materials
Article
Path Generation Strategy and Wire Arc Additive Manufacturing
of Large Aviation Die with Complex Gradient Structure
Jiansheng Zhang 1,2 , Guiqian Xiao 1,2 , Jie Peng 1 , Yingyan Yu 1 and Jie Zhou 1, *
Citation: Zhang, J.; Xiao, G.; Peng, J.;
Yu, Y.; Zhou, J. Path Generation
Strategy and Wire Arc Additive
Manufacturing of Large Aviation Die
with Complex Gradient Structure.
Materials 2022, 15, 6115. https://
doi.org/10.3390/ma15176115
Academic Editor: Frank Czerwinski
Received: 16 August 2022
Accepted: 30 August 2022
Published: 2 September 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Materials 2022, 15, 6115. https://doi.org/10.3390/ma15176115
1 Chongqing Key Laboratory of Advanced Mold Intelligent Manufacturing, College of Materials Science and
Engineering, Chongqing University, Chongqing 400044, China
2 Chongqing Jiepin Technology Co., Ltd., Chongqing 400000, China
* Correspondence: zhoujie635635@163.com
Abstract: To realize automatic wire arc additive manufacturing (WAAM) of a large aviation die with
a complex gradient structure, a new contour-parallel path generation strategy was proposed and
practically applied. First, the planar curve was defined as a vertical slice of a higher-dimensional
surface and a partial differential equation describing boundary evolution was derived to calculate
the surface. The improved Finite Element Method (FEM) and Finite Difference Method (FDM) were
used to solve this partial differential equation. Second, a cross section of a large aviation die was
used to test the path-generation algorithms. The results show that FEM has a faster solving speed
than FDM under the same solving accuracy because the solving domain of FEM mesh was greatly
reduced and the boundary mesh could be refined. Third, the die was divided into three layers: base
layer, transition layer (Fe-based material) and strengthening layer (Co-based material) according to
the difference of the temperature and stress field, and corresponding WAAM process parameters
has been discussed. The optimum welding parameters are obtained as follows: voltage is 28 V, wire
feeding speed is 8000 mm/min and welding speed is 450 mm/min. Finally, the path generation
strategy was practically applied to the remanufacture of the large aircraft landing gear die with a
three-layer structure. The application test on aircraft landing gear dies justified the effectiveness
of the algorithms and strategy proposed in this paper, which significantly improved the efficiency
of the WAAM process and the service life of large aviation dies with complex gradient structures.
The microstructure of the fusion zone shows that the base metal and welding material can be fully
integrated into the welding process.
Keywords: WAAM; contour-parallel path generation strategy; large aviation die; application test
1. Introduction
The service life of traditional large aviation dies is extremely low due to severe working
conditions with high temperatures and heavy loads. WAAM technology is gradually
applied to the manufacturing and repairing of large aviation dies. The core tasks of
WAAM are the welding path generation and application experiment, which directly affect
the quality of additive manufacturing [1,2]. There is much research on welding path
generation. Direction-parallel path and contour-parallel path are two basic methods for
filling arbitrary plane boundaries. The contour-parallel path is more widely used for some
complex boundaries’ WAAM because of its smoother filling trajectory and fewer turning
points [3–5]. In addition, there are many other methods that have certain limitations and
cannot be used for arbitrary boundaries to generate the filling path, such as raster [6],
zigzag [7], spiral [8], fractal space-filling curves [9,10], and hybrid tool path planning
approaches [11,12]. In the contour-parallel path, the contour offsetting will encounter
many polygon operation problems [13], such as ambiguous normal (Figure 1a), singular
point (Figure 1b), self-intersection (Figure 1c), change in topology (Figure 1d) and offsets
intersection of pocket and island boundary (Figure 1e). Many approximate methods, such
https://www.mdpi.com/journal/materials
Materials 2022, 15, x FOR PEER REVIEW 2 of 14

polygon operation problems [13], such as ambiguous normal (Figure 1a), singular point
Materials 2022, 15, 6115 2 of 13
(Figure 1b), self-intersection (Figure 1c), change in topology (Figure 1d) and offsets inter-
section of pocket and island boundary (Figure 1e). Many approximate methods, such as
iterative methods for spline offsetting [14,15], Voronoi diagram-based offsetting [16,17],
as iterative
pixel methods
simulation for spline
and image offsetting [14,15],
processing-based Voronoi[18]
offsetting diagram-based offsetting
and Laplace-based [16,17],
formula-
pixel simulation and image processing-based offsetting [18] and Laplace-based formulation
tion offsetting [19] have been proposed. In iterative methods for spline offsetting, the de-
offsetting
tection [19] have been
of singularities andproposed. In iterative
redundancies is verymethods for spline
complicated offsetting,
to carry the detection
out. Although com-
of singularities and redundancies is very complicated to carry out. Although complex
plex detection can be avoided in Voronoi diagram-based offsetting, two-dimensional
detection can be avoided in Voronoi diagram-based offsetting, two-dimensional Boolean
Boolean set operations take a long time to compute. Pixel simulation and image-pro-
set operations take a long time to compute. Pixel simulation and image-processing-based
cessing-based offsetting require complex pre-processing and post-processing. Laplace-
offsetting require complex pre-processing and post-processing. Laplace-based methods are
based methods are suitable for smooth and continuous path generation [20]. However,
suitable for smooth and continuous path generation [20]. However, the part is sometimes
the part is sometimes overmachined, which depends on the shape and the distance be-
overmachined, which depends on the shape and the distance between two neighboring
tween two neighboring contours [11].
contours [11].

(a) (b) (c) (d) (e)

Various
Figure1.1.Various
Figure problems
problems encountered
encountered in contour
in contour offsetting
offsetting [11]: (a)[11]: (a) Ambiguous
Ambiguous normal;
normal; (b) Sin-
gular point; (c)point;
(b) Singular Self-intersection; (d) Change
(c) Self-intersection; (d)inChange
topology;
in (e) offset intersection
topology; of pocket and
(e) offset intersection of island
pocket
boundary.
and island boundary.

InInaddition
additiontotofilling
filling algorithms,
algorithms, many
many scholars
scholars havehave conducted
conducted some some application
application re-
research
search on the
on the WAAM WAAM process.
process. Lu proposed
Lu proposed a method
a method for repairing
for repairing forging forging diesman-
dies with with
manual
ual WAAM WAAM [21].surface
[21]. The The surface
strength strength and hardness
and hardness of the
of the die are die are strengthened
strengthened by cover- by
covering
ing the high-strength
the high-strength materialmaterial on theofsurface
on the surface the die.ofAnthe die. AnWAAM
automatic automatic WAAM
technology
istechnology
proposed is byproposed
Hong et al. by [22]
Hong toetrealize
al. [22]the
to gradient
realize theadditive
gradientremanufacturing
additive remanufacturing
of ultra-
of ultra-large hot forging dies. The experimental results show that the
large hot forging dies. The experimental results show that the service life of the hot forging service life of the hot
forging dies repaired by the automatic WAAM technology is
dies repaired by the automatic WAAM technology is significantly increased, the materialsignificantly increased, the
material is reduced by more than 50% and the production efficiency
is reduced by more than 50% and the production efficiency is increased by more than 50%. is increased by more
Tothan 50%. Tothe
improve improve
servicethe service
life of hotlife of hot forging
forging dies, additive
dies, additive manufacturing
manufacturing algorithms
algorithms and
and additive
additive manufacturing
manufacturing devices devices
for diefor die remanufacturing
remanufacturing have developed
have been been developed [23].
[23]. The
The results show that using the techniques, welding materials
results show that using the techniques, welding materials can be saved by more than 50%can be saved by more than
50%machining
and and machining costsbecan
costs can be saved
saved by more by more than The
than 60%. 60%.experimental
The experimental resultsresults
show show
that
that WAAM technology is very promising for mold repair, and
WAAM technology is very promising for mold repair, and its core lies in path planning.its core lies in path planning.
ToToovercome
overcomethe theshortcomings
shortcomingsofofthe theabove
abovepathpathgeneration
generationmethods,
methods,a anew newcontour-
contour-
parallel path generation strategy based on the Level Set Function (LSF) is proposed.First,
parallel path generation strategy based on the Level Set Function (LSF) is proposed. First,a
Partial Differential Equation (PDE) is constructed to describe
a Partial Differential Equation (PDE) is constructed to describe the movement of the the movement of the bound-
aries. The equidistant
boundaries. filling of
The equidistant arbitrary
filling boundaries
of arbitrary can be transformed
boundaries into the equidistant
can be transformed into the
movement of the initial boundary in which the moving fronts are expanded or contracted.
equidistant movement of the initial boundary in which the moving fronts are expanded
Second, an improved Finite Element Method (FEM) and an improved Finite Difference
or contracted. Second, an improved Finite Element Method (FEM) and an improved Finite
Method (FDM) are developed to solve the partial differential equation of boundary value.
Difference Method (FDM) are developed to solve the partial differential equation of
Compared with FDM, FEM can obtain better solution accuracy with fewer nodes be-
boundary value. Compared with FDM, FEM can obtain better solution accuracy with
cause the mesh of FEM can approximate the complex boundary with a smaller number
fewer nodes because the mesh of FEM can approximate the complex boundary with a
of elements. Finally, a large aviation die is used to verify the new contour-parallel path
smaller number of elements. Finally, a large aviation die is used to verify the new contour-
generation strategy. The WAAM of the landing gear die proves that the algorithm and
parallel path generation strategy. The WAAM of the landing gear die proves that the al-
gradient structure proposed in this paper can improve the service life of the large aviation
gorithm and gradient structure proposed in this paper can improve the service life of the
die and can improve the efficiency of the WAAM process simultaneously.
large aviation die and can improve the efficiency of the WAAM process simultaneously.
2. Contour-Parallel Path Generation Strategy
2.2.1.
Contour-Parallel Path Generation Strategy
Mathematical Foundation
2.1. Mathematical Foundation
In general, a closed curve in a plane can be defined as S(k) = [ x (k), y(k )] T , in which k
is a In general, a closed
parametrization curve in
variable. a plane
The offset can beof
curve defined as 𝑆(𝑘)
the closed = is𝑥(𝑘),
curve 𝑦(𝑘)as ,follows:
defined in which
k is a parametrization variable. The offset curve of the closed curve is defined as follows:
→
S0 (k ) = S(k ) + n d (1)
 𝑆 (𝑘) = 𝑆(𝑘) + 𝑛⃗𝑑 (1)

Materials 2022, 15, 6115 where, 𝑛⃗ is the unit normal vector to the original curve at parametric point k. As dis-
3 of 13
cussed in the introduction, many problems arise with this approach, such as ambiguous
normal, rarefaction fan at corner point, self-intersection, generation of singular point,
→
change in topology
where, and normal
n is the unit intersection
vector of
to offsets of pocket
the original and
curve at island boundary
parametric point k. As(Figure
discussed1).
Another importantmany
in the introduction, way problems
to definearise
a closed curve
with this in a plane
approach, such is
as the intersection
ambiguous normal,of a
rarefaction
surface fan at
and a plane, ascorner point, self-intersection, generation of singular point, change in
follows:
topology and intersection of offsets of pocket and island boundary (Figure 1).
Another important way to define𝜓(𝑥, 𝑦) = curve
a closed 0 in a plane is the intersection of a(2)
surface and a plane, as follows:
where, 𝜓 is an implicit function, also known as the level set function.
ψ( x, y) = 0 (2)
As shown in Figure 2a, the workpiece is cut to obtain a cross section where the inner
where,
boundary ψ isouter
and an implicit function,
boundary also known
are defined as the leveland
as clockwise set function.
anticlockwise arrangements,
As shown in Figure 2a, the workpiece is cut to obtain a cross
respectively. In particular, as shown in Figure 2b, the inner and outer section where the inner
boundaries can be
boundary and outer boundary are defined as clockwise and anticlockwise arrangements,
represented by the 0-value contour of a three-dimensional surface ψ. The key task is how
respectively. In particular, as shown in Figure 2b, the inner and outer boundaries can be
to construct a calculation method for this surface.
represented by the 0-value contour of a three-dimensional surface ψ. The key task is how
to construct a calculation method for this surface.

(a) (b)
FigureFigure 2. Cross
2. Cross section
section and and boundary
boundary of aoflanding
a landing geardie:
gear die:(a)
(a)Cross
Cross section;
section; (b)
(b)Boundary.
Boundary.

It is impossible to obtain the level set function directly. A desirable method is to

Itdetermine
is impossible to obtain the level set function directly. A desirable method is to de-
the differential equation of the surface function. The time differential expression
termine the differential
of Equation (2) is as equation
follows: of the surface function. The time differential expression
of Equation (2) is as follows: ∂ψ
+ F ||∇ψ|| = 0 (3)
𝜕𝜓
∂t
+ 𝐹‖∇𝜓‖
where F = 1 − εγ and the γ is the curvature of the=corresponding
0 point on the curve. The(3)
𝜕𝑡
ε is the smoothing term coefficient (for equidistant filling, ε = 0). F is the front speed in
wheretheF =direction
1 − εγ and the γtoisthe
normal theboundary.
curvatureThen,
of the corresponding
Equation point onHamilton-Jacobi
(3) is a standard the curve. The ε
is the equation
smoothing [24].term
The coefficient (for equidistant
initial conditions filling,
on ψ are defined ε = 0). F is the front speed in the
as follows:
direction normal to the boundary. Then, Equation q (3) is a standard Hamilton-Jacobi equa-
tion [24]. The initial conditions
ψ( x, y,on
t =ψ0are
) = defined
±min (as x− x ) + ( y − y 0 )2
0 2
follows: (4)

where x and y represent 𝜓(𝑥, 𝑦, the𝑡 horizontal

= 0) = 𝑚𝑖𝑛 (𝑥 − 𝑥 ordinates
and vertical ) + (𝑦 −of𝑦 any ) point in the plane,(4)
0 0
respectively. x and y represent the horizontal and vertical ordinates of any point on the
whereboundary,
x and y represent
respectively. theIthorizontal and vertical
is worth noting that when ordinates
the point ofisany point
on the in the plane,
boundary, the
respectively. 𝑥 and 𝑦 represent the horizontal and vertical ordinates
function value ψ is equal to zero. As shown in Figure 3, the value of ψ( x, y, t = 0) is of any point
moreon
the boundary, respectively.
than zero when the pointItisisinworth noting
the cross that
section andwhen the zero
less than point whenis onthe
the boundary,
point is outsidethe
function
the value ψ is equal to zero. As shown in Figure 3, the value of 𝜓(𝑥, 𝑦, 𝑡 = 0) is more
cross section.
If point
than zero when theP00point
( x00 , y00is) in
is the
the closest point toand
cross section pointlessP(than
x, y),zero
the closest
when point to any
the point point
is outside
on the 0 0
line P P is P . Then, the following formula can be derived:
the cross section.
 q  q 
0 2 0 2 0 2 0 2
∂ ( x − x ) + (y − y ) ∂ ( x − x ) + (y − y ) 
∇ψ( x, y, t = 0) =  ,  (5)
 ∂x ∂y 
Materials 2022, 15, x FOR PEER REVIEW
Materials 2022, 15, 6115 4 of 13

Figure 3. Initial value distribution of the level set function ψ.

If point 𝑃 (𝑥 , 𝑦 ) is the closest point to point 𝑃(𝑥, 𝑦), the closest point to any point
on the line 𝑃 𝑃 is 𝑃 . Then, the following formula can be derived:

𝜕 (𝑥 − 𝑥 ) + (𝑦 − 𝑦 ) 𝜕 (𝑥 − 𝑥 ) + (𝑦 − 𝑦 )
∇𝜓(𝑥, 𝑦, 𝑡 = 0) = , (5)
𝜕𝑥 𝜕𝑦
Figure 3. Initial value distribution of the level set function ψ.
Figure
If the3.initial
Initialvalue
value
of distribution
ψ is defined asofEquation
the level(4),
set itfunction
is easy toψ.get ‖∇𝜓(𝑥, 𝑦, 𝑡 = 0)‖ =
1 fromIf Equation
the initial (5). Therefore,
value Equation
of ψ is defined (3) can be
as Equation (4),simplified asget
it is easy to ||∇ψ( x, y, t = 0)|| = 1
follows:
from Equation (𝑥 , 𝑦 ) Equation
𝑃 Therefore,
If point (5). is the closest
(3) canpoint
𝜕𝜓 to point
be simplified 𝑃(𝑥, 𝑦), the closest point to an
as follows:
on the line 𝑃 𝑃 is 𝑃 . Then, the following + 𝐹 = 0 formula can be derived: (6)
𝜕𝑡
∂ψ
+F=0 (6)
where F is the distance from which 𝜕the boundary
(∂t𝑥 − 𝑥 )moves
+ (𝑦in−unit
𝑦 ) time.
𝜕 (𝑥 − 𝑥 ) + (𝑦 − 𝑦 )
∇𝜓(𝑥, 𝑦, 𝑡 = 0) =
where F is the distance from which the boundary𝜕𝑥
,
2.2. Solution Algorithms
moves in unit time. 𝜕𝑦
2.2. The mathematical
Solution Algorithmsfoundation of the contour-parallel path generation strategy was
introduced
If the in
initial valuesection.
the previous
of ψ is defined as Equation (4), it is easyinto get ‖∇𝜓(𝑥, 𝑦, 𝑡
The mathematical foundationThe boundary
of the is usually path
contour-parallel non-analytical actual con-
generation strategy was
1 from Equation
tour-parallel path (5). Therefore, Equation (3) can bealgorithm
simplified as follows:
introduced in thegeneration. Therefore,
previous section. Thethe non-analytical
boundary is usually non-analyticalof Equation (6)
in actual
needs to be derived. Generally, there are
contour-parallel path generation. Therefore, the 𝜕𝜓 two methods for the numerical solution of
non-analytical algorithm of Equation (6) partial
differential
needs to beEquation (6): the Finite
derived. Generally, thereDifference Method
are two methods +for
𝐹(FDM)
= 0numerical
the and thesolution
Finite Element
of partial
Method (FEM). 𝜕𝑡
differential Equation (6): the Finite Difference Method (FDM) and the Finite Element
whereAs shown
Method F(FEM). in Figure
is the distance 4, the solution
from which domain
the is different for
boundary FDM and
moves FEM.time.
in unit The solu-
tion domain of the FDM is the maximum envelope rectangle
As shown in Figure 4, the solution domain is different for FDM and FEM. The with boundaries in solution
which
the solution
domain domainFDM is the maximum envelope rectangle with boundaries in which of
is
of the Algorithms discretized into a square lattice of n × n. The solution domain the
2.2. Solution
FEM can be an irregular
solution domain is discretized boundary, intoin which the
a square solution
lattice of n ×domain is discretized
n. The solution domainintoofmul-
FEM
tiple beThe
can triangles. mathematical
an irregular solving foundation
Whenboundary, the
in same theofsolution
whichfilling the contour-parallel
problem, the solution
domain pathinto
domain
is discretized generation
of FDM is
multiple strate
larger than When
introduced
triangles. thatin ofsolving
FEM.
the From
previous thissection.
the same point
fillingof view,boundary
The
problem, FEM is more
the solution efficient
domain than
is usually FDM.
non-analytical
of FDM The in act
is larger
space
than complexity
that of FEM. of From
FDM is O(n2).
this pointThe space FEM
complexity
of Therefore,
view, isthe
more of efficient
FEM is O(n).than Therefore,
FDM. The from
spaceof Equa
tour-parallel path generation. non-analytical algorithm
the perspective
complexity of of
FDMsolving spaceThe
is O(n2). andspace
mesh refinement,
complexity the efficiency
of FEM is O(n).of FEM will befrom
Therefore, higher
the
needs
than that
to
of
be
FDM.
derived. Generally, there are two methods
perspective of solving space and mesh refinement, the efficiency of FEM will be higher
for the numerical solution o
differential
than that of FDM. Equation (6): the Finite Difference Method (FDM) and the Finite E
Method (FEM).
As shown in Figure 4, the solution domain is different for FDM and FEM. T
tion domain of the FDM is the maximum envelope rectangle with boundaries in
the solution domain is discretized into a square lattice of n × n. The solution do
FEM can be an irregular boundary, in which the solution domain is discretized in
tiple triangles. When solving the same filling problem, the solution domain of
larger than that of FEM. From this point of view, FEM is more efficient than FD
space complexity of FDM is O(n2). The space complexity of FEM is O(n). Therefo
the perspective of (a)solving space and mesh refinement,(b) the efficiency of FEM will b
than that of
Figure4.4.Solution
Figure FDM.
Solutiondomain
domainand
andinitial
initialvalue
valueofoflevel
levelset
setfunction
functionψ:ψ:(a)
(a)FDM;
FDM;(b)
(b)FEM.
FEM.

The solution domain of FDM may be significantly larger than that of FEM for some
special filled boundaries. In the case shown in Figure 4, the solution domain of FDM is
nearly twice as large as that of FEM. Next, we explore the iterative formula in the solution
 Materials 2022, 15, 6115 5 of 13

domain to solve Equation (6). The iterative formulas of FEM and FDM differ. In our
previous research [25], the iterative formula of FDM has been derived as follows:



ψ x i , y j , t k +1 − ψ x i , y j , t k
+F=0 (7)
∆t
where, tk+1 = tk + ∆t, xi+1 = xi + ∆x and yi+1 = yi + ∆y. ∆t, ∆x and ∆y are increments in
time, x and y, respectively. The smaller these values are, the higher the solution precision.
In addition, i = 1, 2, . . . ., M. j = 1, 2, . . . , N. k = 1, 2, . . . , w. The initial value of the level
set function ψ can be calculated using Formula (4). xi , y j and tk are abscissa, ordinate, and
time of grid point (i, j), respectively. Similarly, the iterative formula of FEM can be derived
as follows:
ψ ( x i , y i , t k +1 ) − ψ ( x i , y i , t k )
+F=0 (8)
∆t
where, xi , yi are abscissa and ordinate of node i (i = 1, 2, . . . ., N), respectively. N represents
the total number of nodes in the solution domain. As we can see from Formula (8), there is
no coupling relationship between nodes. Then, the iteration formula is carried out only in
the time dimension of each node. To simplify and improve the calculation speed, the linear
shape function is used in FEM.
Materials 2022, 15, x FOR PEER REVIEW2.3. Computation Procedures 6 of 14

The solution flow of the FDM is shown in Figure 5a. The solution process can be
divided into the following steps:

(a) (b)
Figure 5. Flow chart of the new contour parallel path generation algorithm using different methods:
Figure 5. Flow chart of the new contour parallel path generation algorithm using different methods:
(a) FDM; (b) FEM.
(a) FDM; (b) FEM.
Step 1: Input the ∆x, ∆y, ∆t, F, Fill spacing and boundaries of the cross-section.
Similarly, the solution
Step 2: Generate flowgrid
the FDM of FEM is shown
according to theinboundaries.
Figure 6b. The solution process can
be divided into the following steps:
Step 3: Calculate the initial value of the level set function ψ by using Equation (4).
Step
Step1:4:Input the time
Calculate increment
the level ∆t, the
set function moving
value speed
ψ at the nextF, theby
time element size, and(7).
using Equation the
boundaries of the cross section.
Step 2: Generate FEM mesh with or without boundary refinement.
Step 3: Calculate the initial value of the level set function ψ for each node by using
Equation (4).
Step 4: Calculate the level set function value ψ at the next time by using Equation (8).
 Step 1: Input the time increment ∆t, the moving speed F, the element size, and the
boundaries of the cross section.
Step 2: Generate FEM mesh with or without boundary refinement.
Step 3: Calculate the initial value of the level set function ψ for each node by using
Materials 2022, 15, 6115 Equation (4). 6 of 13
Step 4: Calculate the level set function value ψ at the next time by using Equation (8).
Step 5: Judge whether the actual filling spacing is equal to the required filling spac-
ing. If not, jump to step 4. If yes, continue to the next step.
Step 5: Judge whether the actual filling spacing is equal to the required filling spacing.
Step 6: Calculate 𝜓 𝑥 , 𝑦 , 𝑡 = 0 by using the slicing method [26].
If not, jump to step 4. If yes, continue
linesto(𝜓the𝑥 next step.
Step
Step7:6:Reassemble
Calculate ψthe xi ,filling
y j , ti = 0 by the ,contour
𝑦 , 𝑡 =calculation
0) and stores the contour-parallel
method.
path. Step 7: Reassemble the filling lines (ψ x , y , t
= 0) and store the contour-parallel path.
i j i
Step
Step8:8:Judge
Judgewhether
whetherthe theequidistant
equidistantfilling
fillingisiscompleted.
completed.IfIfyes,
yes,continue
continuetotothe
thenext
next
step.
step.IfIfnot,
not,jump
jumptotostep
step2 2and
andreplace
replacethetheboundaries
boundarieswith withthe
thecontour
contourparallel
parallelpath.
path.
Step
Step9:9:Output
Outputthetheequidistant fillingcontour
equidistant filling contour(contour-parallel
(contour-parallel path)
path) andand end
end thethe pro-
program.
gram. Similarly, the solution flow of FEM is shown in Figure 6b. The solution process can be
divided into the following steps:

Materials 2022, 15, x FOR PEER REVIEW 7 of 14

(a) (b) (c)

(d) (e) (f)

Contours
Figure6.6.Contours
Figure andand grids
grids with
with different
different offset
offset distances
distances generated
generated by using
by using FDM:FDM:
(a) 10(a)
mm;10 mm;
(b)
30(b) 30 mm;
mm; (c) 50(c)
mm;50 mm;
(d) 70(d) 70 mm;
mm; (e) 90(e) 90 mm;
mm; (f)mm.
(f) 110 110 mm.

Stepcontour
The 1: Inputrepresented by 𝜓 𝑥 , ∆t,
the time increment 𝑦 , 𝑡the=moving
0 can be speed F, the element
calculated size, and
with a mature the
STL
boundaries of the cross section.
model slicing algorithm. Using the value of 𝜓 𝑥 , 𝑦 , 𝑡 as the coordinate component in
Step 2: Generate FEM mesh with or without boundary refinement.
the height direction of the triangular mesh node in a plane and a new STL model, which
Step 3: Calculate the initial value of the level set function ψ for each node by using
can be sliced to get the zero-value contour, is obtained. Triangles close to the boundary
Equation (4).
can be refined to improve calculation accuracy with an equal number of triangles in the
Step 4: Calculate the level set function value ψ at the next time by using Equation (8).
FEM. In FDM, the size of the quadrilateral mesh cannot be refined locally.
Step 5: Judge whether the actual filling spacing is equal to the required filling spacing.
As shown in Figure 6, the contour-parallel path of a landing gear die has been gen-
If not, jump to step 4. If yes, continue to the next step.
erated by using FDM. The x-axes
and y-axes describe the sections to be filled, the color
Step 6: Calculate ψ xi , y j , ti = 0 by using the slicing method [26].
represents the value of the level set function, and
the solid red line indicates the filling
Step 7: Reassemble the filling lines (ψ xi , y j , ti = 0) and stores the contour-parallel path.
line after the offset. After generating the paths, they are used as a new cross-section
Step 8: Judge whether the equidistant filling is completed. If yes, continue to the next
boundary to generate a new FDM grid. With the increase in offsetting times, the grid area
step. If not, jump to step 2 and replace the boundaries with the contour parallel path.
remains the same. Therefore, the computation time is not changed.
Step 9: Output the equidistant filling contour
(contour-parallel path) and end the program.
As
The shown in Figure
contour 7, the contour-parallel
represented by ψ xi , y j , ti =path
0 canof abelanding gear die has been gen-
erated by using FEM. After generating the paths, they
are used as a with
calculated
new
a mature STL
cross-section
model slicing algorithm. Using the value of ψ xi , y j , ti as the coordinate component in the
boundary to generate
height direction of thea triangular
new FEM mesh
mesh.nodeWithinthe increase
a plane andina offsetting times, the
new STL model, whichmesh
can
area
be sliced to get the zero-value contour, is obtained. Triangles close to the boundary canisbe
becomes smaller and smaller until it disappears. Therefore, the computation time
shorter
refinedand shorter. calculation accuracy with an equal number of triangles in the FEM. In
to improve
FDM, the size of the quadrilateral mesh cannot be refined locally.
As shown in Figure 6, the contour-parallel path of a landing gear die has been gen-
erated by using FDM. The x-axes and y-axes describe the sections to be filled, the color
represents the value of the level set function, and the solid red line indicates the filling line
after the offset. After generating the paths, they are used as a new cross-section boundary
to generate a new FDM grid. With the increase in offsetting times, the grid area remains
the same. Therefore, the computation time is not changed.
As shown in Figure 7, the contour-parallel path of a landing gear die has been gen-
erated by using (a) FEM. After generating the (b)paths, they are used as a new (c) cross-section
boundary to generate a new FEM mesh. With the increase in offsetting times, the mesh area
 line after the offset. After generating the paths, they are used as a new cross-section
boundary to generate a new FDM grid. With the increase in offsetting times, the grid area
remains the same. Therefore, the computation time is not changed.
As shown in Figure 7, the contour-parallel path of a landing gear die has been gen-
Materials 2022, 15, 6115
erated by using FEM. After generating the paths, they are used as a new cross-section 7 of 13
boundary to generate a new FEM mesh. With the increase in offsetting times, the mesh
area becomes smaller and smaller until it disappears. Therefore, the computation time is
shorter and shorter.
becomes smaller and smaller until it disappears. Therefore, the computation time is shorter
and shorter.

(a) (b) (c)

(d) (e) (f)

Contours
Figure7.7.Contours
Figure and
and mesh
mesh with
with different
different offset
offset distances
distances generated
generated by using
by using FEM:FEM:
(a) 10(a)
mm;10 mm;
(b)
30(b) 30 mm;
mm; (c) 50(c)
mm;50 mm;
(d) 70(d) 70 mm;
mm; (e) 90(e) 90 mm;
mm; (f)mm.
(f) 110 110 mm.
Materials 2022, 15, x FOR PEER REVIEW 8 of 14
Thecomparison
The comparisonofofthe thepath
pathgeneration
generationprocesses
processesofofFEM FEMand andFDM FDM(Figures
(Figures6 6and and7)7)
showsthat
shows thatFEMFEMcan cansignificantly
significantlyreduce
reducethe thecalculation
calculationtime.time.There
Thereare aretwo
tworeasons
reasonswhy why
theFEM
the FEMalgorithm
algorithm is is faster
faster than
than the
theFDMFDMalgorithm.
algorithm.First,
First,the
thesolution
solution domain
domain of of
FEMFEMcan
FEMbe can be refined
greatly reduced atwith
the boundary,
an increase which
in the can ensure
number of that fewer
offsets (only elements
in the are used
filled area), with
while
can be greatly reduced with an increase in the number of offsets (only in the filled area),
thethe
same accuracy compared with not
FDM, in whichwiththe element cannot be refined at the(in
whilesolution domain
the solution of FDM
domain of does
FDM does decrease
not decrease anwith
increase in the number
an increase in the of offsets
number of
boundary.
the largest rectangular area of the entire cross-section). Second, the mesh of the FEM can
offsets (in the largest rectangular area of the entire cross-section). Second, the mesh of the
be In addition,
refined at thepolygon
boundary, splitting,
whichdisappearing
can ensure that and merging
fewer withare
elements boundary
used with migration
the same
can evolve compared
accuracy automatically withwithout
FDM, inspecial
which the processing
element by using
cannot be these
refined two solving
at the algo-
boundary.
rithms.In Asaddition,
the example shown in Figures 6 and 7, connected domains
polygon splitting, disappearing and merging with boundary migration evolve from three
to can
one,evolve
then to five, then towithout
automatically six andspecial
finally processing
disappear.by It is verythese
using complicated
two solving to deal with
algorithms.
this
Asevolution
the example withshown
conventional
in Figuresgeometric
6 and 7, offsetting
connected algorithms.
domains As we can
evolve fromseethree
fromto Fig-
one,
ure 8, there
then are subtle
to five, then todifferences between
six and finally the equidistant
disappear. It is veryoffset filling contours
complicated to deal generated
with this
byevolution
FDM andwith FEM. conventional
The contours geometric
obtainedoffsetting
by thesealgorithms.
two algorithms As we cancan see the
meet from Figure 8,
require-
thereofare
ments thesubtle
WAAM differences
process.between
FEM canthe equidistant
reduce the solvingoffsetdomain
filling contours
and thus generated
has higherby
FDM andHowever,
efficiency. FEM. The the contours
FDM obtained by these
grid is simple andtwoeasy
algorithms can meet
to generate. ThetheFEM requirements
mesh is
of the WAAM process. FEM can reduce the solving domain
slightly more complex and mature mesh generation algorithms can be used. Although and thus has higher efficiency.
However,
many types ofthe FDMare
paths grid
usedis simple
in WAAM, and such
easy asto raster,
generate. Thespiral,
zigzag, FEM mesh fractalisspace-filling
slightly more
complex
curves, andand mature
hybrid meshplanning
toolpath generation algorithmsthe
approaches, cancontour-parallel
be used. Although pathmanyis thetypes
most of
paths are used
appropriate in WAAM,
for this process.such as raster, zigzag,
Simultaneously, spiral, fractal
the continuous space-filling
smooth curves, and
contour-parallel
hybrid
path toolpath
can reduce planning
drastic changesapproaches, the contour-parallel
in the velocity and acceleration path is the
of the most appropriate
robot.
for this process. Simultaneously, the continuous smooth contour-parallel path can reduce
drastic changes in the velocity and acceleration of the robot.

(a) (b)
Figure
Figure 8. The
8. The equidistant
equidistant offset
offset filling
filling contours
contours byby using
using different
different algorithms:
algorithms: (a)(a) FDM;
FDM; (b)(b) FEM.
FEM.

3. Application Test
3.1. Gradient Structure Design
This section introduces a practical application example of the proposed algorithms
 (a) (b)
Materials 2022, 15, 6115 8 of 13
Figure 8. The equidistant offset filling contours by using different algorithms: (a) FDM; (b) FEM.

3. Application Test
3. Application Test
3.1. Gradient Structure Design
3.1. Gradient Structure Design
This section introduces a practical application example of the proposed algorithms
This section introduces a practical application example of the proposed algorithms
in the repair process of a failed landing gear die using WAAM technology. The repair
in the repair process of a failed landing gear die using WAAM technology. The repair
process includes modeling, slicing, single-layer contour-parallel path generation, robot
process includes modeling, slicing, single-layer contour-parallel path generation, robot
code generation, simulating and WAAM. The failed die is repaired accurately with differ-
code generation, simulating and WAAM. The failed die is repaired accurately with different
entmaterials
materials for different areas, thus realizing the equal-life design concept.
for different areas, thus realizing the equal-life design concept.
AsAs
shown
shown in in
Figure
Figure9, 9,
different areas
different of of
areas thethe
landing
landinggear
geardiedie
have
havedifferent tempera-
different tempera-
tures and
tures stress
and stressdistributions
distributions during
during service.
service.Therefore, it is
Therefore, necessary
it is necessaryto to
design
design different
different
materials for
materials fordifferent
differentareas
areasbefore
beforethe
therepair
repair process, which is
process, which is also
alsocalled
calledgradient
gradientstructure
struc-
ture design (GSD). Due to the high temperature and high stress of the surface
design (GSD). Due to the high temperature and high stress of the surface layer, Co-based layer, Co-
based materials are chosen. The transition layer can be deposited with Fe-based
materials are chosen. The transition layer can be deposited with Fe-based materials. The materials.
The material
material of of
thethe
diedie base
base is 5GrNiMo.
is 5GrNiMo.

Materials
x FOR PEER REVIEW2022, 15, x FOR PEER REVIEW 9 of 14 9 of 14

(a) (b)

(c) (c)
Figure
Figure
Figure 9. Stress and temperature 9. Stress
9.
field and
Stress
and temperature
and temperature
gradient field and
andgradient
fielddesign:
structure gradient structure
field; design:
structure
(a) Stress design:
(b) (a)
(a)Stress
Tempera- Stress field;
field; (b) Tempera-
(b) Temperature
ture field;
ture field; (c) Gradient structure. (c) Gradient structure.
field; (c) Gradient structure.

3.2.3.2.
3.2. WAAM Process Parameters WAAM
WAAM Process Parameters
Process Parameters
TheThe quality
qualityof
The quality of a single weld pass is onea ofa
ofsingle weld
single
the weld
core pass is one
pass
factors of of
thethe
isaffecting
one core factors
thecore
entire affecting
factors affecting
quality thethe
entire quality
entire quality
of the
of the welding plane. Aofsimple welding
the welding plane.
welding plane.A simple
test isAshown welding
simpleinwelding test
Figure 10, is shown
testinis which in
shownthe Figure
in Figure 10,
welding in which the welding
10, in which the welding
voltage is 28 V and thevoltage
voltage
welding is 28 V and
iswire
28 thethe
Vdiameter
and welding
welding
is wire
1.6 mm. diameter
wire
As shown is
diameter in1.6
is mm.
1.6
Figuremm. AsAs
11, shownthein in
shown
with Figure 11,11,
Figure with thethe
with
wire
wire feeding speed (WFS) feeding
wiregetting
feeding speed
speed
larger (WFS)
or(WFS) getting larger
gettingspeed
the walking or the
larger(WS)
or thewalking
walking
getting speed (WS)
speedthe
smaller, getting smaller,
(WS) getting smaller, thethe
weld
weld pass becomes wider. weld pass
If passbecomes
higherbecomes wider.
welding wider.If higher
If higher
accuracy welding
welding
is required, accuracy
the widthisof
accuracy required,
is the weldthethe
required, width
width of of
thethe
weld
weld
pass
passshould
should be smaller,
be smaller,which
which will reduce
will reduce the welding
the welding
pass should be smaller, which will reduce the welding efficiency. In this paper, the size of efficiency.
efficiency. In this
In paper,
this the
paper, size
the of of
size
the aviation
the aviation die to
die be
to repaired
be repaired is large
is and
large to
and improve
to
the aviation die to be repaired is large and to improve the WAAM efficiency, a larger WFSimprove the WAAM
the WAAM efficiency, a
efficiency, larger
a WFS
larger WFS
is selected. is selected.
is selected.

(a) (a) (b) (b) (c) (c)

(d) (d) (e) (e) (f) (f)

(g) (g) (h) (h) (i) (i)

Figure
Figure 10. Weld passes Figure
obtained 10.
10.with
WeldWeld passes
passes
different obtained
obtained
process withwith different
different
parameters: process
(a) process
WS: 200 parameters:
parameters:
mm/min (a)200
(a) WS:
WFS: 4000 WS: 200 mm/min
mm/min WFS:
WFS: 4000
mm/min;
4000 (b)
mm/min;WS: 450
(b) mm/min
WS: 450 WFS:
mm/min 4000
WFS:mm/min;
4000 (c)
mm/min;
mm/min; (b) WS: 450 mm/min WFS: 4000 mm/min; (c) WS: 700 mm/min WFS: 4000 mm/min; (d)WS: 700
(c) mm/min
WS: 700 WFS:
mm/min 4000
WFS: mm/min;
4000 (d)
mm/min;
WS:
WS: 200 mm/min WFS: 6000 (d)200 mm/min
WS:
mm/min; 200(e) WFS:
mm/min 6000
WS: 450 mm/min;
WFS:
mm/min6000 (e) 6000
WS: 450
mm/min;
WFS: mm/min
(e) WFS:
WS:(f)450
mm/min; WS: 6000
mm/min
700 mm/min; (f) WS:
WFS: 6000
mm/min 700 mm/min
mm/min; (f) WS:
WFS: 6000 mm/min; (g)WFS:WS: 6000
200 mm/min;
mm/min (g)
WFS: WS: 200
8000 mm/min
mm/min; WFS:
(h) WS: 8000
450 mm/min;
mm/min (h)
WFS:WS: 450
8000 mm/min
700 mm/min WFS: 6000 mm/min; (g) WS: 200 mm/min WFS: 8000 mm/min; (h) WS: 450 mm/min WFS: 8000
mm/min;
mm/min; (i) WS: 700 mm/min (i) WS: 700 mm/min WFS: 8000 mm/min.
WFS:WFS:
8000 8000 mm/min.
mm/min; (i) WS: 700 mm/min WFS: 8000 mm/min.
 (d) (e) (f)

(g) (h) (i)

Materials 2022, 15, 6115
Figure 10. Weld passes obtained with different process parameters: (a) WS: 200 mm/min WFS: 4000
mm/min; (b) WS: 450 mm/min WFS: 4000 mm/min; (c) WS: 700 mm/min WFS: 4000 mm/min; (d)
WS: 200 mm/min WFS: 6000 mm/min; (e) WS: 450 mm/min WFS: 6000 mm/min; (f) WS: 700 mm/min 9 of 13
WFS: 6000 mm/min; (g) WS: 200 mm/min WFS: 8000 mm/min; (h) WS: 450 mm/min WFS: 8000
mm/min; (i) WS: 700 mm/min WFS: 8000 mm/min.

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

Figure
Figure 11. Morphology
11. Morphology of weld
of weld passes passes
obtained withobtained with different
different process process
parameters: (a) WS:parameters:
200 (a) WS:
mm/min
200 WFS: 4000 mm/min;
mm/min (b) mm/min;
WFS: 4000 WS: 450 mm/min WFS:
(b) WS: 4504000 mm/min;
mm/min (c) WS:
WFS: 700mm/min;
4000 mm/min WFS:
(c) WS: 700 mm/min
4000 mm/min; (d) WS: 200 mm/min WFS: 6000 mm/min; (e) WS: 450 mm/min WFS: 6000 mm/min;
Materials 2022, 15, x FOR PEER REVIEW 10 of 14
WFS: 4000 mm/min; (d) WS: 200 mm/min WFS: 6000 mm/min; (e) WS: 450 mm/min WFS:
6000 mm/min; (f) WS: 700 mm/min WFS: 6000 mm/min; (g) WS: 200 mm/min WFS: 8000 mm/min;
(h) WS: 450 mm/min WFS: 8000 mm/min; (i) WS 700 mm/min WFS: 8000 mm/min.
(f) WS: 700 mm/min WFS: 6000 mm/min; (g) WS: 200 mm/min WFS: 8000 mm/min; (h) WS: 450
mm/min WFS: 8000 mm/min; (i) WS 700 mm/min WFS: 8000 mm/min.
As shown in Figure 12a–c, when the wire feeding speed and the welding voltage
are As
constant,
shown inwith the12a–c,
Figure increase
whenof thewelding speed,
wire feeding theand
speed linear energyvoltage
the welding densityaredecreases and
the cooling rate of the molten pool increases, forming fine columnar
constant, with the increase of welding speed, the linear energy density decreases and the crystals, and the
dendrite
cooling ratespacing decreases,
of the molten which forming
pool increases, leads tofine
thecolumnar
reduction of the
crystals, andcolumnar crystal size and
the dendrite
spacing decreases,
the dendrite which leads
spacing. to the reduction
Conversely, as shownof the
incolumnar crystalwhen
Figure 12b,d, size and
thethewelding
den- speed and
drite spacing. Conversely,
the welding voltage areasconstant,
shown inwith
Figure
the12b,d, wheninthe
increase wirewelding speed
feeding and the
speed, coarse columnar
welding voltage are constant, with the increase in wire feeding speed, coarse columnar
crystals are formed and the dendrite spacing increases. Then, from the perspective of
crystals are formed and the dendrite spacing increases. Then, from the perspective of mi-
microstructure, when the wire feeding speed and voltage are fixed, the larger the welding
crostructure, when the wire feeding speed and voltage are fixed, the larger the welding
speedinin
speed a certain
a certain range,
range, the better
the better the microstructure.
the microstructure.

(a) (b)

(c) (d)
Figure12.
Figure Microstructure
12.Microstructure of of
thethe weld
weld pass
pass with
with different
different linear
linear energy
energy densities:
densities: (a)200
(a) WS: WS: 200 mm/min
WFS: 8000
mm/min WFS:mm/min;
8000 mm/min;(b)(b)WS:
WS: 450 mm/min
450 mm/min WFS:WFS: 8000 mm/min;
8000 mm/min; (c) WS: 700(c) WS: 700
mm/min WFS:mm/min WFS:
8000
8000mm/min;
mm/min;(d) WS: 450 mm/min
(d) WS: WFS: 9000
450 mm/min mm/min.
WFS: 9000 mm/min.
To guarantee WAAM efficiency and error control in the height direction, weld height
and walking speed should be chosen moderately. In this paper, the wire feeding speed is
set to 8000 mm/min and the walking speed is set to 450 mm/min (Figures 10h and 11h).
According to corresponding research [22,27], the weld spacing is set to 1/2 of the weld
width.
Materials 2022, 15, 6115 10 of 13

To guarantee WAAM efficiency and error control in the height direction, weld height and
walking speed should be chosen moderately. In this paper, the wire feeding speed is set to
8000 mm/min and the walking speed is set to 450 mm/min (Figures 10h and 11h). According
to corresponding research [22,27], the weld spacing is set to 1/2 of the weld width.

3.3. Repair Process of the Landing Gear Die

The steps to get the target WAAM model are scanning to obtain the points cloud, en-
capsulating the points cloud into a triangular mesh, converting the mesh into a parametric
Materials 2022, 15, x FOR PEER REVIEW 11 of 14
surface, making a difference with the standard model to obtain the temporary model, and
cutting the temporary model according to the simulation results to get the target WAAM
model. Similar steps have been explained in other research [22]. Then, the target model
target
is slicedmodel
and the is sliced
WAAM and theisWAAM
path generated path
by is
thegenerated
algorithms byproposed
the algorithms
above. proposed
A six-axis
above.
gantry A six-axis
robot gantry
has been robot hasfor
developed been developed
repairing largeforhotrepairing largeinhot
forging dies, forging
which dies, in
a three-axis
which a three-axis has been used to weld and the other three-axis has
has been used to weld and the other three-axis has been used to hammer. After the welding been used to ham-
mer.
of oneAfter
layer theiswelding of one
completed, thislayer
layerisiscompleted,
hammeredthis layer is hammered
to strengthen to strengthen
the welding the
material and
welding
eliminatematerial
residualand eliminate residual stress.
stress.
The
The repair
repair process of the the landing
landinggear geardie
dieisisshown
shownininFigure
Figure1313and
andCo-based
Co-based weld-
welding
ing
wirewire is used
is used for the
for the strengthening
strengthening layer.layer.
The The
WAAM WAAM parameters
parameters arefollows:
are as as follows: volt-is
voltage
28 V,iswire
age 28 V,feeding speed is
wire feeding 8000 is
speed mm/min and welding
8000 mm/min speed isspeed
and welding 450 mm/min. As shown
is 450 mm/min. As
in Figure 13b, the welding process can achieve near no-splash, which
shown in Figure 13b, the welding process can achieve near no-splash, which enhances the enhances the fusion
effect between
fusion welding
effect between materials
welding and theand
materials diethe
base.
die base.

(a) (b)

(c) (d)
Figure13.
Figure Repairprocess
13.Repair processof
ofthe
theaircraft
aircraftlanding
landinggear
geardie
dieusing
usingWAAM:
WAAM:(a)
(a)Remove
Removethe
thefailure
failurearea;
area;
(b) During WAAM; (c) Die after CNC machining; (d) Service
(b) During WAAM; (c) Die after CNC machining; (d) Service site. site.

As shown
As shown inin Figure
Figure 13,
13, by
by using
using the
the contour-parallel
contour-parallel welding
weldingpath,
path,higher
higherWAAM
WAAM
accuracy, less machining allowance and materials consumption are realized. In addition,
accuracy, less machining allowance and materials consumption are realized. In addition,
as shown in Figure 13c, the machined surface is smooth without welding defects such as
as shown in Figure 13c, the machined surface is smooth without welding defects such as
pores, slag inclusions and cracks, which can meet the needs of die repair.
pores, slag inclusions and cracks, which can meet the needs of die repair.

3.4. Service Life Analysis

The high-temperature performance of the die was significantly improved after using
Co-based material on the strengthening layer. For traditional dies, the flash land has ob-
Materials 2022, 15, x FOR PEER REVIEW

Materials 2022, 15, 6115 11 of 13

3.4. Service Life Analysis

The high-temperature performance of the die was significantly improved after using
Co-based material on the strengthening layer. For traditional dies, the flash land has
obvious collapse and wear defects after twice forging (Figure 14a). On the contrary, for
the repaired die with a gradient structure, the die surface is bright and no collapse and
wear defects were found after 20 times of forging (Figure 14b). The service life of the
Materials 2022, 15, x FOR PEER REVIEW 12 of 14
aircraft landing gear die has been significantly improved by using WAAM technology with
a gradient structure.

(a) (b)
Figure 14. Appearance of traditional homogeneous die and repaired die with gr
Appearance of traditional homogeneous die after forging twice; (b) Appearanc
ture die after forging 20 times.

As shown in Figure 15, the fusion effect between the transition laye
(a) (b)
strengthening
Appearanceofof
layer is good,
traditional
and
homogeneous
the microstructure presents a certain tra
Figure14.
Figure 14.Appearance traditional homogeneous diedie
andand repaired
repaired die die
withwith gradient
gradient structure:
structure: (a)
The
(a) grain
Appearance
Appearance size
of gradually
traditional
of traditional homogeneous
homogeneous increases
die after
die after from
forgingforging the
twice; (b)die
twice; (b)base to
Appearance
Appearance theoffusion
of gradient struc- line. T
gradient
structure
ture
thedieheatdie forging
after after
input forging
20 of 20
times. times.
the welding influences the grain size of the previous l
theAs weld
As shown
shown dendrites
ininFigure
Figure15, 15,(Figure fusion12)
thefusion
the near
effect
effect between
between thethe fusion
the line
transition
transition areand
layer
layer transformed
and welding into
thewelding
the
strengthening
(Figure 16)
strengthening layer
layerafter is good,
heat
is good, and
and the microstructure
treatment. presents
Thispresents
the microstructure fusiona line a certain
certain transition
indicates structure.
that the mech
transition structure.
Thegrain
The grainsize
sizegradually
graduallyincreasesincreasesfrom from the
the die
die base
base toto the
the fusion
fusion line.
line. This
Thisindicates
indicatesthatthat
between
the heat input theof transition
the welding layer and
influences the the strengthening
grain size of the previous layer
layer. are
In well combi
addition,
the heat input of the welding influences the grain size of the previous layer. In addition,
distribution
the
the weld dendritesnear
welddendrites (Figure
(Figure the12)fusion
12) nearthe
near line
fusionis
thefusion shown
line
line in Figure
aretransformed
are transformed 16.
intoequiaxed
into Duecrystals
equiaxed to the inco
crystals
(Figure16)16) after
after heat
heat treatment.
treatment. This Thisfusion
fusionline
lineindicates
indicatesthat thatthe
the mechanical
mechanicalproperties
properties
tion, the fusion effect between welding material and base metal needs to
(Figure
betweenthe
between thetransition
transitionlayer layerand andthethestrengthening
strengtheninglayer layerare are well
well combined.
combined. The Theelement
element
the welding
distribution near theprocess.
fusion line Cobalt
is shown and
in iron
Figure 16. permeate
Due to
distribution near the fusion line is shown in Figure 16. Due to the inconsistent composi- the each
inconsistent other around the
composition,
shows
the
tion, fusion that
effect
the fusion the
between
effect base
between metal
welding
welding and
material welding
andand
material basebase material
metal needs
metal needstocan
be beconsidered
fully inintegrated
to considered
be the
in
welding process. Cobalt and iron permeate each other around the fusion line. This shows
process.
the welding process. Cobalt and iron permeate each other around
that the base metal and welding material can be fully integrated into the welding process.
the fusion line. This
shows that the base metal and welding material can be fully integrated into the welding
process.

Figure 15. Microstructure of fusion zone between transition layer and strengthening layer.
Figure 15. Microstructure of fusion zone between transition layer and strengthening layer.
Figure 15. Microstructure of fusion zone between transition layer and strengthe

(a) (b)

(a) (b)
(c) (d)
Materials 2022, 15, 6115 12 of 13

Figure 15. Microstructure of fusion zone between transition layer and strengthening layer.

(a) (b)

(c) (d)
Figure16.
Figure Element
16.Element distribution
distribution nearthe
near thefusion
fusionline:
line:(a)(a) Fe;
Fe; (b)
(b) Co;
Co; (c)(c)Mn;
Mn;(d)
(d)C.C.

4.4.Conclusions
Conclusions
(1) InIn
(1) this
this paper,
paper, thethe mathematical
mathematical basisbasis of a equidistant
of a new new equidistant contour-parallel
contour-parallel path
path gen-
eration strategy has been developed, and two types of iterative formulas based onon
generation strategy has been developed, and two types of iterative formulas based
FEM and FDM are proposed to solve the partial differential equation. By comparison,
FEM has a faster solving speed than FDM under the same solving accuracy because
the solving domain of FEM mesh can be greatly reduced, and the boundary mesh
can be refined. However, the FDM grid is easier to generate and program than the
FEM mesh. The two algorithms have their own advantages and disadvantages, and
the selection of which algorithm can be determined according to the shape of the
cross-section.
(2) The die was divided into three layers: base layer, transition layer (Fe-based material)
and strengthening layer (Co-based material) according to the temperature and stress
field distribution of the aircraft landing gear die in service. The corresponding WAAM
process parameters have been discussed. The optimum welding parameters are
obtained as follows: voltage is 28 V, wire feeding speed is 8000 mm/min and welding
speed is 450 mm/min. Finally, the path generation strategy was practically applied to
the remanufacture of the large aircraft landing gear die with a three-layer structure.
(3) The application test of the aircraft landing gear die justifies the effectiveness of the
algorithms and technology proposed in this paper. The metallographic photos show
that a good combination of die base and welding materials can be obtained at the
micro level. In addition, using this new contour-parallel path generation strategy
in the WAAM process can accurately control machining allowances and improve
additive efficiency. The service life of the aircraft landing gear die can be significantly
improved by using WAAM technology with a gradient structure.

Author Contributions: J.Z. (Jiansheng Zhang) and G.X. put forward the relevant algorithms and
ideas and wrote the article. J.P. wrote the software. Y.Y. carried out experiments. J.Z. (Jie Zhou)
guided and reviewed the whole process. All authors have read and agreed to the published version
of the manuscript.
Funding: This research is supported by the National Natural Science Foundation of China (No.52075058),
the Chongqing Special Project for Technological Innovation and Application Development (No.cstc2022ycjh-
bgzxm0224) and the Chongqing Special Fund for Post-doctoral Program (No.2021XMT006).
Institutional Review Board Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: Thank you to all the authors for their contributions to this article.
Conflicts of Interest: The authors declare that they have no known competing financial interests or
personal relationships that could have appeared to influence the work reported in this paper.
Materials 2022, 15, 6115 13 of 13

References
1. Ding, D.; Pan, Z.S.; Cuiuri, D.; Li, H. A tool-path generation strategy for wire and arc additive manufacturing. Int. J. Adv. Manuf.
Technol. 2014, 73, 173–183. [CrossRef]
2. Yaseer, A.; Chen, H. A Review of Path Planning for Wire Arc Additive Manufacturing (WAAM). J. Adv. Manuf. Syst. 2021, 20, 589–609.
[CrossRef]
3. Farouki, R.T.; Koenig, T.; Tarabanis, K.A.; Korein, J.U.; Batchelder, J.S. Path planning with offset curves for layered fabrication
processes. J. Manuf. Syst. 1995, 14, 355–368. [CrossRef]
4. Yang, Y.; Loh, H.T.; Fuh, J.Y.H.; Wang, Y.G. Equidistant path generation for improving scanning efficiency in layered manufactur-
ing. Rapid Prototyp. J. 2002, 8, 30–37. [CrossRef]
5. Li, H.; Dong, Z.; Vickers, G.W. Optimal toolpath pattern identification for single island, sculptured part rough machining using
fuzzy pattern analysis. Comput. Aided Des. 1994, 26, 787–795. [CrossRef]
6. Dunlavey, M.R. Efficient polygon-filling algorithms for raster displays. ACM Trans. Graph. 1983, 2, 264–273. [CrossRef]
7. Park, S.C.; Choi, B.K. Tool-path planning for direction-parallel area milling. Comput. Aided Des. 2000, 32, 17–25. [CrossRef]
8. Wang, H.; Jang, P.; Stori, J.A. A Metric-Based Approach to Two-Dimensional (2D) Tool-Path Optimization for High-Speed
Machining. J. Manuf. Sci. Eng. 2005, 127, 33–48. [CrossRef]
9. Kulkarni, P.; Marsan, A.; Dutta, D. A review of process planning techniques in layered manufacturing. Rapid Prototyp. J. 2000, 6, 18–35.
[CrossRef]
10. Bertoldi, M.; Yardimci, M.A.; Pistor, C.M.; Guceri, S.I. Domain decomposition and space filling curves in toolpath planning and
generation. In Proceedings of the 1998 International Solid Freeform Fabrication Symposium, Austin, TX, USA, 10–12 August 1998.
11. Dwivedi, R.; Kovacevic, R. Automated torch path planning using polygon subdivision for solid freeform fabrication based on
welding. J. Manuf. Syst. 2004, 23, 278–291. [CrossRef]
12. Shen, Y.; Wei, Y.; Liu, R. A Path Generation Method for Wire and Arc Additive Remanufacturing of Complex Hot-forging Dies.
Int. J. Adv. Manuf. Technol. 2021, 117, 1935–1943. [CrossRef]
13. Dhanik, S.; Xirouchakis, P. Contour parallel milling tool path generation for arbitrary pocket shape using a fast marching method.
Int. J. Adv. Manuf. Technol. 2010, 50, 1101–1111. [CrossRef]
14. Klass, R. An offset spline approximation for plane cubic splines. Comput. Aided Des. 1983, 15, 297–299. [CrossRef]
15. Tiller, W.; Hanson, E.G. Offsets of Two-Dimensional Profiles. IEEE Comput. Graph. 1984, 4, 36–46. [CrossRef]
16. Hoschek, J.; Wissel, N. Optimal approximate conversion of spline curves and spline approximation of offset curves. Comput.
Aided Des. 1988, 20, 475–483. [CrossRef]
17. Veeramani, D.; Gau, Y.S. Cutter-path generation using multiple cutting-tool sizes for 2-1/2D pocket machining. IIE Trans. 2000,
32, 661–675. [CrossRef]
18. Stori, J.A.; Wright, P.K. Constant engagement tool path generation for convex geometries. J. Manuf. Syst. 2013, 19, 172–184.
[CrossRef]
19. Ferreira, R.P.; Scotti, A. The Concept of a Novel Path Planning Strategy for Wire + Arc Additive Manufacturing of Bulky Parts:
Pixel. Metals 2021, 11, 498. [CrossRef]
20. Chuang, J.J.; Yang, D. A laplace-based spiral contouring method for general pocket machining. Int. J. Adv. Manuf. Technol. 2007,
34, 714–723. [CrossRef]
21. Lu, S.; Zhou, J.; Zhang, J. Optimization of welding thickness on casting-steel surface for production of forging die. Int. J. Adv.
Manuf. Technol. 2015, 76, 1411–1419. [CrossRef]
22. Hong, X.; Xiao, G.; Zhang, Y.; Zhou, J. Research on gradient additive remanufacturing of ultra-large hot forging die based on
automatic wire arc additive manufacturing technology. Int. J. Adv. Manuf. Technol. 2021, 116, 2243–2254. [CrossRef]
23. Zhang, J.; Wang, Q.; Xiao, G.; Zhou, J. Filling Path Planning and Polygon Operations for Wire Arc Additive Manufacturing
Process. Math. Probl. Eng. 2021, 2021, 6683319. [CrossRef]
24. Crandall, M.G.; Lions, P.L. Viscosity solutions of Hamilton-Jacobi equations. Trans. Am. Math. Soc. 1983, 277, 1–42. [CrossRef]
25. Hong, X.Y.; Xiao, G.Q.; Zhang, Y.C.; Zhou, J. A New Path Planning Strategy Based on Level Set Function for Layered Fabrication
Processes. Int. J. Adv. Manuf. Technol. 2021, 119, 517–529. [CrossRef]
26. Zhang, J.S.; Zhou, J.; Wang, Q.Y.; Xiao, G.Q.; Quan, G.Z. Process planning of automatic wire arc additive remanufacturing for hot
forging die. Int. J. Adv. Manuf. Technol. 2020, 109, 1613–1623. [CrossRef]
27. Lam, N.; Johannes, B.; Markus, B. Multi-bead Overlapping Models for Tool Path Generation in Wire-Arc Additive Manufacturing
Processes. Proc. Manuf. 2020, 47, 1123–1128.