Injection Mold Design Technology to Locate Weld Lines Away from Highly Loaded Structural Areas

Abstract

This article presents the technology of automated placement of an injection molding gate based on a parametric optimization algorithm with technological constraints consideration. The algorithm is based on the modification of the genetic algorithm using the criterion of maximum equivalent stresses on the weld line as an optimization criterion. The proposed software’s modular structure combines the authors’ modules that implement a new optimization algorithm with the ANSYS 2022R1 and Moldflow calculation kernels called via API interfaces. This structure provides an opportunity to implement developed technology to solve industrial problems using standard mesh generation tools and complex geometric models due to the flexibility of modules and computing kernel scalability. The consideration of the technological constraints allows us to reduce the population size and optimization problem solution computational time to 1.9 times. The developed algorithms are used to solve the gate location optimization problem using the example of an aerospace bracket made of short-reinforced composite material with a nonzero genus surface and a weld line. The use of the proposed technology made it possible to increase the strength of the studied structure by two times.

1. Introduction

Injection molding technology is one of the most common technologies for producing polymer products, accounting for more than 30% of the total volume of plastic products produced [1]. The main advantages of the technology, such as high productivity, the possibility of mass production and the creation of complex configurations, and precise transfer of shape geometry, make it possible to obtain products with high dimensional accuracy and surface quality, making it the optimal choice for the mass production of similar polymer products [2,3]. Speeding up the time to market requires reducing the mold design and manufacturing time [4], including using modern CAE systems [5,6].

A weld line is a defect that inevitably occurs during injection molding when producing polymer parts with a nonzero genus surface [7,8]. A weld line is a thin line formed when a mold is filled at the junction of several melt streams. Weld line problems occur in many industrial applications, including aerospace [9], microelectronics, micro-injection molding [10], and multifunctional structures in embedded electronic components [11], automotive industries, telecommunication systems, and medical engineering [12]. Many factors affect the weld line and material mechanical properties, including the process mode [13], gate location [14], and process conditions of the part [15,16,17]. The presence of a weld line significantly reduces the strength of the product [18,19,20]. This property is particularly evident in manufacturing parts from a material containing a fibrous filler [21]. Many researchers have used the weld line factor to assess the strength of a weld line, which is “the strength of specimens with a weld line/the strength of specimens without a weld line” [22,23]. The following technological parameters of plastic parts’ molding characterize the highest WL factor: high holding pressure, high melt temperature, and low mold temperature [24]. External changes in the product can also be observed at the weld line location: loss of gloss, change in color, and appearance of grooves or scratches. External defects can be eliminated, for example, by induction heating of the part surface [25]. The prediction of the weld line location has become possible due to the development of numerical methods and the emergence of software packages, such as Moldflow and Moldex 3D [9,26,27]. Modeling polymer product molding allows the selection of the optimal technological parameters of molding and shifting the weld line to less loaded parts [28].

Optimization methods are also well suited for solving the problem of constructing an optimal molding process considering the weld line location [29]. The most popular methods are the Tauguchi method [30], finite element methods [31], genetic algorithms [14,32], and machine learning methods [33,34]. The choice of the optimization criterion that considers the molding process’s specifics affects the solution’s convergence [35]. Equivalent stresses are often used to determine the strength of weld-line-containing structures [36]. Therefore, the maximum equivalent stress is the main parameter for selecting the objective function. In [14], the half-sum of the average and maximum equivalent stresses on the weld line is also considered an optimization criterion. This study compares the most frequently used optimization criteria and assesses the convergence of the genetic algorithm for each of them. The molding optimization problem requires a multidisciplinary approach that considers molding hydrodynamics and solves the strength problem to improve the quality of mold products and minimize the weld line’s negative impact on strength.

The scientific novelty of this study is the development of technology for the automated design of a gating system based on parametric optimization algorithms, considering technological constraints. It allows the calculation of the optimal melt injection points into the cavity part to increase the load-bearing capacity of products manufactured by molding from thermoplastic material at the design stage. The modular structure of the software, which allows for the flexible connection of CAE solvers convenient for the end user, is a distinctive feature of the technology implementation. An experimental assessment of the strength of material specimens and structures containing a weld line is performed. It is experimentally confirmed that the developed technique can be used to manufacture aerospace structures and allows for increasing the strength of the product by two times.

2. Materials and Methods

2.1. Materials

The article considers several polymer fibrous materials: a 50% glass-fiber-reinforced polyamide 6, denoted as Armamid PA6 GF 50-1 (in article S50) [37], along with its associated model, as presented in [38], and a 30% carbon-fiber-reinforced polyamide 6, denoted as Gamma Plast UPA6—30 M (in article U30) [39], with its corresponding model [40].

The influence of the weld line on the strength characteristics of the material was assessed using the specimens as an example (Figure 1). The tensile curves of each material are shown in Figure 2. Mechanical tests of the molded specimens were conducted using a universal servo-hydraulic machine (MTS 322) until the structure failed. For each of the considered materials, 1BA type specimens with and without a weld line were produced by injection molding according to ISO 527-2 [41] (Figure 3).

The results confirm the influence of the weld line on the strength characteristics of the material. The presence of the weld line reduced the strength of the specimen by 2.3 and 2.1 for the carbon-fiber-reinforced thermoplastic and glass-fiber-reinforced thermoplastic, respectively, as shown in Figure 2. An assessment of the failure locations of the specimens (Figure 3) shows that specimens without a weld line were of equal strength and failed in different locations, whereas specimens containing a weld line failed at the weld line’s location.

2.2. Methods

2.2.1. Metaheuristic Optimization Algorithm

A modified version of the genetic algorithm [14] was used to determine the optimal point for injection of the melt into the part cavity. Unlike the classical implementation of the genetic algorithm, when the standard ga function in MATLAB is used [42], the modified genetic algorithm has several differences:

-

a new generation is created based on the unit of best solutions, mutation, crossover, and random seeding sets,

-

each set is generated by the genetic operator’s parallel execution,

-

removing repeated nodes’ function is implemented to clear each generation from solutions with coordinates so close that they are projected onto the same node.

The novelty of the proposed implementation of the genetic algorithm is the consideration of technological constraints by selecting nodes available for projection. In addition to this, the use of integer numbers of nodes inside the customized algorithm allows for solving the problem of duplicating solutions that provide a gate location to rounding into the same node of the computational mesh.

Increasing the load-bearing capacity of structures requires that the weld line be at the least stressed location. The stress field of the structure depends mainly on the loading pattern and the geometry of the product. The calculation was performed using an isotropic formulation before optimizing the sprue position. The stresses on the weld line are calculated once on the elements of the stress–strain calculation mesh and can be interpolated to the molding hydrodynamics calculation mesh on which the objective function is calculated at each optimization iteration.

The most appropriate criterion for solving the optimization problem is minimization of maximum stresses on weld lines:

$$\begin{array}{l}minimize\hspace{0.33em}{f}_2={\sigma }_{\max }\\ by\hspace{0.33em}varying\hspace{0.33em}\left(x\right)\in S_{part}\end{array}$$

where $S_{part}$—outer surface of the product,

x—radius vector of the melt entry point,

${\sigma }_{\max }=\underset{i\in N_{\mathrm{weld}}}{\max }\left\{{\sigma }_i\right\}$—maximum value of equivalent Mises stresses on the weld lines,

$N_{\mathrm{weld}}$—number of nodes on the weld line.

The averaging and determination of maximum values by the nodes of the weld line were implemented in APDL in ANSYS.

The proposed algorithm does not require the calculation of the derivatives of the objective function and performs a global search for an extremum. Tests on various multidimensional mathematical functions, such as the Ackley function, demonstrate the reliable detection of optimal solutions, even in the presence of multiple extrema [43]. This suggests that the convergence rate of the algorithm is weakly dependent on the part’s geometric complexity. The search space’s dimensionality has the most significant influence on the number of iterations. Thus, the proposed algorithm can be directly generalized to multi-gate systems by increasing the number of design variables: three design variables for each melt entry point.

2.2.2. Algorithm Implementation

The algorithm for optimizing the melt injection point [44] is implemented as a modular code structure (Figure 4).

The modular structure of the code allows flexible adjustment of the models applied to the direct problem solution. If available, debugged mesh generators and solvers are used for modeling the stress–strain state of products and the technological process of injection molding. The implementation of the modular structure with the connection of the ANSYS and Moldflow calculation kernels with the development of missing modules in the form of proprietary computer codes allows, on the one hand, the use of the required mathematical models and modified optimization algorithms and, on the other hand, the possibility of implementing the developed methodology for solving industrial problems due to the use of standard mesh generation tools and work with complex geometric models.

The initial data for the algorithm are as follows:

(1)

Geometric model of the manufactured product.

(2)

Mechanical characteristics of the part material required for strength (Young’s modulus and Poisson’s ratio) and technological (viscosity model) calculations.

(3)

Operating conditions determining the calculation of the stress–strain state (loading and fastening method).

(4)

Technological parameters of the injection molding process (injection rate, melt temperature, and mold temperature).

The proposed optimization algorithm assumes that the shape of the part and loading conditions determines the equivalent stress field of the structure. It does not depend on the melt injection location and is constant across the optimization cycle iterations. Therefore, the stress–strain state calculation module is performed during the optimization cycle preparation stage. The module is implemented in the ANSYS system and can use the ANSYS mechanical interface and involves the following operations:

-

setting boundary conditions—methods of fixing and loading the part,

-

setting technological constraints—selecting the surface on which the melt entry points can be located as Named Selection,

-

generating a strength calculation mesh,

-

calculating the stress–strain state,

-

exporting the strength calculation mesh (static_structural.db) and the equivalent stress field (static_structural.rst).

The injection molding process is simulated with Autodesk Moldflow connected in batch mode by Synergy API scripts and includes calling several modules. The hydrodynamic mesh generation module is called on the basis of the specified geometry. The mesh is saved in .udm format. After the strength calculation and hydrodynamic mesh generation, the stress field interpolation module is used, which transfers the values of the equivalent von Mises stresses in the structure under study to the hydrodynamic mesh nodes. The objective function is calculated using the hydrodynamics module, implemented by the Moldflow batch mode and controlled by the Synergy API. The number of nodes of the hydrodynamic mesh elements containing the weld line is the result of the module’s operation.

The modified genetic algorithm proposed in the work is implemented in the authors’ module for the optimization of the melt entry point in the APDL language, with an external call of individual functions from the GateOptWeld3D control code in the MATLAB language [14]. The module includes operators of random seeding, elitism, crossing, and mutation, which are described in detail below. The melt entry point’s optimal coordinates were obtained from the optimization module operation, ensuring that the weld line was located in the lowest equivalent stress zone.

The proposed optimization technology can be implemented as a plug-in or add-on based on any commercial CAE solver (including Simulia Abaqus) or open-source solvers (such as OpenFoam). A console interface or API that provides access to the data needed to calculate the objective function is the main requirement.

2.2.3. Molding Mesh Generation Module

In the first stage of code implementation, the calculation hydrodynamic mesh is constructed in Autodesk Moldflow and launched in batch mode by Synergy API commands recorded as a vbs file generated by the GateOptWeld3D control code [14]. The geometric model of the structure in the .stp format is used as the input data. The hydrodynamic mesh of Moldflow is recorded in the .udm format and is based on linear tetrahedral elements. Tetrahedral meshes have the advantage of allowing for a fairly accurate approximation of the product’s complex arbitrary geometry. The properties of the calculation mesh are defined by the Synergy.ImportOptions and Synergy.MeshGenerator classes.

The call of the calculation mesh generator is performed by the StudyDoc.MeshNow function with the False key of execution in batch mode without issuing GUI queries to the user. Then, the success of the mesh execution is checked by a cyclic status query. If the mesh generation is not yet complete, the algorithm waits 5000 s and repeats the query again. The resulting mesh (Figure 5) is saved in the .udm format (Figure 6), and its image is saved as a gif file.

Further processing of operations on the finite element mesh is performed using the ANSYS kernel through calls in the APDL language. The mesh in the .udm format is first converted to the format .ans (Figure 6). Computational meshing was performed using C programming language.

2.2.4. Strength Modulus

The strength module is implemented in the ANSYS system. The loading conditions and finite element mesh parameters are recorded using macros. The equivalent stresses in each mesh node were calculated.

2.2.5. Interpolation Module

The result of the liquid molding calculation is the number of hydrodynamic mesh nodes on the weld line. A preliminary transfer of the stress field from the strength mesh to the molding calculation mesh is performed to determine the stress values at these nodes using the interpolation function of the APDL language module \map GateOptWeld3D code [44] (codes\map\master.txt, codes\map\get_stress.txt).

The interpolation is performed by the *moper command, which is called with the intp and sget keys. In the first phase of stress interpolation, the number of elements of the original mesh within which the nodes of the new mesh lie are determined. In the second phase of stress interpolation, the stress values on the new mesh are recalculated based on the stress values in the elements of the original mesh to which they belong and the shape function of the elements of the original mesh. In this case, the error in stress interpolation is reduced to a minimum due to the fact that the mechanical and hydrodynamic computational meshes have practically the same element size.

The array of interpolated values of the stress field on the hydrodynamic mesh is saved as a file and is used in all optimization iterations to determine the stresses at the weld-line-containing nodes.

The file consists of two columns with the node number in F12.4 format and the equivalent stress values in MPa, written in F12.4 format, for example:

82651.0000 4.7293.

Values of equivalent stresses in a structure, under a given operating condition loading case, are interpolated onto the product molding calculation mesh and saved in a seqv.txt file.

2.2.6. Considering Technological Constraints

The mold design assumes the possibility of bringing the melt injection point to certain surfaces of the part under study. The possible runner geometry and gate location depend on the complexity of the mold shape and technological features of the injection molding process, including limitations on gate diameter, issues of mold deformation, and organization of part ejection. In cold runner gating systems, the melt entry points are often on the part’s side surfaces, which facilitates mold opening. Code implementation of accounting for process constraints is carried out in the \manufc module of the GateOptWeld3D code [14,44] (\manufc\master.txt, \manufc\write_ext_nodes.txt) by selecting surface geometry elements located on the Named Selection with the name “sel_nodes” of the static_structural.db project using the function. The Named Selection can be assigned to the ANSYS workbench mechanical interface (Figure 7) during the strength calculation stage.

2.2.7. Molding Calculation Module

The molding calculation is carried out in Autodesk Moldflow, launched in batch mode by Synergy API by VBS scripts in Visual Basic, automatically generated by the GateOptWeld3D control code [14]. After the calculation, the results of the weld line location are exported in the form of a gif image, xml, and nod formats (Figure 8).

2.2.8. Optimization Module

A typical block of the optimization algorithm is the definition of melt injection nodes on the hydrodynamic molding mesh. This block is called when a random initial seeding is determined, and the objective function is calculated on a list of nodes with specified coordinates. The nsel, s, ext APDL command is used to select a set of surface nodes of the calculation mesh, which selects nodes on the part’s surface [14]. Then, solid elements are excluded from consideration, and the product’s external nodes are covered with a mesh of surface elements. As a result, a transition occurs from the volumetric mesh of the part to a surface mesh, allowing the melt injection node positions and injection direction to normalize.

The normal to the product surface at the melt injection point is determined by averaging the normals of the elements associated with the melt injection point node and selected by the enextn APDL command [14]. The normals of the selected element nodes are obtained using the APDL commands normnx, normny, and normnz.

In the GateOptWeld3D code [14,44], two seeding determination variants are implemented: random seeding, which is used for generating the first generation, and seeding based on the list of coordinates of melt injection points (used when calculating the generation objective function in the case of using a third-party optimization algorithm). Random seeding is performed using the ANSYS system’s random number generator rand, selecting a list of available random nodes on the melt injection surface (with possible technological constraints).

To initialize the random number generator, its idle run skips the first pseudo-random sequence values. The number of idle runs is determined by specifying the seeding number, if repeated comparisons of algorithms are necessary, or randomly, based on the wall time values. During seeding, the obtained position is projected onto the surface elements of the computational mesh using the APDL command node (x, y, z) when the node coordinates are determined. Because of the \initial_seed subroutine, a file is generated with a list of nodes and normals for calculating injection molding in .txt format:

*vwrite, inlet_nodes(1), node_nx(1), node_ny(1), node_nz(1)

(F8.0, F8.4, F8.4, F8.4).

The example of the seed nodes txt file line is:

4913 − 0.0367 0.4264 0.9038.

The convolution of the equivalent stress values on the weld line is used as an optimization criterion. The convolution function can be the maximum, average value operators, or their half-sum. For each injection molding calculation case, a cycle is called over the weld line nodes (Algorithm 1). In each node on the weld line, the equivalent stress is found from the previously interpolated values, and then the optimization criterion is calculated based on the selected convolution function (maximum, average value, or half-sum of the average and maximum values).

Algorithm 1: Determining the values of the objective function

*dim, inlet_quality, array, inlet_nodes_num,6 *do,i,1,inlet_nodes_num   stress_max= 0.0   stress_avg = 0.0   stress_avg_p = 0.0   inlet_quality(i, 1) = spay_nodes%i%(1)   *do,1,2, spay_nodes_num%i% (1)+1     id=map_ids (spay_nodes%i%(j))     *if,s_eqv(id), ge, stress_max, then       stress_max=s_eqv(id)     *endif     stress_avg = stress_avg+s_eqv(id)     stress_avg_p stress_avg_p+ s_eqv(id)**3   *enddo stress_avg = stress_avg spay_nodes_num%i%(1) stress_avg_p = (stress_avg_p)**(1/3) inlet_quality(1,2)= 0.5*stress_max + 0.5*stress_avg inlet_quality(1,3)= stress_max inlet_quality(i,4)= stress_avg

The value of the objective function is exported to a txt file (

Table 1. Example of calculating the objective function for a population of 30 individuals.

Node_ID	Criteria
158.0000	9.8645
45035.0000	10.1049
4481.0000	10.1457
48508.0000	10.3422
4913.0000	11.0257
421.0000	11.2119
39557.0000	15.7199
2658.0000	40.3989
25681.0000	41.1663
7341.0000	42.5090
17508.0000	42.8299
23992.0000	48.2619
17886.0000	51.0144
31201.0000	51.3535
17938.0000	51.6137
175.0000	51.7236
17580.0000	52.0905
17854.0000	52.7470
25592.0000	53.1188
23999.0000	53.4022
36147.0000	60.2192
17293.0000	60.7792
17961.0000	62.7508
663.0000	62.8980
9664.0000	64.1144
49651.0000	66.8427
5113.0000	69.7721
31734.0000	70.6678
36026.0000	82.6672
32619.0000	96.8143

).

3. Results

3.1. Optimal Gate Location Mold Design

The study object was a bracket whose design represents a nonzero genus topology. The initial data are the dimensions of the brackets: 70 × 70 × 25 mm, the thickness of the bracket belts is 4 mm, the wall thickness is 3 mm, and the gate channels are 8 mm wide. Figure 9a shows the design and loading scheme of the bracket. This test piece was chosen because it reflects the key features of aerospace components. It contains tension/compression flanges and shear web, relief openings, and the resulting weld line. The advantage of its simple form is its focus on the weld line issue under investigation, with minimal influence from other factors, which facilitates the validation process. The test piece can also be manufactured in flat molds without slides.

The calculations of the stress–strain state carried out in ANSYS showed that the values of equivalent stresses in the structure do not strongly depend on the fiber orientation in the product (Figure 9b,c). The orientation of fibers near the weld line is determined by the hydrodynamics of melt front convergence: fibers near the weld line lie in the weld plane, which is one of the reasons for the reduction in material strength [22,45,46]. Accounting for this effect does not require recalculating the equivalent stress field but can be accomplished by reducing the permissible value of the equivalent stress limit.

The GateOptWeld3D code implements a heuristic algorithm, which is a modification of the genetic algorithm, in the form of a MATLAB control code. The algorithm is based on a population of melt entry points, which allows the melt entry points to be optimized, regardless of the manufactured product’s shape. The population size was selected as 30 individuals, of which 6 were defined as the best, 6 were used for mutation, 3 were set randomly, and 15 were obtained by crossing.

The algorithm was run with different population sizes: 7, 12, and 30 individuals. To study the convergence of the solution, the calculation of the objective function was considered with and without technological constraints. Considering technological constraints allows us to reduce the search area for solutions from the entire surface of the part to selecting points along its perimeter, thereby excluding solutions that cannot be implemented when molding products. The use of constraints allows us to reduce the population size and computational time required to solve the optimization problem. Solving the problem without considering the constraints requires at least 54 objective function calls. However, technological constraints can reduce the number of calls to 28. Figure 10 shows that reducing the population size affects the algorithm’s accuracy. Therefore, to solve this problem, considering the constraints, the optimal number of individuals is 12, while without considering the technological constraints, 30 individuals are needed. Mutation rates analysis was carried out for the consideration of technological constraints and 12 individuals in the population case (Figure 11). The relative diameter, d, of the mutation (as a ratio to the part size) and the number, Nmut, of mutated individuals in population are varied. In the case studied, they showed good results, with d = 0.4 and Nmut = 1. In all cases, convergence is achieved by the 3rd generation, so that the influence of the mutation is limited by the spread of individuals in the second iteration.

Figure 12 and Figure 13 show the evolution of the solution for different population sizes from the 1st to the 7th iteration using the criterion of minimizing maximum stresses on the weld lines.

The design feature of the part involves thickening in the bracket lug, which leads to uniform melt distribution and a similar molding flow when the gate is located along the perimeter of the lug, regardless of the injection side in the area of the central lug (Figure 14). When the melt is injected into the lug area, the value of the objective function remains close because the weld line location does not change, resulting in a small change in the response surface and the equivalence of the proposed solutions in the central lug area. Therefore, during bracket production, the melt was fed to the center of the lug.

A mold that includes two bracket molding options has been designed for molding a three-dimensional spatially loaded structure with different melt entry points. The first option is the case of the weld line coinciding with the most loaded place, and the second is the optimal option for the gate location, with the lowest stresses on the weld line. Tensile molding of specimens according to the ISO 527 standard [41] was also performed, including and not including the weld line. Figure 15 shows the three-dimensional geometry of the mold in the NX.

In the Moldflow system, the material and expected process mode characteristics were set for both gating system arrangements. Figure 16 shows the molding front, which allows the evaluation of the quality of the future molded product, and the locations of the weld lines.

3.2. Test Study Brackets’ Molding

The verification calculation showed that the selected process mode ensures complete molding of the product, with the weld lines located in the expected places. The mold is made of St-3 steel plates, processed by grinding. Manufacturing was carried out by milling on a CNC machine with cutters with a diameter of 4 mm, with subsequent fastener refinement and mold assembly. The heating of the mold was performed by heating elements (on each half of the mold), and temperature measurement was performed by installed thermocouples. The area near the sprue bushing had an additional temperature control channel. The manufacturing of the mold was the final stage before further molding of the product. The molding of brackets from short-reinforced composite materials was performed on an electric injection molding machine, Negri Bossi VE 210-1700 (Figure 17).

More than 20 brackets were made of carbon-filled polyamide Gamma-plast UPA 6 30 M and more than 20 brackets were made of glass-filled polyamide Poliplastic Armamid PA6 SV 50-1. The mold temperature was 80 °C (in the regular zone) and 100 °C (near the sprue bushing). The melt temperature was 250 °C for glass-filled polyamide and 230 °C for carbon-filled polyamide. The volumetric flow rate of the mold filling was 28 cm³/s, and the mold filling time was 3.5 s. Figure 18 show the brackets made of different materials and manufactured on an injection molding machine.

3.3. Mechanical Tests

Mechanical tests of brackets were conducted using a universal servo-hydraulic machine (MTS 322) until the structure failed. A test fixture was manufactured to ensure fastening and loading of the product in accordance with the designed load (sealing of the bracket base and the shear lateral force load of its lug). Tests were conducted until failure of 20 brackets (10 carbon- and 10 glass-filled polyamide brackets, including 5 of each material with a side gate and front gate). Tests were also conducted on 30 witness specimens (15 of each material, of which 5 had a weld line and 10 did not). Measuring the thickness of the products and weighing the brackets made of the same material showed a deviation in size and weight not exceeding 0.5%, indicating the stability of the molding program. The mechanical test results are shown in Figure 19.

Tensile curves with different gate locations were plotted to evaluate the influence of the proposed manufacturing technology on the strength of structures made of short-reinforced composite materials. The brackets after testing are shown in Figure 20.

3.4. Microstructure Evaluation

Using a Tescan Vega electron microscope, a comparison was made between the orientation of the fibers at the fracture sites of bracket samples manufactured with different gate locations, as well as a comparison of the orientation of the fibers at the fracture sites of the witness samples (Figure 21). Tests showed that fracture of brackets molded from the side entrance of the melt occurred along the weld line. The side-molded bracket fracture was located along the weld line, and the fibers lay predominantly in the fracture plane. It is one of the reasons for the reduced strength of the brackets with weld line problems. The front-molded bracket did not contain a weld line at the fracture site and the fibers in fracture place were predominantly perpendicular to the fracture plane (along the molding). A similar pattern was observed in witness specimens molded with and without a weld line: in the case of weld line specimens, the fibers were in the plane of the mold front closure, while in the specimens without a weld line, they were aligned with the molding direction. Electron microscopic examination of the fracture surfaces of the brackets and test specimens revealed no delamination of the material. This is due to the short reinforcement of the material and the presence of fibers in various directions. Moldflow calculations of the fiber orientation tensor were consistent with electron microscope data, which validates the modeling methodology.

4. Discussion

4.1. Injection Molded Model Validation

Comparison of the molding front and the calculation of the orientation of the reinforcing fibers with the experimentally obtained molding fronts also showed good agreement (Figure 22).

Thus, a comparison of the results of calculations and experiments showed that the manufacture of products from composite materials, according to the proposed technology, allows the weld line to be located away from highly loaded areas and significantly increases the strength of the structure.

4.2. Weld Line Influences the Part Stiffness and Strength

Because of the high molding stability, the weight and thickness of the brackets were not standardized. The variation coefficient is calculated as the ratio of the standard deviation to the average value and does not exceed 6% in determining the strength of the bracket. The strength of brackets containing the weld line in the loaded area of the structure was 0.44 (for carbon composite) and 0.6 (for glass composite) for the strength of brackets with an optimally located melt entry point. The glass composite bracket lost less strength (up to 0.6 from initial) than the witness specimen (up to 0.43 from initial), which can be caused either by higher plasticity of the glass composite or by weld line deviation from maximum stress in the structure. The double loss of bracket strength when the weld line coincides with the area of maximum equivalent stress indicates the need for melt injection point optimization so that the weld line is located away from the most loaded areas.

According to the results of the mechanical tests, the failure of the brackets in which the sprue was on the side occurred along the weld line. The fibers lay along the fracture line in this case. The failure site did not contain a weld line in the second case of bracket molding, and the fibers were mainly located along the molding direction. The same picture was observed when the failure sites of the witness specimens were studied. A comparison of the experimental results and the results of calculating the stress–strain state of brackets with different gating system locations in the ANSYS was also performed (Figure 23).

Comparison of the calculation results and mechanical tests confirmed that the weld line falling in the areas of highly loaded places of spatial structures slightly reduced the rigidity (by 12% in the experiment and by 5% in the calculation) and significantly reduced their strength (by 56% in the experiment and by 32% in the calculation).

5. Conclusions

This article described the technology of automated design of a gating system based on parametric optimization algorithms. The proposed modular structure of the code made it possible to combine the well-known ANSYS and Moldflow software packages to solve a joint optimization problem of describing the optimal point of the melt injection point into the cavity part, considering the weld line location, and the computation time could be reduced by modifying the genetic algorithm. The consideration of the technological constraints allowed us to reduce the population size and optimization problem solution computational time to 1.9 times. The developed design technology was applied to the problem of designing a complex topology bracket made of short-reinforced composite material. A full modeling cycle, including molding calculation and calculation of the stress–strain state of products, and a manufacturing cycle, including cavity part design and injection molding of products on an injection molding machine, were performed. The technology developed standard FEM-based design methods for optimizing process parameters [47] by using geometric operations to search gate location nodes on the computational mesh in the crossover and mutation operators of the proposed metaheuristic algorithm. The developed technology can be used in multi-gate molding cases by increasing the number of design variables, adding three design variables for each melt entry point. Mechanical tests of the brackets were conducted, and the results showed that the proposed design technology increased the strength of brackets made of short-reinforced composite material by two times, while ensuring the weld line’s location in the zone of the least equivalent stresses. The obtained results confirm the adequacy and applicability of the methods used to improve the quality of short-reinforced composite materials by injection molding. The use of the proposed technology is particularly relevant in the aerospace and automotive industries, as it would provide significant economic benefits from saving the weight of structures in these areas of application.