Design of Experimentation, Artificial Neural Network Simulation and Optimization for Integrated Bamboo Processing Machine
0 likes198 views
This research investigates the design and optimization of an integrated bamboo processing machine for splitting and slicing using experimentation and artificial neural network (ANN) simulation. The study identifies key parameters affecting production rate, quality, and efficiency, optimizing the production rate through a mathematical model. The maximum production rate achieved is 2.66 slices per second, validated through ANN techniques.
![P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29
www.ijera.com 23 | P a g e
Design of Experimentation, Artificial Neural Network Simulation
and Optimization for Integrated Bamboo Processing Machine
P. G. Mehar1
, Dr. A. V. Vanalkar2
, Dr. S. S. Khandare3
1
Research Scholar
2
Professor, Deptt. of Mech. Engg. K. D. K. College of Engg. Nagpur (M.S.), India
3
Ex-Principal, B. D. College of Engg. Sevagram, Wardha (M.S.), India
Abstract
In this research work experimentation on integrated bamboo processing machine for splitting and slicing of
bamboo has been carried out. This paper presents the experimental investigation of some parameters of
integrated bamboo processing machine. In this research paper simulation of experimental data using artificial
neural network is carried out. An attempt of minimum-maximum principle has been made to optimize by range
bound process for maximizing production rate of integrated bamboo processing machine.
Key Words: - Bamboo, Splitting, slicing, experimentation, ANN, optimization.
I. INTRODUCTION
The initial process includes Splitting, External
and Internal Knot Removing, Slicing, Bamboo
sticking making, Stick length setting, Stick Polishing
[13]. The initial processes carried out on a bamboo to
make it as a useful product is called as bamboo
processing. Bamboo and bamboo splits are used as
the fencing material and for making various types of
tool handles, ladders and scaffolding. Splits as well
as slivers are used to make a wide range of products
such as baskets, the core of incense-sticks, kites and
toys, flutes and a large number of handicraft items
[1]. Traditionally the bamboo is processed in
different steps and for each step a different machine
is required. The main aim is to develop an integrated
bamboo processing machine to reduce the number of
steps and also to reduce the number of machines
required to complete the desired work. So an
integrated bamboo processing machine is fabricated
which can perform splitting and slicing on a single
machine.
II. EXPERIMENTAL SETUP
Traditionally bamboo slicing is done manually,
or by using a manually operated machine, there was
always a need of machine which will gives slices of
bamboo without splitting operation. Integrated
bamboo processing machine reduces the cost as well
as time for processing bamboo in to slices. In this
research work dependent and independent variables
are identified then experimental data is collected [3].
The experimental set up is as shown in figure 1. It
shows the Integrated Bamboo processing Machine,
load cell, stop watch and energy meter.
Fig. 1: Experimental Set-up along with measuring devices
a) Load cell b) Stop watch c) Energy meter
III. Experimental Investigation
The data of Bamboo processing is not known to
the professionals involved in this type of operations.
The quantitative relationship based on logic is not
possible and hence the only alternative is of
formulating experimental data based model.
The approach adopted for formulating
generalized experimental data based model is
suggested by Hilbert Schenck [8] as given below.
RESEARCH ARTICLE OPEN ACCESS](https://image.slidesharecdn.com/e511012329-151121060127-lva1-app6892/85/Design-of-Experimentation-Artificial-Neural-Network-Simulation-and-Optimization-for-Integrated-Bamboo-Processing-Machine-1-320.jpg)
![P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29
www.ijera.com 24 | P a g e
1. Identification of independent and dependent
variables or quantities.
2. Reduction of independent variables adopting
dimensional analysis.
3. Design of experimental set up.
4. Calibration of an instrument.
5. Measurement of experimental data.
6. Model Formulation
7. Optimization and validation
8. ANN simulation
The formulated model is an approximate
generalized model. The Identification of dependent
and independent variables of the phenomenon is to be
carried out based on known qualitative physical
phenomenon.
IV. IDENTIFICATION OF VARIABLES
The parameters [9] which affect the production
rate, quality and efficiency of machine are selected
are tabulated in the below table 1.
Sr. No. Variable Types of Variable Symbol Dimensions
1. Tool Hardness Independent HT -
2. Relief Angle Independent -
3. Condition of Bamboo Independent BT -
4. Rake Angle Independent α -
5. Outer Diameter of bamboo Independent Do L
6. Inner Diameter of bamboo Independent Di L
7. Force Independent F MLT-2
8. Velocity Independent Vp LT-1
9. Input Energy Independent IE ML2
T-2
10. Work Done Independent Wd ML2
T-2
11. Production Rate Dependent PR T-1
12. Quality Dependent Q -
13. Efficiency Dependent η -
Table 1: Parameters affecting bamboo processing
V. TABULATION OF DATA
The collected data from the experimentation is tabulated in a proper format as given below in table 2.
Input Variable Output Variable
Sr.No.
ToolHardness,
HT(BHN)
ReliefAngle,ø
Conditionof
Bamboo,BT
RakeAngle,α
Outerdiameter
ofBamboo,do
(mm)
Innerdiameter
ofBamboo,di
(mm)
Force,F(N)
Velocity,Vp
(m/sec)
InputEnergy
(KJ)
WorkDone(KJ)
ProductionRate
(No.of
Slice/sec.)
Quality(degree)
Efficiency,(%)
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 Y1 Y2 Y3
1 260 25 2 0 50 30 122.63 0.53 0.197 0.176 2.21 160 89.35
2 260 25 2 0 53 21 147.15 0.54 0.259 0.226 1.49 165 87.05
3 260 25 2 0 52 25 141.26 0.56 0.238 0.212 1.56 155 89.18
4 260 25 2 0 45 27 117.72 0.54 0.174 0.149 1.55 155 85.91
5 260 25 2 0 45 0 153.04 0.56 0.224 0.194 0.97 170 86.42
6 260 25 2 0 43 0 152.06 0.51 0.207 0.182 0.88 180 87.99
7 260 25 2 0 51 0 159.90 0.56 0.266 0.235 0.96 180 88.12
8 260 25 2 0 53 0 163.83 0.54 0.282 0.251 0.94 180 89.02
9 260 25 1 0 51 30 126.55 0.56 0.212 0.186 1.70 140 87.44
10 260 25 1 0 53 21 147.15 0.56 0.259 0.226 1.49 145 87.05
Table 2: Experimental data sample readings](https://image.slidesharecdn.com/e511012329-151121060127-lva1-app6892/85/Design-of-Experimentation-Artificial-Neural-Network-Simulation-and-Optimization-for-Integrated-Bamboo-Processing-Machine-2-320.jpg)
![P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29
www.ijera.com 25 | P a g e
VI. ARTIFICIAL NEURAL NETWORK
(ANN)
The ANN simulation [6][12] is carried out for
the experimental data to validate the model for large
number of readings. The simulation also improves
the model which is obtained by multivariable linear
regression model. The simulated model ensures more
accurate prediction of values.
6.1 SIMULATION BY USING MATLAB
The ANN simulation is carried out by using
MATLAB software. The simulation for 1296
readings is carried out on artificial neurons.
6.2 PRODUCTION RATE (Y1)
ANN is used for validating the input data and
output data (Y1).
Fig. 2: Network size
Figure 2 shows that ready to create a network
and train it. It is tried for a two layer network, with
sig-mod transfer function in hidden layer and a linear
function in an output layer. As an initial guess, here
25 hidden neurons in hidden layers are used. The
network has 10 inputs and 1 output.
Here Leven berg – Marquartt algorithm for training is
used. The network is trained for 20 iterations only
and three targets, training, validations and testing of
data samples.
Fig. 3: Network train for data validation
Network train for data validations is shown in
figure 3 results in progress of 20 epochs, time
required training the network and various
performances of parameters. It clearly shows the size
validation checks.
Fig. 4: Performance of the learning algorithm train
over 20 epoch](https://image.slidesharecdn.com/e511012329-151121060127-lva1-app6892/85/Design-of-Experimentation-Artificial-Neural-Network-Simulation-and-Optimization-for-Integrated-Bamboo-Processing-Machine-3-320.jpg)
![P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29
www.ijera.com 28 | P a g e
Fig. 9: Linear regression performance fitness curve
VII. OPTIMIZATION
The manual optimization is performed by using
the Range Bound Optimization method [4] of
operations research. The optimization is done for
maximum production rate and minimum power
consumption.
7.1 MAXIMIZATION OF PRODUCTION RATE
In order to obtain maximum production rate the
objective function of production rate is used.
Max Y1 = 1.154651 - 6.2E-05X1 + 0.000347X2 -
0.00445X3 - 0.00196X4 - 0.02525X5 + 0.028952X6 -
0.00284X7 + 1.165921X8 + 0.27392X9 +
3.523711X10
Subjected to constraints of different variables present
in table 2,
210 ≤ X1 ≤ 269; 15 ≤ X2 ≤ 25; 1 ≤ X3 ≤ 2; 0 ≤ X4 ≤ 4;
40 ≤ X5 ≤ 56; 0 ≤ X6 ≤ 36;
98.10 ≤ X7 ≤ 168.73; 0.48 ≤ X8 ≤ 0.58; 0.127 ≤ X9 ≤
0.408; 0.114 ≤ X10 ≤ 0.276
For obtaining maximum production rate, the input
variables with negative sign in model are chosen with
lowest value and positive sign in model are chosen
with highest value.
X1 = 210; X2 = 25; X3 = 1; X4 = 0; X5 = 40; X6 = 36;
X7 = 98.10; X8 = 0.58; X9 = 0.408; X10 = 0.276
Substituting above optimal values, maximum
production rate is given by,
Y1 = 1.154651 - 6.2E-05×210 + 0.000347×25 -
0.00445×1 - 0.00196×0 - 0.02525×40 + 0.028952×36
- 0.00284×98.10 + 1.165921×0.58 +
0.27392×0.408X9 + 3.523711×0.276
The maximum production rate is,
Y1 = 2.66 slices/sec.
VIII. CONCLUSION
The set up is developed for experimentation. The
readings are tabulated in a proper format. The ANN
validation is carried with the help of MATLAB for
production rate, quality and efficiency.
The mathematical model is developed by
multivariable linear regression model for production
rate.
The optimization is carried out by using range
bound method for production rate. The maximum
production rate is found to be 2. 66 slices/sec.
REFERENCES
[1] Aniket Baksy, “The Bamboo Industry in
India”, CCS working Paper # 283, July
2013.
[2] C. S. Verma, V. M. Chariar and R. Purohit,
Cleavage Analysis of Bamboo: A Natural
Composite, International journal of
Engineering Research and Application
(IJERA), ISSN 2248-9622, Vol. 2, Issue 2,
Mar-Apr 2012, pp. 1265-1268.
[3] C. N. Sakhale, P. M. Bapat, M. P. Singh and
J. P. Modak, Design of Experimentation,
Formulation of Mathematical Model and
Analysis for Bamboo Cross Cutting
Operation, International Journal of
Multidisciplinary Research & Advances in
Engg. (IJMRAE), ISSN 0975-7074, Vol. 2,
No. I, April 2010, pp 61-83
[4] R. S. Kadu, Dr. G. K. Awari, Dr. C. N.
Sakhale and Dr. J. P. Modak, Formulation of
Various Mathematical Models for the
Investigation of Tool Life in Boring Process
using Carbide and CBN Tools, International
Journal of Application or Innovation in
Engineering & Management (IJAIEM),ISSN](https://image.slidesharecdn.com/e511012329-151121060127-lva1-app6892/85/Design-of-Experimentation-Artificial-Neural-Network-Simulation-and-Optimization-for-Integrated-Bamboo-Processing-Machine-6-320.jpg)
![P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29
www.ijera.com 29 | P a g e
2319-4847, Volume 3, Issue 3, March 2014,
pp. 86-97
[5] S. K. Undirwade, Dr. M. P. Singh, Dr. C. N.
Sakhale, V. N. Bhaiswar, V. M. Sonde,
Experimental and Dimensional Analysis
Approach for Design of Human Powered
Bamboo Sliver Cutting Machine,
International Journal of Engineering Science
& Advanced Technology (IJESAT), ISSN:
2250-3676, Volume 2, Issue 5, Sep-Oct
2012, pp. 1522-1527
[6] C. N. Sakhale, S. N. Waghmare, S. K.
Undirwade, V. M. Sonde and M. P. Singh,
Formulation and Comparison of
Experimental based Mathematical Model
with Artificial Neural Network Simulation
and RSM (Response Surface Methodology)
Model for Optimal Performance of Sliver
Cutting Operation of Bamboo, Elsevier, 3rd
International Conference on Materials
processing and Characterisation (ICMPC
2014).
[7] Anu Maria, Introduction to Modeling and
Simulation, Proceedings of the Winter
Simulation Conference (1997), pp7-13.
[8] Hilbert Schenck, Theories of Engineering
Experimentation, Third Edition, Mc-
GrawHill Company, Hemisphere Publishing
Corporation, Washington, 1979.
[9] R. K. Bansal, Fluid Mechanics and
Hydraulic machines, Lakshmi Publications,
Ninth edition, 2011.
[10] S. K. Choudhary, R. D. Askhedkar, J. P.
Modak, Planning of Experimentation to
Optimize Performance of α-configuration
Stirling Cycle Refrigeration System,
International Journal of Multidisciplinary
Research and Advances in Engineering
(IJMRAE), Vol.2, No.1, April2010, pp. 233-
244.
[11] J. P. Modak and A. R. Bapat, Formulation of
Generalised Experimental Model for a
Manually Driven Flywheel Motor and its
Optimization, Applied Ergonomics, U.K.,
Vol. 25, No. 2, pp. 119-122, 1994.
[12] S. R. Ikhar, A. V. Vanalkar, J. P. Modak,
“Simulation and Mathematical Modeling of
a Manual Stirrup Making Activity Using
Field Data Based Model”, International
Journal of Engineering Research and
Industrial applications. ISSN 0974-1518,
vol.4, Issue no.1 (Feb. 2011), pp311-324.
[13] Prashant Bamboo Machine Sales
Corporation, Nagpur
(www.prashantbamboo.com)](https://image.slidesharecdn.com/e511012329-151121060127-lva1-app6892/85/Design-of-Experimentation-Artificial-Neural-Network-Simulation-and-Optimization-for-Integrated-Bamboo-Processing-Machine-7-320.jpg)
Design of Experimentation, Artificial Neural Network Simulation and Optimization for Integrated Bamboo Processing Machine
- 1. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 23 | P a g e Design of Experimentation, Artificial Neural Network Simulation and Optimization for Integrated Bamboo Processing Machine P. G. Mehar1 , Dr. A. V. Vanalkar2 , Dr. S. S. Khandare3 1 Research Scholar 2 Professor, Deptt. of Mech. Engg. K. D. K. College of Engg. Nagpur (M.S.), India 3 Ex-Principal, B. D. College of Engg. Sevagram, Wardha (M.S.), India Abstract In this research work experimentation on integrated bamboo processing machine for splitting and slicing of bamboo has been carried out. This paper presents the experimental investigation of some parameters of integrated bamboo processing machine. In this research paper simulation of experimental data using artificial neural network is carried out. An attempt of minimum-maximum principle has been made to optimize by range bound process for maximizing production rate of integrated bamboo processing machine. Key Words: - Bamboo, Splitting, slicing, experimentation, ANN, optimization. I. INTRODUCTION The initial process includes Splitting, External and Internal Knot Removing, Slicing, Bamboo sticking making, Stick length setting, Stick Polishing [13]. The initial processes carried out on a bamboo to make it as a useful product is called as bamboo processing. Bamboo and bamboo splits are used as the fencing material and for making various types of tool handles, ladders and scaffolding. Splits as well as slivers are used to make a wide range of products such as baskets, the core of incense-sticks, kites and toys, flutes and a large number of handicraft items [1]. Traditionally the bamboo is processed in different steps and for each step a different machine is required. The main aim is to develop an integrated bamboo processing machine to reduce the number of steps and also to reduce the number of machines required to complete the desired work. So an integrated bamboo processing machine is fabricated which can perform splitting and slicing on a single machine. II. EXPERIMENTAL SETUP Traditionally bamboo slicing is done manually, or by using a manually operated machine, there was always a need of machine which will gives slices of bamboo without splitting operation. Integrated bamboo processing machine reduces the cost as well as time for processing bamboo in to slices. In this research work dependent and independent variables are identified then experimental data is collected [3]. The experimental set up is as shown in figure 1. It shows the Integrated Bamboo processing Machine, load cell, stop watch and energy meter. Fig. 1: Experimental Set-up along with measuring devices a) Load cell b) Stop watch c) Energy meter III. Experimental Investigation The data of Bamboo processing is not known to the professionals involved in this type of operations. The quantitative relationship based on logic is not possible and hence the only alternative is of formulating experimental data based model. The approach adopted for formulating generalized experimental data based model is suggested by Hilbert Schenck [8] as given below. RESEARCH ARTICLE OPEN ACCESS
- 2. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 24 | P a g e 1. Identification of independent and dependent variables or quantities. 2. Reduction of independent variables adopting dimensional analysis. 3. Design of experimental set up. 4. Calibration of an instrument. 5. Measurement of experimental data. 6. Model Formulation 7. Optimization and validation 8. ANN simulation The formulated model is an approximate generalized model. The Identification of dependent and independent variables of the phenomenon is to be carried out based on known qualitative physical phenomenon. IV. IDENTIFICATION OF VARIABLES The parameters [9] which affect the production rate, quality and efficiency of machine are selected are tabulated in the below table 1. Sr. No. Variable Types of Variable Symbol Dimensions 1. Tool Hardness Independent HT - 2. Relief Angle Independent - 3. Condition of Bamboo Independent BT - 4. Rake Angle Independent α - 5. Outer Diameter of bamboo Independent Do L 6. Inner Diameter of bamboo Independent Di L 7. Force Independent F MLT-2 8. Velocity Independent Vp LT-1 9. Input Energy Independent IE ML2 T-2 10. Work Done Independent Wd ML2 T-2 11. Production Rate Dependent PR T-1 12. Quality Dependent Q - 13. Efficiency Dependent η - Table 1: Parameters affecting bamboo processing V. TABULATION OF DATA The collected data from the experimentation is tabulated in a proper format as given below in table 2. Input Variable Output Variable Sr.No. ToolHardness, HT(BHN) ReliefAngle,ø Conditionof Bamboo,BT RakeAngle,α Outerdiameter ofBamboo,do (mm) Innerdiameter ofBamboo,di (mm) Force,F(N) Velocity,Vp (m/sec) InputEnergy (KJ) WorkDone(KJ) ProductionRate (No.of Slice/sec.) Quality(degree) Efficiency,(%) X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 Y1 Y2 Y3 1 260 25 2 0 50 30 122.63 0.53 0.197 0.176 2.21 160 89.35 2 260 25 2 0 53 21 147.15 0.54 0.259 0.226 1.49 165 87.05 3 260 25 2 0 52 25 141.26 0.56 0.238 0.212 1.56 155 89.18 4 260 25 2 0 45 27 117.72 0.54 0.174 0.149 1.55 155 85.91 5 260 25 2 0 45 0 153.04 0.56 0.224 0.194 0.97 170 86.42 6 260 25 2 0 43 0 152.06 0.51 0.207 0.182 0.88 180 87.99 7 260 25 2 0 51 0 159.90 0.56 0.266 0.235 0.96 180 88.12 8 260 25 2 0 53 0 163.83 0.54 0.282 0.251 0.94 180 89.02 9 260 25 1 0 51 30 126.55 0.56 0.212 0.186 1.70 140 87.44 10 260 25 1 0 53 21 147.15 0.56 0.259 0.226 1.49 145 87.05 Table 2: Experimental data sample readings
- 3. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 25 | P a g e VI. ARTIFICIAL NEURAL NETWORK (ANN) The ANN simulation [6][12] is carried out for the experimental data to validate the model for large number of readings. The simulation also improves the model which is obtained by multivariable linear regression model. The simulated model ensures more accurate prediction of values. 6.1 SIMULATION BY USING MATLAB The ANN simulation is carried out by using MATLAB software. The simulation for 1296 readings is carried out on artificial neurons. 6.2 PRODUCTION RATE (Y1) ANN is used for validating the input data and output data (Y1). Fig. 2: Network size Figure 2 shows that ready to create a network and train it. It is tried for a two layer network, with sig-mod transfer function in hidden layer and a linear function in an output layer. As an initial guess, here 25 hidden neurons in hidden layers are used. The network has 10 inputs and 1 output. Here Leven berg – Marquartt algorithm for training is used. The network is trained for 20 iterations only and three targets, training, validations and testing of data samples. Fig. 3: Network train for data validation Network train for data validations is shown in figure 3 results in progress of 20 epochs, time required training the network and various performances of parameters. It clearly shows the size validation checks. Fig. 4: Performance of the learning algorithm train over 20 epoch
- 4. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 26 | P a g e The training stops after 20 iterations because validations error increased as shown in figure 4. It is useful diagnostic tool to plot the training, validations and test error to check the progress of training. The results are shown in figure 4. The test error and validation set error have similar characteristics and does not appear that any significant over fitting has occurred. The goal is to design the production rate and having minimum errors. The best validation performance is 10-3 at 14 epochs. Fig. 5: Linear regression performance fitness curve The next step is to perform some analysis of the network response. Put the entire data set through the network (training, validation and test) as shown in figure 5, and perform a linear regression between network outputs and the corresponding targets. First calculate the network outputs, in this case there are single outputs and three targets. As shown the result of first three figures, the regression values around 0.9 to achieve the targets. 6.3 QUALITY (Y2) ANN is used for validating the input data and output data (Y2). Fig. 6: Performance of the learning algorithm train over 17 epoch The training stops after 17 iterations because validations error increased as shown in figure 6. It is useful diagnostic tool to plot the training, validations and test error to check the progress of training. The results are shown in figure 6. The test error and validation set error have similar characteristics and does not appear that any significant over fitting has occurred. The goal is to design the quality and having minimum errors. The best validation performance is 101 at 11 epochs.
- 5. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 27 | P a g e Fig. 7: Linear regression performance fitness curve The next step is to perform some analysis of the network response. Put the entire data set through the network (training, validation and test) as shown in figure 7, and perform a linear regression between network outputs and the corresponding targets. First calculate the network outputs, in this case there are single outputs and three targets. As shown the result of first three figures, the regression values around 0.9 to achieve the targets. 6.4 EFFICIENCY (Y4) ANN is used for validating the input data and output data (Y4). Fig. 8: Performance of the learning algorithm train over 43 epoch The training stops after 43 iterations because validations error increased as shown in figure 8. It is useful diagnostic tool to plot the training, validations and test error to check the progress of training. The results are shown in figure 8. The test error and validation set error have similar characteristics and does not appear that any significant over fitting has occurred. The goal is to design the efficiency and having minimum errors. The best validation performance is 10-1.5 at 37 epochs. The next step is to perform some analysis of the network response. Put the entire data set through the network (training, validation and test) as shown in figure 9, and perform a linear regression between network outputs and the corresponding targets. First calculate the network outputs, in this case there are single outputs and three targets. As shown the result of first three figures, the regression values around 0.9 to achieve the targets.
- 6. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 28 | P a g e Fig. 9: Linear regression performance fitness curve VII. OPTIMIZATION The manual optimization is performed by using the Range Bound Optimization method [4] of operations research. The optimization is done for maximum production rate and minimum power consumption. 7.1 MAXIMIZATION OF PRODUCTION RATE In order to obtain maximum production rate the objective function of production rate is used. Max Y1 = 1.154651 - 6.2E-05X1 + 0.000347X2 - 0.00445X3 - 0.00196X4 - 0.02525X5 + 0.028952X6 - 0.00284X7 + 1.165921X8 + 0.27392X9 + 3.523711X10 Subjected to constraints of different variables present in table 2, 210 ≤ X1 ≤ 269; 15 ≤ X2 ≤ 25; 1 ≤ X3 ≤ 2; 0 ≤ X4 ≤ 4; 40 ≤ X5 ≤ 56; 0 ≤ X6 ≤ 36; 98.10 ≤ X7 ≤ 168.73; 0.48 ≤ X8 ≤ 0.58; 0.127 ≤ X9 ≤ 0.408; 0.114 ≤ X10 ≤ 0.276 For obtaining maximum production rate, the input variables with negative sign in model are chosen with lowest value and positive sign in model are chosen with highest value. X1 = 210; X2 = 25; X3 = 1; X4 = 0; X5 = 40; X6 = 36; X7 = 98.10; X8 = 0.58; X9 = 0.408; X10 = 0.276 Substituting above optimal values, maximum production rate is given by, Y1 = 1.154651 - 6.2E-05×210 + 0.000347×25 - 0.00445×1 - 0.00196×0 - 0.02525×40 + 0.028952×36 - 0.00284×98.10 + 1.165921×0.58 + 0.27392×0.408X9 + 3.523711×0.276 The maximum production rate is, Y1 = 2.66 slices/sec. VIII. CONCLUSION The set up is developed for experimentation. The readings are tabulated in a proper format. The ANN validation is carried with the help of MATLAB for production rate, quality and efficiency. The mathematical model is developed by multivariable linear regression model for production rate. The optimization is carried out by using range bound method for production rate. The maximum production rate is found to be 2. 66 slices/sec. REFERENCES [1] Aniket Baksy, “The Bamboo Industry in India”, CCS working Paper # 283, July 2013. [2] C. S. Verma, V. M. Chariar and R. Purohit, Cleavage Analysis of Bamboo: A Natural Composite, International journal of Engineering Research and Application (IJERA), ISSN 2248-9622, Vol. 2, Issue 2, Mar-Apr 2012, pp. 1265-1268. [3] C. N. Sakhale, P. M. Bapat, M. P. Singh and J. P. Modak, Design of Experimentation, Formulation of Mathematical Model and Analysis for Bamboo Cross Cutting Operation, International Journal of Multidisciplinary Research & Advances in Engg. (IJMRAE), ISSN 0975-7074, Vol. 2, No. I, April 2010, pp 61-83 [4] R. S. Kadu, Dr. G. K. Awari, Dr. C. N. Sakhale and Dr. J. P. Modak, Formulation of Various Mathematical Models for the Investigation of Tool Life in Boring Process using Carbide and CBN Tools, International Journal of Application or Innovation in Engineering & Management (IJAIEM),ISSN
- 7. P. G. Mehar et al. Int. Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 5, Issue 11, (Part - 1) November 2015, pp.23-29 www.ijera.com 29 | P a g e 2319-4847, Volume 3, Issue 3, March 2014, pp. 86-97 [5] S. K. Undirwade, Dr. M. P. Singh, Dr. C. N. Sakhale, V. N. Bhaiswar, V. M. Sonde, Experimental and Dimensional Analysis Approach for Design of Human Powered Bamboo Sliver Cutting Machine, International Journal of Engineering Science & Advanced Technology (IJESAT), ISSN: 2250-3676, Volume 2, Issue 5, Sep-Oct 2012, pp. 1522-1527 [6] C. N. Sakhale, S. N. Waghmare, S. K. Undirwade, V. M. Sonde and M. P. Singh, Formulation and Comparison of Experimental based Mathematical Model with Artificial Neural Network Simulation and RSM (Response Surface Methodology) Model for Optimal Performance of Sliver Cutting Operation of Bamboo, Elsevier, 3rd International Conference on Materials processing and Characterisation (ICMPC 2014). [7] Anu Maria, Introduction to Modeling and Simulation, Proceedings of the Winter Simulation Conference (1997), pp7-13. [8] Hilbert Schenck, Theories of Engineering Experimentation, Third Edition, Mc- GrawHill Company, Hemisphere Publishing Corporation, Washington, 1979. [9] R. K. Bansal, Fluid Mechanics and Hydraulic machines, Lakshmi Publications, Ninth edition, 2011. [10] S. K. Choudhary, R. D. Askhedkar, J. P. Modak, Planning of Experimentation to Optimize Performance of α-configuration Stirling Cycle Refrigeration System, International Journal of Multidisciplinary Research and Advances in Engineering (IJMRAE), Vol.2, No.1, April2010, pp. 233- 244. [11] J. P. Modak and A. R. Bapat, Formulation of Generalised Experimental Model for a Manually Driven Flywheel Motor and its Optimization, Applied Ergonomics, U.K., Vol. 25, No. 2, pp. 119-122, 1994. [12] S. R. Ikhar, A. V. Vanalkar, J. P. Modak, “Simulation and Mathematical Modeling of a Manual Stirrup Making Activity Using Field Data Based Model”, International Journal of Engineering Research and Industrial applications. ISSN 0974-1518, vol.4, Issue no.1 (Feb. 2011), pp311-324. [13] Prashant Bamboo Machine Sales Corporation, Nagpur (www.prashantbamboo.com)