Professional Documents
Culture Documents
The manuscript was received on 15 March 2006 and was accepted after revision for publication on 8 January 2007.
DOI: 10.1243/0954406JMES342
Keywords: spur gears, bending strength, contact pressure, pitting resistance, optimization,
boundary element analysis, photoelasticity
JMES342 © IMechE 2007 Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
480 V Spitas and C Spitas
forms for increasing the thickness at the fillet [7]. 2 NON-DIMENSIONAL GEAR TOOTH
Although promising, this approach produces non- MODELLING
interchangeable gears, which require special tooling
for their generation. Additionally, these gears exhibit Consider the gear pair shown in Fig. 1: gear 1 being
lower contact ratio and are sensitive to errors in their the pinion and gear 2 the wheel. The law of gear-
centre distance. Also, since only nominal stresses ing [1, 12] requires that both members of the pair
are calculated during the optimization procedure, should have the same nominal pressure angle αo and
ignoring stress concentration phenomena, the results the same module m in order to mesh properly. In the
are only indicative and cannot be used literally in general case, the gears are considered to have adden-
practical gear manufacturing. dum modifications x1 , x2 , respectively, and therefore
In the recent years, the works of Litvin et al. [8] and their pitch thickness is calculated from the following
Kapelevich [9] have introduced asymmetric gear teeth relationship [12]
for increasing root bending strength. These teeth are
composed of a standard 20◦ standard involute flank
on the driving side and of a higher pressure angle soi = csi mπ + 2xi m tan αo = soiu m (1)
(i.e. 25◦ ) involute flank on the coast side. Although
asymmetric involute gears are indeed stronger, they
too present some limitations regarding (a) equally where csi is the thickness coefficient of gear i, (i =
efficient use in both directions of motion and (b) need 1, 2), which in the general case is cs1 = 0.5 = cs2 , and
for special tooling. soiu is the pitch thickness of the corresponding non-
In this paper, non-standard involute gears with dimensional gear for which the module m and the
optimized tooth geometry for minimum root stress face width b are both equal to unity.
are presented. The independent free geometrical The centre distance O1 O2 is calculated using the
design parameters for the gear pair are the thick- following formula
ness coefficients for both gears (i.e. pinion and wheel)
and the individually applied addendum modifica- z1 + z2
tions. The active constraints include all the basic a12 = m + (x1 + x2 )m = a12u m (2)
2
geometrical requirements such as non-zero top land
thickness, the basic kinematic requirements such
The actual operating pitch circle rbi of gear i, (i = 1, 2)
as lack of interference or backlash/thickness con-
should verify the law of gearing and therefore be
flict, and the basic manufacturing issues such as
equal to
ability for the gear teeth to be generated using
standard cutting tools (i.e. hobs). In order for the
optimal gears to be fully interchangeable with their zi
rbi = a12u m = rbiu m (3)
non-optimal standard counterparts, additional con- z1 + z2
straints are the preservation of the nominal centre
distance (S0 corrected gears [1]) and the preserva-
tion of the maximum contact pressure (pitting resis- Let us now consider gears 1 and 2 revolving about
tance) at the lowest point of single tooth contact their centres O1 and O2 , respectively, and meshing
(LPSTC). along the path of contact AB as illustrated in Fig. 1.
In this optimization process, the real maximum ten- During meshing there are two pairs of gear teeth in
sile stresses developed at the root fillet due to bending contact along the segments AA and BB , thus sharing
are considered instead of the nominal stresses. To the total normal load, while there is only a single such
achieve this, the maximum stresses at each iterative pair when tooth contact takes place along the central
step of the algorithm are calculated from interpola- region A B , carrying the total normal load. Point B
tion of tabulated values obtained through BEA runs is the highest point of single tooth contact (HPSTC)
on selected gear tooth models. In order to reduce the for gear 1 and its position, defining the radius rB ,
number of the independent variables, the concept is [10]
of non-dimensional tooth modelling introduced by
Spitas and Costopoulos [10] is used. rB = O1 B
The results obtained by use of the complex
algorithm [11] indicate that a reduction of the max-
imum root stress up to 8.5 per cent can be achieved. = 2
rk1 + (ε − 1)tg (ε − 1)tg − 2 rk1
2
− rg21 (4)
The optimal design was verified experimentally by
using two-dimensional photoelasticity on suitable
machined single-tooth specimens from birefringent Dividing by the module of the pair, the above equation
polycarbonate material. yields its equivalent in terms of non-dimensional
Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science JMES342 © IMechE 2007
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
Optimizing involute gear design 481
2 2 z 2
rB z1 z1
rB u = = + x1 + ck + (ε − 1)π cos αo (ε − 1)π cos αo − 2 + x1 + ck −
1
cos αo (5)
m 2 2 2
From the above equation it is evident that the posi- The non-dimensional modelling can incorporate
tion of the HPSTC of a gear depends only on its the tooth-related stress field by defining a non-
geometry and on the contact ratio of the pair, in dimensional stress σu (z, x, cs , ε), which is produced by
which all the characteristics of the mating gear are applying a unitary load PNu = 1 at the HPSTC. This
incorporated in a condensed form. is related to the actual stress σ through the following
Therefore, the mechanical behaviour of every gear equation
can be modelled only by using its own geomet-
rical characteristics z, x, cs and the contact ratio PN
σ = σu (6)
ε of the pair instead of using all of the geometrical bm
characteristics of the mating gear. Also, the use of non-
dimensional teeth further simplifies the problem, as
every dimension of the transverse section of a full- 3 CALCULATION OF STRESSES USING BEA
scale gear tooth can be derived by multiplying the CONCEPT OF STRESS TABLES
corresponding dimension of the transverse section of
the non-dimensional gear tooth with the module of The established analytical methods rely on the cal-
the pair. culation of the nominal tensile bending stress at the
JMES342 © IMechE 2007 Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
482 V Spitas and C Spitas
Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science JMES342 © IMechE 2007
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
Optimizing involute gear design 483
ε = 1.2
x = −0.2 5.730 4.818 4.141 3.621 3.211
x = +0.0 4.965 4.246 3.696 3.262 2.912
x = +0.2 4.502 3.887 3.406 3.021 2.708
x = +0.4 4.032 3.510 3.094 2.758 2.484
x = +0.7 3.729 3.257 2.873 2.561 2.306
ε = 1.4
x = −0.2 4.906 4.184 3.642 3.222 2.889
x = +0.0 4.214 3.652 3.218 2.874 2.596
x = +0.2 3.800 3.321 2.945 2.642 2.397
x = +0.4 3.413 3.005 2.680 2.418 2.203
x = +0.7 3.054 2.702 2.415 2.183 1.993
ε = 1.6
x = −0.2 4.391 3.793 3.339 2.986 2.703
x = +0.0 3.741 3.283 2.927 2.643 2.412
x = +0.2 3.282 2.911 2.618 2.382 2.189
x = +0.4 2.942 2.629 2.380 2.177 2.012
x = +0.7 2.602 2.339 2.125 1.952 1.809
ε = 1.8
x = −0.2 3.978 3.485 3.109 2.812 2.573
x = +0.0 3.356 2.990 2.703 2.472 2.284
x = +0.2 2.928 2.640 2.411 2.225 2.072
x = +0.4 2.616 2.379 2.188 2.033 1.904
x = +0.7 2.253 2.070 1.921 1.800 1.699
JMES342 © IMechE 2007 Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
484 V Spitas and C Spitas
respectively, when loaded at their corresponding dedendum coefficients, respectively. The radial
HPSTC. clearance constraint is shown schematically in Fig. 3.
Naturally, geometrical, manufacturing and kine-
matical constraints should be applied because the Constraint 4: Interference. If the tip radius rki of gear
optimum teeth should still fulfil certain opera- i revolving about Oi exceeds a maximum value rki max
tional criteria. There are nine different constraints so that the intersection of the tip circle of the gear with
described subsequently and in order to include them the common path of contact at point U defines on the
in the optimization procedure, the following form mating gear j a radius which is lower than its form
of the objective function is adopted using weighted radius rjs , then interference occurs, since the tooth
residuals part below the form radius has a trochoidal and not an
involute form. Consequently, it should always be rki
9
rki max , where rki max = Oi U. The interference constraint
min f (x1 , x2 , cs1 , cs2 ) = max(σ1 , σ2 ) + wi ci (15)
i=1
is shown schematically in Fig. 3.
The penalty functions ci and the weighting coef- Constraint 5: Minimum tip thickness. In common
ficients wi employed in equation (15) are defined gear practice, the tip thickness is never below 0.2 times
below. In general, the penalty functions either take the module, otherwise fracture of the top land of the
very big values when variables exceed the permissi- tooth could occur. In a non-dimensional gear, the tip
ble design boundaries in order to exclude them from thickness should always be sku 0.2.
the next iterative steps (penalty), or are related to the
actual variable values in order to facilitate smooth Constraint 6: Allowable contact ratio. In order to
convergence. In the latter case, the values of the ensure smooth running, the contact ratio should
weighting coefficients are chosen as to provide prod- exceed 1.2. A usual upper limit is 1.8, which in 20◦
ucts wi xi comparable to the non-dimensional stresses standard or shifted spur gears is never surpassed. Sim-
within the objective function and thus streamlining ilarly to the constraints 1 and 2, the contact ratio ε of
the process. the gear pair should lie in the range defined in the
stress tables, therefore big penalties are applied at the
Constraint 1: Allowable addendum modification. boundaries.
The addendum modification coefficient for gear i
is restricted between two values xi min and xi max , Constraint 7: Allowable backlash. The backlash B
depending on the number of teeth z. These values are of a gear pair should always be positive and usually
dictated by common gear practice and manufacture. optimized designs require that this is kept minimum,
In each stress table, there is a range of addendum since thicker teeth are subjected to lower root stress.
modification coefficients that must not be exceeded. Although zero backlash is never actually desirable
for power transmissions, the presence of a minimum
Constraint 2: Allowable thickness coefficients. For backlash does not seriously reduce the tooth thick-
technical reasons, the cutting tool producing the ness, hence the root stresses, and therefore in order to
gears (rack cutter, pinion cutter, or hob) cannot simplify the calculations the optimum backlash can
have infinitely thick or infinitely thin teeth, therefore be considered zero. This can be expressed in terms of
imposing a constraint on the resulting thickness of the a penalty function, suitably big beyond the permis-
generated gear. Thus the thickness coefficient should sible boundaries. The backlash constraint is shown
range between the values cs min and cs max , which are schematically in Fig. 3.
the limit values specified in the stress tables.
Constraint 8: Same contact pressure. Since the pit-
Constraint 3: Minimum radial clearance. In order to ting resistance of a gear pair is proportional to the
ensure that the conjugate gears operate without the developed contact pressure at the LPSTC [1], then
risk of seizure, there should be a minimum allow- the application of a penalty function equal to c8 =
able radial clearance cr min m, where cr min = 0.25. For ((p/pmax ) − 1) will ensure that the pitting resistance
the non-dimensional gear i, it is calculated from the provided by the optimal solution is not decreased
equation when compared with the initial value.
The maximum contact pressure at the LPSTC is
cri = a12 − rkiu − rf iu = cf − ck = cr given by Niemann and Winter [18]
Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science JMES342 © IMechE 2007
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
Optimizing involute gear design 485
Fig. 3 Mesh constraints affecting the design decisions and the optimization process
where ρ is the equivalent radius of curvature of the 4.2 The Complex optimization algorithm
tooth flanks in contact at the LPSTC and it is calcu-
lated from the local radii of curvature of gears 1 and 2 After the formulation of the objective function, the
(ρ1 , ρ2 , respectively) from the equation complex iterative algorithm, as described later, is
executed.
The objective function as per equation (15) is con-
1 1 1 structed and the problem is defined as an uncon-
= + (17) strained minimization problem (the constraints have
ρ ρ1 ρ2
been incorporated within the objective function)
Positive profile shifting usually plays a detrimental min f (x), where x = (x1 , x2 , . . . , xn )T
role in pitting resistance due to the decrease in contact The solution procedure is as follows
ratio, hence the movement of the LPSTC to regions Generate a complex of m + 1 random vectors
with smaller local curvature. x 0 , x 1 , x 2 , . . . , x m . The mean value of this complex is
1
m
Constraint 9: Same centre distance. The applica- x= xi (18)
tion of profile shifting on both gear members leads m + 1 i=0
to a change in their nominal centre distance by
the amount of (x1 + x2 )m. Therefore it is impossi- The following norm is used as the convergence crite-
ble to achieve the same centre distance with shifted rion
gear teeth, unless the positive shift of one of the
1 m
gears equals the negative shift of the other gear f (x i ) − f x < ε (19)
(S0 corrected gears). A suitable penalty function is m + 1 i=0
c9 = |ao − a|, where ao is the nominal centre dis-
tance and a is the current centre distance, respec- where the tolerance ε is a sufficiently small positive
tively. real number.
JMES342 © IMechE 2007 Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
486 V Spitas and C Spitas
4.2.1 Reflection
Gear teeth
Pinion teeth 18 22 28 50
Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science JMES342 © IMechE 2007
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
Optimizing involute gear design 487
of the two gears after the application of the mod- that frictional forces at the contact point were not
ifications, as illustrated in Fig. 5. The pinion tooth present. The experimental results are in excellent
always ends with increased thickness at its root as agreement with the numerical predictions (maxi-
opposed to the mating gear tooth. After the modifi- mum deviation of 2.8 per cent) and the new design
cation, both gear members present the same resis- offers a decrease of the maximum fillet stress rang-
tance to bending, as anticipated from an optimal ing from 1.6 per cent (15 tooth pinion in mesh with
design. a 18 tooth pinion) to 8.5 per cent (15 tooth pinion in
The optimal design has been experimentally veri- mesh with a 50 tooth wheel) for the examined gear
fied using two-dimensional photoelasticity. The spec- combinations. The measured difference between the
imens corresponding to the above-calculated tooth pinion and the mating gear fillet stress never exceeds
profiles were made of special birefringent polycar- 1.6 per cent.
bonate material. The fillet stresses were experimen-
tally measured on a circular plane polariscope under
monochromatic sodium light Fig. 6. The loading was 6 CONCLUSIONS
exerted on the specimen with a special mechanism
constructed in the laboratory in order to ensure In this paper, non-standard involute gears with opti-
that the load was always normal to the profile and mized tooth geometry for minimum root stress were
presented. The constraints used for the optimization
included apart from the basic kinematic, manufac-
turing and geometrical limitations constant contact
pressure to assure same pitting resistance and same
centre distance in order for the optimum gears to
be fully interchangeable with the standard gears in
a given housing. Non-dimensional gear tooth mod-
elling was used to obtain generalized results and BEA
was employed to calculate the maximum fillet stresses
for various loading conditions. The results were tabu-
lated in stress tables, allowing significant decrease in
computational time.
The optimal gear designs obtained indicated that
reduction of the maximum root stress up to 8.5 per
cent can be achieved depending on the number of
teeth of the pinion and mating gear. These find-
ings were further confirmed experimentally using
two-dimensional photoelasticity.
REFERENCES
JMES342 © IMechE 2007 Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015
488 V Spitas and C Spitas
Proc. IMechE Vol. 221 Part C: J. Mechanical Engineering Science JMES342 © IMechE 2007
Downloaded from pic.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 18, 2015