Entropy-TOPSIS appraisal of brake friction linings developed from composite agricultural wastes using optimum manufacturing process parameters

In this study, the optimum process parameters for the manufacture of brake friction linings (BFLs) from palm kernel shells (PKS), periwinkle shells (PWS), and coconut shell (CNS) composites were established using signal–to–noise ratio based on the Taguchi technique. The L9(34) orthogonal array was set up for the investigation of the performance metrics (coefficient of friction, wear rate, and hardness) synergized by multiple criteria evaluation. The manufacturing parameters considered were molding pressure, molding temperature, curing time, and heat treatment time. Consequently, the optimized parameters were utilized for the production of different BFL composites of the PKS/PWS/CNS mix. Finally, entropy and TOPSIS techniques were employed to isolate the best composite for comparative analysis. The results show that the optimum process parameters are 29 MPa (molding pressure), 120 °C (molding temperature), 6 min (curing time), and 2 h (heat treatment time). ANOVA using Minitab 21.1.0.0 shows that the effects of the molding pressure and curing time are statistically significant at α = 0.05, with a total contribution of 94.45%. The entropy-TOPSIS analysis gave sample S2pkpc with a composition of 12% PKS, 15% PWS, and 18% CNS as the best composite. Compared to the asbestos BFL, the composite shows an improvement in friction coefficient (45.7%), wear rate (66%), density (60.2%), and oil and water absorption (233%) (542.8%) respectively. The live test on a Peugeot 301 using S2pkpc BFL confirms the satisfactory performance of the composite. However, an increased wear rate was observed at vehicle speeds above 90 km/h.


Introduction
Automobile brakes are energy conversion devices that convert the kinetic energy of a rotating disc or drum into thermal energy. The result is vehicle deceleration [1]. The friction lining is the key component, as it provides the contact surface with the rotating disc. The development of friction lining has evolved for more than 110 years. Asbestos was introduced in 1908 because it possesses stable physical and chemical properties over a wide range of temperatures, presenting a significant advantage over Frood's textile-rubberimpregnated linings developed in 1897 [2]. However, there is an established health risk in the use of asbestos. Diseases associated with its use include asbestosis, mesothelioma, lung, ovarian, laryngeal cancer, and atelectasis [3]. Hence, there is ongoing research into the production of brake friction linings (BFL) from environmentally friendly agricultural waste to meet the demand in automotive applications. This waste utilization is not only economical, but can also result in foreign exchange revenues and pollution control [4].
Ibhadode and Dagwa [2] used palm kernel shells (PKS) to develop eco-friendly brake pads. Optimal formulation and manufacturing constraints were achieved by the Taguchi technique (molding pressure of 16.74 MPa, molding temperature of 160 °C, curing time of 8 min, and heat treatment time of 2 h). Wear and effectiveness tests were carried out on the developed brake pad using a car, and satisfactory performance was reported at low vehicle speeds. Therefore, the study suggested investigating the effect of particle size on wear rate at high vehicular speeds. Mohammed et al. [5] investigated the effect of PKS particle sizes when used as a friction material for non-asbestos brake pads. Particle sizes of 100 μm, 200 μm, and 350 μm were investigated. Hardness, flame resistance, wear, compressive strength, porosity, and water absorption tests were performed on three samples. The result compared favorably to the commercial brake pad of a pathfinder jeep. The optimal performance was achieved with the grain size formulation of 100 μm. This affirms that the grain size has a significant effect on the braking characteristics of PKS brake friction lining.
The design mix and process parameters for the production of eco-friendly brake pads from periwinkle shells (PWS) and fan palm shells were optimized by Amaren [3]. Signalto-noise ratio technique was used to optimize the coefficient of friction performance metric. An optimal formulation of 35% resin and 65% periwinkle or fan palm shell was established. The control factors at optimal settings were a molding pressure of 41 kPa, molding temperature of 150 °C, curing time of 10 min, and heat treatment time of 1 h. The BFLs developed from 125 μm particle size showed improved mechanical properties with respect to the coefficient of friction (8.57% average), hardness (15.54%), and compressive strength (33.6%).
Gray relational analysis (GRA) was used by Abutu et al. [6] to optimize the production process for coconut shell (CNS) BFL. The supplementary mixture contained epoxy resin, graphite, and aluminum oxide. The optimized process values, as determined by the analysis, are molding pressure of 14 MPa, molding temperature of 140 °C, curing time of 8 min, and heat treatment time of 5 h. Heat treatment time had the most significant effect on the wear rate and friction coefficient, while the hardness and ultimate tensile strength were most affected by the curing time. Bashar et al. [7] investigated the suitability of CNS powder for automotive BFL applications with a grain size of 710 μm. The effect of varying the composition of CNS and epoxy was investigated. It was established that a higher quantity of CNS increases the brittleness of the friction material developed. In each case, the samples with 50% and 60% epoxy and 10% reinforcement had better properties and were recommended for the production of brake pads. In addition, CNS powder was noted to have good brake properties, with an advantage in wear resistance over asbestos. The presence of cast iron fillings in this study was found to cause corrosion of samples and was identified as a significant drawback.
In this research, the optimum process parameters for producing eco-friendly BFL from palm kernel shell, periwinkle shell, and coconut shell composites (PKS/PWS/CNS mix) were established using the signal-to-noise ratio technique. Consequently, various composites of BFLs developed were appraised using entropy and the technique for ordered preference by similarity to the ideal solution (TOPSIS) in order to find the best composite. A comparative analysis of the mechanical and wear properties of the best composite was examined for advantages in relation to asbestos, PKS, PWS, and CNS BFLs.

Materials and method
The materials used for the production of the BFL samples are PKS, PWS, CNS (fillers), epoxy resin (matrix), powdered graphite (lubricant), alumina (abrasive), and bamboo fiber (reinforcement) (Fig. 1). The use of agricultural waste was prioritized, and the particle size of 100 μm was adopted in accordance with recommendations from previous literature [4,5,8].

Sample preparation
The samples were produced by powder metallurgy. The PKS, PWS, CNS, and bamboo fiber (30 kg each) were sundried for 7 days, then transferred to an electric oven (model: Memmert, Western Germany) for further drying at 105 °C for 5 h. The dried materials were transferred to a hammer crusher (type: 000T, PUISSANE: 1.5kv, No: 13634) in order to reduce the size to 2-4 mm. Further crushing to powder was achieved using a ball milling machine (model: 87002 Limoges, France). Thereafter, a BS 410 standard sieve was used to achieve a consistent particle size (100 μm). A homogeneous mixture was obtained by using a mixer (model: 89.2 Ridsdale & Co Ltd, Middlesbrough, England) to mix the powdered materials for 20 min at the percentage composition indicated in Table 1. The composition was fixed in accordance with the range of values presented in the literature, which resulted in improved characterization results [9][10][11]. The powered materials were transferred to a container of epoxy resin and hardener mixture (2:1 proportion), stirred until a past-like substance was formed, and then transferred to a mold for cold pressing.
Cold compaction was performed using a cylindrical die cavity (mold) closed at one end with a diameter of 30 mm at room temperature and a pressure of 15 MPa provided by a uniaxial hydraulic hand press. Another mold was formed to take the shape of a Peugeot 301 brake pad (Fig. 2). The objective of cold press is to eliminate deformations, such as creep and diffusion, associated with high temperature compacting. Hot pressing (compacting) was carried out to ensure the rigidity of the BFL samples. The manufacturing parameters adopted (molding pressure, molding temperature, curing time, and heat treatment time) were set up using the orthogonal Taguchi L 9 (3 4 ) array as shown in Table 2. After the hot press, the samples were allowed to cool at room temperature. Post-curing was performed on the test samples in an oven to ensure they were fully cured. The process flow for producing the test samples is summarized in Fig. 3.

Design of experiment (DoE)
As outlined by Ibhadode and Dagwa [2], the most important process parameters for the manufacture of brake pads are molding pressure, molding temperature, curing time, and heat treatment time. Therefore, Taguchi L 9 (3 4 ) orthogonal array ( Table 2) was set up to investigate the optimum process parameters at three respective levels (low (−), medium (0), and high (+)) of the performance metrics. The level   values are obtained by creating equal intervals between them, as is often done in the manufacture of brake pads [3].
The Taguchi method provides a means of conducting fewer experiments than any experimental design method based on a statistical approach. A performance measure known as the signal-to-noise (S/N) ratio is used to control experimental noise or variability. Generally, the noise factors are responsible for experiment variability, while signal is a representation of the desired target for a good result. Thus, a high S/N ratio idealistically represents the optimized process value from the selected control levels. Using the percentage composition (wt.%) in Table 1 and the DoE in Table 2, nine samples of BFLs were produced for the optimization experiment.
Consequently, ninety-eight (98) samples were produced using the optimized process parameters. Table 3 shows the experimental design for the samples produced. The design varied the percentage composition of the filler materials (PKS, PWS, CNS) by +/− 3%, about a mean value of 15 for the 45% required in the composite.

Product characterization
The samples were characterized for the coefficient of friction, wear rate, hardness, compressive strength, density, and oil and water absorption.
The static friction coefficient was determined using the inclined plane method. The angle of inclination of a wedge at 90° was gradually varied until the specimen was just about to slide. The coefficient of friction (μ) was determined using Eq. (1).
where θ is the angle of inclination. The wear test was conducted using the pin-on-disc testing apparatus as per ASTM G99-05. Test samples were made to slide over a cast iron surface (64 HRC) with a counter surface roughness of 0.3 μm under a load of 20 N, speed of (1) = tanθ 250 rev/min, and distance of 2000 m. The initial and final weights of the samples were measured with a single pan electronic weighing machine accurate to 0.01 g. The difference in weight indicates the mass loss. The wear rate was determined using Eq. (2) [12].
Where ΔW is the change in weight of the sample before and after the test, S is the sliding distance, N is the radial speed (rpm), D is the brake disc diameter, and t is the time taken to expose the specimen to wear.
Hardness tests were carried out on the samples with a diameter of 30 mm using Brinell hardness testing equipment to BS240 on the tensometer (M500-25KN, Gunt Hamburg Hardness Tester, WP300). Based on the ASME specification, a hardened steel ball of 10 mm diameter was pressed into the test sample, and the load P was kept at 500 kgf. The diameter of indentation d was measured with an optical micrometer screw gauge. The test was repeated three times, and the mean value was taken and incorporated into Eq. (3) to obtain the Brinell hardness number (BHN) [13].
The tensometric machine (M500-25KN, Gunt Hamburg, WP300) was used for the compressive strength test. The samples were gradually loaded in compression until failure occurred. The load at which failure occurred was recorded. The density of the samples was determined using the Archimedes' principle. According to this principle, the volume of an irregular object corresponds to the volume of fluid it can displace. The weights of the samples were determined by weighing them on a digital weighing machine. The volume of water displaced was calculated by finding the difference between the mass of the object in air and its effective mass when submerged in water (density 1 g/cm 3 ) [14]. Thus, the density was calculated according to Eq. (4). Soak tests were used to determine the effect of oil and water on the weight of the samples. The samples were soaked separately in oil and water for 24 h. The weights of the samples were recorded before and after the soak test. The absorption rate was calculated according to Eq. (5).
Where Wo is the initial weight of the sample and W1 is the weight of the sample after soak.

Optimization of the manufacturing process parameters
The manufacturing process parameters were optimized for the performance metrics of coefficient of friction, wear rate, and hardness. The metrics were synergized by multiple criteria evaluation technique (Eq. (6)). Consequently, an overall evaluation criteria (OEC), which factors the relative weight of the individual criterion (metric), was developed [15,16].
Where X 1 , X 2 , and X 3 represent the average readings for the friction coefficient, wear rate, and hardness, respectively, for each sample. The first and third terms of the equation are unitless expressions for the coefficient of friction and hardness metrics, respectively (larger is better). The second term is an expression for the wear rate (smaller is better), which is converted into larger is better by subtracting its value from one (1). All terms multiply the objective weight obtained by the entropy method (Sect. 2.6).
Taguchi (SN ratio) optimization technique was subsequently used on the OEC with larger is better quality characteristics (Eq. (7)) [17,18]. Hence, the optimum manufacturing process parameters for the synergized criteria were obtained and utilized for the production of BFLs from the PKS/PWS/ CNS mix.

(4)
Density of test sample = weight in air Mass in air − Effective mass in water Where n is the number of observations in the array and y is the average of the reading

Analysis of variance
Minitab 21.1.0.0 was used in the analysis of variance (ANOVA) to investigate the effect and significance of each factor (molding pressure, molding temperature, curing time, and heat treatment time) on the performance metrics of the BFL. The analysis was carried out using the Fisher test (F-test) at a 5% significance level (α = 0.05). The process parameter/factor is considered significant if the p-value is smaller than the F0.05 [19].

The entropy-TOPSIS analysis and appraisal of developed BFL composites
The entropy-TOPSIS method, as detailed by Zlatko and Vedran [20], was used to determine the optimal composite with respect to the mechanical and tribology properties examined (coefficient of friction, wear rate, hardness, compressive strength, density, oil and water absorption). Thereafter, a comparative analysis of the optimal composite with BFLs of PKS, PWS, CNS, and asbestos was carried out. The standard stepwise procedure was adopted in the analysis, as outlined below.
i. Vector normalization of the decision matrix (matrix of response data for performance metrics) was done using Eq. (8) by dividing each criterion value x ij by the square root of the criteria sum of squares The operation is represented in Eq. (8).
ii. The objective weight of each criterion was determined using the entropy method by finding the weight vector w j of the criterion using Eq. (9).
iii. The weighted normalized decision matrix v i, j was determined using Eq. (12). The weight vector w j is multiplied by each criterion in the normalized matrix.
iv. The positive ideal solution v + j (extreme performance on each criterion) and negative ideal solutions v − j ) (reverse extreme performance on each criterion) are identified. Consequently, the separation of each alternative from the positive ideal solution ( d + i ) and negative ideal solution ( d − i ) was obtained by (Eqs. 13 and 14) respectively for the n-dimensional matrix.
v. Finally, the relative closeness to the positive ideal solution (R i ) was evaluated using Eq. (15). Hence, the Therefore, h = 1 In (7) = 0.514 1 − e j = d j = the degree of diversification e j is the entropy value (12) v i,j = w j n ij for i = 1, … , m;j = 1, .., n where w j is the weight of the jth criterion respective rank of the specimen was utilized in isolating the ideal (optimal) composite.

Vehicle live test
A live test was conducted using a Peugeot 301 car. The developed brake pads (S2 pkpc ) and the standard (commercial) brake pads were installed simultaneously on the vehicle. This ensures that both BFLs are subjected to the same conditions during brake force application. The test was carried out by bringing the vehicle to a stop after driving for 5 min at the following speeds (km/h): 60, 80, 100, and 120. A comparative wear analysis of the BFLs was conducted in each instance.

Result and discussion
The average coefficient of friction, wear rate, and hardness values obtained from the optimization experiments and the SN ratio of the respective performance metrics are presented in Table 4. Table 5 shows the values of OECs and the respective SN ratios based on larger-is-better quality characteristics. Fig. 4 depicts the optimum manufacturing process parameters for producing BFL of PKS/PWS/CNS mix as  [21]. ANOVA (Tables 6 and 7) revealed that the most significant factor influencing the synergized quality characteristic is the molding pressure. This factor has a percentage contribution of 55.51% and is closely followed by the curing time (38.94%). Therefore, high molding pressure and less curing time will improve the overall quality of the BFL. Although the curing time is a monotonic function of time, insufficient curing time will lead to an unacceptable hard BFL, which can cause disc surface scoring and must be avoided. Finally, the molding pressure and curing time are statistically significant, as their p-values are less than 0.05.

Analysis of the developed composites
The results for the mechanical and wear properties of the PKS/PWS/CNS (pkpc) composites are shown in Table 8. The average of two tests was used to improve the accuracy of the results. Hence, fourteen (14) samples were tested to report the values of each performance metric represented in Figs. 5, 6, 7, 8, 9, and 10.

Coefficient of friction
The coefficient of friction of the composites is represented by Fig. 5. All recorded values for the composites exceed the minimum value of 0.3 recommended for automotive brake friction lining by the Standard Organization of Nigeria (NIS:323,1997) [22]. It can be seen from Fig. 5 that sample S2 pkpc gave the best value of friction coefficient (0.51). This sample has a filler composition of 12%PKS, 15%PWS, and 18%CNS. Figure 6 represents the wear rate of the composites. It is observed that sample S2 pkpc (12%PKS, 15%PWS, and 18%CNS) has the lowest wear rate in all composites investigated. This is followed by samples S4 pkpc and S7pkpc, which have equal values for wear.

Hardness of the composites
The hardness test results for the composites are presented in Fig. 7. It is observed that all composites except S3 pkpc meet the minimum hardness value (27BHN) recommended by the Standard Organization of Nigeria for automotive BFLs (NIS 323:1997) [23]. Sample S2 pkpc with a composition of 12%PKS, 15%PWS, and 18%CNS recorded the best hardness value of 38.23.

Compressive strength of the composites
The compressive strengths of the PKS/PWS/CNS composites are represented in Fig. 8. Sample S4 pkpc (18% PKS, 12% PWS, and 15% CNS) gave the best value of compressive strength. Figure 9 represents the densities of the composites. The density values for all composites are lower than the density of commercial brake pads (1.89 g/m 3 ) (Table 11). This shows that brake pads made from PKS/PWS/CNS BFL are lighter and contribute to better fuel efficiency. From Fig. 9, Sample S4 pkpc (18%PKS, 12%PWS, and 15%CNS) recorded the best density value of 1.08.

Oil and water absorption
The absorption of the samples is represented in Fig. 10. The data suggest that the sample S6 pkpc gave the best value for oil absorption, while S2 pkpc gave the best value for water absorption. Sample S6 pkpc is the optimal composite in swell rate, considering the averages of oil and water absorption.

Appraisal of the optimum BFL composite by entropy-TOPSIS analysis
The optimal composite with regard to the combined quality characteristic examined in this study was isolated by the entropy-TOPSIS technique (Sect. 2.6). The objective weights has the highest rank (R i = 0.791). Accordingly, the S2 pkpc is the optimal composite of this study.

Comparative analysis of the optimum composite (S2 pkpc ) with alternatives
A comparative analysis of S2 pkpc with BFLs of PKS, PWS, CNS, and asbestos is presented in Table 11. It is observed that the coefficient of friction, wear rate, and oil and water absorption in S2 pkpc is generally improved over PKS, PWS, CNS, and asbestos BFLs. Specifically, the coefficient of friction is improved by 15   density is improved by 39.8% (PKS), 5.1% (PWS), and 60.2% (asbestos). Table 12 summarizes the deviation (%) of the properties of S2 pkpc from the alternatives. Table 13 and Fig. 12 compare the thickness loss of the asbestos brake pad with the optimal brake pad (S2 pkpc ) (Fig. 11) obtained during the live test at different vehicle speeds. It was observed that the optimal brake pad gave better wear properties at vehicle speeds below 90 km/h. Above this speed, the wear rate was observed to increase slightly. S2 pkpc competes favorably during the test. However, the wear rate increases by 7.14% and 13.3% at 100 km/h and 120 km/h respectively. The decomposition of the matrix (epoxy resin) begins at 149 °C, while the disc-pad interface temperature at normal driving speed (less than 50 mph) is within the range of Table 11 Comparative analysis of optimum PKS/PWS/CNS composite with PKS, PWS, CNS, and asbestos BFL a Adeyemi et al. [12] b Fono-Tamo and Kayo [24] c Yawas et al. [11] d Elakhame et al. [13] e Aku et al. [25] f Pathmanaban et al. [10] g Abutu et al. [6] h Adaokoma et al. [23] i Singh et al. [ 28]. Since the optimal and asbestos BFLs were subjected to the same braking conditions during the test, it can be concluded that the higher wear rate of the optimal BFL at 90 km/h is attributed to the matrix decomposition due to the high temperature generated at this speed. Epoxy resin was used in the present research due to its local availability and low cost.

Conclusion
The following conclusions are drawn from this research: 1. The process parameters for the manufacture of ecofriendly brake friction lining have been optimized by the SN ratio technique. The parameters are respectively 29 MPa (molding pressure), 120 °C (molding temperature), 6 min (curing time), and 2 h (heat treatment time).
These parameter sets can safely be used to produce BFLs of different compositions of PKS/PWS/CNS mix within the process window. 2. The effect of the molding pressure and curing time are statistically significant at α = 0.05, with a total contribution of 94.45%. 3. The optimal composite (S2 pkpc ) established by the entropy-TOPSIS technique has a composition of 12% PKS, 15% PWS, and 18% CNS. Compared to asbestos BFL, the composite shows an improvement in friction coefficient (45.7%), wear rate (66%), density (60.2%), oil absorption (233%), and water absorption (542.8%) respectively. 4. The results of this study indicate that eco-friendly brake pads can be effectively produced from a PKS/PWS/CNS mix.
Acknowledgements The authors thank the Federal Institute of Industrial Research Oshodi, Nigeria, and PAN Nigeria Ltd for providing Author contribution E.D Columbus: writing original draft, project administration, investigation and formal analysis. T.I Ogedengbe: conceptualization, methodology, supervision, and validation.

Data availability
The data used to support the findings of this study are available from the corresponding author upon a reasonable request.

Declarations
Competing interests The authors declare no competing interests.