betaFeedback

WorksPhysics and Astronomy

Optimal Design of Spur Gears using Particle Swarm Optimization

Abstract

Optimizing the design of spur gears, regarding their mass or failure reduction, leads to reduced costs. The proposed work is aimed at using Particle Swarm Optimization (PSO) to solve single and multiple-objective optimization problems concerning spur gears. Pinion, number of teeth, the module, the face width, and the profile shift coefficients of both the pinion and wheel. Mass, gear loss factor, specific sliding, contact ratio, and safety factors are variables considered for the formulation of the objective function. The results show that those variables are reduced when compared to the results of the literature. Mass reduction, for instance, was possible to be achieved due to the reduction of the gear module and face width, increasing the number of teeth to achieve the required working center distance of the gear.

Introduction

The efficiency of gears has been carrying a great focus due to the existing concerns related to the reduction of carbon dioxide [1], market competition, quality, and regulation laws [2]. Gears are essential to machine elements in transmission systems, and their efficiency has been linked to reduced costs, increased life [3], and reduced failures [4].

Increasing the efficiency of gears leads to optimization problems. Optimization of gears has been a frequent topic of interest in the literature [5, 6, 7, 8]. From recent publications, various algorithms and different objectives have been conducted in several investigations. Miler et al. [9] tested the influence of the profile shift on the problem of the minimization of the gear weight. Results have shown the increase of the module value and profile shift coefficients reduce the volume of the gears. Moreover, profile shift coefficients are shown to be more important than using solely a balance between gear face width and gear module. Conclusions remark the greatest influence of the profile shift, which proves the gap in some previous studies, see, e.g. [10] about the non-existence of profile shift coefficients. Korta and Mundo [11] aimed to model a population-based meta-heuristic approach for optimizing teeth profile on micro-geometries, using Response Surface Methodology (RSM) for the quickness of the function evaluation. Daoudi and Boudi [12] solved the minimization problem of the gear weight for four different materials, using a genetic algorithm. Using multi-objective optimization, the authors in ref. [8] experimented, considering volume and efficiency as objectives. The results show that the power loss is lower with higher gear module values. Moreover, profile shift coefficient values for the wheel, when combined with those of the pinion and all the experiments, converge towards a higher number of teeth, even when optimizing the volume.

Hofstetter et al. [13] exposed a gearbox design optimization approach, where load, lifetime, and package requirements are given. After applying the differential-evolution process, adequate design parameters are given for the optimized efficiency, package, and cost. The results show that bearings influence efficiency. Deep groove ball bearings lead to more comprehensive solutions but are cost-effective, and both cylindrical and taper roller bearings are, in the opposite way, more costly but more compact solutions. Other materials can be used to extend the present study in order to obtain more promising results. In the following year, the authors in ref. [3] applied the non-dominated sorting genetic algorithm method (NSGA-II) [14], which incorporates fast elitism and removes sharing parameters of genetic algorithm operators. A decision is made regarding a number of options: Shannon Entropy, Linear programming technique for multidimensional analysis of preference, and technique of order preference by similarity to an ideal solution. The gear module, teeth number, and transmission ratio are considered design variables, and center distance, bearing capacity, and meshing efficiency are the design objectives. In 2020, Atila et al. [15] compared five known metaheuristics that solve two different weighting minimization problems: one introduced in ref. [16] used the width of the gear, the diameters of both the pinion and wheel, the number of teeth of the pinion, and the gear module as the design parameters.

The other problem is an adaptation of the first one by Savsani et al. [17], all of those parameters add to the hardness of the material the gears are made of is given. Artificial algae algorithm and Artificial Bee Colony (ABC) are methods that have minimized the gear weight; Particle Swarm Optimization (PSO) technique is very close to the results of the two previous ones, but its computational cost is considerably lower when compared to the other two. In the same year, Ebenezer et al. [18] compared different nature-inspired and Teaching Learning Based Optimization (TLBO) methods to find the minimum gear weight, including PSO and Cuckoo Search (CS). These two were found to be the methods in which performance was the best; the weight is minimized by 11.41% when comparing to the gear without profile shift; computation time of Simulated Annealing (SA) is shown to be lower than both CS and PSO. Moreover, in 2021, Guilbault and Lalonde [19] applied PSO and a modified Firefly algorithm to minimize the number of initiation points due to cracks appearing along the use of spur gears, by means of a fatigue point of view. Two objectives were set: the optimization of fatigue life concerning the reduction of dynamic loads and dynamic transmission errors (along a set of pinion speeds); the testing of the influence of certain variables as the starting point, the correction amount, and the curvature radius on the optimization of the dynamic transmission errors and dynamic factors. In the same year, Tomori [20] used a sample gear pair to optimize the choice of the profile shift coefficients according to different criteria. Tavčar et al. [21] aimed to optimize polymer gears using center distance, cost, life span, transmission ratio, torque, and speed as input values, and, after material selection, it optimizes tooth root stress, flank pressure, flash temperature, tooth wear and deformation, cost, and volume.

Still, the exposed publications frequently look to minimize the weight of both the pinion and wheel or the distance between their respective centers. Increasing the expected output power and efficiency are other considered objectives, and there are others that are not frequently used. Tomori [20] uses other possible objectives as methods for the computation of the correspondent profile shift coefficients, such as the minimum friction loss on the tooth profile, the optimal bending stress at the root of the pinion, optimal Hertzian stress, optimal Almen product, the optimal magnitude of the resulting flash temperature, optimal lubricant film thickness, or optimal linear wear. However, it does not allow the variation of other gear design parameters. Also, Atila et al. [15] state that control parameters are difficult to set up onto gear design problems. However, convergence is mainly not shown or explained, and parameters have not varied in recent publications. Moreover, Tavčar et al. [21] refer that the the resulting framework has a lack of consideration factors such as tooth profile modifications, thermal response, and the influence of fillers on gearing tribology; [3] makes also reference to the lack of multi-optimization solutions; Miler et al. [9] has referenced to the optimization of the specific sliding and its use as a constraint, due to gearing longevity, which was then studied by Rai and Barman [22] in helical gears.

The proposed work aims to use the PSO technique to study a variety of multi-optimization problems on spur gears resulting from the combination of the following objectives: gear loss factor (HVL), contact ratio, specific sliding, flank pressure safety factor, root stress safety factor and weight. Increasing contact ratio and safety factors decrease mechanical failure risk due to bending and contact stresses [23], problems in which adequate gear designs and lubrication are demanded [20]. Decreasing the maximum specific sliding reduces the risk of sudden breakage of the lubrication film, metal contact, and seizure damage. Also, increasing efficiency leads to energy savings due to the lower power supply to obtain the same output power. The gear mass reduction is carried out in some of the literature since a lighter gear prevents dealing with higher costs; therefore, the gear mass will be also considered as an optimization objective. The adequate spur gear is expected to accomplish most of the objectives. The design variables are the pinion number of teeth, the module, the face width, and the profile shift coefficients of both the pinion and wheel. Also, it aims to investigate and compare different PSO variants and population sizes to obtain the most suitable metaheuristic configuration.

From previous results from the literature [5], it can be hypothesized that, for the same center distance, the profile shift coefficients and the face width values may change considerably since objectives such as the specific sliding, safety factors, and contact ratio is added. In contrast, the number of teeth and modules will remain approximately the same to maintain the compromise between the high safety factors and the remaining objectives. A literature review [24] indicated that lower values for the module, when combined with the use of higher values for the number of teeth resulted in the minimization of weight, or using higher values for the module but lower values for the face width. It is also expected, for the considered transmission ratio and center distance, difficulty in finding more gear designs since the set of the possible number of teeth for the pinion is lower, resulting in a more limited set of possible designs.

This paper is structured as follows: the methodology section presents the adopted framework on which the numerical search is based; then, numerical results are presented, as well as all the constraints and constant values, e.g., of the material properties; conclusions have to The most important points from the study carried out in this work are as follows:

Framework concepts

Multi-objective optimization

Let NN be the number of design variables, defined by the x=(x1,…,xN)\mathbf{x} = (x_1,\ldots,x_N) and f(x)f(\mathbf{x}) an objective function. The number of inequality constraints is given by ngn_g and the functional inequality constraints are gi(x)g_i(\mathbf{x}), i=1,…,ngi = 1,\ldots,n_g. Moreover, the number of equality constraints is given by nhn_h and hj(x)h_j(\mathbf{x}), j=1,…,nhj = 1,\ldots,n_h, are the functional equality constraints. x∈SN\mathbf{x} \in S^N, where SN=[x1L,x1U]×…×[xNL,xNU]S^N = [x_{1L},x_{1U}] \times\ldots\times[x_{NL},x_{NU}] is the search space. Each interval in this Cartesian product is associated to the side constraints of the optimization problem formulation. Moreover, let there have MM objectives, one may be referred to as multi-objective optimization problems, given as follows:

Minimize:F(x)=(f1,…,fM)Subject to:gi(x)≤0,i=1,…,nghj(x)=0,j=1,…,nhxk,min⁡≤xk≤xk,max⁡,k=1,…,N\begin{aligned} \text{Minimize:}\quad&\mathbf{F}(\mathbf{x}) = (f_1,\ldots,f_M)\\ \text{Subject to:}\quad&g_i(\mathbf{x}) \le0,\quad i = 1,\ldots,n_g\\ &h_j(\mathbf{x}) = 0,\quad j = 1,\ldots,n_h\\ &x_{k,\min} \le x_k \le x_{k,\max},\quad k = 1,\ldots,N \end{aligned}

According to Gunantara [24] and António [25], two main methods divide the way one deals with multi-objective optimization: Pareto dominance and scalarization. In Pareto dominance, a set of solutions called the Pareto set is defined as solutions that dominate all the other solutions within the search space, i.e., the solutions that are not dominated. Therefore, no element of the Pareto set can dominate any of the other candidate elements. Dominance can be defined by the mathematical statement in the next equation, where x1\mathbf{x}_1, x2∈SN\mathbf{x}_2 \in S^N [26].

x1≺x2:∀i=1,…,M, fi(x1)≤fi(x2)∧∃i∈{1,…,m}, fi(x1)<fi(x2)\mathbf{x}_1 \prec\mathbf{x}_2 : \forall i = 1,\ldots,M,\ f_i(\mathbf{x}_1) \le f_i(\mathbf{x}_2) \land\exists i \in\{1,\ldots,m\},\ f_i(\mathbf{x}_1) < f_i(\mathbf{x}_2)

The Pareto set is a set with multiple valid solutions. An adequate solution is left for the user to choose after optimization. In scalarization, the various objectives are combined in a single fitness function to transform the problem into a single optimization problem [25].

The current approach, explained in § 3, only considers the scalarization method.

Particle Swarm Optimization

PSO is a bio-inspired population-based technique with plenty of applications in engineering optimization problems. It was inspired by social behaviors like bird flocking and fish schooling [27, 28]. PSO has a generalized flowchart represented in Figure 1. After the generation of the starting population, which is usually done at random, the fitness of the solutions, also called particles, is evaluated with respect to the problem it aims to optimize. Therefore, the particles’ velocity and positions are updated based on mathematical expressions and dependently on the fitness values of the particles. Velocities and positions of each population particle ii, at each generation tt, and for each dimension dd, as it is shown as follows:

Flowchart for the Classic PSO Algorithm / proposed RPSO Algorithm

Figure 1. Flowchart for the Classic PSO Algorithm / proposed RPSO Algorithm.

vd,t+1i=ωvd,ti+ϕ1R1d,ti(pd,ti−xd,ti)+ϕ2R2d,ti(qd,ti−xd,ti)(1)v^i_{d,t+1} = \omega v^i_{d,t} + \phi_1 R^i_{1d,t}(p^i_{d,t} - x^i_{d,t}) + \phi_2 R^i_{2d,t}(q^i_{d,t} - x^i_{d,t}) \tag*{(1)}
xd,t+1i=xd,ti+vd,t+1i(2)x^i_{d,t+1} = x^i_{d,t} + v^i_{d,t+1} \tag*{(2)}

In the equations above, vv is the velocity of a particle, x\mathbf{x} is the position of a particle, ω\omega is the inertia weight coefficient, which has then been introduced in later publications [29], ϕ1\phi_1 and ϕ2\phi_2 are cognitive and social coefficients, respectively, ptip^i_t is the best position of a particle ii among all the generations to the generation tt, g\mathbf{g} is the best position among all particles and generations to the generation tt, and R1R_1 and R2R_2 are values sampled at random.

PSO was first implemented in 1995. This has led the scientific community to verify some issues, resulting in the most known premature convergence and consequently developing variations of PSO. In 1999, Clerc [30] has introduced a constriction factor related to ϕ1\phi_1 and ϕ2\phi_2. In [31], discrete-time PSO is generalized to a continuous-time PSO from an analytical point of view. The constriction factor is multiplied by the right-hand side of (1). Random PSO, abbreviated as RPSO [32], is another PSO variation that is aimed at finding the global minimum. For that, a particle is chosen and sampled at random. In 2006, Van der Bergh [32] demonstrated that PSO is a local but not a global search algorithm. However, RPSO has been proven to be a global search method. Since particles are sampled randomly into the search space, the final solution is not dependent on the starting population. The same RPSO algorithm that has been used in different applications [33, 34, 35] is now used for the optimization of spur gears.

Methodology

The proposed work consists of optimizing gear design parameters, given the minimization function, FF, and inequality and equality constraints that will be defined later on. This work considers that the transmission ratio and the working center distance, uu and a′a', respectively, are given, which leads to choosing between multiple gears that respect the given uu and a′a'. Due to the natural discretization of the number of teeth of both pinion and wheel, z1z_1 and z2z_2, respectively, one may not find suitable or even possible gear combinations, leading to the consideration of an admissible slight variation of the transmission ratio. That variation shall be not higher than a given tolerance tolu\mathrm{tol}_u. In its turn, a′a' does not come with any possible tolerance. The total number of inputs is given in Table 1.

ParameterDefinition
uuTransmission ratio
tolu\mathrm{tol}_uTolerance of the transmission ratio
a′a'Working axis distance

Table 1. Requirement parameters for the possible gears to consider

The transmission ratio is expressed as follows:

u=z2z1(3)u = \frac{z_2}{z_1} \tag*{(3)}

where z1z_1 and z2z_2 are the number of teeth for the pinion and the wheel, respectively. Depending on tolu\mathrm{tol}_u, z2z_2 can be a direct function of z1z_1 or not. This dependency is, in turn, dependent on the value of tolu\mathrm{tol}_u, being z2z_2 value restricted according to eq. (4).

tol⁡u≥∣z2z1−u∣(4)\operatorname{tol}_u \ge\left|\frac{z_2}{z_1}-u\right| \tag*{(4)}

Here, z1z_1 is also restricted by tol⁡u\operatorname{tol}_u, as it is shown in eq. (5) since it is necessary for at least one possible z2z_2. Note that ⌊⋅⌉\left\lfloor\cdot\right\rceil in (5) refers to the nearest integer notation.

tol⁡u≥⌈⌊z1⋅u⌉z1−u⌉(5)\operatorname{tol}_u \ge\left\lceil\frac{\left\lfloor z_1\cdot u\right\rceil}{z_1}-u\right\rceil \tag*{(5)}

The axis distance without any profile shift is calculated using eq. (6).

a=r1+r2=z1m2+z2m2=a(z1,z2,m)(6)a=r_1+r_2=\frac{z_1m}{2}+\frac{z_2m}{2}=a(z_1,z_2,m) \tag*{(6)}

where r1r_1 and r2r_2 are the pitch radii, respectively, and mm is the profile module, which is equal for both pinion and wheel. The working axis distance is dependent on gear standard parameters and the optimized profile shifts. Two conditions may be necessary to know this incognita: the base pitch is the same whichever the profile shift is, and the pitch is the sum of the thicknesses of the pinion and wheel when considering the working conditions, as according to eq. (7).

{inv⁡α′=inv⁡α+2⋅tan⁡α⋅Σxz1+z2a⋅cos⁡α=a′⋅cos⁡α′Σx=x1+x2(7)\begin{cases} \operatorname{inv}\alpha'=\operatorname{inv}\alpha+2\cdot\tan\alpha\cdot\frac{\Sigma x}{z_1+z_2}\\ a\cdot\cos\alpha=a'\cdot\cos\alpha'\\ \Sigma x=x_1+x_2 \tag*{(7)} \end{cases}

In eq. (7), the involute function inv⁡(x)=tan⁡(x)−x\operatorname{inv}(x)=\tan(x)-x gives the involute of the angle xx. The function dependence represented in eq. (8) must be in the count to accomplish the initial constraint.

α′=α′(Σx,α,z1,z2,a(z1,z2,m))=α′(Σx,α,z1,z2,m)(8)\alpha'=\alpha'(\Sigma x,\alpha,z_1,z_2,a(z_1,z_2,m))=\alpha'(\Sigma x,\alpha,z_1,z_2,m) \tag*{(8)}

Eq. (8) remarks an important conclusion: a′a' will be not dependent of x1x_1 or x2x_2 individually, since only their sum is explicitly written in eq. (7).

Therefore, and because a′a' is given, the sum of the profile shifts may be turned on the only unknown of eq. (7), being x2x_2 the only parameter to be calculated from the rest of the gear parameters.

The profile shift is considered in this work to accomplish the axis distance value; still, this is typically done to accomplish the equality sliding speeds, which is also intended to minimize in this work.

There are many constraints that may be considered due to the real-scenario application. Firstly, α\alpha and mm are normalised values. The module is one of the possible values from any of the listed sets LA\mathbf{L}_A and LB\mathbf{L}_B (according to DIN 780 Part 1), being LA\mathbf{L}_A the most preferable.

LA={0.05, 0.06, 0.08, 0.1, 0.12, 0.16, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1, 1.25, 1.5, 2, 2.5, 3, 4, 5, 6, 8, 10, 12, 16, 20, 25, 32, 40, 50, 60}\mathbf{L}_A=\{0.05,\ 0.06,\ 0.08,\ 0.1,\ 0.12,\ 0.16,\ 0.2,\ 0.25,\ 0.3,\ 0.4,\ 0.5,\ 0.6,\ 0.7,\ 0.8,\ 0.9,\ 1,\ 1.25,\ 1.5,\ 2,\ 2.5,\ 3,\ 4,\ 5,\ 6,\ 8,\ 10,\ 12,\ 16,\ 20,\ 25,\ 32,\ 40,\ 50,\ 60\}
LB={0.055, 0.07, 0.09, 0.11, 0.14, 0.18, 0.22, 0.28, 0.35, 0.45, 0.55, 0.65, 0.75, 0.85, 0.95, 1.125, 1.375, 1.75, 2.25, 2.75, 3.25, 3.5, 3.75, 4.25, 4.5, 4.75, 5.25, 5.5, 5.75, 6.5, 7, 9, 11, 14, 18, 22, 27, 28, 30, 36, 39, 42, 45, 55, 70}\mathbf{L}_B=\{0.055,\ 0.07,\ 0.09,\ 0.11,\ 0.14,\ 0.18,\ 0.22,\ 0.28,\ 0.35,\ 0.45,\ 0.55,\ 0.65,\ 0.75,\ 0.85,\ 0.95,\ 1.125,\ 1.375,\ 1.75,\ 2.25,\ 2.75,\ 3.25,\ 3.5,\ 3.75,\ 4.25,\ 4.5,\ 4.75,\ 5.25,\ 5.5,\ 5.75,\ 6.5,\ 7,\ 9,\ 11,\ 14,\ 18,\ 22,\ 27,\ 28,\ 30,\ 36,\ 39,\ 42,\ 45,\ 55,\ 70\}

Also, the values of the profile shifts are restricted, as in eq. (9).

Smin⁡≤∑i=12xi≤Smax⁡x1,min⁡≤x1≤x1,max⁡x2,min⁡≤x2≤x2,max⁡(9)\begin{aligned} S_{\min}&\le\sum_{i=1}^{2}x_i\le S_{\max}\\ x_{1,\min}&\le x_1\le x_{1,\max}\\ x_{2,\min}&\le x_2\le x_{2,\max} \tag*{(9)} \end{aligned}

where Smin⁡=−1S_{\min}=-1, Smax⁡=2S_{\max}=2, x1,min⁡=0x_{1,\min}=0, x1,max⁡=1x_{1,\max}=1, x2,min⁡=−0.5x_{2,\min}=-0.5, x2,max⁡=1x_{2,\max}=1 for this work. The minimum profile shifts for pinion and wheel are limited by the non-occurrence of undercut, written in eq. (10).

ximin⁡=z′−ziz′,i=1,2(10)x_{i\min}=\frac{z^{\prime}-z_i}{z^{\prime}},\qquad i=1,2 \tag*{(10)}

where z′=2(sin⁡α)−2z^{\prime}=2(\sin\alpha)^{-2}. Also, xi,min⁡∗x_{i,\min}^{*}, i=1,2i=1,2 are restricted according to eq. (11).

xi,min⁡={0,if i=1∧xi,min⁡<0xi,min⁡,otherwise(11)x_{i,\min}= \begin{cases} 0, & \text{if }i=1\land x_{i,\min}<0\\ x_{i,\min}, & \text{otherwise} \end{cases} \tag*{(11)}

Using the previous equations, a proposed algorithm presented in Figure 2 is aimed to search for a valid gear. Given the variables z1∗z_1^{*}, z2∗z_2^{*}, m∗m^{*}, x1∗x_1^{*}, α\alpha, bˉ\bar{b}, where bˉ\bar{b} is the normalized face width of the gear, the calculation of a valid gear is done based on a function called FS\mathbf{F}_S. This function is then aimed at transforming a given set of initial parameters into another set that accomplishes the aforementioned constraints of center distance and transmission ratio. The output of the proposed algorithm corresponds to the vector of the design variables d=(z1,z2,x1,x2,m,α,b)\mathbf{d}=(z_1,z_2,x_1,x_2,m,\alpha,b). In the algorithm, z1z_1, z2z_2, x1x_1 and mm are modified around a fixed starting point (z1∗,z2∗,m∗,x1∗)(z_1^{*},z_2^{*},m^{*},x_1^{*}), and the rest of the missing parameters, i.e., x2x_2 and b=5m(bˉ+1)b=5m(\bar{b}+1), where bb is the face width of the gear, are directly calculated once it is known that the resulting gear is compatible with the given inputs for the transmission ratio, its tolerance, and the working axis distance.

Proposed flowchart for the optimization process

Figure 2. Proposed flowchart for the optimization process.

The scalar functions that will be used later to defined multi-objective functions are now defined. These are the contact ratio (f1f_1), the mass of the gear (f2f_2), the root safety factor (f3f_3), the flank safety factor (f4f_4), the gear loss factor (f5f_5) and the difference between specific sliding (f6f_6).

Considering rbi=ricos⁡αr_{bi}=r_i\cos\alpha and rai′=ri+m(1+xi)r_{ai}^{\prime}=r_i+m(1+x_i), i=1,2i=1,2, the contact ratio is defined as follows:

f1=f1(d)=εα=ra12−rb12+ra22−rb22−a′⋅sin⁡α′p⋅cos⁡α(12)f_1=f_1(\mathbf{d})=\varepsilon_{\alpha}=\frac{\sqrt{r_{a1}^{2}-r_{b1}^{2}}+\sqrt{r_{a2}^{2}-r_{b2}^{2}}-a^{\prime}\cdot\sin\alpha^{\prime}}{p\cdot\cos\alpha} \tag*{(12)}

The adopted mass used in the present work is according to eq. (13). In the equation, bb is the face width of the gear, ρ\rho is the specific gravity of the material of the gear, ra1=r1+x1mr_{a1}=r_1+x_1m and ra2=r2+x2mr_{a2}=r_2+x_2m.

f2=f2(d)=W=ρπb(ra12+ra22)(13)f_2=f_2(\mathbf{d})=W=\rho\pi b\left(r_{a1}^{2}+r_{a2}^{2}\right) \tag*{(13)}

The root (SHS_H) and flank (SFS_F) safety factors are accordingly to eq. (14) and eq. (15), respectively.

f3=f3(d)=SH=σHaσH(14)f_3=f_3(\mathbf{d})=S_H=\frac{\sigma_{Ha}}{\sigma_H} \tag*{(14)}
f4=f4(d)=SF=σFaσF(15)f_4=f_4(\mathbf{d})=S_F=\frac{\sigma_{Fa}}{\sigma_F} \tag*{(15)}

The contact pressure σH\sigma_H is defined:

σH=ZDσH0∏Ki(16)\sigma_H=Z_D\sigma_{H0}\sqrt{\prod K_i} \tag*{(16)}
σH0=Ft2r1bu+1u∏Zi(17)\sigma_{H0}=\sqrt{\frac{F_t}{2r_1b}\frac{u+1}{u}\prod Z_i} \tag*{(17)}

where ZDZ_D is the pinion single pair tooth contact factor, ZiZ_i are other contact factors, such as the zone factor, the contact ratio factor and the elasticity factor, KiK_i are correction factors, such as the application factor, the dynamic factor and the face and the transverse load factors for contact stress, and FtF_t is the tangential load. σHa\sigma_{Ha} is defined as the allowed contact pressure. Moreover, the flank pressure σF\sigma_F is defined as follows:

σF=Ftbm∏Yi∏Ki(18)\sigma_{F}=\frac{F_{t}}{bm}\prod Y_{i}\prod K_{i} \tag*{(18)}

where YiY_i and KiK_i are correction factors that depend on system characteristics such as the lubricant, roughness, or the numerical method being used for their calculation. σFa\sigma_{Fa} is defined as the allowed flank pressure. The gear loss factor HVLH_{VL} is defined as follows:

f5=f5(d)=HVL=1pb∫ABFN(x)vg(x)Fbtvtb dx(19)f_{5}=f_{5}(\mathbf{d})=H_{VL}=\frac{1}{p_{b}}\int_{A}^{B}\frac{F_{N}(x)v_{g}(x)}{F_{bt}v_{tb}}\,dx \tag*{(19)}

In (19), pb=πmcos⁡αp_b=\pi m\cos\alpha is the base pitch, A and B are the gearing starting and ending points, respectively, vg(x)v_g(x) is the sliding speed on a point xx, FN(x)F_N(x) is the normal load in relation to the base radius, FbtF_{bt} is the tangential load in relation to the base radius and vtbv_{tb} is the tangential component of the sliding speed in relation to the base radius. The difference of the specific sliding speeds Δg\Delta g is given as follows:

f6=f6(d)=Δg=∥∣1−z1z2⋅ra2′2−rb22a⋅sin⁡α−ra2′2−rb22∣−z2z1⋅ra1′2−rb12a⋅sin⁡α−ra1′2−rb12−1∥(20)f_{6}=f_{6}(\mathbf{d})=\Delta g=\left\lVert\left|1-\frac{z_{1}}{z_{2}}\cdot\frac{\sqrt{r_{a2}^{\prime2}-r_{b2}^{2}}}{a\cdot\sin\alpha-\sqrt{r_{a2}^{\prime2}-r_{b2}^{2}}}\right|-\frac{z_{2}}{z_{1}}\cdot\frac{\sqrt{r_{a1}^{\prime2}-r_{b1}^{2}}}{a\cdot\sin\alpha-\sqrt{r_{a1}^{\prime2}-r_{b1}^{2}}}-1\right\lVert \tag*{(20)}

The functional constraints are represented in eq. (21) and eq. (22).

g1(d)=tolu(d)−toluMg2(d)=ϵαM−ϵα(d)g3(d)=SHM−SH(d)g4(d)=SFM−SF(d)(21)\begin{aligned} g_{1}(\mathbf{d})&=tol_{u}(\mathbf{d})-tol_{u}^{M}\\ g_{2}(\mathbf{d})&=\epsilon_{\alpha}^{M}-\epsilon_{\alpha}(\mathbf{d})\\ g_{3}(\mathbf{d})&=S_{H}^{M}-S_{H}(\mathbf{d})\\ g_{4}(\mathbf{d})&=S_{F}^{M}-S_{F}(\mathbf{d}) \tag*{(21)} \end{aligned}
h1(d)=α′(d)−αg′h2=u−ug(22)\begin{aligned} h_{1}(\mathbf{d})&=\alpha^{\prime}(\mathbf{d})-\alpha_{g}^{\prime}\\ h_{2}&=u-u_{g} \tag*{(22)} \end{aligned}

In the current work, ϵαM=1.45\epsilon_{\alpha}^{M}=1.45, SHM=1.4S_{H}^{M}=1.4 and SFM=2.0S_{F}^{M}=2.0 are the allowed minimum contact ratio, minimum root safety factor and minimum flank safety factor, respectively. The maximum transmission tolerance ratio toluMtol_{u}^{M}, the exact working axis distance ae′a_{e}^{\prime} and the desired transmission ratio ueu_{e} are defined for each case study later on.

Algorithm 1 Gear Search Algorithm (FsF_s)

Given: α\alpha, uu, a′a', tol⁡u\operatorname{tol}_u (Optional variables: z1∗z_1^*, z2∗z_2^*, m∗m^*, x1∗x_1^*, bˉ\bar{b})

Initialize: z1∗z_1^*, z2∗z_2^*, m∗m^*, x1∗x_1^*, bˉ\bar{b} (if not previously given)

  1. z10z_1^0, z1∗←int⁡(z1∗)z_1^* \leftarrow\operatorname{int}(z_1^*); discretize m∗m^* and call it m0∗,m∗m_0^*,m^*;

  2. With α\alpha, use (10) to limit x1∗x_1^*.

  3. Tz1∗∗←{m0∗}T_{z_1^*}^* \leftarrow\{m_0^*\}. Evaluate the minimum constant kz1∗∈Nk_{z_1^*} \in\mathbb{N}: z2∗=(z1∗+kz1∗c)(u+εtol⁡u)∈N,∀c∈Z∗∧∃ε∈[−1,1]⊂Rz_2^* = (z_1^* + k_{z_1^*}c)(u+\varepsilon\operatorname{tol}_u) \in\mathbb{N}, \forall c \in\mathbb{Z}^* \land\exists\varepsilon\in[-1,1] \subset\mathbb{R}, where Z∗⊂Z\mathbb{Z}^* \subset\mathbb{Z} accordingly to (5). Z←{c}\mathbb{Z} \leftarrow\{c\}.

  4. According to the condition defined in Step 3, define a set Ez1∗E_{z_1^*} of all possible ε\varepsilon. If Ez1∗≡∅E_{z_1^*} \equiv\varnothing, go to Step 11.

  5. Apply (3) to find z20z_2^0, z2∗←z1∗(u+ε∗tol⁡u)z_2^* \leftarrow z_1^*(u+\varepsilon^*\operatorname{tol}_u), ε∗:min⁡{∣ε∣, ε∈Ez1∗}\varepsilon^*: \min\{|\varepsilon|,\ \varepsilon\in E_{z_1^*}\}. Ez1∗∗←{ε∗}E_{z_1^*}^* \leftarrow\{\varepsilon^*\}

  6. With z1∗z_1^*, z2∗z_2^*, m∗m^*, α\alpha and z′z' calculate the limits of x1∗x_1^* and x2∗x_2^* using (9); use (11). flag←0\mathit{flag} \leftarrow0

  7. Use (7) to calculate ∑x\sum x; x2∗←∑x−x1∗x_2^* \leftarrow\sum x-x_1^*.

  8. If the results on Steps 6 and 7 are according to (9), then z1←z1∗z_1 \leftarrow z_1^*, z2←z2∗z_2 \leftarrow z_2^*, m←m∗m \leftarrow m^*, x1←x1∗x_1 \leftarrow x_1^*, x2←x2∗x_2 \leftarrow x_2^* and bˉ=5m(bˉ+1)\bar{b}=5m(\bar{b}+1) and exit the function. Else, and if flag←0\mathit{flag}\leftarrow0, then x1∗←rand⁡(0,1)x_1^* \leftarrow\operatorname{rand}(0,1), flag←1\mathit{flag}\leftarrow1 and go back to Step 7. Else, go to Step 9.

  9. If Tz1∗∗≡LA∪LBT_{z_1^*}^* \equiv L_A \cup L_B go to Step 10. Otherwise, m∗←min⁡{∣mTz1∗∗−m0∗∣}m^* \leftarrow\min\{|m_{T_{z_1^*}^*}-m_0^*|\}, where Tz1∗:=LB∖Tz1∗∗⇐Tz1∗∗⊂LA∧LA∖Tz1∗∗⇐Tz1∗∗⊄LAT_{z_1^*}:=L_B\setminus T_{z_1^*}^* \Leftarrow T_{z_1^*}^*\subset L_A \land L_A\setminus T_{z_1^*}^* \Leftarrow T_{z_1^*}^*\not\subset L_A. Tz1∗∗←{m∗}∪Tz1∗∗T_{z_1^*}^* \leftarrow\{m^*\}\cup T_{z_1^*}^*. Go back to Step 6.

  10. If Ez1∗∗≡Ez1∗E_{z_1^*}^* \equiv E_{z_1^*} go to Step 11. Else, z2∗←z1∗(u+ε∗tol⁡u)z_2^* \leftarrow z_1^*(u+\varepsilon^*\operatorname{tol}_u), ε∗:min⁡{∣ε∣, ε∈Ez1∗∖Ez1∗∗≠∅}\varepsilon^*: \min\{|\varepsilon|,\ \varepsilon\in E_{z_1^*}\setminus E_{z_1^*}^* \ne\varnothing\}; Ez1∗∗←{ε∗}∪Ez1∗∗E_{z_1^*}^* \leftarrow\{\varepsilon^*\}\cup E_{z_1^*}^* and go back to Step 6.

  11. c∗←min⁡{∣z1∗+kz1∗c−z10∣, c∈Z∗∖Z}c^* \leftarrow\min\{|z_1^*+k_{z_1^*}c-z_1^0|,\ c\in\mathbb{Z}^*\setminus\mathbb{Z}\}. Z←{c∗}∪Z\mathbb{Z}\leftarrow\{c^*\}\cup\mathbb{Z}. z1∗←z1∗+kz1∗c∗z_1^* \leftarrow z_1^*+k_{z_1^*}c^*. Tz1∗∗←{m0∗}T_{z_1^*}^*\leftarrow\{m_0^*\}. Go back to Step 4.

Return: z1,z2,x1,x2,m,α,bˉz_1,z_2,x_1,x_2,m,\alpha,\bar{b}

The flowchart that represents the optimization procedure is represented in Figure 2. Initially, gear configurations are generated at random as a standard procedure of PSO. Then, F∗\mathbf{F}^* is used to correct the randomized configuration, and that obeys the conditions of center distance, transmission ratio, and tolerance. After the procedure, the resulting gears are evaluated, and the best is considered. PSO and RPSO are used for the search for new potential candidate configurations. The process repeats, including the evaluation of F∗\mathbf{F}^* until the stopping criteria are met.

Due to the significant number of possible variables that can be optimized and the number of functions that can be minimized, a total of three comparative studies of the designed algorithm have been conducted:

  • Study 1. variation of the tolerance tol⁡uM\operatorname{tol}_u^M: the tolerance related to the transmission ratio variation is set to two different values, 0.05 and 0.2.

  • Study 2. Variation of the objective function: using tol⁡uM=0.2\operatorname{tol}_u^M=0.2, a single-objective function F1F_1 and total of three multi-objective functions (F2F_2, F3F_3 and F4F_4 are aggregation functions) have been considered:

    • F1F_1: Mass minimization (also conducted in studies 1 and 3):

Minimize:F1(d)=f2(d)Subject to:gi(d)≤0hk(d)=0di,L≤di≤di,u, i=1,…,Nd(23)\begin{aligned} \text{Minimize:}\quad&F_1(\mathbf{d})=f_2(\mathbf{d})\\ \text{Subject to:}\quad&g_i(\mathbf{d})\leq0\\ &h_k(\mathbf{d})=0\\ &d_{i,L}\leq d_i\leq d_{i,u},\ i=1,\ldots,N_d \tag*{(23)} \end{aligned}
  • F2F_2: Mass minimization and maximization of SHS_H and SFS_F:

Minimize:F2(d)=f2(d)2+(11+f3(d))2+(11+f4(d))2Subject to:gi(d)≤0hk(d)=0di,L≤di≤di,u, i=1,…,Nd(24)\begin{aligned} \text{Minimize:}\quad F_2(\mathbf{d})&=\sqrt{f_2(\mathbf{d})^2+\left(\frac{1}{1+f_3(\mathbf{d})}\right)^2+\left(\frac{1}{1+f_4(\mathbf{d})}\right)^2}\\ \text{Subject to:}\quad&g_i(\mathbf{d})\leq0\\ &h_k(\mathbf{d})=0\\ &d_{i,L}\leq d_i\leq d_{i,u},\ i=1,\ldots,N_d \tag*{(24)} \end{aligned}
  • F3F_3: Minimization of the gear mass and HVLH_{\mathrm{VL}};

Minimize: F3(d)=f2(d)2+f5(d)2F_3(\mathbf{d}) = \sqrt{f_2(\mathbf{d})^2 + f_5(\mathbf{d})^2}

Subject to:

gi(d)≤0g_i(\mathbf{d}) \le0
hk(d)=0h_k(\mathbf{d}) = 0
di,L≤di≤di,u, i=1,…,Nd(25)d_{i,L} \le d_i \le d_{i,u},\ i = 1,\ldots,N_d \tag*{(25)}
  • F4F_4: Minimization of the gear mass, HVLH_{\mathrm{VL}} and differences of the specific sliding at the roots and maximization of SHS_H, SFS_F and contact ratio.

Minimize: F4(d)=∑i=2,5,6fi(d)2+∑j=1,3,4(11+fj(d))2F_4(\mathbf{d}) = \sqrt{\sum_{i=2,5,6} f_i(\mathbf{d})^2 + \sum_{j=1,3,4} \left(\frac{1}{1+f_j(\mathbf{d})}\right)^2}

Subject to:

gi(d)≤0g_i(\mathbf{d}) \le0
hk(d)=0h_k(\mathbf{d}) = 0
di,L≤di≤di,u, i=1,…,Nd(26)d_{i,L} \le d_i \le d_{i,u},\ i = 1,\ldots,N_d \tag*{(26)}
  • Study 3. Variation of the PSO configurations: using the mass minimization as the objective and tolerance toluM=0.2tol_u^M = 0.2, the following variations have been considered so that the computational time is equal for all of the variations; therefore, a fair comparison between them is possible:

    • M1: RPSO with a population of 10 particles (#POP = 10) and 10000 generations

    • M2: PSO with a population of 10 particles (#POP = 10) and 10000 generations

    • M3: RPSO with a population of 50 particles (#POP = 50) and 2000 generations

    • M4: PSO with a population of 50 particles (#POP = 50) and 2000 generations

    • M5: RPSO with a population of 500 particles (#POP = 500) and 200 generations

    • M6: PSO with a population of 500 particles (#POP = 500) and 200 generations

Results and discussion

The results of spur gear optimization are documented in this section. For reasons related to the comparison with literature studies, two case studies are driven and used to execute the algorithm and to obtain the results. The dependent variables related to the center distance, transmission ratio, lubricant, material properties, input power, and speed of the pinion are listed in Table 2, according to the selected gear from each literature study. Properties such as input power, transmission ratio, and center distance, which are significant inputs of the proposed algorithm, were calculated based on the gear variables of Table 3.

General Properties
Oil: PAO ISO VG 150; Material: 20MnCr5; Material Young Modulus: 210 GPa; Material Poisson ratio: 0.3, Tlub=70∘CT_{\mathrm{lub}} = 70^{\circ}\mathrm{C}
Ref.AuthorsAbbr.Specific properties
[9]Miler et al.C1Material specific gravity: 7.84; Input power: 10.053 kW; Speed of the pinion: 960 rpm; Center distance: 199.845 mm; Transmission ratio: u=3.55u = 3.55
[5]Maputi et al.C2Material specific gravity: 8; Input power: 12.480 kW; Speed of the pinion: 1500 rpm; Center distance: 130.625 mm; Transmission ratio: u=4u = 4

Table 2. General and specific gear properties considered in this work for comparison

VariableC1C2
Design variables
α\alpha [º]2025
z1z_12319
z2z_28276
x1x_10.6990
x2x_20.1360
mm [mm]3.752.75
bb [mm]22.534.84
Objective functions
ϵα\epsilon_\alpha1.4581.488
HVLH_{\mathrm{VL}}0.1410.129
Δg\Delta_g0.6280.870
Mass [kg]14.34710.160
SHS_H2.131.69
SFS_F7.656.93

Table 3. Selected gears from two literature references

The conducted research on gear optimization has played an important role in the exposition of a large set of results from different numerical experiments. The existent commercial gear tooth contact analysis software KISSsoft was used for validating the results of the model implemented in the present study.

In Table 4, results on the minimization of the gear mass are represented. Several conclusions can be taken from it. Firstly, the mass has been lower than the literature for the studied cases [5], [9]. This reduction corresponds to from 51% to 55%. The reduction of the masses had been previously hypothesized since both safety factors of the gears of each of the original gears were not near the allowable values. With the decrease of the module, the minimum face width value of 10 mm has been achieved for case C1. In C2, the face width value is not decreased to its minimum since the limit of the safety factor SHS_H has been reached.

Moreover, it is possible to verify that no significant difference between tolerance values exists. However, a tolerance of 0.2 is preferred since it takes a lower computational time. With the decrease of the mass, HVLH_{\mathrm{VL}} and the difference in the maximum specific sliding have also been reduced. The contact ratio is increased.

VariableC1 (F1F_1)C1 (F1F_1)C2 (F1F_1)C2 (F1F_1)
toluM=0.05tol_u^M = 0.05toluM=0.2tol_u^M = 0.2toluM=0.05tol_u^M = 0.05toluM=0.2tol_u^M = 0.2
Design variables
α\alpha[º]20202525
z1z_144442626
z2z_2156156104104
x1x_10.31750.19170.42630.4259
x2x_2−0.3946−0.2688−0.1104−0.1100
mm[mm]2222
bb[mm]101016.99916.988
Output variables / objective functions
ϵα\epsilon_\alpha1.7641.7881.4451.445
HVLH_{\mathrm{VL}}0.0820.0770.1050.105
Δg\Delta_g0.2130.0380.2460.246
SHS_H1.561.581.401.40
SFS_F2.392.412.962.96
Mass [kg]6.4246.4384.9094.906

Table 4. Optimized gears for each of the case studies Ck_k for different tolerances.

In Table 5, the results of the second study are shown. For the objective function F2F_2, it has been possible to decrease the mass by 3% and 25% while still increasing the safety factor SFS_F. Module values and face width are now higher, so the root stress is decreased and the safety factor SFS_F is increased. For the objective function F3F_3, the mass has been significantly reduced, as occurred in the situation where mass reduction is the single objective. The HVLH_{\mathrm{VL}} decreases as the module decreases and the number of teeth increases. This reduction in the module and the increase of the number of teeth promotes a possible reduction in width and, consequently, in the gear mass. The module and face width values are similar to the single objective.

VariableC1 (toluM=0−2tol_u^M = 0 - 2)C2 (toluM=0.2tol_u^M = 0.2)
F2F_2F3F_3F4F_4F2F_2F3F_3F4F_4
Design variables
$\alpha1$202020252525
z1z_1224430182618
z2z_2781561036910470
x1x_10.45060.08230.25090.17250.26820.0528
x2x_2−0.4891-0.4891−0.1594-0.1594−0.1350-0.1350−0.1308-0.13080.0477−0.4998-0.4998
m[mm]m[\mathrm{mm}]423323
b[mm]b[\mathrm{mm}]21.8661025.05226.51717.69726.095
Output variables / objective functions
ϵα\epsilon_\alpha1.5981.8071.6961.4491.4671.506
HVLH_\mathrm{VL}0.1530.0760.1050.1340.0960.139
Δg\Delta g0.5750.1260.0410.3330.0180.649
Mass[kg]13.9096.45015.9447.5905.1337.509
SHS_H2.071.572.111.591.401.56
SFS_F8.552.426.696.313.076.40

Table 5. Optimized gears for each of the case studies Ck_k for different multi-objective problems.

Moreover, reductions of 80% and 98% on the value of Δg\Delta g, when comparing to the literature [5], [9], are observed even if it was not aimed to minimize it. When it comes to trying to minimize and maximizing all the output variables of interest, using the objective function F4F_4, one can observe that some objectives are indeed improved, but, as a consequence, the others have been penalized. For instance, in C1, a reduction of 94% of the Δg\Delta g, an increase in 16% of the contact ratio and a reduction of 26% of the HVLH_\mathrm{VL} are obtained, but the mass is now higher, and safety factors are lower than the original gear. In C2, mass and Δg\Delta g are improved, but the same does not happen with the other objectives.

Lastly, the mass reduction optimization results for different selected configurations of PSO variations (of population size and random search) have been obtained. Since the resulting gear of the optimization process for the mass reduction is already known, the objective of this study is to observe the behavior of the different configurations along with the iteration and the selection of the most suitable for this problem. The plot corresponding to the average (on ten simulation repetitions) of the minimum gear mass for each case and generation is represented in Figure 3.

Gear mass variation along the number of evaluations

Figure 3. Gear mass variation along the number of evaluations.

According to Figure 4, in case C1, the gear mass is reduced and stabilized in less than 10000 iterations for most configurations. When using the PSO local search with ten particles, the difficulty of finding the global minimum is evident. However, for case C2, no gear mass stabilization is apparent, even for 100000 iterations. Using PSO with 10 particles continue to be not suitable for the problem since the average gear mass is always higher than the rest of the configurations for any of the number of generations stopping criteria. However, using RPSO configurations is the most suitable option, especially for a smaller number of particles, resulting from the complexity of the problem of gear mass reduction.

Contact ratio variation for Cases 1 and 2

Figure 4. Gear contact ratio variation along with the number of evaluations. In Case 1, the contact ratio is higher than 1.8 if the method is well chosen; in Case 2, the contact ratio tends to the constrained value of 1.45.

Moreover, the figures in Appendix A (5 to 14) have resulted from the variation of other relevant gear variables with the iterations. Concerning both safety factors, for both case studies, a decrease is observed. As previously observed, when minimizing the mass, the safety factor SHS_H is also minimized until its convergence to the constraint. Also, the minimum specific sliding and HVLH_{\mathrm{VL}} are decreased in general. However, the contact ratio may decrease or increase depending on the case study. In C1, the evident increase in the contact ratio is observed, and its convergence happens around no more than 1.85. In opposite, the contact ratio decreases in C2, in which the value stabilized around the constraint value of 1.45.

Concerning the design variables of the gear, it is possible to observe that the number of teeth of the pinion has a slight tendency to increase. In contrast, the module tends to decrease to satisfy the gear’s required center distance. On average, the module’s value is never no more than 2 mm. Moreover, in C2, this value is only achieved at the end of several iterations. The face width is also decreasing at the end of the iterations, and the pinion and wheel shift coefficients in C1. These coefficients tend to increase in C2.

Conclusions

The present work is aimed at solving the design problem of spur gears under the usual transmission ratio and center distance requirements. The optimized designs have been compared to selected gears from the recent literature. Single and multiple-objective problems have been considered: 1. gear mass reduction; 2. gear mass reduction and gear loss factor HVLH_{\mathrm{VL}}; 3. gear mass reduction and increase of safety factors SFS_F and SHS_H; 4. gear mass, specific sliding and HVLH_{\mathrm{VL}} reductions and increase of SFS_F, SHS_H, and contact ratio. The main geometrical parameters of the spur gear are the design variables: pinion and wheel number of teeth and profile shift coefficients, module and face width. There are constraints considered in the present work: SHS_H and SFS_F may not be lower than 1.4 and 2, respectively, and the contact ratio may not be lower than 1.4. Also, the face width is considered between 6 and 12 times the value of the module. Moreover, module values are standard. The main results are as follows:

  • Gear mass reductions of 51% to 55% have been achieved compared to the literature [5, 9] due to the reduction of the module’s value and, consequently, the face width.

  • The reduction of the gear mass led to safety factors near their possible minima.

  • The difference between specific sliding decreases by up to 98%, compared to the literature [5, 9], in multi-objective optimization.

  • The gear mass increases when more objectives are considered.

However, further investigations need to be carried out concerning the optimization problems of spur gears. Firstly, solving the optimization problem is costly, and the considered feasible search space is significantly big. Therefore, a different approach regarding reducing the feasible search space to a search space near the safety factor constraints can be considered in the future. Also, the significant amount of mechanical and geometrical constraints regarding the profile shift coefficients may have been driven to an almost random search; the PSO algorithm controls it if other design variables such as the number of teeth and the module do not change. Therefore, the proposed algorithm may also be a topic of interest for further improvements in future works.

The significant contribution of this work is an optimization algorithm that proposes a suitable spur gear for specific objectives and with requirements of transmission ratio and center distance. In the future, the development of an interface for choosing the optimized spur gear based on objectives that the user weights would be a topic of interest for the scientific community’s support.

Annex A

HVL variation for Cases 1 and 2

Figure 5. Gear loss factor variation along with the number of evaluations. In general, HVLH_{\mathrm{VL}} tends to have a lower value for a reduced mass.

Variation of the pinion number of teeth along with the number of evaluations

Figure 6. Variation of the pinion number of teeth along with the number of evaluations. In general, the number of teeth in the pinion tends to increase to achieve a reduced gear mass.

Gear module variation along with the number of evaluations

Figure 7. Gear module variation along with the number of evaluations. In general, the module tends to decrease to achieve a reduced gear mass.

Pinion shift coefficient variation along with the number of evaluations

Figure 8. Pinion shift coefficient variation along with the number of evaluations. In Case 1, the shift coefficient tends to have a lower value; in Case 2, the value is increased.

Gear face width variation plots for Case 1 and Case 2

Figure 9. Gear face width variation along with the number of evaluations. In general, the face width is reduced for the achievement of a reduced gear mass.

Wheel shift coefficient variation plots for Case 1 and Case 2

Figure 10. Wheel shift coefficient variation along with the number of evaluations. In Case 1, this shift coefficient tends to have a minimum value corresponding to the side constraint of the problem; in Case 2, the coefficient tends to decrease.

Flank press safety factor variation plots for Case 1 and Case 2

Figure 11. Flank press safety factor variation along with the number of evaluations. In general, since the face width decreases, the pressure increases, reducing the safety factor.

Root stress safety factor variation along with the number of evaluations

Figure 12. Root stress safety factor variation along with the number of evaluations. In general, since the face width decreases, the pressure increases, reducing the safety factor.

Specific sliding variation along with the number of evaluations

Figure 13. Specific sliding variation along with the number of evaluations. In general, the specific sliding decreases with the decrease of the gear mass.

References

  1. [1]E. Maier, A. Ziegltrum, T. Lohner, and K. Stahl, ‘Characterization of TEHL contacts of thermoplastic gears’, Forsch. im Ingenieurwes., vol. 81, no. 2, pp. 317–324, 2017, doi: 10.1007/s10010-017-0230-4.DOI
  2. [2]D. J. Politis, N. J. Politis, and J. Lin, ‘Review of recent developments in manufacturing lightweight multi-metal gears’, Prod. Eng., vol. 15, no. 2, pp. 235–262, 2021, doi: 10.1007/s11740-020-01011-5.DOI
  3. [3]Q. Yao, ‘Multi-objective optimization design of spur gear based on NSGA-II and decision making’, Adv. Mech. Eng., vol. 11, no. 3, pp. 1–8, 2019, doi: 10.1177/1687814018824936.DOI
  4. [4]S. Li and A. Kahraman, ‘A scuffing model for spur gear contacts’, Mech. Mach. Theory, vol. 156, p. 104161, 2021, doi: https://doi.org/10.1016/j.mechmachtheory.2020.104161.
  5. [5]E. S. Maputi and R. Arora, ‘Multi-objective spur gear design using teaching learning-based optimization and decision-making techniques’, Cogent Eng., vol. 6, no. 1, p. 1665396, Jan. 2019, doi: 10.1080/23311916.2019.1665396.DOI
  6. [6]Maputi, Edmund S. and Arora, Rajesh, ‘Design optimization of a three-stage transmission using advanced optimization techniques’, Int. J. Simul. Multidisci. Des. Optim., vol. 10, p. A8, 2019, doi: 10.1051/smdo/2019009.DOI
  7. [7]Z. Chen, Y. Jiang, Z. Tong, and S. Tong, ‘Residual Stress Distribution Design for Gear Surfaces Based on Genetic Algorithm Optimization’, Materials , vol. 14, no. 2. 2021, doi: 10.3390/ma14020366.DOI
  8. [8]D. Miler, D. Žeželj, A. Lončar, and K. Vučković, ‘Multi-objective spur gear pair optimization focused on volume and efficiency’, Mech. Mach. Theory, vol. 125, pp. 185–195, 2018, doi: https://doi.org/10.1016/j.mechmachtheory.2018.03.012.
  9. [9]D. Miler, A. Lončar, D. Žeželj, and Z. Domitran, ‘Influence of profile shift on the spur gear pair optimization’, Mech. Mach. Theory, vol. 117, pp. 189–197, 2017, doi: https://doi.org/10.1016/j.mechmachtheory.2017.07.001.
  10. [10]H. Abderazek, D. Ferhat, I. Atanasovska, and K. Boualem, ‘A differential evolution algorithm for tooth profile optimization with respect to balancing specific sliding coefficients of involute cylindrical spur and helical gears’, Adv. Mech. Eng., vol. 7, no. 9, pp. 1–11, 2015, doi: 10.1177/1687814015605008.DOI
  11. [11]J. A. Korta and D. Mundo, ‘A population-based meta-heuristic approach for robust micro-geometry optimization of tooth profile in spur gears considering manufacturing uncertainties’, Meccanica, vol. 53, no. 1, pp. 447–464, 2018, doi: 10.1007/s11012-017-0737-7.DOI
  12. [12]K. Daoudi and E. M. Boudi, ‘Genetic Algorithm Approach for Spur Gears Design Optimization’, in 2018 International Conference on Electronics, Control, Optimization and Computer Science (ICECOCS), 2018, pp. 1–5, doi: 10.1109/ICECOCS.2018.8610520.DOI
  13. [13]M. Hofstetter, D. Lechleitner, M. Hirz, M. Gintzel, and A. Schmidhof­er, ‘Multi-objective gearbox design optimization for xEV-axle drives under consideration of package restrictions’, Forsch. im Ingenieurwesen/Engineering Res., vol. 82, no. 4, pp. 361–370, 2018, doi: 10.1007/s10010-018-0278-9.DOI
  14. [14]K. Deb, S. Agrawal, A. Pratap, and T. Meyarivan, ‘A Fast Elitist Non-dominated Sorting Genetic Algorithm for Multi-objective Optimization: NSGA-II BT - Parallel Problem Solving from Nature PPSN VI’, 2000, pp. 849–858.
  15. [15]Ü. Atila, M. Dörterler, R. Durgut, and İ. Şahin, ‘A comprehensive investigation into the performance of optimization methods in spur gear design’, Eng. Opt., vol. 52, no. 6, pp. 1052–1067, Jun. 2020, doi: 10.1080/0305215X.2019.1634702.DOI
  16. [16]T. Yokota, T. Taguchi, and M. Gen, ‘A solution method for optimal weight design problem of the gear using genetic algorithms’, Comput. Ind. Eng., vol. 35, no. 3, pp. 523–526, 1998, doi: https://doi.org/10.1016/S0360-8352(98)00149-1.
  17. [17]V. Savsani, R. V Rao, and D. P. Vakharia, ‘Optimal weight design of a gear train using particle swarm optimization and simulated annealing algorithms’, Mech. Mach. Theory, vol. 45, no. 3, pp. 531–541, 2010, doi: https://doi.org/10.1016/j.mechmachtheory.2009.10.010.
  18. [18]N. Godwin Raja Ebenezer, S. Ramabalan, and S. Navaneethasanthakumar, ‘Design optimisation of mating helical gears with profile shift using nature inspired algorithms’, Aust. J. Mech. Eng., pp. 1–8, May 2020, doi: 10.1080/14484846.2020.1761007.DOI
  19. [19]R. Guilbault and S. Lalonde, ‘Tip relief designed to optimize contact fatigue life of spur gears using adapted PSO and Firefly algorithms’, SN Appl. Sci., vol. 3, no. 1, p. 66, 2021, doi: 10.1007/s42452-020-04129-4.DOI
  20. [20]Z. Tomori, ‘An Optimal Choice of Profile Shift Coefficients for Spur Gears’, Machines , vol. 9, no. 6. 2021, doi: 10.3390/machines9060106.DOI
  21. [21]J. Tavčar, B. Černe, J. Duhovnik, and D. Zorko, ‘A multicriteria function for polymer gear design optimization’, J. Comput. Des. Eng., vol. 8, no. 2, pp. 581–599, 2021, doi: 10.1093/jcde/qwaa097.DOI
  22. [22]P. Rai and A. G. Barman, ‘Tooth Profile Optimization of Helical Gear with Balanced Specific Sliding Using TLBO Algorithm BT - Advanced Engineering Optimization Through Intelligent Techniques’, 2020, pp. 203–210.
  23. [23]J. Wang and I. Howard, ‘Finite element analysis of High Contact Ratio spur gears in mesh’, J. Tribol., vol. 127, no. 3, pp. 469–483, 2005, doi: 10.1115/1.1843154.
  24. [24]N. Gunantara, ‘A review of multi-objective optimization: Methods and its applications’, Cogent Eng., vol. 5, no. 1, p. 1502242, Jan. 2018, doi: 10.1080/23311916.2018.1502242.DOI
  25. [25]C. Antônio, Otimização de Sistemas em Engenharia: Fundamentos e algoritmos para o projeto ótimo, Quântica E. Porto, 2020.
  26. [26]Y. Sun, Y. Gao, and X. Shi, ‘Chaotic multi-objective particle swarm optimization algorithm incorporating clone immunity’, Mathematics, vol. 7, no. 2, pp. 1–16, 2019, doi: 10.3390/math7020146 M4 - Citavi.DOI
  27. [27]R. Eberhart and J. Kennedy, ‘A new optimizer using particle swarm theory’, in MHS’95. Proceedings of the Sixth International Symposium on Micro Machine and Human Science, Oct. 1995, pp. 39–43, doi: 10.1109/MHS.1995.494215.
  28. [28]D. ping Tian, ‘A Review of Convergence Analysis of Particle Swarm Optimization’, Int. J. Grid Distrib. Comput., vol. 6, no. 6, pp. 117–128, 2013, doi: 10.14257/ijgdc.2013.6.6.10 M4 - Citavi.DOI
  29. [29]Y. Shi and R. Eberhart, ‘A modified particle swarm optimizer’, in 1998 IEEE International Conference on Evolutionary Computation Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98TH8360), May 1998, pp. 69–73, doi: 10.1109/ICEC.1998.699146.
  30. [30]M. Clerc, ‘The swarm and the queen: towards a deterministic and adaptive particle swarm optimization’, in Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat. No. 99TH8490(6), Jul. 1999, vol. 3, pp. 1951-1957 Vol. 3, doi: 10.1109/CEC.1999.785513.
  31. [31]M. Clerc and J. Kennedy, ‘The particle swarm - explosion, stability, and convergence in a multidimensional complex space’, IEEE Trans. Evol. Comput., vol. 6, no. 1, pp. 58–73, Feb. 2002, doi: 10.1109/4235.985692.DOI
  32. [32]F. van den Bergh, ‘An Analysis of Particle Swarm Optimizers’, no. November, p. 315, 2001.
  33. [33]R. Fitas, ‘Optimal Design of Composite Structures using the Particle Swarm Method and Hybridizations’, University of Porto, 2022.DOI
  34. [34]C. Fitas, Ricardo; Hesseler, Stefan; Wist, Santino; Greb, ‘Kinematic Drawing Simulation Optimization of a Composite B-Pillar Geometry using Particle Swarm Optimization’.
  35. [35]R. Fitas, G. das Neves Carneiro, and C. Conceição António, ‘An elitist multi-objective particle swarm optimization algorithm for composite structures design’, Compos. Struct., vol. 300, p. 116158, 2022, doi: https://doi.org/10.1016/j.compstruct.2022.116158.

Paper details

vlnc:W_bctqar2x
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents