Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

https://doi.org/10.1007/s40430-019-1685-z

TECHNICAL PAPER

Modelling and multi‑objective optimization of ultrasonic inserting

parameters through fuzzy logic and genetic algorithm
K. Anand1 · S. Elangovan1

Received: 21 December 2018 / Accepted: 14 March 2019

© The Brazilian Society of Mechanical Sciences and Engineering 2019

Abstract
As the usage of plastic components has increased in various industries, the methods for fastening have increased rapidly.
When the plastic components are fastened by self-tapping screws or bolts, failure occurs because of stripped threads or plastic
creep. In these circumstances, threaded metal inserts provide improved joint performance and ability to assemble and disas-
semble the components without degrading them. Even though many techniques such as insert moulding, thermal insertion
and cold insertion are available for joining thermoplastic material with metal insert, ultrasonic insertion is one of the most
preferred processes because of the shorter cycle time usually less than a second, possibility of simultaneous installation of
the multiple inserts and large-scale automation possibilities for higher production operations. The technical problems faced
by the industries in ultrasonic insertion process are poor insertion quality which affects the function of the product. These
problems arise because of the improper selection of insertion parameters. The objective of this paper is to optimize the
ultrasonic insertion parameters for improving the quality of joint through non-traditional optimization techniques. Response
surface methodology (RSM) is used to design the experiments, and then pullout strength and stripping torque are measured.
Data obtained from the measurement are utilized to develop a nonlinear equation between the responses and predictors, and
optimal combinations of insertion parameters are found out by fuzzy logic and genetic algorithm (GA) approach. From the
confirmatory test, it was observed that the fuzzy logic yields better output results than GA.

Keywords  Ultrasonic insertion · Pullout strength · Stripping torque · Response surface methodology · Fuzzy logic · Genetic
algorithm

1 Introduction thermoplastic component. Thermoplastic component may be

amorphous or semi-crystalline category. Mostly, amorphous
Method of fastening the plastic components is important in is used because of their random orientation of molecules.
a wide range of industries. When the plastic parts are fas- The ultrasonic plastic welding machine is used as ultra-
tened by self-tapping screws, there is a chance of failure in sonic insertion machine with few alterations in the horn
assembly after some time due to continuous usage. To pre- and work holding fixture. Ultrasonic insertion is performed
vent failure in plastic assembly, metal insert is introduced. either by horn contact with metal insert or by horn contact
Metal insert is a cylindrical component with internal threads with thermoplastic component. The horn is made of alu-
and external features. The features are knurls, serrations, minium alloy, titanium alloy or mild steel. Mounting hole
grooves or undercuts. Straight or taper inserts might be used. in thermoplastic component guides the insert. When ultra-
Taper insert is used for better alignment, and straight insert sonic horn is pressed against the metal insert, then ultrasonic
with pilot is used for alignment along the axis of hole in the vibrations create friction between metal insert and thermo-
plastic component. Friction causes plastic to melt and insert
sinks into the hole. Some advantages of ultrasonic insertion
Technical Editor: Lincoln Cardoso Brandão.
process are short cycle time, elimination of the possibility of
* K. Anand mould damage, avoiding the insert falling into mould during
[email protected] insert moulding process, minimal induced stress, high pro-
ductivity and high repeatability. The schematic representa-
1
Department of Production Engineering, PSG College tion of ultrasonic insertion is shown in Fig. 1.
of Technology, Coimbatore, Tamil Nadu 641 004, India

13
Vol.:(0123456789)
 188   Page 2 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

the developed models are coupled with GA to optimize the

ultrasonic metal welding parameters.
Mukherjee et al. [7] have used different non-traditional
optimization techniques to optimize the wire electrical dis-
charge machining (WEDM) process. The results of these
non-traditional optimization techniques are compared. The
results of their study show that biogeography-based opti-
mization algorithm outperformed the other optimization
techniques.
Saber et  al. [8] have developed and utilized particle
swarm optimization and genetic algorithm to minimize the
total machine load. The proposed hybrid particle swarm
optimization and genetic algorithm (HPSOGA) gives better
results compared to other algorithms.
From the literature survey, many researchers have used fuzzy
logic and genetic algorithm for optimizing different processes
Fig. 1  The schematic representation of ultrasonic insertion [3]. 1. such as ultrasonic metal welding, ultrasonic insertion, GMAW
Ultrasonic insertion horn, 2. Metal insert, 3. Thermoplastic compo-
welding, wire electrical discharge machining and abrasive water
nent, 4. Anvil. A - Amplitude of the horn vibrations, PS - Static pres-
sure force jet machining. It seems that limited works have been carried out
for modelling and optimizing (multi-objective) the ultrasonic
insertion process. So, modelling and multi-objective optimiza-
The fuzzy logic, design of experiments (DOE) and tion of ultrasonic insertion parameters using fuzzy logic and
genetic algorithm are used in many areas for optimizing genetic algorithm have been attempted in this work.
different processes. Some of the findings from the litera-
ture survey are given below.
Satpathy et al. [1] have optimized the ultrasonic metal 2 Materials and methods
welding parameters for dissimilar materials using fuzzy
logic and genetic algorithm. Their study concludes that 2.1 Design of metal insert and thermoplastic test
fuzzy logic is economical and a better method than genetic specimen
algorithm for optimizing the process parameters.
Marijayaprakash et al. [2] investigated the failures of The design of metal insert and dimensions of thermoplastic
boilers during cogeneration process. The statistical tool, test specimen is decided by functional requirement of the
viz. failure mode and effect analysis (FMEA), and Taguchi plastic assembly. The metal insert used in this study is made
DOE have been applied to optimize the process param- of brass with two grooves and helical flutes. An insert has to
eters. Since conventional FMEA and Taguchi DOE are be designed to provide both pullout and torsional resistance.
not able to yield better solution, fuzzy logic and genetic So, the insert is designed with 45˚ flute angle (to resist both
algorithm techniques have been used to optimize the pro- pullout and torsional load) and two grooves. The dimensions
cess parameters. of the brass insert are shown in Fig. 2.
Anand et al. [3] have used RSM–GA integration tech- Thermoplastic test specimen (thermoplastic mounting
nique for optimizing the ultrasonic insertion process to max- hole) used in this work is made of natural polycarbonate.
imize pullout strength. The authors revealed that RSM–GA Generally, optimum insertion performance is achieved when
integration technique yields better results than RSM. the boss outside diameter (OD) is twice the insert diam-
Correia et al. [4] compared the optimized results obtained eter. The mounting hole in the plastic component is usually
from GA and RSM in GMAW welding process. Their study 0.38–0.51 mm smaller than the insert OD, and can be either
showed that GA yields better results than RSM technique, straight or have an 8-degree taper. The hole should be deeper
in an irregular experimental data. than the insert length to provide a well for excess melt and
Azlan et al. [5] have optimized the process parameters prevent “bottoming out” of the insert [9]. The dimensions of
in the abrasive water jet machining using simulated anneal- the polycarbonate test specimen are shown in Fig. 3.
ing (SA) and genetic algorithm (GA) to minimize surface
roughness value. 2.2 Experimental details
Elangovan et al. [6] have developed mathematical models
for ultrasonically welded joint strength using RSM in terms In this study, ultrasonic insertion is performed with ultra-
of pressure, weld time and amplitude of vibrations. Then, sonic plastic welding machine (1500 W, 20 kHz). Stepped

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 3 of 20  188

Fig. 4  Ultrasonic plastic welding machine. 1. Al Stepped horn, 2.

Anvil, 3. Control panel
Fig. 2  Dimensions of brass insert

2.3 Identification of control factors

The ultrasonic insertion process involves a number of param-

eters which may influence the performance of ultrasonically
inserted joint. From experimental trials, three important
parameters such as pressure (P), inserting time (IT) and
holding time (HT) were selected. By trial experiments, the
working range for each parameter was set in such a way that
sound insertion can be obtained. Pressure, inserting time and
holding time have been divided into three levels. The factors
with their levels are shown in Table 1.

2.4 Response surface methodology

The influence of all input variables on the responses is inves-

tigated by response surface methodology. RSM is a statisti-
cal technique which is useful for modelling and analysing
the problems in which responses of interest are influenced by
several input parameters, and the objective is to either maxi-
Fig. 3  Dimensions of thermoplastic test specimen mize or minimize the responses [10]. In RSM central com-
posite design (CCD) for three parameters and three levels,
the number of observations is 20, which is less when com-
horn has been employed for this study. The horn is made of pared to full factorial design and error degrees of freedom
aluminium alloy because it offers excellent acoustical prop- is 10 which is also less. When safe limits are not known,
erties; also it is used where the low amplitude is required. the central composite designs are an excellent choice. The
The maximum amplitude at the tip of the horn is 60 µm. The design of 20 runs is shown in Table 2.
vertical movement of horn is done pneumatically. The ultra- To reduce the variability in experimental results rand-
sonic plastic welding machine is shown in Fig. 4. Experi- omization, blocking and repetition have been considered.
mental set-up for ultrasonic insertion process and fixtures Totally 80 experiments were conducted, 20 experiments
used in insertion process are shown in Fig. 5. for pullout strength with two replications and 20 runs for

13
 188   Page 4 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 5  Experimental set-up for

ultrasonic insertion process

Stepped Al Horn

Brass Insert

Fixture for holding

Poly carbonate test specimen
Test Specimen

Table 1  Ranges of input parameters and their levels 2.5 Pullout strength test

Input parameters Level 1 Level 2 Level 3
Pullout strength of the joint is measured at room tempera-
Pressure (P) 3 bar 3.25 bar 3.5 bar ture using computerized tensile testing machine (TKG EC
Inserting time (IT) 2.5 s 3 s 3.5 s 10-KN) with ASTM D 1761 standard. During this test,
Holding time (HT) 3 s 3.25 s 3.5 s straining rate was 15 mm/min. A special fixture was made
for this testing. The pullout strength values of all 20 runs
are shown in Table 2. Tensile testing machine and initial
stripping torque with two replications. The photograph of
set-up for testing are shown in Fig. 7. Specimen after pull-
ultrasonically inserted polycarbonate test specimens before
out test is shown in Fig. 8.
test is shown in Fig. 6.

Table 2  Experimental results and predicted value from Fuzzy logic

Test runs Pressure (bar) Inserting Holding time (s) Pullout Stripping Normalized Normalized FMPI
time (s) strength torque (Nm) pullout strength stripping torque
(MPa)

1 3 3.5 3 14.91 20.56 0.7470 0.3796 0.483

2 3 2.5 3.5 16.25 16.25 0.9452 0.20864 0.224
3 3.25 3 3.25 14.51 25.87 0.6878 0.5902 0.462
4 3.5 2.5 3 14.7 15.65 0.7159 0.1848 0.206
5 3.5 3.5 3.5 16.62 35.84 1 0.9857 0.926
6 3.25 3 3.25 12.59 25.51 0.4038 0.5759 0.385
7 3.25 3 3 13.33 20.45 0.5133 0.3752 0.385
8 3.25 3.5 3.25 12.32 33.46 0.3639 0.8913 0.275
9 3.5 3 3.25 16.32 36.2 0.9556 1 0.925
10 3.25 3 3.25 13.24 28.05 0.5 0.6767 0.385
11 3.25 3 3.5 16.19 27.66 0.9363 0.6612 0.531
12 3.25 3 3.25 13.26 26.23 0.5029 0.6045 0.385
13 3.25 2.5 3.25 10.78 25.68 0.1360 0.5827 0.0769
14 3 3 3.25 16.11 10.99 0.9245 0 0.194
15 3.5 3.5 3 16.13 34.86 0.9275 0.9468 0.924
16 3 3.5 3.5 16.16 17.53 0.9319 0.2594 0.351
17 3.25 3 3.25 12.82 27.96 0.4378 0.6731 0.385
18 3 2.5 3 9.86 14.91 0 0.1554 0.0786
19 3.25 3 3.25 13.56 27.54 0.5473 0.6564 0.385
20 3.5 2.5 3.5 16.35 35.98 0.9601 0.9912 0.925

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 5 of 20  188

2.6 Stripping torque test Threads with M10 size were machined on the drill bit, and
it was mounted on spindle of the drilling machine. Thermo-
Stripping torque is the torque at which metal inserts are plastic test specimen was fixed with three-jaw chuck which
stripped out from the thermoplastic component. Drilling is connected to digital control panel of dynamometer. When
machine and drilling tool dynamometer were employed to spindle rotates, M10 threads assemble with metal insert, and
measure stripping torque of ultrasonically inserted joint. then, metal insert will be stripped out from plastic part at

Fig. 6  Polycarbonate test speci-

mens after insertion

Fig. 7  Tensile testing machine

and initial set-up for testing

Poly carbonate
Test Specimen

Holding Fixture

Fig. 8  Polycarbonate specimens
after pullout testing

13
 188   Page 6 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

maximum torque [11]. The stripping torque values for all the Step 1 Normalize the response in order to remove cer-
experiments are listed in Table 2. Torque-testing facility and tain degree of uncertainty. Other reason behind normaliza-
initial set-up for testing are shown in Fig. 9. tion is that membership function (MF) curve depends on
the mapping of input variables in its range of 0–1 [1]. The
2.7 Methodologies formula used for normalization is
Y(i) − YMAX (i)
2.7.1 Fuzzy logic X (i) = , (1)
YMAX (i) − YMIN (i)
Fuzzy logic is the form of logic that is approximate rather where X(i) is the normalized value, Y(i) is the observed
than precise. It is modelling of human intelligence in lin- value, YMAX(i) is the maximum value of ith response, YMIN
guistic form. In contrast to classical logic theory, “either (i) is the minimum value of ith response.
the statement is true or false (0 or 1)”, fuzzy logic has the Step 2 Simultaneous maximization of pullout strength
capacity to handle the concept of partial truth, where the and stripping torque are the objective of analysis. All the
truth value is in the range between complete truth and normalized input values are given to inference engine, and
complete false. It consists of three main steps: fuzzifica- FMPI output is generated by MATLAB fuzzy logic tool-
tion, mamdani inference engine and defuzzification. The box. For maximizing the FMPI, the optimum combination of
flow chart for fuzzy logic unit is presented in Fig. 10. In process parameters has to be determined. The fuzzy values
fuzzification, the membership function is used to fuzzify are defined by membership function. The linguistic values
the normalized values. The mamdani inference engine uses of input and output are expressed in analysis. So, divide the
IF–THEN rules to perform fuzzy reasoning and defuzzifi- input into small (S), medium (M), large (L) and output is
cation to convert the fuzzy values into single fuzzy multi- expressed as very small (VS), small (S), medium (M), large
performance index (FMPI). This FMPI value is predicted (L) and very large (VL). The sharing of values is given in
for every run in the experiment, and corresponding FMPI Table 3 and Fig. 11 which shows the graphical representa-
value is found. tion of membership function for inputs and FMPI. Various
There are four steps which are to be followed, and they
are

Fig. 9  Set-up for testing strip-

ping torque

M10 Bolt

Drilling Tool
Dynamometer

Fig. 10  Flow chart for fuzzy

logic unit Membership function Fuzzy rules

X1 Inference engine
Fuzzification (Mamdani) Defuzzification Y
X2

X1 – Normalized pullout strength X 2 – Normalized stripping torque Y – MPCI

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 7 of 20  188

degrees of membership of the fuzzy values are calculated The relative importance among the inserting parameters
based on the values of X1, X2 and Y. should be known for getting the highest value of FMPI. For
Step 3 In fuzzification, triangle membership function this, optimum combination of inserting parameter levels
is most commonly used. Next step is to establish relation- can be determined by calculating average FMPI values.
ship between input and output variables by IF–THEN rules. From Table 4, pressure is at rank 1 and it is the most influ-
Mamdani inference engine is used to evaluate these rules, encing parameter. The main effect plot for FMPI mean is
and the minimum memberships of various fuzzy sets are shown in Fig. 13. From the graph, the optimum combination
taken into account. The mathematical notation is of parameter is P (3.5 bar), IT (3.5 s) and HT (3.5 s).
𝜇AB(X,Y)=MIN(𝜇A(X) ,𝜇B(Y) ) (2) 2.7.2 Development of mathematical model
where 𝜇A(X) and𝜇B(Y) are membership functions of different
fuzzy sets. As ultrasonic insertion is a nonlinear process, a second-
These rules are combined together to form the final out- order mathematical model was used for the accurate pre-
put. These fuzzy output values can be obtained in the form diction of the response for getting the relationship between
of linguistic terms, when the fuzzy implication and fuzzy input process parameters and the output responses. It can
aggregation are united. The fuzzy logic reasoning is shown be expressed as FMPI = f (P, IT, HT), where FMPI value
in Fig. 12. In this study, for each rule, the two inputs are was obtained from fuzzy logic and it was considered as the
assigned in the fuzzy subsets and the first two columns rep- output response or yield. The equation expressed to find out
resent it. The number of rules gained from the present study the second-order polynomial is given by Eq. (4)
is 20.
Step 4 Finally, the defuzzifier converts the fuzzy values 3
∑ 3
∑ 3
∑
into non-fuzzy crisp values through centre of gravity or cen- Y = 𝛽0 + 𝛽i Xi + 𝛽ii Xi2 + 𝛽ij Xi Xj , (4)
troid method. So, inference output 𝜇y is transformed into i=1 i=1 i,j=1i≠j

defuzzifier value Y0 . Figure 12 shows the output for relevant

where 𝛽0 mean of all responses, 𝛽i and 𝛽ij are the coefficients
inputs. The output is calculated with the help of Eq. (3)
of main and interaction of variables.
∑20 ∗ � � The analysis of data and coefficient calculations were
i=1 x 𝜇y
Y0 = ∑20 � � (3) done using MINITAB 17.1. Here, 95% confidence level was
i=1 𝜇y used for analysis. By back-elimination method, insignificant
coefficients were removed without affecting the accuracy of
the model. The second-order mathematical model developed
for the ultrasonic insertion process is shown in Eq. (5)
Table 3  Fuzzy value ranges for inputs and FMPI variables FMPI = 0.3809 + 0.2577(P) + 0.1448(IT) + 0.0881(HT)
Parameters Linguistic variables Fuzzy ranges + 0.2045(P)2 − 0.1795 (IT)2 + 0.1025 (HT)2
Input S 0–0.3 + 0.0886(P ∗ HT) − 0.1244(IT ∗ HT)
(5)
M 0.3–0.7
L 0.7–1.0 From the regression analysis, R2  , adj R2 and stand-
Fuzzy multi-performance VS 0–0.165 ard error results were obtained. Here, R2 value is 0.9479
index (FMPI) S 0.165–0.385 which means the model can expect 94.79 variations in
M 0.385–0.605 FMPI. If F-calculated is greater than F-tabulated, then the
L 0.605–0.825 parameter or interaction is significant. From Table 5, pres-
VL 0.825–1.0 sure is the most significant parameter and interaction of

Fig. 11  Membership function
for inputs and FMPI

13
 188   Page 8 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 12  Fuzzy logic reasoning procedure for the neutral position

Table 4  Main effects on FMPI Factors Levels Difference Rank Optimum level

values
1 2 3

Pressure 0.2661 0.3655 0.7816 0.5155 1 P3

Inserting time 0.3023 0.4423 0.5918 0.2895 2 IT3
Holding time 0.4153 0.3859 0.5916 0.2057 3 HT3

pressure and holding time is the least significant param- pressure, inserting time and holding time. Fitness of all
eter. HT * HT is the non-significant parameter. The same chromosomes is determined by the use of rank scaling
result can be obtained for p value interpretation. function. Now, these members are called as parents. The
parents were selected based on roulette wheel function for
2.7.3 Genetic algorithm reproduction. The elite count, crossover rate, crossover
fraction, crossover function, mutation function and muta-
Genetic algorithm is an unconventional optimization tech- tion rate are specified in reproduction options to create
nique used for solving both constrained and unconstrained children for the next generation.
problems based on natural selection “survival of the fit- For moving the individuals between sub-populations
test” strategy. A double-vector random chromosome popu- based on the condition that the population size should be
lation was first initialized. The genes are decoded, namely set as a vector length of more than 1, this process could

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 9 of 20  188

Fig. 13  Main effects plot for

FMPI means

Table 5  ANOVA results for Source DF Adj SS Adj MS F-Cal P value F-Tab Significance
FMPI
Model 8 1.35436 0.169296 25.03 0 2.95 Significant
Linear 3 0.95059 0.316865 46.84 0 3.59 Significant
P (bar) 1 0.66327 0.663269 98.05 0 4.84 Significant
IT (s) 1 0.20982 0.209815 31.02 0 4.84 Significant
HT (s) 1 0.07751 0.0775110 11.46 0.006 4.84 Significant
Square 3 0.21758 0.072526 10.72 0.001 3.59 Significant
P (bar)*P (bar) 1 0.11465 0.114648 16.95 0.002 4.84 Significant
IT (s)*IT (s) 1 0.08848 0.088476 13.08 0.004 4.84 Significant
HT (s)*HT (s) 1 0.02899 0.028995 4.29 0.063 4.84 Non-significant
2-Way interaction 2 0.18619 0.093196 13.76 0.001 3.98 Significant
P (bar)*HT (s) 1 0.06259 0.062587 9.25 0.011 4.84 Significant
IT (s)*HT (s) 1 0.12360 0.123604 18.27 0.001 4.84 Significant
Error 11 0.07441 0.006764
Lack-of-fit 6 0.06947 0.011578 11.72 0.008
Pure error 5 0.00494 0.000988
Total 19 1.42877

be controlled by migration parameters such as migration and holding time 3 s, respectively. The best individual and
direction, migration interval and migration fraction. In this best fitness are plotted and shown in Fig. 14.
operation, the worst individuals are replaced with the best
individuals in another sub-population. After all these opera-
tions, the correct population was replaced with the children 3 Results and discussion
to form the next generation. The algorithm stops when it
meets the specified number of generations as the stopping 3.1 Effect of insertion parameters on pullout
criteria [1]. strength
The MATLAB optimization toolbox was used to maxi-
mize FMPI value as given in Eq. 5. The insertion parameters, Figures 15, 16 and 17 show the interaction effect of pres-
search ranges and the parameter setting for genetic algorithm sure and inserting time on pullout strength by keeping hold-
are given in Table 6. From the mean fitness plot, the conver- ing time as constant. It was observed that pullout strength
gence was observed and best fitness plot variables with best increases from inserting time 2.5–3 s, and the reason is
fitness were revealed. From this plot, process parameters for plastic flows around the insert and settled in the external
optimum conditions are pressure 3 bar, inserting time 3.5 s features of insert. Pullout strength decreases from inserting

13
 188   Page 10 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Table 6  Parameter settings for genetic algorithm 3.2 Effect of insertion parameters on stripping

Types of operations and parameters Functions or param- torque
eters value used
Figures 24, 25 and 26 show the effect of insertion time and
Population type Double vector
pressure on stripping torque by keeping holding time as con-
Population size 50
stant. It was understood that stripping torque increases with
Creation function Constraint dependant
increase in inserting time for any value of pressure. The
Scaling function Rank
reason is, when the pressure is increased, the frictional force
Selection function Roulette
acting over the thermoplastic test specimen increases. Also,
Reproduction
more insertion time gives sufficient time for plastic material
 Elite count 2
to flow and settle around the external features of the insert.
 Crossover fraction 0.8
Figures 27, 28 and 29 show the effect of pressure and
 Mutation function Uniform
holding time on the stripping torque by keeping insertion
 Mutation rate 0.007
time as constant. It was seen that stripping torque increases
 Crossover function Heuristic
with increase in pressure and holding time, and it is due to
 Crossover rate 0.8
an increase in pressure which gives more frictional force
Migration
on plastic material and increase in holding time that avoids
 Direction Forward
back-out of insert from plastic part after horn retracts.
 Fraction 0.2
Figures 30, 31 and 32 show the effect of insertion time
 Interval 20
and holding time on stripping torque by keeping pressure as
Stopping criteria
constant. From these figures, it was observed that stripping
 Generations 100
torque increases when holding time increases until 3.25 s,
 Stall generation 50
this is because molten plastic has sufficient time to solidify
 Function tolerance 1.00E − 06
and form strong bond.
Search range
 Pressure 3–3.5 bar
 Inserting time 2.5–3.5 s
3.3 Conformation test
 Holding time 3–3.5 s

After evaluating the optimum combination parameters from

fuzzy logic and genetic algorithm, their responses and cor-
time 3–3.5 s due to excessive installation time that affects responding FMPI values are compared. These results are
the existing joint. tabulated in Table 7. The optimum values of pullout strength
Figures 18, 19 and 20 show the interaction effect of and stripping torque obtained from fuzzy logic are better
pressure and holding time on pullout strength by keep- than the optimum values obtained from genetic algorithm.
ing insertion time as constant. From these figures, it was From this study, FMPI value of 0.926 obtained from fuzzy
understood that pullout strength decreases from pressure logic is better than the FMPI value of 0.485 obtained from
3–3.25 bar because of the pressure applied from the horn genetic algorithm.
is not sufficient to plasticize the material. Pullout strength
increases from pressure 3.25–3.5 bar because there is an
increase in pressure which gives greater strain and fric- 4 Conclusions
tional force over the plastic material and it flows around
the external features of the insert and increases pullout Considering the case of ultrasonic insertion, the process
performance. parameters which affect the quality of the joint (pullout
Figures 21, 22 and 23 show the interaction effect of strength and stripping torque) were identified and selected
inserting time and holding time on pullout strength by by trial experiments. In this study, the experiments were
keeping the pressure as constant. It was observed that conducted as per the response surface methodology on
pullout strength is minimum at lower-level holding time ultrasonic insertion set-up and the following inferences
(3  s); this is because of insert backing out from ther- have been arrived.
moplastic test part after horn retracts. Pullout strength
is maximum at higher holding time; the reason is that • By using fuzzy rule-based model, the multi-responses
molten plastic has sufficient time to cool and solidify such as pullout strength and stripping torque were taken
around insert. as two inputs and combined into one response, i.e. FMPI.

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 11 of 20  188

Fig. 14  Best fitness and best

individual plot

Fig. 15  Interaction effects of pressure and inserting time on pullout strength (HT = 3 s)

13
 188   Page 12 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 16  Interaction effects of pressure and inserting time on pullout strength (HT = 3.25 s)

Fig. 17  Interaction effects of pressure and inserting time on pullout strength (HT = 3.5 s)

The reason is that the correlation of response can be pressure of 3.5 bar, inserting time of 3.5 s and holding
prevented and individual priority weights need not be time of 3.5 s are the optimum combinations to get best
assigned. insertion performance using this method.
• Based on its main effects results, pressure which occu- • FMPI data were used to develop a second-order math-
pies rank 1 followed by inserting time and holding time ematical model, and analysis was performed to test
are the most influencing parameters on the response. A the accuracy of the model with the experimental value

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 13 of 20  188

Fig. 18  Interaction effects of pressure and holding time on pullout strength (IT = 2.5 s)

Fig. 19  Interaction effects of pressure and holding time on pullout strength (IT = 3 s)

using ANOVA. This model can explain the variation in was obtained. From the best individual plot, inserting
FMPI up to 94.79% time was the most influencing parameter followed by
• The developed mathematical model was used as fitness pressure and holding time. The optimum condition
function for genetic algorithm. After certain iterations, is pressure 3  bar, inserting time 3.5  s and holding
the optimum combination of the process parameters time 3 s.

13
 188   Page 14 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 20  Interaction effects of pressure and holding time on pullout strength (IT = 3.5 s)

Fig. 21  Interaction effects of inserting time and holding time on pullout strength (P = 3 bar)

• From experimental investigation, the maximum value reason was higher pressure gives greater strain and
of pullout strength and stripping torque was obtained frictional force on the plastic material that allows the
when the input parameters were at a higher level. The plastic to the external features of the insert. Higher

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 15 of 20  188

Fig. 22  Interaction effects of inserting time and holding time on pullout strength (P = 3.25 bar)

Fig. 23  Interaction effects of inserting time and holding time on pullout strength (P = 3.5 bar)

inserting time gives sufficient time for plastic material • Finally, a comparison of fuzzy logic and genetic algo-
to flow and settle around the external features of the rithm was made to check which one is accurately opti-
insert and higher holding time squeezes molten plastic mizing the parameters to get the maximum FMPI value.
to cool and solidify around insert. From the results, it was observed that fuzzy logic tech-

13
 188   Page 16 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 24  Interaction effects of pressure and inserting time on stripping torque (HT = 3 s)

Fig. 25  Interaction effects of pressure and inserting time on stripping torque (HT = 3.25 s)

nique gives high FMPI value than genetic algorithm. • In the present study, the insertion parameters such as
So, fuzzy logic is better and economical way of predict- pressure, inserting time and holding time were opti-
ing the process variables. mized to improve the quality (pullout strength and
stripping torque) of the joint. Since the profile of the
ultrasonic horn also influences the quality of the joint,

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 17 of 20  188

Fig. 26  Interaction effects of pressure and inserting time on stripping torque (HT = 3.5 s)

Fig. 27  Interaction effects of pressure and holding time on stripping torque (IT = 2.5 s)

13
 188   Page 18 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 28  Interaction effects of pressure and holding time on stripping torque (IT = 3 s)

Fig. 29  Interaction effects of pressure and holding time on stripping torque (IT = 3.5 s)

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188 Page 19 of 20  188

Fig. 30  Interaction effects of inserting time and holding time on stripping torque (P = 3 bar)

Fig. 31  Interaction effects of inserting time and holding time on stripping torque (P = 3.25 bar)

13
 188   Page 20 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2019) 41:188

Fig. 32  Interaction effects of inserting time and holding time on stripping torque (P = 3.5 bar)

Table 7  Conformation test results for optimization 2. Mariajayaprakash A, Senthilvelan T, Gnanadass R (2015) Opti-
mization of process parameters through fuzzy logic and genetic
Optimal insertion conditions algorithm—A case study in a process industry. J Appl Soft Com-
put 30:94–103
Level From fuzzy logic From genetic algorithm
3. Anand K, Elangovan S (2017) Optimizing the ultrasonic inserting
P3 IT3 HT3 P1 IT3 HT1 parameters to achieve maximum pull-out strength using response
surface methodology and genetic algorithm integration technique.
Pullout strength (Mpa) 16.62 14.91 Measurement 99:145–154
Stripping torque (Nm) 35.84 20.56 4. Correia Davi Sampaio, Goncalves Cristiene Vasconcelos, Simoes
Sebastiao, da Cunha Jr, Ferraresi Valtair Antonio (2005) Compari-
FMPI 0.926 0.485
son between genetic algorithms and response surface methodol-
ogy in GMAW welding optimization. J Mater Process Technol
160:70–76
the study can be extended to design and optimization 5. Zain Azlan Mohd, Haron Habibollah, Sharif Safian (2011) Opti-
of the horn profile in the future. mization of process parameters in the abrasive waterjet machining
using integrated SA–GA. Appl Soft Comput 11:5350–5359
6. Elangovan S, Anand K, Prakasan K (2012) Parametric optimiza-
tion of ultrasonic metal welding using response surface methodol-
Acknowledgements  The authors express their sincere thanks to the ogy and genetic algorithm. Int J Adv Manuf Technol 63:561–572
Management and the Principal, PSG College of Technology, Coim- 7. Mukherjee Rajarshi, Chakraborty Shankar, Samanta Suman
batore, for providing the necessary support and infrastructure to carry (2012) Selection of wire electrical discharge machining process
out this work. parameters using non-traditional optimization algorithms. Appl
Soft Comput 12:2506–2516
Compliance with ethical standards  8. Mir Mir Saber Salehi, Rezaeian Javad (2016) A robust hybrid
approach based on particle swarm optimization and genetic algo-
Conflict of interest  The authors declare that they have no conflict of rithm to minimize the total machine load on unrelated parallel
interest. machines. Appl Soft Comput 41:488–504
9. Troughton MJ (2008) Handbook of plastics joining a practical
guide. PDI Publications, New York
10. Mathews Paul G (2005) Design of experiments with MINITAB.
References ASQ Quality Press, Milwaukee
11. Plastic design library (1995) Fatigue and tribological properties
1. Satpathy Mantra Prasad, Moharana Bikash Ranjan, Dewangan of plastics and elastomers. William Andrew Inc, Norwich
Shailesh, Sahoo Susanta Kumar (2015) Modeling and optimi-
zation of ultrasonic metal welding on dissimilar sheets using Publisher’s Note Springer Nature remains neutral with regard to
fuzzy based genetic algorithm approach. Eng Sci Technol Int J jurisdictional claims in published maps and institutional affiliations.
18:634–647

13