Abstract
High specific strength, strength-to-weight ratio, cheap cost, and other advantages, nanofillers are now the subject of most research on natural fibers. The current research’s main goal is to combine the Taguchi and artificial neural networks (ANN) approaches to maximize the mechanical characteristics of nanocomposites. The parameters: (i) nano-SiO2 wt%, (ii) banana fiber wt%, (iii) compression pressure in MPa, and (iv) compression molding temperature in °C were selected to achieve the objectives above. An L16 orthogonal array was used to optimize the process parameters based on the Taguchi technique. According to the intended experiment, mechanical characteristics, such as tension, bending, and impact strength, were assessed. The ANN was used to forecast outcomes that were optimized. The fiber mat thickness of banana fiber and the weight ratio of nano-SiO2 showed a considerable improvement in the mechanical characteristics of hybrid composites. According to the Taguchi technique, the most significant mechanical characteristics were 47.36 MPa tensile, 64.48 MPa flexural, and 35.33 kJ of impact under circumstances of 5% SiO2, 19 MPa pressure, and 110 °C. With 95% accuracy, ANN-predicted mechanical strength. The ANN forecast was more accurate than the regression model and experimental data. The above nanobased hybrid composites are mainly employed to satisfy the needs of the contemporary vehicle sector.
1. Introduction
Due to the overuse of plastic materials, which results in numerous pollution problems and health problems for all living things, the entire world is at risk. Utilizing bio-based natural materials like jute, hemp, wheat, pineapple, banana, coco fiber, wool, ramie, and roselle can significantly minimize this [1, 2]. The concentration of lignocellulosic biomass, hemicellulose, and lignin determines the qualities of natural fibers, which vary depending on the production region. These fiber combinations are utilized for various purposes, such as packing, thermos insulators, vibration-dampening materials, and building materials [3, 4]. It demonstrated lower mechanical properties than synthetic fibers like fiberglass, graphene, glass, and obsidian. The mechanical performance of polymeric materials is increased by the hybridization of natural and synthetic fibers [5, 6]. Hybridization enhanced the composite’s characteristics because it increased their interfacial adherence. Owing to their water-absorbing nature, such natural fabrics have inferior characteristics, which results in poor adherence when using hydrophilic polymers [7]. The addition of up to 35% fiber improved the mechanical properties of the pineapple/flax-based composite materials significantly [8]. That compressed molding technique offered a compressive pressure of 14 MPa and a temperature of 120°C. At a wt% of 30%, sisal/pineapple epoxy matrix composites showed good tension, bending, and shock characteristics. Ribbed and simple type weave designs were utilized to manufacture banana/kenaf polymer composites using the hand lay-up method [9, 10]. A simple type arrangement was used to display the ultimate compressive characteristics.
In addition to mechanical qualities, thermodynamic and morphological durability, flame and chemical resistance, photoelectrochemical characteristics, and other characteristics, polymer nanocomposites constitute a novel class of composites [11]. Due to their unique qualities and multiple uses in advanced technologies, nanocomposites containing mineral fillers have garnered considerable interest [12]. Inorganic nanoparticles and polymeric matrices are mainly combined to provide the qualities of polymeric composites. Due to their high thermal properties, environmental resilience (durability), electrical, toxicological, and mechanical behavior, conductive polymers can be used as a nanocomposite matrix [13, 14]. Specific polymers like epoxy matrix are widely recognized as highly fragile. Due to this drawback, such polymers can only be used in goods with great shock and breakage resistance. The polymer nanocomposites’ mechanical properties were enhanced by adding fillers to the matrix material [15]. Nanofillers contain a lot of surface area, which makes them chemically active and facilitates easier matrix interaction. To lower production costs, address some of the drawbacks of polymers, and broaden their uses, stiff fillers can strengthen polymeric materials in various ways. The type of the polymers and the filler’s fraction determines whether additives affect such polymers’ properties [16, 17]. Several aspects of polymers, including their mechanical, thermodynamic, electromagnetic, and physical properties, can be changed by fillers. To create polymer materials and comprehend their behavior, it is crucial to have the proper weight, chemical composition, semicrystallinity, chemical solubility, and heat resistance of the polymer as well as the surface energy, chemical composition, and dispersion of the nanoparticle [18, 19]. Alternate polymerization reactions, in vivo polymerization processes, direct mixing, solution dispersal, the sol–gel technique, melt compounding, melt extruding, and injection molding are all methods for producing thermoplastic matrix composites [20, 21]. Although each method is unique, the ultimate morphologies of the nanocomposites are crucial. The thermoplastic interaction that supports effective dispersal and dispersion of the nanoparticles in the polymer matrices also has a role in determining the shape. Due to the substantial surface area between the polymer matrices and nanofillers, composite materials offer better physical and mechanical characteristics than host polymers [22, 23].
A model based on recently observed measurements must be created to forecast material quality. This might drastically reduce the additional experimental labor needed to create lightweight structures. For this aim, artificial neural networks (ANN) have recently been created in this discipline. ANN employs linked nodes (known as neurons), in which the connections are “evaluated” to simulate the functionality of a human brain [24, 25]. The biological nervous system inspired ANN. Such a network is capable of learning from mistakes and figuring out how to solve functional relationships that are rather complicated, convoluted, and multifunctional [26, 27]. Since there are no prescribed rules concerning the nature of the issue, the network defining the connection is created immediately by instances of this methodology, which distinguishes it from traditional processes. The neural network can independently establish the relationship between causes (input data) and effects (output data) throughout development [28, 29]. ANN has also been demonstrated to be an effective analytical tool for solving issues in various material science and technology fields [30, 31].
As the literature shows, selecting the proper mechanical properties parameters for nanocomposites related to natural fibers is problematic. This may be resolved by utilizing the Taguchi-based technique, which aids in determining the optimal multiresponse outcomes for bending, shock, and tension characteristics.
2. Experimental Works
2.1. Materials
Organic banana fibers from the GVR Enterprises in Madurai, India, were employed as a reinforcing material. As the matrix material, epoxy grade LY556, with lower density, excellent corrosion resistance, superior mechanical performance, and specific stiffness, was combined with curing agent HY951. The nano-SiO2 particles are supplied by Naga Chemical Industries in Chennai, Tamil Nadu, India. In Table 1, the composition and characteristics of fiber reinforcements are described. Photographic image of banana fiber, and nano-SiO2, epoxy matrix materials are shown in Figure 1(a)–1(c), respectively.
The banana fibers were extracted in the following manner. The inside peel of a banana plant is used to harvest the banana fiber. The banana stem is physically skinned in the apparatus used to harvest banana fibers. The outer peel, which has been removed and is dark and greenish, is wasted. The stem now has a white-colored snarl. The banana fiber is extracted from this stem bark. Its interior layer, which has been removed, is handled via a machine with compression crushers and a mechanism for sorting fibers. Because the white inner bark has a significant quantity of humidity inside of it, the peel must be free of such humidity. Compression rollers aid in reducing the moisture content of fibers. The white peel is removed and run through the rollers nipped to remove moisture. Humidity purges like a sugarcane juice press. The peeling stem is fed as long as possible between the roller nip by the machine operator while holding one end of a peeling stump.
2.2. Fabrication of Hybrid Composites
To improve the fibers’ crystalline nature and compliance with the matrix composites, the exterior of the fibers was treated with 5% NaOH to eliminate waste materials, wax components, certain hemicellulose, and lignin. The alkali-treated fibers underwent a thorough cleaning, were dried in a warm environment, and were then roasted in an oven at 70°C for 3 hr. The epoxy resin was thoroughly mixed with different weight proportions of nanofillers using a glass steel rod. The mixing process was then repeated with a mechanical stirrer at 30 RPM. In the end, a 10 : 1 proportion of epoxy resin and hardener was applied to the nanomixture. Banana fiber reinforcement ranging from 150 to 300 gsm-based thick fiber mat was inserted in a steel mold to perform compression molding (150× 150 ×3 mm3). The desired material in wt% is poured into the mold with reinforcing fibers at temperatures and pressures ranging from 10 to 19 MPa and 100 to 130°C. The following section provides further information on how Taguchi-based analysis was used throughout the procedure. Table 1 reveals the parameters and their levels used in the present research.
2.3. Taguchi Approach
The typical investigative plan continues to be overly complicated and impractical because it calls for several investigations. Fewer experiments are used in the Taguchi approach. The effects of process factors on mechanical qualities are shown by S/N ratio and analysis of variance (ANOVA). The Taguchi approach to the experimental process is a helpful procedure to rationally explain, investigate, and optimize various process elements to get the intended outcome. Using this method, the primary restrictions are identified by converting the outcome of the investigation into an S/N proportion. According to the compelling argument for improving the critical feature, the S/N ratio features were divided into three categories: (i) more excellent was best, (ii) nominally better, and (iii) smallest was best. Irrespective of the performance characteristic set, the performance characteristic is larger than the S/N ratio. Therefore, the level with the highest S/N ratio is the best level for the variable [17]. The composites’ flexural, tensile, and impact strengths were evaluated throughout the testing process, utilizing the more excellent principle. Since more excellent is better, S/N ratios of prominent qualities are written as follows:
2.4. Mechanical Characterization
The samples underwent mechanical testing following ASTM requirements following the compression molding process. The ASTM D 638-03 protocol was used to perform tensile testing on a universal testing machine with the highest capability of 15,000 kgf. Following the ASTM D 790 protocol, Flexural testing was conducted on the same universal testing machine. Following the ASTM D 256 standard, an Izod impact tester was utilized for hardness tests. To verify the measurements, every outcome was examined again.
3. Result and Discussion
3.1. S/N Ratio Indications
To examine the process improvements and to choose defect-free, max strength nanocomposite from a wide variety of system variables, mechanical characteristics for the L16 OA were evaluated. In this study, the distinctive feature of observed tension, bending, and impacts depend on selected features. The statistics for the tests conducted in the Taguchi experimental L16 factorial design were done using Minitab 17 software. Following the “Bigger is Better” choice, the S/N ratio determined the mechanical qualities. The best hybrid composite factors have the highest S/N ratios [32]. L16 OA and S/N ratios for mechanical characteristics are shown in Table 2. The reaction rate mean was determined for each of the 16 experiment conditions. The effect and ideal situation for the elements under consideration were determined via ANOVA and S/N ratio. S/N ratio did the most fantastic job of condensing the diversity in excellent qualities. For this investigation, larger-the-better was taken into account [33, 34].
The effect plot (bigger is better) on the S/N ratio for mechanical characteristics is shown in Figure 2(a)–2(c). The ideal conditions for all mechanical qualities were 5 wt% SiO2, 300 gsm of banana fiber mat, 19 MPa pressure, and 110°C molding temperature. The larger and better S/N reaction exhibited the most excellent mechanical properties. Tables 3–5 display the effects of the control variable on the tension, bending, and impact properties with the S/N ratio reaction.
(a)
(b)
(c)
3.2. Analysis of Variance
To evaluate the importance of discontinuous functional units, an ANOVA was performed. The process parameter known as F-test is what is thought to have an influence on the tension, bending, and impact characteristics. The proportion of contributions for every processing parameter is shown in Tables 6–8 under Column F. The contribution % for each element is shown in Figure 3. The contributing proportion is the portion of the trial’s overall difference that took every significant effect into account [35, 36].
(a)
(b)
(c)
The percentage contribution of the process variables on mechanical behavior is shown in Figure 3(a)–3(c). To reach the maximum tension, Table 6’s F-value and contributing % function as controlled variables. The p-value specifies the likelihood of recurring factors. Banana fiber mat has contributions from nano-SiO2 of 35.68%, 46.94%, pressure, and temperature of 7.52%, and 8.77%, respectively. Thus, it is abundantly evident that banana fiber mat is the main factor in achieving excellent mechanical properties. According to Table 6, banana fiber mat plays a significant role in achieving flexural strength; it provided 43.17% and 39.56% of the nano-SiO2, respectively, with pressure and temperature coming in second and third place with 2.16% and 13.28%. Banana fiber mat gives 41.14% impact strength, followed by nano-SiO2, which contributes 39.07%, molding pressure, which contributes 4.56%, and temperature, which provides 14.31%. The findings indicate that the banana fiber mat and nano-SiO2 are the main factors in the current study’s achievement of the most significant mechanical strength properties.
3.3. Regression Equation
The regression equation model was established using four control conditions and the four levels considered for this investigation. By lowering the sum of the squared royalties, the traditional regression approach was utilized to create the formulas for tension, bending, and impact. The resulting regression models are as follows:
The regression equation’s variables that exhibit favorable indications cause the mechanical characteristics to be worth more, while those that generate negative indications for the mechanical properties to be worth less. The ideal conditions for all mechanical qualities were 5 wt% SiO2, 300 gsm of banana fiber mat, 19 MPa pressure, and 110°C of molding temperature. The weight ratio of SiO2 nanofiller boosted the strength of natural nanocomposites by up to 5%. When the weight ratio of the SiO2 nanofiller was raised further, the matrix and reinforcement interaction deteriorated [37]. This is a result of the particle and resin being mixed improperly. The banana’s gsm fiber has beneficial effects on all levels. It proves that stronger banana weave will increase nanocomposite strength. It is because the existence of heavier, high cellulose components reduces the effects of vacancies. The results of the tests also showed that mechanical properties decreased as the heating rate (130°C) rose. When the fiber was exposed to intense compressive pressure, it was easy to see how quickly its mechanical qualities degraded (19 MPa). This decline in mechanical properties may be due to the thermal deterioration of natural fibers (banana) at high temperatures [27]. The lowest strength may also come from poor matrix-fiber interface strength at low compressive pressures. An optimal pressure with increased strength was discovered at a pressure of 19 MPa, which may be regarded as the greatest pressure that can be used to increase mechanical properties without affecting the fiber structure, indicating that complete reinforcing occurs [38, 39]. The surface plot of mechanical characteristics with regard to different parameters is shown in Figure 4.
(a)
(b)
(c)
(d)
(e)
(f)
3.4. Artificial Neural Networks
To match and resolve the research data gathered through various methods, ANN are utilized. This model has three layers: inputs, outputs, and hiding. The output layer informs the operators, the hiding layer processes data, and the input layer is often used for data entry. The ANN technique is used to forecast the mechanical performance in MATLAB R2015. Figure 5 shows the ANN structure of the current research. Inputs included nano-SiO2, banana fiber, molding pressure, and temperature; outputs were tensile, bending, and impact strength. A backpropagation using feed-forward for the prediction of mechanical characteristics was used. An ANN using a Levenberg–Marquardt training tool was also used. There are four neurons in each of the hidden layers. Three groups of input and output data are created using the ANN; these categories are training (60%), tests (20%), and verification (20%) [24].
Ten of the 16 measurements were selected for training, three for verification, and three for testing. The artificial neural toolkit received all of the input and output information. It serves as a stand-alone test for the networks. Four neurons are connected to four outcomes established in the output nodes, whereas one or two neurons are defined in the hidden state [40, 41]. Four neurons respond to input factors established in the input nodes. Specified input and output data sets were introduced into the 4-4-4-3 ANN design after constructing the networks. It is seen in Picture 5. The correlation analysis and the proportional confidence interval were used to evaluate the model. The comprehensive flow chart for the ANN prediction is shown in Figure 6. The developed ANN system calculates the % of expected error by using Equation (5).
Table 9 displays the statistics’ learning, verification, and assessment sets with the estimated error percentage. The average % of projected error (0.9232) was less than 3%, and the correlation for all categories was 0.9718 for tensile, 0.9567 for flexural, and 0.9620 for impact. Figure 7(a)–7(c) demonstrates the predictive accuracy of the ANN model 7(c). Figure 8(a)–8(c) illustrates the research observations with a significant link between network responsiveness. Experimental validation is done whenever the measured variables are within their ideal ranges. It is employed to assess the accuracy of the experimental findings.
(a)
(b)
(c)
(a)
(b)
(c)
The outcomes of experimental, predictive, and ANN are described in Figure 8. It has been determined that both the Taguchi and ANN approaches produce trustworthy findings based on the study of experimental data. In comparison to experimental values acquired by Taguchi, the findings of the ANN approach were higher, resulting in more excellent predictions with 95% reliability. The ANN optimum model aids in incorporating the process factors that result in the desired cast composites. Given that it saves money and time, this is ideal for natural fiber-based automobile sectors. Table 10 exhibits the optimum outcomes of mechanical properties based on the experimental, regression, and ANN results.
4. Conclusion
L16 OA was extensively used in experimental trials to evaluate the mechanical properties of the manufactured composites, such as tension, bending, and impact. The Taguchi technique’s predicted optimal parameter for nano-SiO2/banana-based hybrid composites were compared with ANN in this work. Hemp fiber mat in gsm, molding pressure, temperature, and nano-SiO2 wt% were the four parameters and levels considered for this study. The following findings were drawn from the study of the data using the conceptual Taguchi’s optimization approach, S/N ratio, ANOVA, and regression analysis:(1)The Taguchi analysis of the present research shows five wt% of nano-SiO2, 300 gsm of banana fiber mat, 19 MPa of molding pressure, and 110°C of compression molding temperature show the optimum parameters.(2)The parameters mentioned above exhibit the highest values of mechanical properties, like 47.36 MPa of tensile, 64.48 MPa of flexural, and 35.33 kJ of impact.(3)Up to 95% accuracy was achieved in predicting the mechanical performance of hybrid nanocomposites using ANN and regression models. Regression and ANN were expected to be more reliable based on the outcome of mechanical properties compared to the actual.(4)Compared to experimental results, the regression prediction shows 0.94% improvement in tensile, 1.05% in flexural, and 0.61% in impact.(5)Compared to experimental results, the ANN prediction demonstrates a 1.29% improvement in tensile, 2.71% in flexural, and 1.47% in impact.(6)As per the ANOVA analysis, the woven banana contributes above 40% to the mechanical strength compared to other parameters because the banana fiber mat acts as a primary stress transmission medium in the current research. It effectively transfers the load to the matrix.
Data Availability
The data used to support the findings of this study are included in the article. Should further data or information be required, these are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.