ABSTRACT
This study investigated the modelling of soft sensors for the prediction of turmeric thermal properties using data-driven methodology. The study examined the effect of drying time, drying temperature and air velocity during turmeric drying using exhaustive search technique; estimated the thermal properties of dried turmeric rhizome using existing empirical relations; developed soft sensors using Artificial Neural Network (ANN), Regression Tree (RT), Support Vector Machine (SVM), Gaussian Process Regression (GPR) method for the prediction of the thermal properties; and statistically compared the goodness of the models and select a model with better prediction. Proximate composition analysis was conducted for each of the dried samples of the turmeric to determine the nutritional composition. The soft computing methods were deployed in estimating specific heat, thermal conductivity, and thermal diffusivity of the dried turmeric using four input variables time, temperature, air velocity, and relative humidity individually and collectively. Two hundred and ninety-five (295) data set out of the three hundred data set obtained from the experiment, were used to develop, train and test the models using five-fold cross-validation with five (5) of the remaining data set aside and used for independent validation of the predictive model result. The average nutritional composition of the dried turmeric rhizomes were crude fibre (2.9%), crude protein of 4.22, and carbohydrate of 33.56%. Other nutrients include nitrogen 4.22%, ash 1.6%, and fat 2.9%, with a moisture content of 4.4% and 40.4% dry matter. The result of the model indicated that the square exponential of the GPR models has the best convergence for specific heat with the combination of all the input variables. Quadratic SVM have the best prediction for thermal conductivity with the combination of all input variable. Matern S/2 with all inputs is the model with the best estimation of specific heat, having an MSE of 0.000164 and R2 of 1. Quadratic SVM with all inputs best estimate the thermal conductivity with R2 of 0.98 and MSE of 0.0000864. Fine Gaussian SVM is the model with the best estimate for ther9mal diffusivity having using the input variables of Time, Air velocity and temperature having MSE and R2 values of 0.00037461 and 0.09, respectively. The study concluded that ANN has the best prediction for thermal properties for a single input, whereas, for all input variants, the models differ in their estimation capabilities.
TABLE OF CONTENTS
Cover page
Title page i
Declaration ii
Dedication iii
Certification iv
Acknowledgement v
Table of contents vi
List of tables x
List of figures xi
Abstract xii
CHAPTER 1
INTRODUCTION
1.1 Background of the study 1
1.2 Statement of problem 3
1.3 Aim and objectives 4
1.3.1 Aims of the study 4
1.3.2 Objectives of the study 4
1.4 Justification of the study 4
1.5 Scope of the study 5
CHAPTER 2
LITERATURE REVIEW
2.1 Turmeric rhizome 6
2.2 Drying 7
2.2.1 Proximate composition analysis 10
2.2.2 Thermal properties 10
2.2.2.1 Specific heat 11
2.2.2.2 Thermal conductivity 13
2.2.2.3 Thermal diffusivity 14
2.3 Soft sensors 15
2.3.1 Artificial neural network (ANN) 16
2.3.1.1 Neuron modelling 17
2.3.1.2 Architecture 18
2.3.1.3 Learning process 20
2.3.3 Root mean square error (RMSE) 21
2.3.4 Support vector machine (SVM) 22
2.3.4.1 Support vectors 23
2.3.4.2 Kernel Machines 23
2.4 Gaussian regression model 24
2.5 Regression
tree model
29
2.6 Statistical analysis 30
2.6.1 Statistical error 30
2.6.2 Cross- validation 32
2.6.3 Data set split 33
CHAPTER 3
MATERIALS AND METHOD
3.1 Simple collection 35
3.2 Experimental drying procedure 35
3.3 Proximate composition analysis 36
3.3.1 Determination of crude fat 37
3.3.2 Determination of ash content 37
3.3.3 Determination of crude fibre 38
3.3.4 Determination of total moisture content 39
3.3.5 Determination of crude protein 39
3.4 Thermal properties calculation using
empirical relationships 41
3.4.1 Specific Heat 41
3.4.2 Thermal conductivity 41
3.4.3 Thermal diffusivity 41
3.6 Model development and validation 42
3.6.1 Artificial neural network (ANN) 42
3.6.2 Regression tree (RT) 43
3.6.3 Support vector machine (SVM) 43
3.6.4 Gaussian process regression (GPR) 45
CHAPTER 4
RESULTS AND DISCUSSION
4.1 Proximate composition of dried turmeric 47
4.2 Summary of experimental statistic 48
4.3 Soft sensor prediction for specific heat 49
4.4 Thermal conductivity 51
4.5 Thermal diffusivity 54
CHAPTER 5
CONCLUSION AND RECOMMENDATION
5.1 Conclusion 58
5.2 Recommendation 59
5.3 Contribution of knowledge 59
REFERENCE
APPENDIX
LIST OF TABLES
Table Title Page
4.1 ` Proximate
analysis of dried turmeric rhizome 45
4.2 Summary
of experimental statistics 46
4.3 Prediction
of specific heat using artificial neural network 47
4.4 Prediction
of specific heat using regression tree 47
4.5 Prediction
of specific heat using support vector machine 47
4.6 Prediction
of specific heat using Gaussian process regression 48
4.7 Best
and least five rank models for estimation of specific heat 49
4.8 Prediction
of thermal conductivity using artificial neural network 50
4.9 Prediction
of thermal conductivity using regression tree activity 50
4.10 Prediction
of thermal conductivity using support vector machine model 50
4.11 Prediction
of thermal conductivity using Gaussian process regression 50
4.12 Best
and least five rank models for estimation of thermal conductivity 52
4.13 Prediction
of thermal diffusivity using artificial neural network 52
4.14 Prediction
of thermal diffusivity using regression tree model 52
4.15 Prediction
of thermal diffusivity using support vector machine 53
4.16 Prediction
of thermal diffusivity using Gaussian process regression (GPR) 53
4. 17 Best
and least three models for estimation of thermal diffusivity 54
LIST
OF FIGURES
Figure
Title Page
2.1 Turmeric
rhizome 6
2.2 Typical
drying rate curve under constant drying condition 8
2.3 Schematic
diagrams of biological neurons 16
2.4 Mathematical
modeling of a neural network 17
2.5 NN
architecture of feed forward and feedback neural network 18
2.6 SVM
Hyper plane separation 21
2.7 Support
vector machine architecture 22
3.1 Diagrammatic
representation of the tray dryer 34
CHAPTER 1
INTRODUCTION
1.1 BACKGROUND OF THE STUDY
Turmeric is from the ginger family of Zingiberaceae,
primarily cultivated in tropical and sub-tropical regions of the world
(Jeevarathinan and Pandiarajan, 2016), with cultivation increasingly notice in
Nigeria (Nasri et al., 2014). The
rhizome mainly grows in tropical and sub-tropical areas such as India, Jeva,
China, Taiwan, Bengal, Australia and Thailand (Suresh et al., 2009). Turmeric is one species that can act as an antioxidant
and anti-carcinogenic substance that can prevent cardiovascular diseases
(Prathapan et al. 2009). The ability
of turmeric to fight diseases is due to the presence of the curcuminoids in the
rhizome (Maheshwari et al. 2006;
Abdel et al., 2010). Hence, research
has shown that curcuminoids are non-toxic even at a higher dosage, and
therefore a safe ingredient in medicines and cosmetics (Goel et al. 2008).
The presence of moisture in the rhizome,
which is about 5-7% (Jeevarathinan and Pandiarajan, 2016), leads to the
development of microorganisms that can deteriorate the rhizome reducing the
effectiveness of the rhizome due to bioactive compound depletion. It is
imperative to preserve the bioactive compound through the oldest means of
preservation. Drying is primarily the removal of moisture content because of
phase changes using thermal energy.
Drying as
a complex operation is one of the oldest unit operations consisting of
subliminal processes such as a change in quality, shrinkage, multicomponent
moisture transport, chemical/ biochemical reactions, phase change, mass
transfer (Mujumdar, 2011). The use of hot air drying such as tray dryers,
microwave ovens and many more have become the solution of choice to most
agricultural drying operation for better product quality, reduction in weight
and volume of product, better preservation, ease of packing, storage and
transport (Samira et al., 2015).
Thermophysical properties are the properties
of a material exhibited by a material when the heat is passed through it. They
are those properties of a material calculated to determine the performance
parameters such as the material's heat transfer coefficient and energy
efficiency (Mahbubul, 2019). The fundamental thermophysical properties are
specific heat capacity, thermal conductivity, thermal diffusivity, coefficient
of linear thermal expansion, vaporisation heat, and heat of combustion (Neikov,
2019). The understanding of thermal conductivity, thermal diffusivity, and
specific heat of materials allow process engineers to design better pulverising
and drying equipment (Jeevarathinan and Pandiarajan, 2016). It also allows for
understanding material performances when subjected to the heating process. Food
substances thermal property understanding is also vital for the design of heat
transfer, dehydrating and sterilising equipment (Kaleemullah and Kailappan,
2006).
Thermal properties such as specific heat can
be used to calculate the heat load imposed on processing equipment(Shah
et al., 2018; Brian et al., 2001). Thermal Conductivity shows that heat flow through a
material depends on the material's temperature, porosity, and composition (Denis,2010),
indicating that material structure dramatically affects the materials' thermal Conductivity
(Rao et al., 2005).
Transient movement of heat measures the capacity of a
material to conduct thermal energy relative to its ability to store thermal
energy (Shah et al., 2018).
Physically, thermal diffusivity measures how fast the temperature of a
materials changes when heated or cooled (Dennis, 2010).
Soft sensors are valuable inferencing tools that use
easy-to-measure variables (temperature, pressure, time and many more) to inference
process output (Saptoro, 2013). Hence, it is essential in the mathematical
modelling of processes. There are two classifications of soft sensors:
model-driven sensors, also known as first principle models, these types of
models are based on fundamental and dependent on complete physicochemical
knowledge of the process, which is often unavailable(Saptoro,2013). In
contrast, the data-driven soft sensors depend on data obtained within the
system. The goal of developing soft sensor models is to predict
difficult-to-measure process variables. Hence recognised operating modes could be
modelled with appropriate local models (Ge et
al. 2011; Xiong et al. 2005).
Common
soft sensors used by researchers in modelling by various authors include principal
component analysis (PCA), partial least squares, artificial neural networks,
Neuro-Fuzzy .systems, Gaussian Process Regression (GPR) (Grbi'c et al. 2013), Particle Swarm
Optimization (PSO), Regression Tree and Support Vector Machines (SVM) (Kadlec et al. 2009). This study modelled the
thermo-physical properties of dried turmeric in a tray dryer using a data-driven
approach. The following data-driven soft sensors are employed in this study;
i.
Artificial Neural Network
(ANN),
ii.
Regression Tree (RT),
iii.
Support Vector Machine (SVM)
iv.
Gaussian Process Regression
1.2 STATEMENT OF PROBLEM
The
thermo-physical properties of materials, especially food material, are of
considerable importance in food process industries. These thermo-physical
properties control the transfer of heat in food material. These properties are
relevant for designing and predicting heat transfer operations during handling,
processing, canning, and food distribution. When introduced in controlling
these properties during food processing, human errors can cause losses and
sometimes litigations.
Every
process requires a controller in order to regulate the operation of the system.
The use of conventional controllers has been in existence since the ancient
times, for the control of unit operations in the industries. These conventional
controllers were developed using rigrous mathematical equations which are
cumbersome, rigid and may not adapt to the system which can result to
re-modelling of the mathematics. Thus in order to reduce the rigrous nature of
developing a controller, fuzzy based controllers (soft senors) were developed.
Soft
sensors models are developed to be used for artifical based controllers. Soft
sensors models include Artifical Neural Network (ANN), Adaptive Neuro Fuzzy
Inference System (ANFIS), Support Vector Machine (SVM), Regression Tree (RT), Gaussian Process Regression
(GPR). These models are adaptable
to user language and flexible. Hence the use of soft sensors can be adopted to
eliminate errors involved in the estimation of these properties.
1.3 AIM AND OBJECTIVES
1.3.1 Aim of the Study
To
develop soft sensor models that predicts the thermal properties of dried
turmeric rhizome in a tray dryer.
1.3.2 Objectives of the Study
a.
To investigate the effect of drying time, drying temperature and air
velocity during turmeric drying using exhaustive search technique
b.
To estimate thermal properties of dried turmeric rhizome using existing
empirical relations
c.
To develop soft sensors using Artificial Neural Network (ANN),
Regression Tree (RT), Support Vector Machine (SVM), Gaussian Process Regression
(GPR) method for the prediction of the thermal properties.
d.
To statistically compare the goodness of the models and select a model
with better prediction.
1.4 JUSTIFICATION OF THE STUDY
The use of conventional controller for the
control of processes has always involved cumbersome mathematics which are
rigrous, ridig and not adaptable. The motivation for this study is to develop flexible
and easy soft sensor models for the prediction of thermal properties of
turmeric rhizome in tray dryers using easy to measure process parameters such
as drying time, temperature, air velocity. The adoption of these artifical
based controller for the control of process. The most common model variants are
selected to show that no specific specialisation is needed to fit these models related
to process engineers.
1.5 SCOPE OF THE STUDY
This study investigated
the effect of drying
time, drying temperature and air velocity on turmeric drying. The effect of the
drying process on the nutritional composition of the turmeric was also carried
out using proximate composition analysis. The empirical relationship developed
from the drying parameters and the nutritional composition were used to model
soft sensors; Artificial Neural Network (ANN), Gaussian Process Regression (GPR),
Support Vector Machine (SVM) and Regression Tree (RT) for the prediction of
specific heat, thermal conductivity and thermal diffusivity of the dried
turmeric rhizome.
Login To Comment