1. Introduction
A gearbox is a crucial part of a mechanical drive system. It is employed to improve the torque and decrease the speed from the input to the output of the gearbox. In reality, there are a wide variety of forms of gearboxes: worm, planetary, bevel, and cylindrical gearboxes (with spur or helical gears). Because they operate more smoothly and silently than spur-gear gearboxes, helical gearboxes are highly common and exceed other types of gearboxes in terms of efficiency [
1]. This is the rationale behind the efforts of numerous academics to optimize the helical gearbox.
Up to now, there has been quite a lot of research on optimizing a helical gearbox. Research on helical gearbox optimization has been fairly extensive up to this point on both single-objective and multi-objective optimization problems. The single-objective optimization problems have been conducted with different objectives, including increasing gearbox efficiency [
2,
3], reducing radiated noise [
4,
5,
6], gear box length [
7,
8,
9,
10], gearbox height [
11], gearbox volume [
8,
9,
10,
11,
12], gear mass [
11,
13,
14], gearbox cross-section area [
15,
16,
17,
18], and gearbox cost [
19,
20,
21]. The research was conducted for two-stage [
2,
3,
7,
8,
9,
10], three-stage [
14,
18,
22], and four-stage gearboxes [
9,
21]. Solving the MOOP for the helical gearbox, in particular, is highly studied as, in practice, a gearbox can only be satisfied by a large number of objective functions.
Various techniques have been employed to address the multi-objective optimization issue with a gearbox. Wang, Y., et al. [
23] provided an optimal multi-objective analysis of a cycloid pin gear planetary reducer. The study focused on analyzing the reducer volume, turning arm bearing force, and pin maximum bending stress using Pareto optimal solutions in order to minimize all three of these objectives. The study’s findings demonstrate that the modified algorithm can generate Pareto optimal solutions that surpass those generated by the regular design. Chong, T.H., et al. [
24] conducted a study on multi-objective optimization problem (MOOP) to reduce the size of the gears and minimize the vibration caused by their meshing in cylindrical gear pairs. The design method minimized the size and vibration of cylindrical gear pairs by taking into account the strength and geometric limitations. The method’s validity will be verified by the design findings presently employed in an elevator reduction drive. Park C.I. [
25] reported the outcomes of their study on addressing the MOOP for reducing helical gear noise. The study employed the design of an experiment and the response surface method as the approach to address the MOOP. The MCDM method was utilized by Chrystopher, V.T., et al. [
26] to determine the optimal gear material for a gearbox with the goal of improving the wear resistance and surface fatigue. The objective of this study is to improve the efficiency of the surface fatigue resistance when it is applied to gearboxes. Jawaz Alam and Sumanta Panda [
27] developed an approach for spur gear set design optimization that aims to reduce the gear weight, contact stress throughout the contact path, and suitable film thickness at the contact point. This work combined particle swarm optimization, particle swarm optimization-based teaching learning optimization, and Jaya methods to ensure a significant decrease in the weight and contact stress of a profile-modified spur gear set with sufficient film thickness at the site of contact. The results of the study show that the gear design is significantly better with optimum addendum coefficient values inside the design space than with traditional designs. M. Patil et al. presented a novel multi-objective optimization of a two-stage spur gearbox using the NSGA-II approach [
28]. This attempt involved two aim functions: minimum gearbox power losses and minimum gearbox volume. The results of the study indicate that there is a high probability of wear failure in solutions that are obtained using single-objective minimization. Furthermore, the total power loss is reduced by half when multi-goal optimization is used instead of single-objective optimization.
Using the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) approach, A. Parmar et al. [
29] carried out an optimization analysis of a planetary gearbox while taking into account considerable regular mechanical and tribological constraints. When comparing the results of single-objective optimization with and without tribological limits, using the study’s recommendations led to a significant reduction in the weight and power loss. With the aid of the NSGA-II method, D. Miler et al. [
30] performed a multi-objective optimization of the gear pair characteristics with the aim of lowering the transmission volume and power losses. This method was also used in [
31] to solve a MOOP, including the reduction of both the gearing mass and the flank adhesive wear speed, and in [
32] to solve an optimization problem involving the design of a multi-speed gearbox that has four competing objectives. The NSGA-II was combined with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) approach in [
33] to solve a multi-objective optimization problem of a small gearbox with three objectives: the gearbox volume, the power output, and the center distance. Three distinct techniques were used in the multi-objective optimization of a two-stage helical gear train by R.C. Sanghvi et al. [
34]: genetic algorithms (GAs), NSGA-II, and the MATLAB optimization toolbox. The study had two aims, which were to determine the minimal gearbox volume and the load-carrying capacity. The comparison of the findings shows that the NSGA-II approach produced superior outcomes than the other techniques for both objectives. Sabarinath, P., et al. [
35] looked into the multi-objective design of the transmission using helical gear pairs. The opposing number of overlap ratios and gear volumes served as indicators of the goal functions. The study employed the Parameter Adaptive Harmony Search Algorithm (PAHS) to resolve the optimization problem. Using a mono-objective self-adaptive algorithm, the optimization of tooth modifications for spur and helical gears was resolved in [
36]. The particle swarm optimization (PSO) technology served as the foundation for this tactic. The multi-objective optimization enhanced the maximal contact pressures and the transmission error signal’s root mean square values. A multi-objective micro-geometry optimization of the gear tooth with response surface methods was conducted by J.A. Korta et al. [
37]. H. Wang et al. [
38] also employed the response surface methodology approach to multi-objectively optimize the helical gear in a centrifugal compressor. Liu, S., et al. [
39] presented a MOOP work using two-stage helical gear sets in a helical hydraulic rotary actuator. The optimization was performed via the NSGA-II method, incorporating four permutations of the three objectives: volume, transmission efficiency, and maximum contact stress. The findings suggest that there is a trade-off between the maximal contact stress and the volume in the ideal region. Also, taking into account various combinations enhances the durability of the optimization outcomes. Liu, X., and Pan, Q., [
40] utilized a Discrete Non-Dominated Sorting Genetic Algorithm III (NSGA-III) to optimize a two-stage gearbox. The study investigated three objectives: volume, center distance, and power loss. It also analyzed several design characteristics, such as the number of gear teeth, module, and helix angle.
Gray relation analysis (GRA) and Taguchi techniques were recently applied by X.H. Le and N.P. Vu [
41] to the MOOP of designing a TSHG. The lowest gearbox mass and the highest gearbox efficiency were the two objectives selected for this investigation. The optimal values for the five crucial design components that contribute to making a TSHG were established using the study’s findings. In order to optimize the gearbox efficiency and reduce the gearbox volume for a two-stage bevel helical gearbox, the Taguchi and GRA techniques were also utilized in [
42]. Furthermore, in order to improve the efficiency and lower gearbox bulk, these techniques were applied to the optimization of a TSHG with in [
43]. Hung, T.Q., and Huong, T.T.T., [
44] used Taguchi and GRA methods to solve the MOOP. The aim of this work was to reduce the gearbox height and enhance the gearbox efficiency. The Taguchi and GRA techniques were also applied in [
45] to minimize the gearbox cross-section area and maximize the gearbox efficiency. The TOPSIS method was utilized to address the MOOP of a TSHG with SDGSs by V.T. Dinh et al. [
46]. The objective of this work was to decrease the cross-sectional area of the gearbox and improve its efficiency.
In decision-making procedures, two distinct approaches are employed: MCDM and multi-objective optimization (MOO). MCDM is a highly utilized technique for ranking and selecting options across several domains. It assists in determining the optimal solution that fulfills various requirements from an extensive list of options. On the other hand, the primary aim of MOO is to employ optimization techniques in order to concurrently identify the optimal solutions for numerous separate objectives. MCDM focuses on the process of prioritizing choices, while MOO focuses on identifying the most optimal solutions that maintain a balance between several objectives. In this work, the MAIRCA technique is utilized as a multi-criteria decision-making (MCDM) method to address the multi-objective optimization problem (MOOP). The MAIRCA method is a novel MCDM approach that was first proposed in 2014 by Dragan, P., and colleagues [
47]. This method relies on the evaluation of theoretical and empirical alternative ratings through comparison. The MAIRCA technique provides reliable results and utilizes a robust and well-organized analytical framework for ranking the possibilities. This method exhibits superior stability in comparison to other methods, such as MOORA, TOPSIS, ELECTRE, and COPRAS [
48]. The tool is a straightforward mathematical instrument that exhibits a remarkable level of stability when it comes to variations in the criteria’s nature and characteristics [
49].
Although helical gearbox multi-objective optimization has been thoroughly investigated, the best main design parameters for these gearboxes have not been determined using the MCDM method. Furthermore, the power loss during idle motion, or the power loss resulting from idle rotating gears submerged in lubricant in the event of bath lubrication, was not included in the aforementioned research. The results of a multi-objective optimization study on a TSHG with double gear sets in the second stage are presented in this work. This optimization effort’s two primary goals are to improve the gearbox efficiency and decrease the gearbox volume. Three additional optimal fundamental design factors for the gearbox are examined: the CWFW for both stages and the gear ratio of the first stage. In addition, the MAIRCA method is applied to the optimization task, and the entropy method is used to establish the weights of the criteria. The novelty of this study is that, for the first time, a MOOP involving a TSHG with SDGSs is successfully solved using the MCDM approach (MAIRCA technique). Second, in the process of multi-objective gearbox optimization, the power losses during idle motion have been considered for the first time in determining the gearbox efficiency. Furthermore, the solutions to the challenge outperform those from previous research.
5. Multi-Objective Optimization
To conduct simulation experiments, a computer program has been developed. For the purpose of the investigation, the gearbox ratios of 10, 15, 20, 25, 30, 35, and 40 are all considered. The solutions to this u
t = 35 problem are shown below. For the first 125 testing cycles, this total gearbox ratio is employed (as stated in
Section 3.1). The gearbox mass and efficiency, which are the experiment’s output values, will be entered into the MAIRCA as input parameters in order to solve the MOOP. With each MAIRCA step, the gap between the two levels of each variable will decrease. For example, when u
g = 35 in step 1, u
1 rises from 4.58 to 8.31 (
Table 2). Consequently, the distance between the two levels of u1 is (8.31 − 4.58)/4 = 0.933. Until there is less than 0.02 between the two levels of the variable, this process will be repeated.
Table 3 displays the main design parameters and output responses for u
t = 35 in the MAIRCA experiment’s fourth and final iteration. The weights of the criteria have been determined through the application of the entropy approach (see
Section 3.3). To begin with, obtain the normalized values of p
ij using Equation (43). The entropy value of each indicator (me
j) is found using Equation (44). Lastly, determine the weight of the criteria w
j using Equation (45). For the latest MAIRCA experiment, the weights of V
gb and η
gb are found to be 0.4508 and 0.5492, respectively.
Section 3.2 provides guidelines for using the MAIRCA approach for multi-objective decision-making. Adopting the measures recommended in
Section 2.1, in particular, after the starting matrix is put up, the priority, or criterion PAj, is calculated by applying Equation (2). As a result, V
gb and η
gb have the same priority, which is 1/125 = 0.008. In addition, the value of parameter t
pij is determined using Equation (38) and the weight of the criterion is determined in
Section 2.2. The tp
ij values for η
gb and V
gb are 0.0036 and 0.0044, respectively. The values of tr
ij are then obtained using Equations (39) and (40), and the values of g
ij are then determined using Equation (41). In the end, the values of the criteria functions Qi are determined using Equation (42). The results of the option ranking and the different parameters that the MAIRCA technique computed (for the last run of MAIRCA work) are displayed in
Table 4. The table shows that option 55 is the best option out of all the options provided. Accordingly, u
1 = 7.02,
= 0.25, and
= 0.4 are the ideal values for the primary design features (see
Table 3).
The MOO problem is handled using two additional MCDM techniques, Evaluation by an Area-based Method of Ranking (EAMR) and Measurement of Alternatives and Ranking According to COmpromise Solution (MARCOS), in order to evaluate the reliability of the outcomes that are found (option 55 is the best). To determine the weights, the entropy approach is also used. When both of these methods are used to address the MOO problem, option 55 finally performs the best (see
Table 5). Using all three MCDM methods, the best option, 55, is produced, demonstrating that the best option is independent of the decision-making process selected.
Continuing with the preceding discussion,
Table 6 displays the ideal values for the primary design parameters that match the remaining u
t values of 10, 20, 25, 30, 35, and 40. With the data in this table, the following conclusions can be drawn. Whereas
chooses the greatest value (
= 0.4),
chooses the lowest value (
= 0.25). This is because the smallest gearbox volume can only be produced by a small box with a cross-sectional area (LxH). This requires d
w21 and d
w22 to be roughly equal in order to be accomplished [
55]. In addition, a larger
is needed to reduce the diameter of d
w22 due to the higher torque of the second stage. On the other hand, because the first stage has a lower torque, a smaller
needs to be selected in order to raise d
w21.
In order to evaluate the effectiveness of applying the formula for the power loss in the idle motion when calculating power loss in gears, the calculated values of the gearbox efficiency in this work are compared with the results in [
33]. From
Table 4 [
33], when the total gearbox ratio is 7.5, the gearbox efficiency will be 99.33–99.71% (determined from input power of 10 kW, maximum output power of 9.971 kW and minimum output power of 9.933 kW). In addition, in this work, when u
t changes from 10 to 40, the gearbox efficiency will have the values of 87.42–93.44% (
Table 3). In fact, the efficiency of a helical gear train is 0.93–0.98 (93–98%) and the efficiency of a pair of bearings is 0.99–0.995 (99–99.5%) [
51]. From these data, the efficiency of a TSHG can be determined as follows:
By comparing the above computed values with (1), it is evident that the gearbox efficiency in this work (87.42–93.44%), is fairly close to reality (83–95%). Moreover, the gearbox efficiency in [
33] (99.33–99.71%) is extremely high and not consistent with reality. This proves that using the formula to calculate the power loss in the idle motion in this work is a valuable new point and should be applied.
To evaluate the accuracy of finding the optimal values of the proposed method with other methods, like the Taguchi and Gray Relational Analysis (GRA) methods, the result of finding the optimal gear ratio of the first stage u
1 in this work is compared with that in [
41]. In this work, the solutions to the optimization problem, for example, u1, are found by directly comparing (over multiple steps) 22,500 values (with u
1 = 1–9) to choose from (as mentioned in
Section 3.1). In this case, the step between the variables of u
1 is only 0.02. Meanwhile, when using the Taguchi and GRA methods, the total number of options to choose the optimal value of u
1 is only 25 (
Table 3 [
41]). In this case, the step between the variables is 0.775 (much larger than when using the proposed method). As a result, the value u
1 found using the proposed method will be more accurate than using the Taguchi and GRA methods. In other words, the proposed method is better than the Taguchi and GRA methods.
The association between the ideal values of u
1 and u
t is depicted in
Figure 9. Additionally, it is discovered that the ideal values of u
1 may be determined using the regression equation that follows (with R
2 = 0.9974):
After finding u
1, the optimal value of u
2 is calculated by the following formula: