleland & Cleland, 1996). 

Fans move air by the rotation of a number of aerofoils. Propeller fans are typically used 

for air circulation in chillers (Osbourne, 1977). These fans enable large volumes of air to 

be moved through the evaporator fins relatively economically and have a low capital cost. 

Air enters a propeller fan from all directions and is discharged mainly axially, but there is 

also some radial discharge. The velocity of the air stream drops rapidly as it leaves the 

outlet and spreads out at an angle between 15 and 20 degrees. When a resistance is 

imposed, such as ice formation on the evaporator, the air tends to flow back through the 

impeller. Because of this, propeller fans are not suitable for working against any 

substantial air flow resistance. 



Chapter 2: Literature Review 17 

2.3 IMPROVING AIR DISTRIBUTION WITHIN CHILLERS 

Improvement of air flow di stribution within chillers tends to be situation specific because 

of the large number of vari ations in design and operating conditions. Careful 

consideration should be taken in the early stages of design s ince retrofitting devices to 

chillers is often difficult and expensive. 

2.3. J CHILLER DESIGN 

In general , the poorest air distribution is found in long chillers in which the rails and the 

air flow are ali gned with the length of the room (Figure 2.4a). Air velocities of less than 

0 .1 mis have often been fo und between carcass sides in these circumstances (Garnett , 

1974). This design is considered to be outdated and is almost non-ex istent in the modern 

meat industry. 

In modern chillers, the most common design has the rails posi tioned along the length of 

the room and air is directed at right angles to the rails (Figure 2.4b). This design has the 

advantage of the air maintaining its momentum across the chi ller as the ai r onl y has to 

travel the width of the chill e r. Also air deflection is reduced as carcasses present their 

narrowest profi les to the air flow . Although this design offers some advantages, the 

ove rall air distribution still tends to be poor. In most cases the air flow pattern becomes a 

circular one, with the greatest proportion of air passing between the rails and ceiling in 

one direction and back between the floor and necks in the o ther. This gives rise to a 

'dead' centra l area where there is littl e air flow over the carcasses. In addition, it is the 

neck, not the heavy hindquarter which experiences the highest air flows (Garnett, 1974). 

A common method used to improve the ai r flow pattern produced from the standard 

design, is to use a fa lse wall or duct (Figure 2.4c) . This prevents air short circuiting from 

the evaporator outlet to the inlet, thus bypassing the carcasses. However, the problem 

still exists in that the bulk of fast moving air follow s the path of least resistance around 

the outside of the carcass stack . 



18 Chapter 2: Literature Review 

l• -----------..... --
i w 

Fig. 2.4a: Wall Mounted units 

Air flow along rails 

Fig. 2.4c: Ducted air flow system 

Figure 2.4: Design configurations of chillers 

Fig. 2.4b: Wall Mounted units 

Airflow across rails 

Systems which guide air down onto the rails usually employ a plenum ceiling with slots 

above the rails (Figure 2.5a, 2.5b), or have a series of turning vanes (Figure 2.5c). Both 

systems can lead to hygiene problems and difficulties in ensuring the system is delivering 

an even di stribution of air down between all the carcasses, however, these problems are 

not insurmountable (Garnett, 1974). The advantages of such systems is that the heavy 

hindquarter of the carcass is exposed to the fast moving air flow near the ceiling of the 

chiller giving faster chilling rates and also, when fine tuned, a more relatively air 

distribution can be achieved. 

A recent approach is to reverse the flow of air in a ducted system (Figure 2.5d). Air is 

discharged across the floor and flows up between the carcasses (Willix, 1993). This 

system can produce a substantially more uniform air flow compared to the standard 

design as carcasses create a plenum effect at floor level enabling very even air distribution 

between carcasses. Will ix ( 1993) suggests that a potential problem with this 

configuration could be condensation on the chiller rails and superstructure caused by 

moist, warm air coming off the carcasses. 



Chapter 2: Literature Review 

I 

Fig. 2.Sa: Plenumceilmg system 

Fig . 2.Sc: Wall mounted unit s 

with turning vanes 

Figure 2.5: Design configurations of chillers 

~ D ~ • ~ ' -,,r,-1 -,'1,,- ,-

Fig. 2.Sb: Slotted ceiling system 

Fig. 2.Sd: Reverse ducted sys tem 

19 

Will ix ( 1993) used a scale model to test a number of different chiller design 

configurations (details of this model are outlined in Section 2.4. /). Results suggested that 

the basic design with wall mounted evaporators and no air deflectors gave the poorest air 

distribution of the configurations tested, regardless of room size or room loading. 

Extensive by-passing of air from evaporator inlet to outlet occurred, causing warm air to 

be drawn up throu gh the product instead of the cold air being forced downwards. At 

ceiling level, the air was observed to be drawn back towards the evaporator inlet due to 

the high velocity by-pass effect. This caused considerable recirculation patterns and poor 

air movement between carcasses. 

Willix ( 1993) found that the ducted air flow system, where the air inlet is at floor level 

and air is discharged at ceiling level, showed some improvement to the standard chiller 

configuration. In this case less by-passing of the air from inlet to outlet was evident. 

However, even with this configuration the high velocity by-pass effect was still visible, 

giving rise to poor air circulation between carcasses. 

The plenum system gave the most uniform air distribution in a large IO rail room of the 

configurations that were tested. However, the standard configuration fitted with turning 

vanes gave the most uniform air distribution in a smaller 6 rail room. 



20 Chapter 2: literature Review 

The model showed that a reverse ducted air flow system produced an air flow pattern of 

superior uniformity in comparison to patterns resulting from the standard design with wall 

mounted evaporators and the ducted system. 

Early tests performed by Garnett (1974) at MIRINZ compared a standard chiller beef 

chiller with wall mounted evaporators and a chiller which directed air vertically 

downwards onto the hocks of carcasses using ceiling baffles. Results indicated that the 

average air flow over the carcasses for the vertical air flow configuration was 2-3 times as 

great at hock level and lifted by 1/ 3 between the middle of the sides. The average deep leg 

temperature was reduced by a further 4 °C after 16 hours of chilling and the average 

weight loss was reduced by 5% of the total weight loss . 

Hodgson ( 1971) assumed that a linear relationship existed between the air circulation rate 

and the average air velocity at rump height. Chillers were compared by considering the 

rate of air circulation required to give an air velocity at rump height of 0.75 mis per unit 

floor area. Hodgson found that a slotted ceiling system performed the best of all the 

configurations tested and required an air circulation rate of only 0.166 m3/s per m2, 

whereas, the plenum ceiling system, needed a high air circulation rate of 0.53 m3/s per 

m2. Both the standard configuration fitted with turning vanes and the ducted 

configuration, had an air circulation rate of about half of the plenum ceiling system with 

0.257 and 0.248 m3/s per m2 respectively. Hodgson concluded that, in general, air 

circulation rates above 0.25 m3/s per m2 of floor area are required to keep the air 

temperature values below 5°C during peak load conditions. 

2.3.2 AIR TURBULENCE IN CHILLERS 

Measurement of air velocities in industrial chillers taken by Kondjoyan and Daudin 

( 1995), demonstrated that mean air speed is only a very rough description of the flow. 

The air flow was found to include very rapid and large fluctuations spread over a few 

milliseconds which were superimposed on slow fluctuations spread over some minutes. 

Kondjoyan and Daudin ( 1995) found that if the flow contains only rapid fluctuations, 

turbulence is best described by the term turbulence intensity which is the standard 

deviation of the air speed fluctuations divided by the mean air speed. Turbulence 

intensity was found to affect the values of the heat transfer coefficients as much as the 

mean air speed. Increasing the air turbulence intensity from 6% to 31 % had the same 



Chapter 2: Literature Review 21 

effect on the chilling rate as increasing the air speed from 0.5 mis to 1.5 mis on a 90 kg 

pork carcass hindquarter. Both of these variations reduced chilling times by one hour, for 

an air temperature of I °C and a relative humidity of 80%. They concluded that as fan 

power is roughly proportional to the cube of air speed, it would be more profitable from 

an energy point of view to promote air turbulence rather than increase air velocity. 

2.4 VISUALISING AIR FLOW PATTERNS 

It is important to have a good insight into air flow within existing industrial chillers to be 

able either to modify operating conditions or to later improve the design of new chillers. 

Air flows in refrigerated rooms can be examined in a number of ways. The main methods 

used are: scale modelling (Edwards , 1979; Lovatt et. al., 1993), computational fluid 

dynamics (van Gerwen & van Oort, 1989; Wang & Visser, 1991 ), smoke or bubble 

generation, (Mueller, 1983; Amos, 1993 ; Missenden et. al., 1995), ribbons (Winstanley, 

1983) and anemometers (Hodgson , 1971; Oliver, 1986; Falconer, 1993; Daudin, 1992; 

Baleo, 1995). 

2.<+. J SCALE MODELLING 

The purpose of using small scale models 1s to be able to determine the effect of 

modifications in design or major changes to operating conditions without mcurnng 

unnecessary expense or risk. Scale models enable simple and accurate prediction of 

important criteria for the correct design of a full-scale installation. 

Edwards ( 1979) used the technique of dimensional analysis scale modelling to study air 

flow patterns in carcass freezers. This technique meant that when using one-quarter scale 

models, air velocity had to be four times the full scale speed to maintain dynamic 

similarity with full-size freezers. 



22 Chapter 2: Literature Review 

Willix (1993) and Lovatt et. al. ( 1993) used a 1/10, 2-dimensional scale model, with 

water as the flow medium , to investigate air flows in beef chillers. The model 

represented the cross-section of a chiller up to 10 m wide and 4.5 m high. Visualisation 

was achieved by using water as the flow medium and electrodes to generate hydrogen 

bubbles which were swept along by the flow. The model could be altered to represent a 

standard forced draft chiller with or without air deflectors, ducted and reverse ducted 

systems with or without air deflectors, as well as a plenum ceiling design, each with 

either six or ten rails. Results demonstrated that the model could provide a quick and 

inexpensive way of evaluating a wide range of alternative chiller configurations with a 

high degree of visual impact. 

2.4.2 COMPUTATIONAL FLUID DYNAMICS 

Computational Fluid Dynamics (CFD) is essentially the use of computers to solve the 

relevant fluid flow equations (Navier-Stokes equations) (Gigiel et. al. , 1994). Part of all 

CFD packages are graphic visual di splays of the flow data. These are made up from 

information about the velocity, temperature and pressure of the fluid at every point in a 

computational mesh. Although thi s information can be obtained by measurement on a 

phys ical model, it would take a very long time and the measurements must be repeated 

for each change in external conditions. Although it is time consuming to set up the initial 

model , once a CFD model is obtained for one set of conditions, little effort is required to 

change the model to generate another set of data for different physical conditions. This 

makes CFD a very powerful mathematical tool for the modelling of fluid flows. 

However, difficulti es arise in modelling flow around irregular shaped, 3-dimensional 

objects such as carcasses. In many cases complex objects must be approximated as a 

spheres, cylinders or boxes. Also, CFD is expensive due to the cost of the equipment, the 

software package, training courses and the large number of hours that are needed to 

develop expertise. The computing power needed is also relatively expensive. PC's can be 

used but are slow in many real applications. 

van Gerwen and van Oort ( 1989) produced a advanced computer model using CFD which 

simulated the flow and heat transfer phenomena in cold stores for seasonal storage of 

fruit. The model was ab le to calculate the air velocities and product temperatures , based 

on the cooler air fl ow rate. product properties and the geometry of the cold store. 



Chapter 2: Literature Review 23 

Wang and Visser ( 1991) simulated air flows in closed refrigerated rooms using CFD. 

The model described the flow in a closed room without air inlet and outlet openings, 

including detailed modelling of evaporators with circulating fans and various 

configurations of room loadings with boxes of produce. 

Amraie ( 1993) carried out a small CFD study for a loaded test room which simulated 

conditions inside a chiller. This was a two dimensional model with rectangular sections 

to represent carcasses. Predictions from the model showed that (i) about half the air 

stream short circuited back to the evaporator inlet, (ii) short circuiting took place for the 

evaporator if there was only a short distance between outlet and inlet and (iii) air 

movement amongst the carcasses was small but there was high air movement along floors 

and return walls. These predictions agreed with experimental air flow data obtained in a 

pig pre-chiller. 

2.4.3 SMOKE OR BUBBLE TECHNIQUE 

A common way of monitoring air flow in a system is to use smoke trails. Visualisation of 

air flows by the use of 'smoke' includes methods which make use of a variety of smoke­

like materials such as vapours, fumes and mists. Smoke flow visualisation techniques 

have evolved to a point where quantitative information can be obtained from the images 

of the smoke patterns using computers for image processing and lasers to illuminate the 

smoke (Mueller 1983). Less popular agents such as using bubbles for flow visualisation 

can also be effective (Missenden et. al. 1995). 

Amos et. al. ( l 993) carried out smoke tests to ascertain air flow patterns in a large 

horticultural coolstore. The tests were carried out when the coolstore was empty, the 

doors were closed and all fans were operating at low speed. Results of the tests indicated 

significant back mixing and short-circuiting of air. 

½issenden et. al. (1995) investigated room air mixing in a laboratory test chamber by 

using a bubble generator to visualise air flow patterns (as described in 2.2.3). The test 

room was fitted with bubble generators that had a capacity of 800 bubbles per second and 

produced bubbles of about 0.003 min diameter with neutral buoyancy. 



24 Chapter 2: Literature Review 

2.4.4 RIBBON OR TAPE TECHNIQUE 

Use of smoke or bubble generators is often impractical in premises used for foodstuffs so 

alternative methods must be employed. Attaching tapes or ribbons at strategic positions 

within the chiller makes it possible to build up a picture of overall air movement. 

Winstanley ( 1983) used tapes to detect ventilation deficiencies in a refrigerated room. It 

was found that a 2-3 inch strip of opaque polyethylene, quarter of an inch wide and the 

thickness used in a typical retailers' polybags, are suitable on most occasions. The tapes 

enable operators to see at a glance what is happening to air flow. 

A number of difficulties arise when using this technique. Firstly, the air speed must be 

greater than a threshold value to overcome the gravity force acting on the tape, which 

depends on the thickness of the tape and the type of material used. Secondly, the mean 

position of the tape is often difficult to assess due to the air flow fluctuations commonly 

found in chillers (Daudin and van Gerwen, 1996). 

2.4.5 ANEMOMETRY 

Anemometry is concerned with velocity measurement of fluid flow . There are many 

types of anemometers available for the measurement of air speed such as those 

employing hot-wire, vane, cup, laser or sonic techniques. However, of these types, the 

hot-wire or thermal anemometer has traditionally been the most accepted method for 

measuring the detailed movements of air flow (Fingerson 1994). Hodgson (1971), Oliver 

( 1986), Daudin ( 1992), and Falconer (1993) all successfully used hot-wire anemometers 

to measure air flows in refrigerated rooms in order to obtain an overall picture air flow 

patterns. 

For many years the hot-wire anemometer has been successfully used to measure turbulent 

fluctuations, because it can handle fast fluctuations with ease. In its simplest form, a hot 

wire anemometer consists of a small metallic wire, usually of platinum or tungsten, which 

is heated by an electric current that flows through it. If the wire is cooled by air flowing 

around it, its electrical resistance is reduced and this reduction in resistance can be related 

to the air velocity. 



Chapter 2: Literature Review 25 

The major limitation of most thermal anemometers (which is common to all devices that 

aim for point measurements) is their inability to measure the directional component of 

velocity. Thermal anemometers which can give a directional reading, utilise an array of 

six hot-wires and are thus expensive and cumbersome (Fingerson 1994). 

Three orthogonally mounted vane or propeller anemometers (x-y-z arrangement), can be 

used effectively to measure the speed and direction of air flows (Jamieson, 1996). This 

method requires the use of vectors to calculate the magnitude and direction of the 

resultant, giving the overall air velocity. The main advantage of this technique is its low 

cost. 

Since its introduction in 1964, the use of the laser velocimeter to measure the details of 

flowing fluids has expanded rapidly and has been used to visualise air flow patterns in 

refrigerated rooms. Baleo et. al. ( 1995) visualised the trajectories of injected particles in 

a refrigerated display case air curtain using a laser lighting system. 

The laser velocimeter uses Doppler-shifted light scattered from particles in the flow to 

deduce the velocity of those particles. The laser velocimeter can resolve directional 

components and with frequency shifting, can detect flow reversals (Fingerson 1994). 

However. this method of air flow measurement has a high capital cost. 

2.5 MEASURING AIR VELOCITIES IN CHILLERS 

Measuring air flow within an industrial chiller is often difficult for a number of reasons 

(Daudin & van Gerwen, 1996). Firstly, the air flow is very turbulent due to fans and 

obstacles. This means that the airs direction and speed fluctuates rapidly with time and 

therefore their mean values at one location is difficult to measure. Secondly, the mean 

flow direction varies in space and consequently anemometers that are directionally 

dependent cannot be used. The mean air velocity also varies widely in space, so that a 

known value at one or several points have little meaning, particularly if they are not 

averaged over a long period of time. Thirdly, air speeds within chillers are often very low 

and in some places below the threshold of many anemometers. 



26 Chapter 2: Literature Review 

2.5. J MEASUREMENT DURATION 

Daudin and van Gerwen ( 1996) carried out tests in chillers and dryers to evaluate how an 

average air velocity can be measured in turbulent and unsteady flows. Results suggested 

that if an accuracy of about 0.1 mis is to be obtained, measurements performed at a 

frequency of l Hz must be averaged at least over a 5 minute period of time (Figure 2.6). 

It was therefore concluded that obtaining an air flow profile in a chillroom using a 

standard averaging procedure will be a tedious and time consuming process. 

Daudin and van Gerwen ( 1996) suggest a different approach to assess air velocity 

distributions where the anemometers are moved slowly and measurements are recorded at 

regular intervals. The movement is achieved using either the conveyor in mechanical 

chillers or an independently operated truck. A discrete Fourier Transform is applied to 

the data in order to eliminate time-fluctuations and obtain the mean air velocity versus 

spatial co-ordinates. In this way air velocity can be measured at several thousands of 

points. 

>. 
Ll 1.3 
0 

c3 
> ..... 1. 1 

::: 
;:; 0.9 
E 

"'O 
0 0.7 
~ 

-u 
- 0 -::,::: , .) 
u 

0 60 120 180 240 300 360 420 480 540 600 

Time (s) 

Figure 2.6: Variation of the calculated average air velocity with time at 4 locations in the 

chiller where the air flow was turbulent and unsteady 

(Daudin and van Gerwen, 1996) 



Chapter 2: Literature Review 27 

Missenden ( 1987) used a similar approach to measure air movement within a test room. 

A remotely controlled vehicle was used to carry a pole on which thermistor probes were 

mounted. Air flows were measured at various positions on the track by controlling the 

vehicle and anemometers from outside the room. In this way any disruption to the air 

flo w caused by moving in and out of the room was avoided. Unfortunately, it would be 

difficult to implement this technique in an industrial situation where chillers are 

constantl y being loaded and unloaded, as fixtures needed to enable a tracked vehicle to 

operate would cause process problems. Also, industri al chillers are often tightly packed, 

making it difficult for a mechanical vehicle to move freely between carcasses. 

2.5.2 NUMBER OF MEASUREMENTS 

In an industrial batch chilling situat ion, measurements of air flow are usually performed 

over a fixed chilling period. Once the length of each measurement is decided, the period 

of the chi ll cyc le largely dictates the max imum number of measurements that can be 

obtained. 

Amraie ( 1993) measured air flows in an industri al pig pre-chiller which had a volume of 

87.7 m3. A total of 279 point measurements were taken in the full pre-chiller and 130 

point measurements were taken in the empty pre-chiller to produce an air flow profile of 

the room. 

2.6 DELTA WING VORTEX GENERATORS 

2.6. I VORTEX GENERATION 

A highly swept, thin delta wing shaped aerofoil produces strong vortex interactions off 

the leading edges of the wing. These vortices are used as a form of direct lift control in 

aerodynamics, enabling an aircraft to manoeuvre at high angles of attack without stalling 

(Rom, 1992). The most favourable case for vortex generation is where the axis is nearly 

paralle l to the translational motion and the diameter is large. This occurs over the surface 

of a highly swept wing at incidence to a flow fi e ld. Vortices are formed by the separation 

of the flow from the leading edge as the angle of attack is increased, as illustrated in 

Figure 2 .7 



28 Chapter 2: Literature Review 

Figure 2.7: Structure of flows over delta wings with sharp leading edges (Rom, 1992) 

As with a flat plate inclined to the air flow (Section 2.2.3), the low pressure above the 

wing draws the separated sheet down towards the surface and rotation commences 

(Missenden et. al. 1995). A stagnation point occurs where the vortex loses its structure 

and breaks up, or bursts into large scale turbulence of considerable characteristic length 

and penetration. This creates a strong mixing effect which can be applied to enhance heat 

and mass transfer rates. The position of the burst depends on the angle of incidence, a 

figure below 30° provides bursts beyond the leading edge. 

2.6.2 DELTA WING VORTEX GENERATORS FOR IMPROVED AIR MIXING 

Missenden (1987) investigated the use of thin delta plan aerofoils of small semi angles 

(degree of sweep) at large angles of incidence in an artificially ventilated room to assess 

the influence of strong leading edge vortices upon air movement (Figure 2.8). 

It was determined by computer modelling that the optimum wmg position within a 

ventilated room was within the fast moving diffuser outlet air stream. The model also 

showed that the conversion of the air stream to vortices increased dramatically as the 

wings approached the diffuser. Smaller wings allowed more intensive use of the thin fast 

moving part of the air stream but reduced the bursting and penetration distance of the 

vortex fragments. 



s 

Chapter 2: Literature Review 29 

C -max 

La= Semi angle (degree of sweep) 

S = Trailing edge semi span (m) 

Cmax = Maximum chord length (m) 

Mean Chord length (m), Cm= 2/ 3Cmax 

Aspect Ratio= 2S 

cm 

Figure 2.8: Delta wing geometry (Missenden, 1987) 

Test results from a ventilated room showed that the vortex generators reduced the overall 

measured room air movement, whether assessed as mean air speed, momentum rate or 

power. At the same time, fragments of the vortex were in the order of 3-4 maximum cord 

lengths (0.9-1.2 m) and directed downwards and downstream of the wings trailing edges. 

This effect was maximised with large numbers of small wings situated within the fast 

moving plug flow region of the air stream, close to the ceiling and to the diffuser outlet. 

Larger wings reduced the effect but provided a longer distance spread of the vortex 

fragments. 

Missenden concluded that experimentation together with computer models indicated that 

vortex generators could be expected to reduce the overall room momentum and kinetic 

energy. This would allow for higher air supply rates in air conditioned spaces before a 

condition of discomfort is reached. Also, locally perceived air movement is increased 

due to the fragments of fast moving burst vortices penetrating many cord lengths beyond 

the wings edges. 

Amraie ( 1993) considered improving room air movement and velocity distribution within 

meat chilling rooms by using delta wing vortex generators. He investigated the local air 

velocities inside a commercial chilling room both empty and filled with pig carcasses. 

Results showed that some of the local air velocities around the carcasses were very low 

and were considered to be detrimental to the quality of the meat, indicating substantial 

room for improving air distribution. It was concluded from this set of experiments that 



30 Chaprer 2: Literature Review 

the use of delta wings as baffles to convert the air flow from a unidirectional air stream, 

progressively into rotational and then turbulent multidirectional fragments of moving air 

directed into the occupied part of the room, could substantially reduce chilling times. 

An experimental simulation of the pig carcass chiller was carried out in a laboratory test 

room. The room had similar dimensions to the pig pre-chilling room which was 4.5 m 

wide, 5.5 m long and 3.0 m high. A dummy evaporator produced uniform air flow into 

the room through a baffled diffuser mounted high on one wall. The diffuser was 

constructed from plywood with 348 equally spaced holes and it produced a face velocity 

of 1.57 mis. A total of 40 'dummy' foam pig carcasses were hung in the test room at 

various positions to represent a fully loaded chiller. Air speed and temperature 

measurements were taken by thermistor probes positioned strategically in the test room. 

Two different sizes of wings were used, a small wing with a maximum cord length of 395 

mm and a semi-span of 20° and a larger wing with a maximum cord length of 1000 mm 

and a semi-span of 20°. Wings were trialed in two different configurations. The first 

configuration involved placing 4 small delta wings close to the diffuser, followed by 2 

large wings, then this was followed by two rows of 5 small wings (4, 2, 5, 5). The second 

configuration utilised 2 large wings mounted close to the diffuser, then a row of 4 small 

wings which was followed by two further rows of 5 small wings (2, 4, 5, 5). Different 

positioning of the delta wings indicated that the most effective configuration in room air 

mixing was the latter. Amraie concluded that the delta wings gave quantifiable increases 

in room air movement. 

2.7 RESEARCH OBJECTIVES 

It is evident from this literature survey that differences in opinion still exist regarding the 

optimum conditions required within chillers. However, researchers generally agree on the 

importance of reducing carcass temperatures rapidly once the slaughtering and dressing 

operations have been completed. It is also accepted that velocity and turbulence of air 

within a chillroom have a large influence on overall carcass chilling rates . Studies have 

shown that velocity distributions within chillrooms are generally far from the ideal 

situation of uniform air flow and, in contrast, there are often stagnant areas between 

carcasses where good air circulation is needed most. 



Chapter 2: Literature Review 31 

Low or stagnant local air velocities around carcasses are considered to be detrimental to 

the quality of meat. Air distribution systems that have been developed in the past have 

only been partially successful in that they have tended to cause hygiene problems which 

result from difficulties in cleaning or have involved a high capital cost. Therefore, further 

development of simple, low cost air distribution systems is required to prevent zones of 

low air flow and recirculation of air. 

The potential of delta wings as vortex generators has been shown by Missenden (1987), 

and Amraie ( 1993). However, delta wings have not been tested in industrial meat carcass 

chillers and the effect on rates of chilling and weight loss have not been examined. 

Therefore, the aim of this research is to further extend the work carried out by Amraie 

( 1993) in improving environmental conditions within chillrooms. This will be achieved 

by using delta wings to convert the unidirectional air stream produced by the evaporator 

fans, into turbulent, multidirectional fragments of moving air directed into the occupied 

part of the chiller. These fragments of fast moving air should increase air movement 

between carcasses and thus make the local environment more representative of the 

average room conditions , and in doing so improve the quality of the final product, as well 

as providing energy savings. 

The specific objectives were to: 

(i) Characterise air flow patterns in an empty industrial meat carcass chiller without 

delta wings. 

(ii) Characterise air flow patterns in the fully loaded meat carcass chiller without 

delta wings. 

(iii) Characterise air flow patterns in an empty industrial meat carcass chiller with 

delta wings installed and compare results to those found without delta wings. 

(iv) Characterise air flow patterns in the fully loaded meat carcass chiller with delta 

wings installed and compare results to those found without delta wings .. 

( v) Determine the effect of pitch , size and spacing of delta wings on the air flow 

pattern. 



32 

Chapter 3 

MATERIALS AND METHODS 

3.1 INTRODUCTION 

The aim of this study was to trial delta-wing vortex generators within an industrial meat 

carcass chiller and assess their performance in improving the air flow uniformity and its 

effectiveness adjacent to the product. 

Initial chiller characterisation was achieved by taking extensive air flow measurements 

while the chiller was both empty and fully loaded with carcasses. Delta wings were then 

installed and the modified chiller was re-characterised. Air flow data from the modified 

chiller was compared to data from the initial characterisation to determine if the delta 

wings induced any changes in air flow characteristics. 

3.2 CHILLER SPECIFICATIONS 

3.2. I CHILLER CONSTRUCTION 

Due to project time constraints only one chiller was studied in this research. The chiller 

was chosen to have a design typical of that used in the New Zealand meat industry with 

wall mounted evaporators and sandwich panel for walls and ceiling. The chiller is 

situated at Venison Packers Trading Limited, Fielding and had been constructed in May, 

1992. 

The basic dimensions of the chiller are shown in Figure 3.1. The overall dimensions are 

10 metres long by 5 metres wide and 3.53 metres high. Loading and unloading doors are 

located at opposite ends of the chiller and each are 1.2 metres wide by 2.48 metres high. 

The chiller utilises three evaporator units mounted along the length of the room at ceiling 

height with a distance of 510 mm to the wall on the suction side. Each unit is 360 x 3000 

mm long by 470 mm high with a 240 mm spacing between the units. 



I Loading door 
- - 1 t I 
I i 

~LJ----/ 
I 

le 1.2 ) 
I 5.0 

Chapter 3: Materials and Methods 

. . , 
I 
I 
I 

10.0 

Figure 3.1: Basic chiller dimensions 

33 

v-. 

The chiller contains six carcass rails. Rails are suspended from steel H-beams directed 

across the width of the chill er. The H-beams are supported on perpendicul ar stee l beams 

mounted onto the wa ll s. Full detailed dimensions of the chiller are shown in Appendix 

A. Figure A. I . 

3.2.2 CHILLER EVAPORATOR SPECIFICATIONS 

The ch iller houses three SP8/4/HG/S comprex evaporator un its supplied by Comprex 

Industries Limited, Feilding, New Zealand (Figure 3.2). The units employ six Ziel! EBM 

A350 fans for air circulation which are constructed from four 175 mm blades at a 30° 

pitch. The vol umetric flowrate through each unit is 12,200 m3/hr with fans operating 

with a face velocity of approximately 5 m/s and rotation speed of 1350 rpm. The power 

requirement for each fans motor is 650 watts. 



34 Chapter 3: Materials and Methods 

Figure 3.2: Evaporator Units 

3.2.3 CHILLER OPERATION 

Slaughtering is usually carried out between the hours of 6 a.m and 4 p.m. according to 

demand. Hot carcasses are weighed and loaded into the chiller after being slaughtered 

and dressed. During loading, refrigeration is turned off and doors are left open, but fans 

operate, circulating warm air at IO- l 5°C around the chiller. 

Dressed carcasses can exhibit large weight variations. The dressed weight of red deer 

carcasses range from 35 to 160 kg, whereas fallow deer carcasses range from 17 to 25 kg. 

The maximum capacity of the chiller is 150 medium sized red deer with an approximate 

dressed weight of 50-60 kg or 170 fallow deer with an approximate dressed weight of 

20 kg. However, the chiller can be loaded with a combination of red and fallow deer. 

Once slaughtering is finished, the chiller doors are shut and refrigeration is turned on. 

Under normal operating conditions carcasses are chilled for 14 hours, however in special 

circumstances carcasses might be left in the chiller for up to 62 hours. 



Chapter 3: Materials and Methods 35 

During chilling air temperature in the room is controlled to 2°C which is measured at the 

evaporator air-on position. Fans and refrigerant are turned off for 30 minutes at regular 4 

hour intervals for evaporator defrost. Under nom1al operation, humidity levels are 

initially about 90% relative humidity (RH) at the start of the chill cycle and fall to 

approximately 83 % before carcass unloading commences. During defrost periods, 

relative humidity rises to about 90% for a short time. 

After chilling is complete the cold carcasses are re-weighed and then boned out. 

Carcasses are removed in the same order that they were loaded into the chiller, starting 

from the first rai l and working back to the sixth rail. Fans and refrigeration remain on 

while unloading takes place. 

3.3 AIR FLOW MEASUREMENTS 

To measure air flow, two types of anemometers were used in this study: A Dantec 

FlowMaster 54N60 hot-wire anemometer was used for the purposes of measuring the 

mean, standard deviation , and range of the air speed over a set period; while three 

orthogonally mounted Gill propeller anemometers were used to measure air velocity 

(speed and direction ). 

3.3. I MEASUREMENT METHODS 

Air speed measurements made using the hot-wire anemometer were performed by two 

different methods: 

Method I: Air speed could be measured over a set integration period of 10, 60 or 180 

seconds and the mean and range of the air speed was read directly from the 

anemometers display. 

Method 2: The anemometer could be connected to a Grant 1200 series data logger via a 

RS232 cable which continuously records air speed data as a voltage at a 

frequency of 1 Hz over any measurement period greater than 1 second. The 

mean, range and standard deviation of the air speed was then calculated from 

this data. 



36 Chapter 3: Materials and Methods 

In using the first method, standard deviations for each integrated measurement was 

estimated from the range using tables from Kramer and Twigg ( 1970) (Section 3.3.2.8). 

For the second method, logged voltage data was firstly converted to air speeds, then the 

mean, range and standard deviation of the air speed for each measurement could be 

calculated directly from this data. All instrument testing was performed using the second 

method of data recording. The first method was used for some of the initial work 

performed in the chiller but was later substituted for the second method. 

The method used for measuring air velocity using the orthoganally mounted propeller 

anemometers was to connect each transducer to a Grant 1200 series datalogger. Voltage 

outputs from the anemometers were continuously recorded by the datalogger at a 

frequency of I Hz over the duration of each measurement. Voltage data was converted to 

air speeds which was then used to obtain a time-averaged value of the air velocity 

(direction and magnitude). 

For all air speed and velocity measurements taken between the heights of 1.0 m and 3.0 

m, both the hot-wire and orthoganally mounted propeller anemometers were attached to a 

2.0 m long stainless steel rod mounted on top of a camera tripod. Height adjustment was 

achieved by clamping the anemometers at different positions along the rod. For 

measurements below 1.0 m, anemometers were attached to a small laboratory clampstand. 

The hot-wire anemometer was positioned to the side of the propeller anemometers to 

avoid interference from the blades, but kept at the same height. Both the hot-wire and 

orthogonally mounted propeller anemometers attached to the clampstand are shown in 

Appendix A, Figure A.5. 

3.3.2 AIR SPEED MEASUREMENTS - INSTRUMENT CALIBRATION AND TESTING 

Before the hot-wire was used to characterise air flows within the chiller, a number of tests 

were carried out to assess its performance. 

3.3.2. I Calibration 

The Dantec FlowMaster 54N60 hot-wire anemometer was calibrated against a 

micromanometre (traceable to national standards) at the Meat Industry Research Institute 



Chapter 3: Materials and Methods 37 

of New Zealand (MIRINZ) in the range of 0-8 mis. The anemometer was found to have 

an R2 value of 0.9998 and an offset of +0.002 mis over this range. 

3.3.2.2 Correlation of Voltage Signal With Displayed Air Speed Reading 

In using the second method for the measurement of air speeds (described in Section 

3.3. J), logged voltages needed to be converted to time averaged air speed readings. The 

relationship between the continuous voltage output produced by the anemometer and the 

time averaged values was determined by taking replicate measurements at a range of air 

flow rates in a variable air speed tunnel (Figure 3.3). Measurements were performed 

using a frequency of I Hz over a duration of I minute. Data from this experiment is 

tabulated in Appendix R Table B. l. Additional photos of the air tunnel are shown in 

Appendix A, Figures A.6 and A.7. 

Te:;;t1rn.: 

nrra 

Gnllec! region 

• -----;,_---------
• ~---------•-------; 

~ 11-------~ 

HeatinQ element 

Variable ,-• I 
;~: speed fan 

Figure 3.3: Variable air speed tunnel 

• • • 

Figure 3.4 shows that the voltage signal was found to be directly proportional to the 

displayed air speed reading according to: 

V (mis)= 10.14S(millivolts) (3.1) 

Voltages recorded from the hot-wire anemometer by the Grant 1200 series data logger 

were converted to air speed readings using equation 3.1. 



38 Chapter 3: Materials and Methods 

4.5 
VJ 

E 4.0 
':,JJ 

3.5 c:: 
"t:l 

"' 3.0 2:: 
"t:l 2.5 0 
V 
c:... 2.0 er. 
.... 
"' l.5 

"t:l 
'-.) 
>. l.O 
"' c.. 0.5 VJ 

6 
() () 

() 0.05 0.l (J.l 5 0.2 0.25 0.3 0.35 0.4 0.45 

Voltage output signal (V) 

Figure 3.4: Correlation of hot-wire anemometers displayed air speed reading with time 

averaged values 

3.3.2.3 Transducer Housing Interference 

Air speed measurement errors caused by the protective housing around the anemometer 

sensors were investigated. The hot-wire anemometer transducer contains a velocity 

sensor and a temperature compensator made of nickel wire coils on the end of a 40 cm 

long shaft. Both the temperature compensator and the velocity sensor are encased in the 

framework of a protective metal housing (Figure 3.5). 

Figure 3.5 Hot-wire anemometer transducer 

Tests were carried out in the variable speed air tunnel using a constant air flow of 

approximately l rn/s (Figure 3.3). The transducer housing was rotated through I 80 

degrees with the shaft perpendicular to the direction of the flow. At each change in angle, 

the mean air speed reading was measured over a duration of I minute. 



Chapter 3: Materials alld Methods 39 

As the housing shape is symmetrical there were two possible orientation angles where the 

housing area presented to the air stream was at a minimum. These occmTed 180 degrees 

apart from each other. One of these positions was used as an initial starting point and was 

arbitrarily chosen to be at a O degree orientation angle. By assuming a 'true' air speed 

reading was obtained when the hot-wire filament was perpendicular to the air flow and 

the housing area presented to the air stream was at its least, errors associated with each 

angle of rotation could be determined. Experimental data from these tests is tabulated in 

Appendix B. Table B.2 . 

T ransducer housing angle to air flow 

;on -1: •I- -----+20--•=~-4+0----+~---s'-;-~-1+~0--~, 2-o-- -. - ,4+-J o~---,+~o---, ;o 

~ -20 t "·, 
OJ I 

_:; -30 -
'.J 

E. -40 -
z 

"' -50 -
C 

~ -60 t 
'.J -70 T 

t:f: -so I 
-90 

Figure 3.6: 

\ 
I 

\ 

• 
- - - - Symmetry line 

Anemometer air speed reading interference at various transducer housing 

angles to a ir flow. 

Figure 3.6 shows th at as the housing is turned past an angle of 45 degrees, measurement 

errors become obvious due to the filament being sheltered by the metal housing. Results 

indicate that significant errors in the air speed reading occur between housing angles of 

45 and 135 degrees to the air flow . Similar errors can be expected between angles of 225 

and 3 I 5 degrees due to the housing symmetry. These results were consistent with those 

of Daudin and van Gerwen ( I 996). 

To e liminate thi s potential source of measurement error within the chiller, all 

measurements were perfo1med by orientating the anemometer according to a red cotton 

tread attached to the end of the transducer. The thread gave an indication of the air flow 

direction around the sensors . This enabled the housing to be rotated to a position of least 



40 Chapter 3: Materials and Methods 

interference. Although, this method was relatively subjective it was adequate to enable 

the transducers housing to be kept within the critical range of ±45 degrees to the air flow. 

3.3.2.4 Measurement Time 

Air flow within a chiller is often turbulent and unsteady. In order to gain a better 

representation of air speed at any one position within a chiller, readings can be integrated 

over a set period of time. The longer the duration of the measurement the closer the mean 

air speed reading should be to the tme mean air speed. 

Tests were carried out to determine the measurement duration required to obtain an 

accurate estimate of the mean air speed in a fluctuating air stream. Four positions within 

the chiller with turbulent air flow were chosen and at each of these positions air speed 

readings were taken over a measurement duration of 5 minutes. Air speed data from 

these tests is tabulated in Appendix B, Table B.3 and the mean, range and standard 

deviation of the air speed calculated at various time intervals for each position is given in 

Appendix B, Tables B.4-B.7. 

Figures 3.7 to 3. IO indicate that the air flow in the chiller is made up of very rapid, large 

fluctuations. This agrees with work performed by Kondjoyan and Daudin ( 1995) in the 

chilling and storage of foodstuffs (Section 2.5. l Literature Review). These fluctuations in 

the air flow cause the calculated mean air speed value to vary with time. The longer the 

duration of measurement time, the less impact each fluctuation in air speed has on the 

calculated mean. 

When considering the duration of measurement time for taking mean air speed 

measurements, a 'trade-off has to be made between the length of time for each 

measurement and the quantity of measurements taken. According to results shown in 

Figures 3. 7 to 3. 10, the ideal measurement period should be at least three minutes in order 

to obtain an accurate representation of the true mean air speed at any one position within 

the chiller. However, a 3 minute measurement period would significantly reduce the 

number of readings able to be taken over a chill cycle. Therefore, a compromise had to 

be made. A measurement period of 1 minute was chosen to enable a satisfactory quantity 

of readings to be completed over the limited time available. 



Chapter 3: Materials and Methods 41 

3.0 

2.8 

2.6 

2.4 
Vl 

E 2.2 

-::, 
2.0 C) 

CJ 
c.. . .,, 

1.8 

I 
I 

~~I J ~ 1.I u ¼ ii ~. 

N 
I 

~ I I r , 
r1 

f 
--Air speed 

<( 
1. 6 

--Mean air speed 

1.4 

1.2 

1.0 

0 30 60 90 120 150 180 2 10 240 270 300 

Measurement duration (sec) 

Fi gure 3.7 : Variation of the calculated average air speed and actual air speed with time 

for position I. 

7..7 ·-· 

:2.6 

2.5 

Vl l 
---:::: 

:2.4 i 
-::, 
u 

' V 
i,.r '..J ~ c.. 2.3 Y) --Air speed 

<( --Mean air speed 
') ') 

2.1 

2.0 

0 30 60 90 120 150 180 2 10 240 270 300 

Meas urement duration (s ec) 

Figure 3.8: Variation of the calculated average air speed and actual air speed with time 

for position 2. 



42 Chapter 3: Materials and Methods 

3.4 

3.2 

3.0 

VO 

2 2.8 
-::, 

c.) 
c.) 
c.. 2.6 z 

·-<( 

2.4 

I 

11. A II \ I , \ V ijl'ff ~ ~ n '' 1--Air speed 
i . 1-Mean air speed 

2.2 

2.0 

0 30 60 90 120 150 180 210 240 270 300 

Mcasurcrrent duration (sec) 

Figure 3.9: Variation of the calculated average air speed and actual air speed with time 

for position 3. 

1.0 

0.9 

0.8 

0.7 
~ 

Cf, 

E 0.6 

23 0.5 
c.) 
c.. 
V) 0.4 ;.... 

<C 
(l.3 

0.2 

~ 

'm 
0.1 

0.0 

0 30 60 

i 
I 

I 

l 
I ,, 

V 

i ~ ) ~ I 
90 120 150 180 210 240 270 300 

Mcasurerrent duration (sec) 

1--Air speed 

/-- Menn air speed 

Figure 3.10: Variation of the calculated average air speed and actual air speed with time 

for position 4. 



Chapter 3: Materials and Methods 43 

Figure 3.1 1 shows the maximum errors obtained for the mean, standard deviation and 

range of the air flow for different measurement periods over the four positions (Data is 

given Appendix B, Table B.9) . 

0 .70 
~ 

~ 0.60 
C \\ ;:: 0.50 
i3 
Q.) 

:::l 0.40 
0 
'/} 

0.30 ..c 
"' 
:::l 0.20 
- X 
X: 

"' 0. 10 
2: 

0.00 

0 

• 

60 

• • 

X 

120 180 

Meas urement duratio n (s ec) 

• 

X 

240 300 

X M ean 

• Range 

4 Stdev . 

Figure 3.1 1: Maximum errors in mean estimates of statistical parametres for 4 positions 

within the chiller with turbulent air flow. 

For a measurement duration of I minute, the maximum errors for the mean and standard 

deviation were considered to be acceptable at 0.038 m/s and 0.017 m/s respectively. 

However, the estimated average error for the range over this time was significant at 0.436 

mis. 

3. 3.2.5 Estimarion of Measurem.ent Error Caused By Random. Fluctuations in Air Flow 

Random air flow fluctuations within the chiller cause slightly different estimates of the 

mean, range and standard deviation of air speed to be obtained for replicate 

measurements . The uncertainty associated with each reading was estimated by comparing 

a number of air speed measurements taken in the same location within the chiller. 

Using the same data collected for the estimation of measurement time, the mean, range 

and standard deviation of the air speed were calculated over a I minute measurement 

duration for the four positions discussed in Section 3.3.2.4. This resulted in five l minute 



44 Chaprer 3: Materials and Methods 

readings being obtained for the three parametres at each position. The range of these five 

replicate readings was used as an estimate of the uncertainty associated in the air speed 

measurements. The data from this experiment is tabulated in Appendix B, Table B.9 

and B.10 and is summarised in Table 3.1. 

Position l Position 2 Position 3 Position 4 

Mean Range Stdev. Mean Range Stdev. Mean Range Stdev. Mean Range Stdev. 

(m/s) (m/sJ ( rn/sJ (m/s) (m/s) (m/s) (m/s) (m/s) (mis) (mis) (mis) (mis) 

Range 0.182 0.619 0.094 0.051 0.122 0.024 0.040 0.304 0.040 0.078 0.345 0.072 

Table 3.1: Range between statistical parametres for 5 air speed readings taken m 4 

different positions within the chiller. 

Table 3.1 shows that the maximum range over the 5 replicate mean air speeds readings 

was 0.18 m/s which was considered to be acceptable. It should be noted that this does not 

take into account errors associated with deviations in probe placement. 

3.3.2.6 Attainment of'Steadv State Conditions 

The major concern with the attainment of steady state air flow conditions is not the length 

of time that the chiller is left before measurements commence, since air flow steady state 

conditions are attained very rapidly. Instead the main problem is to ensure that steady 

state is re-established before each measurement is taken, as the air flow is disturbed each 

time the anemometer is re-positioned. 

There are four distinct steps (phases) in taking a measurement in the chiller: 

Step 1 (phase 1 ): Entering the chiller to position transducer ( opening and closing chiller door). 

( 15 seconds approx.) 

Step 2 (phase 2): Positioning the transducer at the correct measurement site within the chiller. 

(30 seconds approx.) 

Step 3 (phase 3): Exiting the chiller (opening and closing chiller door). 

( 15 seconds approx.) 

Step 4 (phase 4): Operating the anemometer from outside the chiller (taking the measurement). 

( 60 seconds) 



Chapter 3: Materials and Methods 45 

Every time the chiller door is opened or the transducer is re-positioned to a new 

measurement site, there is a substantial disruption of the air flow within the chiller. If 

steady state conditions are not quickly re-established again after the disruption, an 

inaccurate estimate of the air flow parameters at the next measurement position will 

result. Therefore, it is critical that air speed measurements are taken while the chiller is 

under steady state conditions. Tests were performed to assess whether steady state 

conditions are re-established before the next measurement commences. Air speed data 

was continuously logged while performing five air flow measurements of 1 minute 

duration at different positions within the chiller. Air speed data is tabulated in Appendix 

B, T able B.11. 

Figure 3.12 indicates that as soon as the transducer is in position and the vicinity around 

the probe is vacated, steady state conditions are almost immedi ately re-established. The 

openi ng and clos ing of the chill er door does not seem to greatly di srupt the air fl ow 

amongst the carcasses, as the carcasses buffer any disturbance. 

I 2 3 -+ 2 3 4 I 2 3 4 I 2 3 4 I 2 3 4 
2.-l 

2.2 

1.0 

1. S 
'JC 1.6 

------
1.-l 

"3 I .2 '.J 
c.. 
,;: 1. 0 --

<( O.S 

0.6 

0.-l 

0. 2 

0.0 
0 30 60 90 120 150 180 2 IO 24 0 270 300 330 360 390 420 450 480 5 10 540 570 600 

Time (s) 

Figure 3.12: Continuous logging over 5 measurements of air flow at different positions 

within the chiller 

I. - Phase I: Entering chiller (opening and closing door) (-15 sec. ) 

2. - Phase 2: Pos itioning of anemometers transducer (-30 sec. ) 

3. - Phase 3: Exiting chiller (opening and closing door) (- 15 sec.) 

4 . - Phase 4: Measurement time (60 sec.) 



46 Chapter 3: Materials and Methods 

3.3.2. 7 Estimation of the Air Speed Standard Deviation 

For air speed data recorded by the first method ( described in Section 3.3.1) the standard 

deviation needed to be estimated from the range by: 

s = standard deviation 

R = Range 

d2 = Factor for estimating s from R 

(3.2) 

The factor c/2, depends on the sample size. Using a measurement duration of 1 minute at 

a frequency of I Hz produces a sample size of 60 data values. The factor d2, for a sample 

size of 60 is 4.639 (Kramer and Twigg, 1970). 

This approximation assumes that the air speed data is normally distributed about the 

mean. This assumption can be tested by investigating the distribution of a number of air 

speed measurements. The distributions of the 60 air speed readings for 4 measurements 

are presented in Appendix B, Figures B.1-B.4. The data for these distributions is taken 

from Appendix B, Table B.3. 

Distribution plots indicate that the air speed readings approximate a normal curve thus 

validating the use of Equation 3.2. However, it was shown in Section 3.3.2.4 that there 

are significant measurement errors involved in estimating the air speed range over a l 

minute period. Therefore, estimations of the standard deviation by this method should be 

treated with caution. 

3.3.3 DIRECTIONAL AIR FLOW MEASUREMENTS - INSTRUMENT CALIBRATION 

AND TESTING 

3.3.3. J Calibration 

The three Gill propeller anemometers used in this study were acquired from Works 

Consultancy Services, Wellington. Each anemometer consisted of a 4 bladed, polystyrene 

propeller with a diametre of 22 cm attached to a 30 cm long, slim generator. 



Chapter 3: Materials and Methods 47 

Anemometers were individually calibrated at Works Consultancy Services Limited in the 

range of 0.5 to 5 mis before use. Each anemometer was adjusted to produce an output 

voltage signal proportional to the speed of the propeller according to: 

Vtrue (ml ) = 17.54 Volts + 0.14 (R2 = 0.9999) (3.3) 

For values below 0.5 mis air speeds were approximated by: 

V(mls)"" 17.82 Volts (3.4) 

3.3. 3.2 Response of a Single Anemomerer at Various Orientation Angles to the Air Flow 

A series of tests were performed to assess whether three orthogonally mounted propeller 

anemometers could be used as a reliable directional air flow indicator. Firstly, the 

response of a single anemometer output signal was investigated at different blade angles 

to the air flow. Ideally, the output response from the anemometer will be equal to the true 

air speed multiplied by the sine of the angle at which the air stream meets the 

anemometers blade. 

[deal response (mis) = (Air speed) sin(blade angle to air flow) (3.5) 

The closer the anemometer response follows the ideal response, the more accurately the 

air flow direction can be calculated by the use of air velocity vectors. 

One of the propeller anemometers was chosen arbitrarily and tested in an air tunnel to 

assess the response of the blades at various angles to an air stream. It was assumed that 

the results of these tests would be applicable to the other two anemometers. The 

anemometer was positioned in the air tunnel and the blade tilted in 10 degree increments 

through 90 degrees. At every change of angle of the blade, the response from the 

anemometer was recorded over a duration of 1 minute. The converted air speed readings 

were checked against the 'ideal' air speed reading that should be obtained at that particular 

blade angle (Equation 3.5). Six experimental runs were done using different tunnel air 

flow rates for each run. Tunnel air speed was adjusted from approximately 0.3 mis 

through to 2.0 mis by a variable speed fan. This corresponded to the range of air flow 



48 Chapter 3: Materials