AC 2007-2729: ADVANCED MODELING IN BIOLOGICAL ENGINEERING

USING SOFT-COMPUTING METHODS

George Meyer, University of Nebraska-Lincoln

GEORGE MEYER, Professor, has taught graduate and undergraduate classes that involve plant
and animal growth and environmental factors, modeling, and instrumentation and controls for
both agricultural and biological systems engineering students for 28 years. He has received
national paper awards and recognition for his work in distance education and has received
university teaching awards. His current research include measurement and modeling of crop
water stress, fuzzy logic controls for turf irrigation management, and machine vision detection,
enumeration, and plant species identification for spot spraying control and precision agriculture.

David Jones, University of Nebraska-Lincoln

DAVID JONES, Professor, has taught graduate and undergraduate classes that involve fuzzy set
theory and soft computing techniques, risk assessment of complex systems, and mathematical
modeling of physical and biological systems for the past 18 years. He also teaches a Heat and
Mass Transfer course to engineering juniors and the senior design classes. He has received
numerous university and national awards for his teaching excellence.

Page 12.184.1

© American Society for Engineering Education, 2007

 Advanced Modeling in Biological Engineering
Using Soft-computing Methods
Abstract

A new engineering graduate course on advanced modeling techniques and

applications provides both basic and practical understanding of techniques for
simulating biological and environmental processes to future scientists and
research engineers. Of particular importance are those models that benefit with
soft computing methods. Simulation of biological and environmental systems
involves the treatment of vagueness, uncertainty, and incomplete information
usually associated with these systems. A primary course emphasis was the
inclusion of fuzzy set theory and the positioning of fuzzy set theory (FST) within
a broader topic of soft computing. At the conclusion of the course, students had
developed their own paradigms and semester projects related to their particular
research interest. Students made use of current literature for theory formation and
hypothesis building related to biological and environmental systems. Future
researchers must effectively use methods to simulate ambiguous systems for
directing limited resources toward the solution of these problems. Principle
course topics included fuzzy variables, inference systems, neural networks, signal
processing, controls, visual simulation, machine vision, and genetic algorithms in
support of modeling. Students were expected to read and critique related journal
articles each week. To enhance communication skills, students lead selected class
sessions by discussing and critiquing refereed articles related to soft computing
and modeling, especially within their chosen research areas. Students learned
practical modeling skills using MATLAB® , MATHCAD®, and LabVIEW®
programming exercises. This paper discusses the course content and topics
presented, and how the course continues to evolve. A summary of student
projects and results are also presented.

Keywords: Courseware, biological systems, modeling, fuzzy systems, optimization

Page 12.184.2
Course Concept

A new 3-credit graduate level course addressing advanced modeling in biological engineering
was approved in 2006 and is now being offered at the University of Nebraska over a 15-week
semester. Preliminary versions of this course had been offered for three years prior as a special
problems class, for only a handful of selected graduate students. This past fall, eleven students
from four discipline areas took the class. Three were masters, while nine were doctoral students.
Ten students completed the class while one took an incomplete. The objective of this course is to
provide students with a basic and practical understanding of the use of modeling techniques for
simulating biological, environmental, and associated engineering systems.

Primary emphasis included the application of fuzzy set theory and the positioning of fuzzy set
theory (FST) within a broader topic of soft computing. At the conclusion of the course, students
were to be able to (a) develop and test FST and apply selected FST models to their particular
research interest, (b) use current modeling literature related to biological and environmental
systems for theory formation and hypothesis building, and (c) effectively use these methods to
simulate systems for understanding and solving new research problems. Supporting topics
included neural networks, machine vision, and genetic algorithms. Table 1 shows the syllabus of
course topics by number. Appendix A gives the university catalog description. Of particular
concern is the treatment of uncertainty and incomplete information associated with biological
and environmental systems.

Fuzzy Logic Modeling

The universe of discourse of available information in the biological world includes both precise
and ambiguous information. An approach is the use of FST. The concept of FST is attributed to
Zadeh1,2, who developed this theory to address uncertainty, ambiguity, and vagueness for pattern
recognition and classification problems. Fuzzy sets are defined with degrees of membership for
linguistic variables. FST is in contrast to crisp or binary memberships in classical set theory
where no uncertainty is taken into account. Fuzzy logic (FL) for solving problems through
modeling uses either a rule-based or fuzzy arithmetic inference systems to map fuzzy
antecedents to fuzzy consequents as modus ponens logic. FL has been demonstrated in many
physical applications to be good enough to simulate or mimic the human decision-making
processes based on imprecise criteria3. For example, machine vision which mimics human vision
and reasoning would certainly fit into this category.

The class begins by actually demonstrating existing fuzzy models. The traditional means of data
representation are usually treated as probabilistic in nature. Research has developed the means
to use FST an alternative to (and in some cases complementary to) a probabilistic approach. The
basis of FL is the concept of a fuzzy set. A fuzzy set is defined as a set without a clearly defined
boundary. Sets may overlap. Such a set can contain elements with only a partial degree of
membership. Zadeh1 noted that the concept of fuzzy sets parallels ordinary set theory in many
ways, but fuzzy sets are more general and potentially have wider applicability, particularly in the
fields of pattern classification and information processing. Ross’s book4 and classes 2-10 were
used to cover the fundamentals.
Page 12.184.3
From the agricultural and biological literature, fuzzy logic has been demonstrated for crop
management and environmental control in greenhouses5,6,7. Schmoldt8 described the simulation
of plant physiological processes using fuzzy logic. Center and Verma9 developed a fuzzy
photosynthesis model for tomato plants. Fuzzy results compared favorably to traditional
modeling techniques. Gottimukkala10 developed a fuzzy logic classifier to detect weeds. Fuzzy
rules for each weed species were established using spectral response data collected using a
spectrometer. The membership functions were derived using a genetic optimization algorithm
(GA). GA was covered in classes 19-21. Tizhoosh11 provided a fundamental overview of the
application of fuzzy logic to image processing. Several modeling techniques were identified
including fuzzy clustering, rule-based systems, fuzzy geometry and arithmetic, and fuzzy
measure theory. Chi et al12 also presented useful algorithms for fuzzy machine vision.

Bezdek13 summarized various imaging methods using neural networks (NN). A neural network
is a technique that seeks to develop an accurate response function or rules based on a model and
input and output data. Ross4 outlined methods for generating fuzzy membership functions from
imprecise image data. Basak et al14 described neural-fuzzy methods of unsupervised training for
determining membership functions for various applications. El-Faki et al15 used fuzzy color
features to establish a simple weed detection method using a color machine vision system.
Fuzzy rules have been shown to mimic complex mathematical functions very accurately14.

Laviolette and Seaman16 and Pal and Bezdek17 presented a mathematical background for using
FST as a means of describing uncertainty. They also indicated that the use of FST is only
appropriate for certain types of data, and should not arbitrarily usurp the use of probabilistic
measures. Furthermore there is a work that combines the use of FST and probabilistic
measures17, 18, 19. A more aggressive approach to data representation is the use of fuzzy
classifier systems, also described as fuzzy clustering. Geyer-Schulz20 reviewed these systems
and concluded that fuzzy genetic-based machine learning has an advantage over crisp
representations. This is important when trying to extract information from highly ambiguous
data21, 22.

Why Fuzzy Set Theory?

Fuzzy logic provides several advantages over crisp mathematical methods:3

1. While fuzzy logic is based on the same theoretical concepts as classical sets, the theory adds
two new laws: the law of excluded middle and the law of contradiction. Otherwise, mapping
of fuzzy sets to functional relationships is similar to classical methods.
2. Fuzzy logic as a modeling tool is flexible. Fuzzy reasoning can be simple or complex to
predict a unique numerical value for the consequence or predict a classification category of a
fuzzy consequence.
3. Fuzzy logic inference can be developed from the experience of a human expert. If the
relationships between input and output data are well understood, rules can be readily
developed to reflect this a priori knowledge. FL can mimic the human thought process to
process and predict imprecise results.
4. While fuzzy logic is tolerant of imprecise data, its precepts allow convergence to classical
sets. Fuzzy logic can model nonlinear functions of arbitrary complexity. Fuzzy systems can
Page 12.184.4

be created to address any set of input/output data. The ability to deal with nonlinear
 relationships is enhanced through use of adaptive techniques such as neural networks (NN).
5. Fuzzy logic can be blended with conventional control techniques. In many control
applications, fuzzy logic may be used to enhance existing systems and simplify their
implementation. Fuzzy logic is based on natural language and fundamental logic. Because
fuzzy reasoning is based on imprecision in the data, rules can be generated in linguistic terms
that are more readily comprehended.

Membership Functions\

A membership function (MBF) is a mathematical relationship that defines how each data point
within an input space is mapped to a membership value µ or degree of membership (0 ≤ µ ≤ 1) .
The input space (X) is referred to as the universe of discourse. Membership functions can take
on various forms of the distribution of the input data, including piecewise linear, Gaussian,
sigmoid, or polynomial curves. In the case of a classical Boolean set, the membership function µ
would be a constant value (either zero or one) for a given input.

Logical Operations

Fuzzy logic can be considered a superset of conventional Boolean logic, rather than a simple
binary truth table based on intersection (AND) and union (OR) functions. Mathematical
operations are described using variables or membership values between zero and one. One
means of implementing this method is the redefinition of the AND operator as a minimum and
the OR operator as a maximum of two membership values. In the case of a fuzzy system, the
NOT operator becomes the additive complement of the input membership value.

Classical Versus Fuzzy Sets and Mapping

Assume X and Y are two separate universes of discourse, with X representing input information
for the model and Y the output classification information. A model will define a mapping
function between an element x within universe X and a corresponding element y within universe
Y, or
f :X → Y (1)
Define two sets A and B on the input universe X. For a classical Boolean model, characteristic
(indicator) functions for the input data are defined by
1 x∈A
χ A ( x) =  (2)
0 x∉A
1 x∈B
χB ( x) =  (3)
0 x∉B
These functions describe “membership” in the respective sets A and B, for each element x within
universe X. The values 1 and 0 represent the possible values for y within the output universe Y
to which the input x can be mapped. Classical set operations can now be defined in terms of
these characteristic functions as follows:
Union: A ∪ B = {x x ∈ A or x ∈ B}
Page 12.184.5

→ χ A∪ B (x ) = χ A ( x ) ∨ χ B ( x ) = max (χ A ( x ), χ B ( x )) (4)
 Intersection: A ∩ B = {x x ∈ A and x ∈ B}
→ χ A∩ B (x ) = χ A ( x ) ∧ χ B ( x ) = min(χ A ( x ), χ B (x )) (5)
Complement: A = { x x ∉ A} → χ A (x ) = 1 − χ A (x ) (6)
Containment: A ⊆ B → χ A (x ) ≤ χ B ( x ) (7)
Next, two fuzzy sets A and B are defined on the same universe X. Membership functions are
~ ~
defined for the two sets as continuous functions:
µ A ( x ) = f 1 (x ) 0 ≤ f 1 (x ) ≤ 1 (8)
~

µ B (x ) = f 2 ( x ) 0 ≤ f 2 (x ) ≤ 1 (9)
~

Set operations can be defined as for classical sets:

Union: (
µ A∪ B ( x ) = µ A ( x ) ∨ µ B ( x ) = max µ A (x ), µ B (x ) ) (10)
(x ) = µ (x ) ∧ µ (x ) = min(µ (x ), µ (x ))
~ ~ ~ ~ ~ ~

Intersection: µ A∩ B A B A B (11)
~ ~ ~ ~ ~ ~

Complement: µ A ( x ) = 1 − µ A (x ) (12)
~ ~

Containment: A ⊂ B → µ A (x ) ≤ µ B (x ) (13)
~ ~ ~ ~

If-Then Inference Rules

If-then rule definitions apply to fuzzy logic as for Boolean logic, with the exception that the crisp
mathematical values are replaced with linguistic notation that reflect the imprecision of the data.
Interpretation of fuzzy rules is a three-part process:

1. First, fuzzify the inputs. Resolve all fuzzy statements in the antecedent to a degree of
membership between zero and one. If there is only one part to the antecedent, this is a degree
of support for the rule.
2. Apply a fuzzy operator to multiple part antecedents. If there are multiple parts to the
antecedent, apply fuzzy logic operators and resolve the antecedent to a single number
between zero and one. This is a degree of support for the rule.
3. Apply the implication method. Use the degree of support for the entire rule to shape the
output fuzzy set. The consequent of a fuzzy rule assigns an entire fuzzy set to that output.
This fuzzy set is represented by a membership function that is chosen to indicate the qualities
of the consequent. If the antecedent is only partially true (i.e., is assigned a value less than
one), then the output fuzzy set is truncated according to the implication method.

In general, a fuzzy logic inference system is composed of two or more rules that define the
relationships between input and output variables. The output of each rule is a fuzzy set. The
output fuzzy sets for each rule are then aggregated into a single output fuzzy set, which is then
defuzzified to resolve the output into a single value.

Fuzzy inference maps a given input to an output using fuzzy logic. Mapping provides a basis
from which decisions can be made or patterns discerned. There are two basic types of fuzzy
Page 12.184.6

inference systems in common use, Mamdani and Sugeno. Mamdani-type inference expects the
output membership functions to be fuzzy sets23. After aggregation, there is a fuzzy set for each
output variable. The output membership function can be defined as a singleton spike (single
value) or as a centroid of a two-dimensional function. Sugeno-type systems can be used to
model any system in which the output membership functions are either linear or constant values.
The Mamdani method is highly intuitive, has widespread acceptance, and is well suited to human
input. The Sugeno is computationally efficient and works well with linear techniques,
optimization, and adaptive techniques. It has a guaranteed continuity of the output surface.
Fuzzy arithmetic can be used to develop useful models. Fuzzy inference for practical controls
was implemented with a hands-on session with the LabVIEW® controls toolkit (National
Instruments, Inc, Austin, TX).

Fuzzy Arithmetic

Fuzzy arithmetic was covered in classes 8 and 9 and Ross, chapter 12. Problems assigned were
usually computed by hand. The instructors found very few existing software examples to
implement the approximate methods of the extension principle: vertex, Dong, Shah and Wong
(DSW) and modified DSW algorithms24,25, so MATHCAD® (Mathsoft, Inc., Needham, MA)
versions were written, with the future expectation of converting them later to MATLAB script.

Fuzzy Clustering

During pattern recognition, the volume of raw data is often prohibitive for development of an
efficient fuzzy inference system26. Therefore, it is necessary to identify natural groupings of data
to produce more efficient representations of the data. This is often accomplished through the use
of clustering. Clustering refers to the identification of distinct subclasses within a data set and
the partitioning of these data into subclasses. Data are assigned to a given cluster based on a
minimum distance from the cluster center in the feature space. The clustering method can be
either hard (crisp classification partitions) or fuzzy (cluster membership is determined by a
membership function). Fuzzy C-means (FCM) clustering has been suggested as a method for
obtaining segmented object membership functions from color images. FCM and the Gustafsen-
Kessel (GK)26 clustering algorithms were covered in classes 7, 11, and 12. Various cluster
validity methods were also discussed and demonstrated28.

Machine vision systems with imaging and fuzzy clustering for example can distinguish grasses
from broadleaf weed plants especially those are 7-21 days old. A machine vision with a plant
species classification system by Camargo Neto and Meyer29, Camargo Neto et al30,31 was
demonstrated in classes 17 and 20. Improved and automated systematic inventorying and
assessment of plants, their condition, competitiveness, and growth stages from images across
natural vegetation could be very important for remote sensing applications.

Neural Networks and Controls

A Neural Network (NN) is an information-processing system with analytical characteristics that

mimic biological neural processes32. A NN adjusts parameters of the membership functions
using either a backpropagation algorithm or in combination with a least squares method. The
Page 12.184.7

goal of a NN is to accurately map input variables to the corresponding output value, while
minimizing the predictive error rate. Bezdek13 summarized the use of neural networks in fuzzy
pattern recognition for generating membership functions, performing fuzzy logic operations, and
derivation of optimal rule sets. NN’s were covered and demonstrated in classes 22 and 23. Fuzzy
control systems utilize rules that are developed or defined arbitrarily, or based on accepted
relationships between input and output variables32. For many applications, especially data
classification, rules often can not be readily defined based on observations of the variables alone.
Moreover, membership functions for data classification often cannot be predefined. In these
instances, an adaptive learning technique such as a neural network approach is used to train or
discover potential membership functions and rules, based on input data33. Ross4 described
methods for generating membership functions using neural networks. This class focused in the
Adaptive Neural Fuzzy Inference System software (ANFIS) as part of the MATLAB Fuzzy
Logic Toolbox® (MathWorks, Inc., Natick, MA).

Students Projects

Ten independent student projects were developed and presented as posters at an annual
Department open house which invited other students, alumni, faculty, administrators and
potential employers (class 29). The final written term paper was graded according to the
instrument shown in Appendix B. A brief summary description is given for each project below.

Project 1: Fuzzy Model for Predicting the Microbial Growth in Shell Eggs
Microbial contamination in eggs is a major cause of concern for the poultry industry. Salmonella
species, in particular is of interest as it causes the widely spread infection- Salmonellosis. It is
estimated that nearly 1.3 million people are affected by salmonellosis every year. Salmonella
Enteritidis is considered to be the main pathogen responsible for causing salmonellosis. One of
the important contributing factors for the growth of this pathogen is temperature. Improper
temperature conditions can lead to the growth of this pathogen in the eggs, causing the infection.
Hence it is very important for the poultry industry to constantly monitor the temperature to keep
a check on the growth of the microorganism. Periodic/random microbial tests on the eggs are
conducted to ascertain the growth of the microorganism. The traditional tests are time consuming
(takes nearly 48 hrs) to get the results. It is difficult for a large poultry industry to hold the stocks
for long duration before they are cleared microbiologically. Hence a need was felt to develop a
quicker and accurate method to predict the growth of the microorganisms. A predictive model
will be very useful for the industry to make decisions regarding the microbiological safety of
eggs. The predictive model will also take into account the fluctuations in the temperature in the
real conditions. Thus a Fuzzy predictive model will enable to predict the growth of the
microorganism at a given temperature and time there by saving time and money of the industry.

Project 2: ANFIS Model for Predicting the Error in Air Temperature Measurement in MMTS
Fuzzy Inference is the process of formulating the mapping from a given input to an output using
the concept of fuzzy set theory. This mapping forms the basis from which rules can be developed
and the contribution of an input against the output can be drawn. In this study the error in air
temperature measurement was predicted through ANFIS and the results were correlated with that
of the actual error in air temperature measurement obtained from the experimental site. By using
ANFIS, the error in air temperature measurement was predicted and was correlated with the
Page 12.184.8

actual error obtained in experimental site. The correlation factor was obtained as 0.97 for training
data set and 0.84 for testing data set. The effect of each input variable on the output variable is
shown graphically with correlation coefficient. The contribution of each input variable on the
output in the ascending order of the contribution was wind speed, solar radiation, clear index,
solar time according to the values of correlation coefficient.

Project 3: Fuzzy and Neural Network Modeling Of Near-Infrared Reflectance Data For Beef
Tenderness Prediction
Beef tenderness is an important quality attribute for consumer satisfaction. The objective of this
study was to implement Adaptive Neuro-Fuzzy Inference System (ANFIS) modelling of Near-
Infrared (NIR) reflectance spectra for beef tenderness prediction. A spectrometer (λ : 1000 –
1800 nm) was used to record the NIR reflectance spectra of beef ribeye steaks (n=294). Slice
shear force (SSF) values were used as tenderness reference. Based on SSF values, three beef
tenderness categories namely tender (SSF=< 21 kg), intermediate (SSF=21.1 to 25.9 kg), and
tough (SSF>=26 kg) were defined. Two canonical variables were derived from reflectance
values using PROC STEPDISC in SAS. A Sugeno Fuzzy Inference System (FIS) with two
inputs (canonical variable 1 and 2) and one output (tenderness category) was developed. To
handle the non-integral nature of the FIS output, Gaussian membership functions were developed
using the mean and standard deviation of training results and solved with maximum membership
criteria. A ten-fold cross validation procedure was followed. Cross-validation and true
validation classification accuracies were 92.6% and 69.4%, respectively. ANFIS models of NIR
reflectance spectra shows promise in predicting beef tenderness.

Project 4: Fuzzy Logic Modeling of Tapioca Starch-polylactic acid-based Nanocomposite Foams

as Affected by Type of Organoclay

Polymer melt-intercalation or exfoliation is a promising approach for the preparation of

nanocomposites since they exhibit greatly improved mechanical, thermal and barrier properties.
The structure of nanoclay platelets in the nanocomposites depends not only on the properties of
polymer matrix and nanoclay, but also on the extrusion conditions. The objective of the present
work is to investigate the effects of different types of nanoclays and extrusion conditions upon
the nanoclay structure in nanocomposites prepared with a twin screw extruder. Tapioca starch
containing 10% polylactic acid (PLA) and 3% nanoclay at 16% moisture content were extruded.
Feed rate, screw configuration, screw speed, barrel temperature, die opening were maintained
constant. Three different types of clay were used to determine the effect on nanoclay dispersion.
Wide-angle X-ray diffraction (WAXD) was used to examine the gallery height or d-spacing in
clay particles to determine nanoclay interaction and exfoliation in the foam. The d-spacing can
be determined by the diffraction peak in the XRD patterns. Although data had not been
completed collected, a fuzzy inference model was prototyped and presented.

Project 5. Physical Activity Detection using Fuzzy C Means Clustering and a Mamdani FIS.

A system for monitoring and classifying physical activity that a person performs daily for
physical therapists and trainer was developed using fuzzy methods. Biaxial accelerometer data
was used to classify different activities, using Fast Fourier Transform (FFT) data from ten young
individuals (5 males and 5 females, age 23.1+3.1 years). Activities included over ground
Page 12.184.9

walking, over ground running, reaching activity, star climbing, stair negotiation, table activity,
treadmill walking, treadmill running, vita glide, and wheel chair propulsion. Amplitude and
frequency parameters were derived from the FFT’s. Model output however, was correct only 42
% of the time.

Project 6. Optimal Distribution of Electrical Power using a Genetic Algorithm.

A genetic algorithm method34 was used to discover a distributed grid (DG) electrical power
system for a standard IEEE 30-bus transmission system. The DG on each bus with a possibility
of seven buses was derived by two genes. Each set of genes formed a chromosome. The Loss of
Load Probability Index (LOLP) was calculated for a possible 16,384 potential systems. The
fitness function was a reliability factor time LOLP) plus Cost. A crossover rate of 0.8 and
mutation rate of 0.05 were used. Less desirable systems were removed, and best systems were
found by observing a LOLP versus Cost cluster diagram.

Project 7. Using Fuzzy Logic to Manage the Harvest of Wildlife Populations in the Face of
Uncertainty.

Decisions about the magnitude and frequency of wildlife harvest are made by wildlife managers.
Population dynamics are poorly understood. Two fuzzy modeling approaches were compared. A
fuzzy knowledge-based model (FKM) was compared to a rule-based fuzzy cluster (FCM) model.
A third model (MSY) harvested the population with a constant rate set at the theoretical
maximum sustainable yield. Each model was tested and compared on whether they held the
population at a specified target size, maintained a predictable average yield, and whether their
use would cause the population to drop below a minimum threshold. This was done for a 50
year sequence. Both the FKM and FCM models met the specified criteria, while MYS model
failed to meet any.

Project 8. In-line Viscosity Control in an Extrusion Process using a Fuzzy Controller.

In an extrusion process, the end product quality will depend on a number of process parameters,
such as screw speed, quality of the feed material, melt temperature, melt pressure, and melt
viscosity. The quality of the feed material and screw speed can be set manually. Melt
temperature, viscosity, and pressure can be measured by in-line transducers. A prototype fuzzy
controller to automatically adjust screw speed was based on two inputs of viscosity -viscosity set
point difference and the change in viscosity with time. Each was divided into seven linguistic
sets of membership values, resulting in 49 rules. Preliminary test on the fuzzy inference systems
suggest an R2=0.78 and optimization might be required to shorten the response time of the
controller.

Project 9. Predicting Drying Requirements with Fuzzy Logic.

A data driven fuzzy logic model was created to portray the characteristics of variability might
occur during extrusion and manufacturing of breakfast cereal at a local company affects the
drying requirement of the process. Several ANFIS models were trained and tested for predicting
moisture content of the cereal from six proprietary measured extruder inputs. Correlation results
Page 12.184.10

(R2) of predicted versus actual output ranged from 0.64 to 0.98.

Project 10. Fuzzy Logic Modeling of Extrusion Expansion of Starch-Based Foams

Biodegradable starch-based packaging foams were extruded with a single screw laboratory scale
extruder. Extrusions were carried out with 18% moisture starch and three polystyrene levels of
15, 20, and 25%. Talc was added at 0, 1,2, and 5 percent. The size of foam pellets was
measured. Fuzzy logic methods including FCM clustering ANFIS rule-generation were used to
map torque and pressure to radial expansion, and torque, pressure, and talc to radial expansion.
The models training and validation were able predict expansion very well with correlation
factors around 0.94.

Student Acceptance

Student reaction to this class was very good. The overall class score was (3.38+0.54 /4.0). The
overall instructor score was (3.67+0.35 /4.0). Students were quite complimentary about the
course. They felt both instructors did a good job. The material was considered very relevant.
Examinations were fair, but long. A common complaint was knowledge of the software used,
and a request to make it a prerequisite or provide some elementary training may be considered
for future offerings. Improvements could be made on the homework. Posters at the open house
were well-done. Term papers were well-written. A couple of the student papers are being
considered for publication.

Bibliography

1. L. Zadeh, “Fuzzy sets,” Information and Control 8(3), pp. 338-353, 1965.
2. L. A. Zadeh, “Outline of a new approach to the analysis of complex systems and decision processes," IEEE
Transactions on Systems, Man, and Cybernetics 3(1), pp. 28-44, 1973.
3. H. X. Li and V. C. Yen, Fuzzy Sets and Fuzzy Decision-Making, CRC Press, Boca Raton, FL, 1995.
4. T. J. Ross, Fuzzy Logic with Engineering Applications, 2nd Edition. McGraw-Hill, Inc., New York, 2005.
5. K. Kurata and N. Eguchi, “Machine learning of fuzzy rules for crop management in protected cultivation,”
Transactions of the ASAE 33(4), pp. 1360-1368, 1990.
6. K. Chao, R. S. Gates, and R. G. Anderson, “Knowledge-based control systems for commercial single stem
rose production,” ASAE Paper No. 97-4032, St. Joseph, MI, 1997.
7. P. J. Salgado, B. Cunha, and C. Couto, “A fuzzy identification and controller for the agricultural
greenhouse,” 7th International Conference on Computers in Agriculture, American Society of Agricultural
Engineers, St. Joseph MI, 1998.
8. D. Schmoldt, “Simulation of plant physiological processes using fuzzy variables,” AI Applications 5 (4),
pp. 3-16, 1991.
9. B. Center and B. P. Verma, “A fuzzy photosynthesis model for tomato,” Transactions of the ASAE 40(3),
pp. 815-821, 1997.
10. S. N. Gottimukkala, N. Zhang, X. Lin, and N. Wang, “An application of fuzzy logic in weed detection,”
ASAE Paper No. 99-3139, St. Joseph, MI, 1999.
11. H. R. Tizhoosh, “Fuzzy image processing: Potentials and state of the art,” IIZUKA 98, 5th International
Conference on Soft Computing, Iizuka, Japan, 1, pp. 321-324, 1998.
12. Z. Chi, H. Yan, and T. Pham, “Fuzzy algorithms with applications to image processing and pattern
Page 12.184.11

recognition,” Advances in Fuzzy Systems - Applications and Theory, World Scientific, Singapore, 1996.
13. J. C. Bezdek, “A review of probabilistic, fuzzy, and neural models for pattern recognition,” Journal of
Intelligent and Fuzzy Systems 1(1), pp. 1-25, 1993.
14. J. Basak, R. K. De, and S. K. Pal, “Unsupervised feature selection using a neural-fuzzy approach,” Pattern
Recognition Letters (Elsevier) 19, pp. 997-1006, 1998.
15. M. El-Faki, N. Zhang, and D. E. Peterson, “Weed detection using color machine vision,” ASAE Paper No.
97-3134, St. Joseph, MI, 1997.
16. M. Laviolette and J. W. Seaman, Jr., “The efficacy of fuzzy representations of uncertainty,” IEEE
Transactions on Fuzzy Systems 2(1), pp. 1-12, 1994.
17. N. R. Pal and J. C. Bezdek, “Measuring fuzzy uncertainty,” IEEE Transactions on Fuzzy Systems 2(2), pp.
107-118, 1994.
18. T. J. Meitzler, L. Arafeh, H. Singh, and G. R. Gerhart, “Fuzzy logic approach for computing the probability
of target detection in cluttered scenes,” Optical Engineering 35(12), pp. 3623-3636, 1996.
19. D. Dubois and H. Prade, “Fuzzy sets, probability and measurement,” European Journal of Operations
Research 40, pp. 135-154, 1989.
20. A. Geyer-Schulz, “Fuzzy classifier systems,” in Fuzzy Logic, R. Lowen and M. Roubens, eds., Kluwer
Academic Publishers, pp. 345-354, 1993.
21. S. D. Lindsey, R. W. Gunderson, and J. P. Riley, “Spatial distribution of point soil moisture estimates using
Landsat TM data and fuzzy-c classification,” Water Resources Bulletin 28(5), pp. 865-875, 1992.
22. T. Furuhashi, K. Nakaoka and Y. Uchikawa, “A study on fuzzy classifier systems for finding control
knowledge on multi-input systems,” in Genetic Algorithms and Soft Computing, Francisco Herrera and Jose
Luis Verdegay, eds., Physica-Verlag, 1996.
23. Mamdani E. H. and S. Assilian, “An experiment in linguistic synthesis with fuzzy logic controller,”
International Journal of Man-Machine Studies 7(1), pp. 1-13, 1975.
24. Dong, W. and Shah, H. 1987. Vertex method for computing functions of fuzzy variables. Fuzzy Sets and
Systems 24:65-78.
25. Dong, W. , Shah, H. and Wong, F. 1985. Fuzzy computations in risk and decision analysis. Civil Engr.
Systems. 2:201-208.
26. Y. W. Lim and S. U. Lee, “On the color image segmentation algorithm, based on the thresholding and the
fuzzy c-means techniques,” Pattern Recognition 23(9), pp. 935-952, 1990.
27. Gustafson, E.E., and W.C. Kessel. 1979. Fuzzy clustering with a fuzzy covariance matrix. IEEE CDC, San
Diego, CA., 761 - 766.
28. Gath, L. and A.B. Geva, 1989. Unsupervised Optimal Fuzzy Clustering. IEEE Transactions on Pattern
Analysis and Machine Intelligence. 11(7):773-781.
29. Camargo Neto, J., Meyer, G.E., 2005. Crop species identification using machine vision of computer
extracted individual leaves. In: Chen, Y.R., Meyer, G.E., Tu S. (Eds.), Optical Sensors and Sensing
Systems for Natural Resources and Food Safety and Quality, Proc. SPIE, Bellingham.WA., Vol. 5996, pp
64-74.
30. Camargo Neto, G.E. Meyer, D. D. Jones, A.K. Samal. 2006. Plant Species Identification using Elliptic
Fourier Analysis. Computers and Electronics in Agriculture (Elsevier), 50:121-134.
31. Camargo Neto, G.E. Meyer , D. D. Jones. 2006. Individual Leaf Extractions from Young Canopy Images
using Gustafson-Kessel Clustering and a Genetic Algorithm. Computers and Electronics in Agriculture
(Elsevier) 51:65-85.
32. L. V. Fausett, Fundamentals of Neural Networks, Prentice Hall, Inc., Englewood Cliffs, NJ, 1994.
33. T. Yamakawa, “A fuzzy logic controller,” Journal of Biotechnology 24(1), pp. 1-32, 1992.
34. Holland, J.H. 1975. Adaptation in natural and artificial system. The University of Michigan Press, Ann
Arbor, MI.

Page 12.184.12
Table 1. Advanced modeling in biological engineering. Schedule of Topics.
Class Topic
Course Introduction:
1
Modeling, Challenges, Ambiguity and Uncertainty
Introduction to soft computing.
2 Introduction to Fuzzy Set Theory (FST)
Classical Set Theory v FST
3 Probability and Statistics v FST
4 Membership Functions, Development and Conversions
5 MATLAB laboratory: Introduction to MATLAB and Toolboxes for Modeling
Fuzzy Modeling Applications: MATLAB Fuzzy Toolbox and Mamdani and Sugeno
6
type models
7 Clustering Strategies for data organization and model development
8,9 Arithmetic with fuzzy numbers (Hand calculations and MATHCAD functions)
Classical and Fuzzy Relations
9,10
Fuzzy Logic and Inference Systems
11,12 MATLAB laboratory: Model development and cluster analysis
13,14 MATLAB laboratory: Cluster analysis and model optimization
15,16 Fuzzy decision making
17 MATLAB laboratory: Fuzzy Image Enhancement and Fuzzy Segmentation
Image Shape and Texture Analysis- Classical, Elliptic Fourier, Haralick, and Gabor
18
Filters
19 Introduction to Genetic Algorithms, Fitness Functions, and Optimization
Genetic Algorithms, Fitness Functions, and Optimization Applications; Machine vision
20
– Plant Species Identification Example
Genetic Algorithms, Fitness Functions, and Optimization Applications; Machine vision
21
– Applications
22,23 Neural Network Systems
24 Introduction to Process Control Models: Comparing PID with Fuzzy Logic Control
Laboratory: Applications of Process Control Models with Fuzzy Logic (integration of
25
fuzzy controllers with LabVIEW instrumentation systems)
Adaptive Modeling: Integration of Instrumentation (LabVIEW) with Real-time Process
26 Script Models – Controlled Environment Turf water use and fuzzy irrigation control
example.
27,28 Student Progress reports
29 Student Poster Presentations
Page 12.184.13
Appendix A

Catalog 951. Advanced Modeling in Biological Engineering (3 cr) Lec 3. Prereq: Graduate
Description: student status in Engineering or permission. Advanced modeling techniques and
applications. Topics are extracted from the current literature and research areas.
Specific topics vary depending on research interest and current literature. Credit:3
hours

Text Book: Fuzzy Logic with Engineering Applications 2nd Edition

Timothy J. Ross
John Wiley and Sons, Ltd.
ISBN: 0-470-86075-8
…and selected excerpts and recent papers from the literature.

Format: The class meets 1.5 hours on each of two week days to be announced. Class material
will be presented via lectures, class discussions, directed readings, homework,
quizzes, and reviews of research papers. The students are expected (with guidance
from the instructors) to access the research literature (mostly published journal
articles) and lead the class in discussions of assigned research papers related to
modeling, soft computing, and their individual research areas. Of particular
importance are those works that use modeling and can apply soft computing in the
students’ research area. Students will be expected to read 1-3 related journal articles
each week. Examinations include only bi-weekly quizzes consisting of 1 or 2
problems or conceptional responses. Students will also have hands-on activities with
MATLAB, LABVIEW, and MATHCAD (no previous programming experience is
expected, but student must have a basic knowledge of the Windows operating
system). More details will be discussed in class, but the course grade depends on the
completion and execution of the paper reviews, student-led class discussions, and a
final term project.

Page 12.184.14
Appendix B. FINAL REPORT - SCORE SHEET - 100 Points

CATEGORIES POSSIBLE POINTS COMMENTS

POINTS GIVEN

REPORT CONTENT AND FORMAT

Summary - Single descriptive paragraph or abstract. 5
Background and Literature Review - Cites useful
literature, previous instrumentation approaches, and 5
describes purpose and impact of the project.
Objectives - Overall objective, and 1-2 sub objectives. 5
Procedure/ Methods - Methods used. Source of data
and/or knowledge. Appropriate equations, algorithms, 20
units, numbered, and cited. Support with pictures,
figures and/or tables, as appropriate.
Results and Discussion - Reports details and results of
development efforts, methods, testing, with table and 25
figure support.
Conclusions - Appropriate conclusions that match
objectives. Suggestions for future work. 10
References (Complete citations - ASAE method). 5
Figures and Tables - Appropriate use of significant
digits, information displayed, correctly justified, labeled, 5
footnotes, and captions in the correct position.
Landscapes attached correctly to the report. Stands by
themselves.
Appendices - Appropriate supporting material.
Examples: computer code used, sample calculations, 5
samples of raw data, etc.
OVERALL REPORT APPEARANCE
Neatness and Writing Quality - Proper spacing,
justification, proper grammar, correct spelling, and page
numbers. Drawings and sketches are neat and properly
labeled. 5
Overall Creativity – Demonstrates mastery of modeling
and the use and interpretation of soft computing 10
techniques.

FINAL TOTAL SCORE 100

Page 12.184.15
Biographical Information

GEORGE MEYER, Professor, has taught graduate and undergraduate classes that
involve plant and animal growth and environmental factors, modeling, and
instrumentation and controls for both agricultural and biological systems
engineering students for 28 years. He has received national paper awards and
recognition for his work in distance education and has received university teaching
awards. His current research include measurement and modeling of crop water
stress, fuzzy logic controls for turf irrigation management, and machine vision
detection, enumeration, and plant species identification for spot spraying control
and precision agriculture.

DAVID JONES, Professor, has taught graduate and undergraduate classes that
involve fuzzy set theory and soft computing techniques, risk assessment of
complex systems, and mathematical modeling of physical and biological systems
for the past 18 years. He also teaches a Heat and Mass Transfer course to
engineering juniors and the senior design classes. He has received numerous
university and national awards for his teaching excellence.

Page 12.184.16