Optimization of the Factors Influencing Biodegradation and Thermal Stability of Banana Pseudo Stem Fibers in Nigeria

ABSTRACT Lignocellulosic fibers, one of natural biopolymers, are derived from abundant banana pseudo stem (BPS) agricultural waste in different parts of Nigeria. In this study, the lignocellulosic fibers of three common banana cultivars, agbagba, paranta, and omini, were investigated to determine their chemical composition and mechanical properties. Furthermore, the experimental investigations were correlated with the corresponding Taguchi L9 orthogonal array design under three factors – fiber treatment, diameter, and cultivar type – to find the optimal factors that are pertinent to the desired biodegradation and thermal stability of the fibers. The optimization results indicated that the fiber treatment followed by diameter and the cultivar type was the most influential of the responses, respectively. However, increased cellulosic content led to higher tensile strength and modulus, while higher lignin corresponded to higher elasticity. Meanwhile, the predictions of the biodegradation and thermal stability derived from the Taguchi design via S/N ratio ANOVA and regression modeling correlated adequately with the corresponding experimental observations. Ultimately, the fibers with optimum factors were T3D1C1 and T3D1C3, where T3, D1, C1, and C3 denote that the associated fiber was treated with acetic acid, had a diameter of 60.77 µm, and belong to the cultivar type of agbagba and omini, respectively.


Introduction
Banana plants have wide varieties that are cultivated in large amounts in different parts of Nigeria. Thus, in many regions of Nigeria, bananas play a pivotal role in terms of their economy and staple food supply and are an integral part of the livelihood of many households (IITA 2021;ProMusa 2020). Notably, Nigeria is the most populous nation in Africa with a total landmass of 45,000 ha available for banana cultivation (ProMusa 2020) and produces approximately 3.09 million metric tons of banana per year, thus making Nigeria the producer of the second highest amount of bananas in Africa after Cameroon (IITA 2021;Kohli, Garg, and Jana 2021). In addition, after harvesting the banana from the banana plant, the remnant banana pseudo stems (BPS), comprising 75% of the plant, are left to decompose naturally in the open space, thereby contributing to environmental pollution. Alternatively, in some regions of Nigeria, BPS fiber is partially used in making rope, sleeping mats, cloth, and agricultural storage bags. Likewise, many reports in the literature have highlighted the potential of BPS fiber as a sustainable source of raw material for several applications (Ferede and Atalie 2022), including polymeric composite making (Abd El-Baky and Attia 2019; Gideon, Atalie, and Nawab 2022;Parre et al. 2020).
In reality, practically, all chemical substances and materials are susceptible to biodegradation, with time being the most important factor. While some materials, like glass and other plastics, take many centuries to decay, things like natural fiber may do so in a matter of days. More than 89% of the original material must be broken down by biological (aerobic bacteria) into CO 2 , water, and minerals within 6 months, according to biodegradability criteria employed by the Europeans (Haider et al. 2019). Biodegradation period of solid wastes can also be investigated by the use of respirometrya process involving CO 2 generation in soil conditions simulation of water, soil, and microbes. Studies on lignocellulosic biomass of banana peels over several days yielded enough CO 2 to validate the degradation process. However, over the same period, similar process for polyethylene plastic failed to yield any CO 2 , which suggests that biodegradation might take several years (Raddadi and Fava 2019). Figure 1 details the expected biodegradation duration of materials in a marine environment as reported by Raddadi and Fava (2019). Based on this reality, there is rapid adoption of natural fiber as the filler in polymer composites in order to reduce environmental pollutions as documented in many studies (Atalie and Gideon 2018;Haider et al. 2019).
Similar to biodegradation, some research suggests that the chemical composition of natural fibers, such as cellulose, hemicellulose, and lignin, affects their thermal behavior. When it comes to thermal stability at high temperatures, natural fiber-reinforced composites perform better than pure natural fibers (Asim et al. 2020;Parre et al. 2020). The treated fibers' thermal properties significantly enhanced because of increased interfacial adhesion between the fibers and matrix. Better fiber-matrix bonds surround the fibers fully and shield them from direct temperature contact, whereas treated fiberreinforced composites often have smaller residual quantities because some lignin was partially washed away during the treatment process (Kohli, Garg, and Jana 2021;Parre et al. 2020).
BPS fiber-made products, such as agricultural storage bags, are sometimes susceptible to microbial infestation from contaminated agricultural goods and other harsh conditions. Thus, thermogravimetric analysis (TGA) and biodegradation investigation of BPS fibers will help find the optimal values of the factors influencing the fiber that would improve the durability and thermal stability of the products at elevated temperatures. Hence, this study conducted chemical composition and mechanical properties on bpsfs and further examined the biodegradation and thermal stability of the BPS fibers of three banana cultivars common in Nigeria through experiments and the Taguchi optimization technique to optimize the factors influencing the fiber and find the most influencing factors. While experimental studies on biodegradation (Kohli, Garg, and Jana 2021) and thermal properties (Parre et al. 2020;Xu et al. 2015) of banana fibers have been sufficiently documented in the literature, validation of such studies through Taguchi robust analysis is yet to be explored as per the knowledge of the authors. For this reason, three BPS fiber factors were considered in the Taguchi optimization to determine their influence and significance on the biodegradation and thermal stability of the fiber. All factors were processed under three levels, i.e., three fiber treatments (NaOH, acetic acid, and no treatment), diameters (60.77, 82.23, and 105.77 µm), and cultivar types (agbagba, paranta, and omini) were considered in the optimization. Finally, while ranking the influence and significance of the factors, their percentage contribution and significance level were considered (Raddadi and Fava 2019).
Subsequently, the BPS of each banana cultivar was made into ribbons (Figure 2 (a)), and the BPS fibers were manually extracted from healthy ribbons by following a procedure similar to the one in the work of Oyewo et al. (2022). The extracted BPS fibers of the banana cultivars are shown in Figure 2 (d)-(f). Furthermore, sodium hydroxide and acetic acid solutions (Figure 2 (b) and (c), respectively) were procured from Bond Chemical Industries Ltd., Lagos, Nigeria.

Chemical composition
According to the procedures outlined in Sanjay et al. (2019), lignin, -cellulose, and other substances that contain fibers were studied in textile materials. Before undergoing chemical treatment, the fibers were heated to 110°C until they attained a constant weight, at which point the moisture content was calculated. The starting weight and ending weight were recorded to calculate the percentage water content. And the E1755-01 method was used to determine the ash content (Sanjay et al. 2019).

Mechanical properties
The tensile strength of bpsfs was measured using a universal tensile machine WMS-2 (Ginan Fsid Universal Instrument, Limited, Beijing, China) at 34°C and 60%, controlled temperature and standard humidity, respectively, according to the work by Xu et al. (2015). Ten samples were used to test each cultivar, and the standard deviation of the data was used to calculate the variations.

Fiber treatment
First, each BPS fiber was immersed in a 5 wt. % sodium hydroxide solution for 2 h and washed generously under running water as described elsewhere (Oyewo et al. 2022) for sodium hydroxide treatment, and the above procedure was followed by replacing sodium hydroxide with 2 wt. % acetic acid for the acetic treatment. Furthermore, 20 samples of the BPS fiber of each banana cultivar, with three different diameters, were observed through a digital microscope, and their average diameters were recorded.

Biodegradation
As detailed in Table 2, nine samples were generated from Taguchi optimization due to different combination of controlling Subsequently, as per the procedure reported in the work of Kohli, Garg, and Jana (2021), 40 g of each fiber sample was buried in a covered container with 25%, 24%, 23%, and 28% of loamy soil, cow dung, sandy soil, and distilled water, respectively, for 90 days at room temperature (29.5 ± 1.1 °C), to incorporate the soil conditions on the sample. Also, the samples were intermittently monitored, and their weight was recorded during burial. At the end of the 90 days period, the samples were recovered, and % weight reduction was calculated.

Thermo gravimetric analysis
Furthermore, the TGA of each sample was carried out by heating 10 mg of the dried sample from 30 to 500 °C with the aid of Saglar TGA/SRT 801 instrument as described in the work of (Parre et al. 2020). So, the thermal stability of the sample through the percentage of leftover (residual mass) from the above heating was calculated by considering the difference between the final and initial mass of the sample as the leftover mass.

Design of experiment and Taguchi method
The control factors affecting BPS fibers and their respective levels selected are presented in Table 1. In addition, using the Minitab 19 software, Table 2 presents the L9 orthogonal array, signal-to-noise ratio (S/N ratio). Additionally, Table 2 presents the percentage mass leftover from thermal stability and biodegradation responses of BPS fibers from the experiments. Notably, the fiber treatment, diameter, and cultivar were designated as T, D, and C, respectively.
Necessarily, the Taguchi method enables experimental work to be validated with corresponding theoretical modeling. In contrast to other validation methods, including the factorial analysis, reduction in experimental quantity for the evaluation/modeling is evident with Taguchi analysis, thereby saving time and the cost of the study. In addition, the noise in the system or S/N ratio of the outputs is also evaluated through the Taguchi method. Generally, a high S/N ratio or noise is unwanted in any system. However, the smaller the better, the normal the better, and the bigger the options in the Taguchi optimization selection for any experiment (Oyewo et al. 2022). In this study, a strong output signal for the fiber responses was desired; hence, the larger the better option was selected and expressed (Equation 1) (Jabbar et al. 2017), the signal À to À noise ratio for larger the better S=N ¼ À 10 log 1 n where n was the observation number(s) and X denoted the observed data of the fiber response.

Analysis of variance
An analysis of variance (ANOVA) was performed on the fiber response data of this study to determine the influential factor(s) among the fiber treatment, diameter, and cultivars. In addition, ANOVA was used to obtain the percentage contribution and significance of each factor. Other means that validated the theoretical modeling of this study included regression analysis, probability plots, confirmation tests, and Pareto charts.

Chemical composition and mechanical properties of bpsf
The mechanical characteristics of bpsfs are presented in Table 3 along with their chemical compositions. Agbagba, paranta, and omini were discovered to have the highest concentrations of cellulose and hemicellulose, respectively; paranta had the highest concentration of lignin. The mechanical characteristics of natural plant fibers are significantly influenced by the cellulose concentration and cellulose crystallinity of the material. As a result, cellulose content increases along with tensile strength as presented in Table 3. A high lignin concentration has been discovered to improve the natural fibers' capacity to stretch. The same result was attained by Petroudy (2017) who found that leafiran fibers have a lignin content of about 25%, resulting in a higher degree of elongation than kenaf, jute, and pineapple, which have lignin contents of 17%, 9%, and 8.3%, respectively.

Biodegradation
Biodegradation test of the samples is presented in Figure 3 (a). Although all the samples underwent degradation gradually until the duration of the simulation experiment, the degree of biodegradation varies in some samples due to different chemical constituents. However, the percentage leftover mass for each sample after 90 days is presented in Table 2. In essence, sample number 7, T3D1C3 (i.e, treatment level 3, acetic acid; diameter level 1, 60.77 µm; and cultivar level 3, paranta) had the optimal response, while the least preferred response was that of sample number 3, T1D3C3 in both tests. In terms of microbial susceptibility to biodegradation, hemicellulose, non-crystalline cellulose, cellulose, and lignin come in that sequence, from least to most susceptible. In plants, lignin's primary function is to safeguard the structural integrity of the cell and bind other components to hemicellulose. Due to the presence of aromatic and heterogeneous cross-linked in its structure, biodegradation of lignin is not easily accomplished. Lignin is therefore the component of natural fiber that takes the longest to break down. The pace at which lignin is susceptible to microbial attacks determines how easily fibers can degrade (Luthra et al. 2020).

Thermo gravimetric analysis
The percentage mass residual obtained from the biodegradation and TG analyses of the BPS fibers are presented in Table 2. According to Figure 3 (b), the degradation of the fibers took place between 31°C and 550°C and had three primary weight loss zones for all samples. Like the biodegradation test, the highest leftover mass residue was found in sample number 7, while the least was found with sample number 3. The degradation temperature is divided into three stages: initial, middle (50%), and final (called fiber mass residue). All three stages can be visualized in Figure 3 (b). Final decomposition temperature covered the breakdown of non-cellulosic materials of lignin, the component of banana fiber that is hardest to break down (Asim et al. 2020). Due to the removal of alkali-sensitive sites with greater alkali concentrations, which are also active sites for moisture absorption, the amount of water absorbed in these decreased. Additionally, improved lignin, wax, and hemicellulose removal from the

Signal-to-noise ratio
The rank and order of the factors influencing the BPS fiber (fiber treatment, diameter, and cultivar) on the experimental fiber response, as well as the optimal level of each factor, were obtained via the S/N ratio. In particular, after finding the difference between the highest and the lowest levels of each factor (known as delta), the factors with the highest positive delta were ranked as the most important factor (Oyewo et al. 2022). Via delta ranking, the order of influential factors is determined as treatment (first), diameter (second), and cultivar type (third). Accordingly, the S/N ratio of the biodegradation response of the BPS fibers is presented in Table 4. Furthermore, Figure 4 presents the optimal level for each factor. The optimal level for biodegradation is T3D1C3 -which can be considered as the treatment at level 3 (acetic acid), diameter at level 1 (60.77 µm), and cultivar level 3 (paranta). Prediction of T3D1C3, as obtained from S/N ratio, satisfactorily corresponds to the results obtained from biodegradation experimentation (sample number 7). Furthermore, the S/N ratio from the TGA of the BPS fibers was used to determine the order of the influential factors (Table 5) and the optimal level of each of these factors ( Figure 5). Similarly, in the order of importance, the S/N ratio of TGA also ranks fiber treatment, diameter, and cultivar type, as the influential factors. Thus, the optimal level for TGA is T3D1C1 -which can be interpreted as the treatment at level 3 (acetic acid), diameter at level 1 (60.77 µm), and cultivar level 1 (omini). Notably, Taguchi optimization analysis prediction of T3 and D1 was the same value obtained with the   biodegradation experimental results. However, cultivar type gave C3 (paranta) in experimentation and C1 (omini) in the optimization.

Analysis of variance (ANOVA)
The results of the ANOVA of the S/N ratios from TGA and biodegradation of BPS fibers are presented in Table 6. The confidence interval for the ANOVA of the S/N ratios of the BPS fiber responses was 95%, and the remaining 5% (0.05) was regarded as the probability of failure (p-value). Thus, a parameter that showed the significance of each level of the factors influencing the fibers was the p-value. So, since a 95% confidence interval was used, a factor that had a p-value less than 0.05 was considered significant, and subsequently, the null hypothesis was discarded. In contrast, when the p-value of a factor was greater than 0.05, the factor was considered insignificant, and the null hypothesis was accepted. Meanwhile, the p-value and the contribution percentage of individual factor  toward the optimization outcome were instrumental in the ANOVA. Table 6 also gives the percentage contribution of the factors obtained from the TGA of the fibers. While the S/N ratio is used to determine the order of influential factors, it cannot be used to evaluate the percentage contribution of each factor. Both tests have the highest percentage contributions in fiber treatment (up 72.26%), followed by diameter (up 23.77%) and cultivar type (2.53%), which supports the order determined by the S/N ratio. Additionally, Table 6 shows the p-values for TGA and biodegradation. The null hypothesis is disapproved in both cases where the fiber treatment and diameter p-values are less than 0.05. The cultivar type, however, is more than 0.05, hence the null hypothesis is rejected.

Regression modeling
Regression equations for the optimization of the influencing factors of the BPS fibers based on the TGA and biodegradation of the fibers are the Equations (2) and (3), respectively. In particular, in the regression modeling of the BPS fiber responses, the confidence of determination (R 2 ) in the TGA responses of the fibers was 93.75%, whereas that of the biodegradation responses of the fibers was 88.59%. In both cases, R 2 was within the acceptable range. Furthermore, for each response, the values of all the factors were inserted into the corresponding regression equation to generate the predicted values, called the fits. The fits were then correlated with the corresponding experimental values to determine the exactness, variation, and accuracy of the modeling and are presented in Table 7. In particular, the percentage modeling error was calculated by dividing the difference between the experimental value and the fit, called the residual, by the experimental value. So, for the TGA and biodegradation, the percentage error was within −0.70-9.98 and −0.46-10.511%, respectively. These errors, within 10%, were acceptable aaccording to Oyewo et al. (2022); (Abd El-Baky and Attia 2019). As such, there was a sufficient correlation between both the experiments and the modeling, given the fiber responses. The regression equation for the TGA of the BPS fibers: 4:253 þ 1:012 fiber treatment À 0:02713 fiber diameter þ 0:045 fiber cultivar (2) The regression equation for the biodegradation of the BPS fibers: 18:02 þ 6:00 fiber treatment À 0:0713 fiber diameter þ 1:77 fiber cultivar (3)

Probability plot and Pareto chart
The normal probability plots of the TGA and biodegradation of the BPS fibers are given in Figure 6 (a) and (b), respectively. Furthermore, the regression equation efficacy and the significance of the factors influencing the fibers in relation to the fiber responses were validated with the help of the probability plots. So, at α equal to 0.05, there was a 5% probability of error between the experimental and the model predictions of the responses, that is, the confidence interval (CI) remained at 95%. Next, at this CI, a linear probability plot was generated between the residual and the percentage probability of the responses. Meanwhile, according to Oyewo et al. (2022) and Arıcı, Çelik, and Keleştemur (2021), the residuals must be adequately fitted with a linear probability plot without many outliers. Here, the residuals in Figure 6 (a) and (b) for the fiber responses from the TGA and biodegradation of the fibers, respectively, were very close to the linear probability plot without undue variations. Furthermore, the Pareto charts for the BPS fiber responses from the TGA and biodegradation of the fibers are shown in Figure 6 (c) and (d). Unlike the linear probability plot, the Pareto chart is used to determine the significance of the influential factors by corroborating the S/N ratios and the ANOVA of the fiber responses. Procedurally, first, a standardized reference line (2.571) was generated automatically in the Pareto chart which is useful to determine the significance of a factor. Then, any bar above this reference line was considered to correspond to a significant factor; otherwise, the bar was considered to correspond to an insignificant factor.

Confirmation test
Validation of the experimental and Taguchi model-predicted BPS fiber responses via a confirmation test was necessary to ensure their usability. As such, an improvement in the S/N ratio of the responses toward finding the optimal influencing factors of the biodegradation and thermal stability of the fibers was achieved using the following Equation 5 .
where α l indicated the mean S/N ratio, α 0 was the mean S/N ratio at the optimal level of the influencing factor, and y represented the number of influencing factors. Therefore, for the confirmation test, the average value representing the fiber responses from the experiment was compared with the corresponding experimental and the model predicted values. Eventually, the TGA and biodegradation responses had the experimental and predicted values compared to the average value, T2D2C3. In addition, since the optimum fiber predicted from the TGA (T3D1C1) was not experimented with before in this study, a new experiment with T3D1C1 was run, and the comparison was made with the best experimental value (T3D1C3) and average value. So, this new test (T3D1C1), i.e., the optimum prediction from Taguchi analysis, showed an improvement of 3.72% with the experimental value and 29.8% with the average value. Equally, a comparison made using regression analysis of experimental, Taguchi prediction, and average value is presented in Table 8. Finally, the confirmation test was performed based on a mean with, where β was the mean value of the response parameter; T, d, and C were the fiber treatment, diameter, and cultivar at the optimal level, respectively; and Y was the mean of the response parameter. The comparison between the average value of the experimental and model-predicted BPS fiber response parameters from the biodegradation and TGA of the fibers is presented in Table 8. In general, a reduction in the S/N ratio improves the efficiency of the optimization process (Oyewo et al. 2022). Following diameter and cultivar selection as determined by ANOVA analysis (% contribution) and s/n ratio, fiber treatments had the greatest impact on all outputs. The thermal stability and biodegradation of natural fibers are thus governed by a number of important elements, some of which include chemical treatments and fiber diameter. In affirmation to this claim, the study of Ezeamaku et al. (2022), investigating the effectiveness of some chemical treatments on banana fiber, also found improvement in tensile strength with acetic acidtreated banana fiber (79.67%) than potassium permanganate (77.78%) and sodium hydroxide (60.71%), the least. Similar to chemical treatments, the effect of lower diameter on natural fiber has been outlined in this study. Additionally, a study by Gangil et al. (2020); Masood et al. (2018) and microscopy research by Jaafar et al. (2019) demonstrate that smaller diameter fibers have better fiber dispersion in the matrix than larger diameter fibers, making them suitable for textile applications , and a potent challenger to synthetic fibers .
In addition to the above results, the results of the validation of the experimental and Taguchi optimization method-predicted responses of the fibers in the form of 2-dimensional contour and 3-dimensional surface plots are shown in the supplementary data section of this article.

Conclusion
The work aimed at studying and analyzing the optimal values of the factors influencing the biodegradation and thermal stability of banana pseudo stem (BPS) of three common banana cultivars of Nigerian origin using theoretical modeling and experiments. All bpsf fibers were first examined for chemical composition and mechanical characterization. The three principal chemical components were cellulose, hemicellulose, and lignin. Agbagba and omini had the highest amounts of cellulose and hemicellulose (68.61% and 65.17%), while omini had the highest lignin content (7.51%). The chemical composition and mechanical characteristics showed a strong association, with higher lignin concentration corresponding to higher elasticity and higher cellulosic content corresponding to higher tensile strength and tensile modulus. In particular, three influencing factors were selected: fiber treatment, diameter, and fiber cultivar in the study/analysis. The subsequent prediction from the optimization using the Taguchi method indicated that the fiber treatment was the most influential factor of thermal stability while fiber treatment and diameter were the most influential factors of biodegradation of the fibers. Next, the optimal factors, found using the S/N ratio and Pareto charts, correspond to samples T3D1C1 and T3D1C3 for the biodegradation and thermal stability of the fibers, respectively; with T3 denoting 5 wt. % NaOH-treated fiber, D1 denoting a fiber diameter of 60.77 µm, and C1 and C2 denoting agbagba and omini, respectively. Finally, the experimental fiber response parametric values were found to be well correlated with the model predicted fiber response parametric values via probability plot and regression modeling, as the percentage error between the experimental and predicted values was less than 10%.