Enhancement of Additive Manufacturing Processes for Thin-Walled Part Production Using Gas Metal Arc Welding (GMAW) with Wavelet Transform

Abstract

Additive manufacturing encompasses technologies that produce three-dimensional computer-aided design (CAD) models through a layer-by-layer production process. Compared to traditional manufacturing methods, additive manufacturing technologies offer significant advantages in producing intricate components with minimal energy consumption, reduced raw material waste, and shortened production timelines. AM methods based on shielded gas welding have recently piqued the interest of researchers due to their high efficiency and cost-effectiveness in manufacturing critical components. However, one of the most formidable challenges in additive manufacturing methods based on shielded gas welding lies in the irregularity of weld bead height at different points, compromising the precision of components produced using these techniques. In this current research, we aimed to achieve uniform weld heights along the welding path by considering the most influential parameters on weld bead geometry and conducting experimental tests. Input parameters of the process, including nozzle angle, welding speed, wire speed, and voltage, were considered. Simultaneously, image processing and wavelet transform were employed to assess the uniformity of weld bead height. These parameters were applied to produce intricate parts after identifying optimal parameters that yielded the smoothest weld lines. According to the results, the appropriate bead for manufacturing the part was extracted. The results show that the smoothest bead line is achieved in 27 V as the highest level of voltage, at a 90° nozzle position and the maximum wire feed rate. Parts manufactured using this method across different layers exhibited no distortions, and the repeatability of production substantiated the high reliability of this approach for component manufacturing.

1. Introduction

The Fourth Industrial Revolution has arisen from the convergence of industrial production and information and communication technologies, with a strong focus on increasing automation and connecting physical and virtual realms. Various cutting-edge tools, including the Internet of Things, cloud computing, big data analytics, cyber-physical systems, and domains like additive manufacturing, have played pivotal roles in driving the development of this industrial revolution [1,2]. Among these, additive manufacturing, with its layer-by-layer production process and the conversion of computer models into physical components, emerges as a critical and practical component of the Fourth Industrial Revolution.

The distinctive capabilities of additive manufacturing, particularly mass customization and the production of lightweight components, offer substantial potential for enhancing efficiency, expanding applications, and creating environmentally friendly products when integrated with the Internet of Things and data in the Fourth Industrial Revolution [3].

Modern additive manufacturing technologies have found extensive applications in various fields, including aerospace, maritime, medical, automotive, and energy sectors, owing to their capacity to generate integrated and complex geometries, high production speed, desirable strength, high density, optimal material utilization, design freedom, and potential for cost reduction [4,5,6,7,8]. Notable techniques among these methods include Selective Laser Sintering (SLS), Direct Metal Deposition (DMD), Laser-Based Additive Manufacturing (LBAM), Directed Light Fabrication (DLF), Electron Beam Additive Manufacturing (EBM), and Wire Arc Additive Manufacturing (WAAM) [9,10,11,12].

In recent years, one particularly notable technology in additive manufacturing for metals has been identified as Wire Arc Additive Manufacturing (WAAM). This method excels at fabricating complex geometric metal components and has been the subject of numerous studies due to its high level of automation [13]. Compared to other additive manufacturing methods, such as Powder Bed Fusion (PBF) using lasers, WAAM exhibits a higher deposition rate for producing large-scale components. While WAAM systems are increasingly employed in industrial and academic domains, several challenges and obstacles still need to be overcome to maintain process stability [14].

In additive manufacturing using the WAAM method to produce components, layer deposition is carried out based on information derived from a 3D model. During the layer deposition process in the additive manufacturing method based on Gas Metal Arc Welding (GMAW), a metal wire is melted using an electric arc and deposited onto the previously formed weld bead, thus forming a new layer. The crucial aspect of this process lies in the non-uniformity of the deposited weld bead [15,16].

Figure 1 provides a schematic representation of the GMAW process. This welding process is highly versatile and can be employed in automatic or semi-automatic modes by connecting it to a CNC machine or a robotic arm. The adjustable parameters in the GMAW process, which significantly influence penetration depth, weld bead geometry, and overall weld quality, include current intensity, voltage, welding speed, wire type, diameter, and wire feed rate [17,18,19].

In particular, when constructing layer-by-layer components, the accuracy of the weld bead is paramount to accurately represent geometry and prevent process irregularities [20]. Due to the high sensitivity of the GMAW process to minor changes in deposition parameters, such as arc current, deposition rate, arc voltage, layer geometry under different heat dissipation conditions, interlayer temperature, quality of previous layer surfaces, and substrate deformation, a critical issue in additive manufacturing based on GMAW is the variation in the height of the deposited layer. This variation results in low and high spots that can accumulate during the additive process, leading to the formation of porosity or an inadequate layer shape.

Moreover, these low and high spots affect the nozzle-to-workpiece distance, with excessive distance leading to inadequate shielding of the pool and shorter distance causing the spatter phenomenon, where droplets adhere to the nozzle or collide with the upper layer’s surface. The deposited substrate exhibits unevenness in the additive manufacturing method based on GMAW, creating significant differences in the substrate’s geometry at the beginning and end of welding paths. Additionally, the intersection of deposition paths may result in partial protrusions, leading to a substantial change in the layer’s height. Furthermore, the connecting surface is not smooth at the junction of welding paths. Overall, the persistence of unevenness throughout the process poses challenges in completing the production process or results in a finished piece with very low dimensional accuracy [17,21,22]. Figure 2 illustrates two uneven and smooth paths in the GMAW process.

One of the challenges in GMAW welding is the inconsistency of weld geometry along the weld line and the significant variation in weld bead height. Consequently, Xiong et al. [23] have developed a new method to enhance components’ surface quality and dimensional accuracy by combining plasma arc deposition and milling techniques. The precision of components produced in the combined CNC process, which eliminates the steps resulting from layer-wise additive manufacturing, is ensured, resulting in parts that closely resemble the original component. Using the optimized set of process parameters obtained, a group of metal components, such as hollow metal containers, has been manufactured. Karunakaran et al. [24] have also presented a combined layer-wise additive manufacturing process using milling techniques to enhance dimensional accuracy and surface quality. Xiong and Zhang [21] have developed an auxiliary visual sensor system to monitor the distance between the nozzle and the upper surface for adaptive control of deposition height in layer-wise additive manufacturing of GMAW-based materials. Deviations in the nozzle-to-upper-surface distance are compensated for by adjusting the working flat displacement and deposition rate in the next layer. After simplifying the controlled process into a linear system, they implemented an adaptive control system to maintain a constant nozzle-to-upper-surface distance. The effectiveness of the controller has been evaluated through the deposition of single-bead multi layer walls, and experimental results have confirmed that the developed controller can improve process stability when applied. Xiong et al. [25] have presented a single-layer neural network-based self-regulating controller for layer width adjustment. In this method, the input parameter is the nozzle velocity, and the output parameter is the layer width. The controller’s performance is valid under deposition conditions, interlayer temperature variations, and layer width changes. The proposed controller significantly enhances process stability. Another study investigated control strategies for arc striking and extinguishing areas in constructing additive manufacturing materials based on multi layer single-pass GMAW. The arc striking and extinguishing areas are placed within the same layer to compensate for height differences in deposited parts. A control strategy for alternate deposition in adjacent layers is employed for open-path parts. Parameter adjustment within the arc striking and extinguishing area uses the same deposition direction [26]. In another investigation, Xiong et al. [27] recommended using GMAW in additive manufacturing for constructing thin-walled parts. The nozzle is oriented perpendicular to the work surface during additive manufacturing. They examined the effects of nozzle-to-workpiece distance, wire transfer speed, and nozzle traverse speed on the produced sloped surface. They demonstrated that an increase in nozzle traverse speed leads to an increase in the surface angle relative to the horizontal plane, while an increase in wire speed results in a reduced angle. Furthermore, they achieved a sloped surface angle of 35 degrees [27]. Considering that the geometry of the weld pool at the start and end of the weld bead is usually non-uniform in comparison to the central region, which significantly influences the formation in the production of GMAW-based enhancements, Hu et al. [28] investigated the reasons and optimization strategies for irregularities in the weld pool in the unstable region. The results indicated that the non-uniform geometry of the weld pool could be attributed to the fluid flow and metal stirring in the weld pool, with the length of the initial bulky region being positively correlated with the slope at the end and the length of the weld pool. Several strategies have been proposed for controlling irregularities in the weld pool by adjusting welding parameters, fill pass options, and path planning patterns. These methods contribute to achieving a continuous and smooth deposition surface, forming the foundation of the GMAW-based additive manufacturing process. Xiong et al. [29] focused on the surface quality of the side sections of the components. They proposed a laser vision-based system to observe the appearance of a surface. The effects of critical process parameters such as interlayer temperature, wire transfer speed, weld nozzle travel speed, and the constant wire speed-to-weld nozzle travel speed ratio on the surface roughness of thin-walled parts were discussed and investigated in detail. Improving the surface quality of thin-walled parts accompanies a decrease in interlayer temperature. Reducing wire speed and weld nozzle travel speed leads to an increase in surface smoothness. Jay Vora et al. [30] investigated the microstructure and mechanical properties of the WAAM process using SS316L metal wire. To assess these properties, tensile, impact, and microhardness tests were conducted on three distinct regions of the fabricated wall: upper, middle, and lower. Their research findings indicate that suitable bonding occurred between the different layers, with complete fusion and the absence of observable oxidation between the layers. Furthermore, the structure created using the WAAM process, based on GMAW, conforms to industrial application standards. Reyazul Warsi et al. [31] studied the mechanical properties of the WAAM process on both conventionally heated and preheated A36 steel substrates. This study used a low-carbon alloy steel wire, E870S-6 filler wire. Subsequently, a single square and composite square bead consisting of nine layers were fabricated. Measurement coordinate data obtained using a Coordinate Measuring Machine (CMM) indicated that the bead on a preheated substrate exhibited more excellent dimensional stability and reduced protrusion. Michal Wieczorowski et al. [32] focused on examining the process parameters of WAAM in fabricating a thin-walled structure using AA5356 material. Tensile testing and computer thermography were employed to assess mechanical properties and analyze porosity in the samples. Their experimental results reveal that a reduction in overall pore volume accompanies increased linear velocity. Additionally, the tensile strength of the samples remains independent of the porosity measured by computer thermography.

Production of additive materials for wire and arc welding-based structures is an up-and-coming technology for large-scale metal construction due to its rapid deposition rate, high energy efficiency, and cost-effectiveness [8]. These components can be fabricated almost identical to the original piece through layer-by-layer deposition of metallic materials using welding processes. A significant challenge in constructing structures with large and thin walls is the inconsistency of the deposited layers. Accumulating layers with minimal smoothness results in noticeable height differences at various locations within a layer, hindering the continuous deposition of multi layered materials. Ma et al. [16] have presented optimization strategies for robotic additive and subtractive manufacturing of large and thin-walled aluminum structures. They proposed three optimization strategies to achieve flat layers: deposition with weaving, arc igniting and arc extinguishing control, and local measuring and milling strategy. Experimental results demonstrate that combining these strategies can enhance surface smoothness and reduce height discrepancies within a layer. Utilizing these solutions, a large piece with thin walls has been fabricated, showcasing the capability of producing substantial metal components in a short time frame using a robotic additive and subtractive manufacturing system.

Additive manufacturing, a layer-by-layer fabrication method, has witnessed remarkable growth in the production of intricate components, offering substantial potential for reducing production time and minimizing material waste compared to traditional manufacturing methods [33]. However, it is crucial to acknowledge that the dimensional accuracy and surface quality of parts created through additive manufacturing processes often lag behind those produced using conventional manufacturing techniques [23].

One of the challenges in additive manufacturing, particularly when employing gas-shielded welding, is achieving uniform layer thickness. Given that the final piece is constructed incrementally, the precision and quality of the produced parts are significantly influenced by the surface irregularities and the heights of the weld beads in each layer. The cumulative impact of errors and inconsistencies in each layer can profoundly affect the final product. Numerous parameters come into play in determining weld bead geometry, directly affecting layers’ formation in thin-walled components. By appropriately adjusting these parameters, it becomes possible to attain a uniform weld bead geometry suitable for additive manufacturing.

This research introduces a novel method based on image processing and wavelet transform to address the challenges related to dimensional accuracy and enhance the height uniformity of weld beads in the additive manufacturing process of thin-walled parts using GMAW. This study investigates influential parameters such as welding voltage, wire feed rate, nozzle travel speed, and nozzle angle to examine the height uniformity of weld beads in the fabrication of thin-walled parts through image processing and wavelet transform. Figure 3 shows the schematic view of the process parameters and their effects on the welding powder geometry.

To evaluate the performance of the proposed method, a three-axis controlled table was employed for arc control and the production of complex geometric components. These components were manufactured based on the results obtained from the wavelet transform. A dedicated program was developed to generate motion codes for the table from the part model in each layer. Throughout the production of components using this method, the uniformity of component height was consistently maintained until the final layer.

2. Image Processing and Wavelet Transform

A substantial body of research exists on arc welding-based additive materials accompanied by gas; however, less attention has been devoted to analyzing surface distortions in welding. Among various machine vision techniques employed to detect surface defects in welding over the past decade, texture analysis has proven to be a powerful tool. This represents a significant starting point in comprehending the surfaces produced by various processes, such as friction stir welding [21]. In some cases, surface distortions can reveal the luminance of the metal surface, and their reflection can be harnessed using a specific model grounded in certain hypotheses and the solution of the image ray equation through specific algorithms [17,23].

Mathematical transformations are employed for signal analysis and extracting information that is not readily discernible from raw signals. Wavelet transform is one of the most efficient mathematical transformations in signal processing. Wavelets are mathematical functions that rapidly analyze sound, images, and vibrations. Describing the application of one-dimensional discrete wavelet transform is straightforward: this transformation decomposes a signal into two other signals, namely approximation, encompassing low-frequency components, and details, encompassing high-frequency components, as depicted in Figure 4A. This decomposition facilitates the reconstruction of the original signal based on these components. In Figure 4A, H₁(z) represents a high-pass filter, and H₀(z) represents a low-pass filter, determined based on an essential characteristic in wavelet transform known as the mother wavelet [34]. Figure 4B illustrates the signal decomposition using this transformation.

The scale and wavelet functions, such as Haar (Haar wavelet) and Daubechies (Daubechies wavelet), among others, are well established. It is possible to approximate a discrete signal in $l^2\left(z\right)=\left\{{\sum }_{n=-\infty }^{\infty } {\left|f\left[n\right]\right|}^2<\infty \right\}$ using Equation (1) [34,35].

$$f\left[n\right]={\displaystyle \frac{1}{\sqrt{M}}}{\sum }_k W_{\varnothing }\left[j_0,k\right]{\varnothing }_{j_0,k}\left[n\right]+{\displaystyle \frac{1}{\sqrt{M}}}{\sum }_{j=j_0}^{\infty } {\sum }_k W_{\psi }\left[j,k\right]{\psi }_{j,k}\left[n\right].$$

(1)

In the above function, $f\left[n\right]$, ${\varnothing }_{j_0,k}\left[n\right]$, and ${\psi }_{j_0,k}\left[n\right]$ are in the discrete functions [0, M−1]. The wavelet transform coefficients are obtained by calculating inner products, as shown in Equations (2) and (3).

$$W_{\varnothing }\left[j_0,k\right]={\displaystyle \frac{1}{\sqrt{M}}} {\sum }_n f\left[n\right]{\varnothing }_{j_0,k}\left[n\right].$$

(2)

$$W_{\psi }\left[j,k\right]={\displaystyle \frac{1}{\sqrt{M}}} {\sum }_n f\left[n\right]{\psi }_{j,k}\left[n\right].$$

(3)

Equation (2) represents the approximation coefficients, while Equation (3) represents the detail coefficients [34].

To extend this concept to two dimensions, consider representing a two-dimensional signal or image as a 2D matrix, denoted as x(m × n).

To acquire the 2D wavelet transform, recalling the two-dimensional Fourier transform is helpful. The basis for the modified function is provided by Equation (4)

$$EXP\left(j\left(w_1{t}_1+w_2{t}_2\right)\right)$$

(4)

Instead of the transform’s EXP(jwt) coefficients, two variable functions give rise to the 2D wavelet transform. The scale and wavelet functions are $\varphi (x,y)$ and $\psi (x,y)$. The basis functions are expressed in Equations (5) and (6).

$${\varphi }_{j,m,n}\left(x,y\right)=2^{\frac{j}{2}}\varphi (2^{j}x-m, 2^{j}y-n)$$

(5)

$${\psi }_{j,m,n}^i\left(x,y\right)=2^{\frac{j}{2}}{\psi }^i\left(2^{j}x-m, 2^{j}y-n\right) i=\left\{H,V,D\right\}$$

(6)

When these functions are defined separately, the analysis of the 2D function becomes more straightforward, allowing for a focus on the design of 1D wavelets and scale functions. The modified decomposition and synthesis equations are presented as Equations (7)–(9) [34].

$$W_{\varphi }\left(j_0,m,n\right)={\displaystyle \frac{1}{\sqrt{MN}}}{\sum }_{x=0}^{M-1} {\sum }_{y=0}^{N-1} f\left(x,y\right){\varphi }_{j0,m,n}\left(x,y\right).$$

(7)

$$W_{\psi }^i\left(j,m,n\right)={\displaystyle \frac{1}{\sqrt{MN}}}\sum _{x=0}^{M-1} \sum _{y=0}^{N-1} f\left(x,y\right){\varphi }_{j,m,n}^i\left(x,y\right). i=\left\{H,V,D\right\}$$

(8)

$$\begin{gathered} f\left(x,y\right)={\displaystyle \frac{1}{\sqrt{MN}}} {\sum }_m {\sum }_n W_{\varphi }\left(j_0,m,n\right){\varphi }_{j0,m,n}\left(x,y\right) \\ +{\displaystyle \frac{1}{\sqrt{MN}}} {\sum }_{i=H,V,D} {\sum }_{j=j0}^{\infty } {\sum }_m {\sum }_n W_{\psi }^i\left(j,m,n\right){\varphi }_{j,m,n}^i\left(x,y\right) \end{gathered}$$

(9)

The 2D wavelet transform is implemented by applying the 1D wavelet transform to each row and then applying the 1D wavelet transform to each column. Figure 4C illustrates the generation of the 2D wavelet transform at one level.

For a deeper understanding of the concept of the 2D wavelet transform, let us examine it in two dimensions. Initially, the rows of the 2D signal undergo a 1D wavelet transform, dividing the signal into low-frequency (L) and high-frequency (H) components, as depicted in Figure 5A. The L component comprises signal parts with gradual changes along the horizontal axis, while the H component includes rapid changes.

In the next stage, by applying a one-dimensional wavelet transform along the columns, these two components are also decomposed into two high-frequency and low-frequency components on the vertical axis. In other words, the L component is further decomposed into high-frequency and low-frequency components, denoted as LL and LH, in the direction of the columns. The LL component encompasses a portion of the signal characterized by slow variations in horizontal and vertical directions, resembling the original signal closely. The LH component contains a portion of the signal with rapid variations in the horizontal direction and slow variations in the vertical direction.

Suppose we have an image with 21 rows, where the first 10 rows are nearly constant in both vertical and horizontal directions, and the eleventh row differs from them, simultaneously displaying constant variations in the horizontal direction.

In that case, the eleventh row will exhibit rapid changes in the vertical direction and slow changes in the horizontal direction, and these changes will manifest in the LH component of the image at corresponding locations. Similarly, image H is decomposed into two components: HH and HL, with the HL component incorporating a portion that exhibits rapid variations in the horizontal direction and slow variations in the vertical direction, while the HH component encompasses a portion that displays rapid variations in both horizontal and vertical directions, or, in other words, rapid changes in the diagonal direction. Figure 5B illustrates the final decomposition of the image into these four components.

As demonstrated, horizontal lines are represented along the HL axis, vertical lines along the LH axis, and diagonal lines along the HH axis. In general, it can be asserted that the LH image accentuates variations along the vertical axis, the HL image emphasizes changes along the horizontal axis, and the HH image spotlights changes along the diagonal axis. Conversely, the LL image essentially serves as the initial image from which high-frequency components and noise have been somewhat removed [34,35]. Figure 6 depicts an image and its decomposition using a two-dimensional wavelet transform.

3. Experimental Setup and Plan

The WAAM method generates parts by depositing weld beads onto each other. This technique produces weld beads utilizing the GMAW process. The research employed GMAW equipment, including the Migatronic Mig 501 power supply and PARS FEED 4520C wire feeding system. A cooling system (Figure 4C), PARS COOL, and a gas preheater, the Neptune, were also employed.

In this research, a constant voltage Joosha welding instrument was used. In this instrument, the voltage can be adjusted and the amperage will change during the welding process. To facilitate the execution of experimental tests during construction, a three-axis controlled table was utilized, adapted from an FP4M milling machine table for movement in three axes at adjustable speeds. Initial experiments were conducted to investigate weld bead geometry on a weld line and determine the optimal geometry for manufacturing thin-walled parts. The nozzle of the welding machine was connected to the milling table for initial assessments and obtaining desirable parameters for creating welds with suitable geometry, as depicted in Figure 7, where GMAW is shown with milling table motion control.

In additive manufacturing, parts are produced layer by layer, each serving as the substrate for the next layer if parameters ensuring a uniform welding surface are not utilized. Ultimately, the produced component may exhibit defects.

Metal parts protected with gas-shielded electric arc welding were employed to obtain an initial image for weld geometry determination. The metal surface was captured using a digital camera. Subsequently, the digital camera lens, specifically the DFK-23GM021, was calibrated according to the specifications in

Table 1. Digital camera specifications.

Amount	Specifications
60 frames per second	Data rate
1280 × 960	Resolution
Ccd	Sensor
Diameter 30 mm and focal length 12 mm	Lens m1614-mp2

.

Upon entering the computer, the captured image was converted into a grayscale image. Then, brightness and contrast were adjusted to enhance the image’s clarity. The selection of appropriate values for the two image attributes, brightness and contrast, depended on the lighting conditions and image acquisition.

The following steps describe the image processing using MATLAB R2024b software for calculating the upper weld edge.

Step One: Initially, the image underwent morphological expansion using the ‘imopen’ function. Morphological expansion of the image allowed for the selective removal of small components from the image, ensuring that no spurious elements interfere with the primary features of the image.

Step Two: In this stage, image contrast was enhanced using MATLAB software’s ‘imadjust’ command to reduce image blur. Additionally, a 2D filter was predefined using the ‘fspecial’ function, and an unsharp mask, as per Equation (10), was applied to enhance image clarity.

$${\displaystyle \frac{1}{\left(\propto +1\right)}}\left[-\alpha \alpha -1-\alpha \alpha -1 \alpha +5 \alpha -1-\alpha \alpha -1-\alpha \right]$$

(10)

Step Three: In this stage, the image’s edges were sharpened radially, resulting in a more distinct contrast between the weld edges and the image background, facilitating the extraction of weld edges. The resulting edge-sharpened image was then converted into a binary image. Subsequently, the binary image underwent morphological erosion, followed by linear dilation at a 45-degree angle using the ‘imdilate’ function to connect regions within the image seamlessly.

Step Four: In this stage, any holes within objects in the binary image were filled using MATLAB software’s ‘imfill’ command. This step was carried out to make the welded regions more distinct and eliminate scattered small components around the image, with components containing fewer than 1000 pixels being removed.

Step Five: A set of connected points was labeled using MATLAB software’s ‘bwlabel’ function during this stage. Noise and unrelated points to the weld, such as markings on the workpiece, were removed. Additionally, two yellow-colored patterns previously marked around the welding area were considered for removal, resulting in the extraction of the welding pattern. The Sobel edge detection algorithm was then applied to detect the edges of objects in the image. Figure 8 illustrates the flowchart of the steps mentioned above.

In the present study, an experimental design was employed, and a complete factorial approach was utilized to investigate the effects of welding voltage parameters (V), wire feed speed (mm/min), nozzle travel speed (mm/min), and nozzle angle (°) on the smoothness of the weld bead’s top surface. Image processing and Wavelet transform techniques were utilized for analysis. For each input parameter, three levels were considered, resulting in 81 experiments.

Table 2. Levels of experimental design parameters.

Parameter	Level 1	Level 2	Level 3
Voltage (V)	17	22	27
Wire feed speed (cm/min)	210	231	252
Nozzle travel speed (mm/min)	200	250	315
Nozzle angle (°)	75	90	105

below presents the levels of these parameters.

Automated control of motion in three axes is crucial for producing intricate components. In pursuit of this goal, this research endeavor involved the construction of a three-axis table. This specialized table was equipped with ball screws, rails, and carriages, offering a generous range of 80 cm for the x-axis and y- axis and 70 cm for the z-axis. The motive force behind these axes was a robust 35 KW stepper motor with a 1.8° motion. Figure 9 provides an overview of the CNC system employed for additive part manufacturing.

To generate toolpaths, converting the three-dimensional design into two-dimensional layers and subsequent definition of the toolpath within them are essential steps. The transformation of the three-dimensional design into two-dimensional layers was carried out using Catia software V5R18, as depicted in Figure 10, which outlines the procedural steps for creating a toolpath for fabricating a curved-blade component based on its three-dimensional model. These steps encompass the following:

Stage 1: Drawing upon experimental tests conducted to ascertain the dimensions of the weld bead geometry and armed with knowledge of the height of each layer created by GMAW, it becomes imperative to partition the desired geometry into evenly sized planes equal to the height of each layer. The central planes within the part are delineated in white.

Stage 2: In the following stage, determining the plane intersections is imperative to derive the cross-sections of the desired layers.

Stage 3: The generated cross-section should be transposed to a new component, with squares of precise dimensions established around this section to preclude any alterations in subsequent stages or other software.

Stage 4: The preservation of the cross-section of the body as a DXF file entails the input of the section into the drafting section of Catia software. To accomplish this, the cross-section must be endowed with volume.

Stage 5: In this phase, the voluminous cross-section is dispatched to the drafting section, and by selecting the front view, the cross-sectional drawing is generated. This drawing can subsequently be stored in DXF format.

Stage 6: The saved DXF file is accessed through AutoCAD software (2016 version), whereby the lines of each layer are meticulously extracted. Finally, the programmed file is executed. Initially, the machining speed or nozzle movement speed must be determined. Subsequently, the initiation and termination points of the toolpath are selected, and the filename housing the CNC machine input code is designated for storage. Toolpath codes are composed in a file based on G01, G02, and similar commands.

The stages of the production of components through additive manufacturing utilizing a CNC machine are elucidated in Figure 11. The workflow commenced with a three-dimensional model and culminated in the creation of a three-dimensional body. As previously elucidated, the three-dimensional model transformed two-dimensional layers using Catia software. Subsequently, the deposition path was devised for each layer in AutoCAD software. The deposition path and pertinent welding parameters were converted into codes for integration into the CNC system. The desired shape was ultimately manufactured, with machining and post-processing conducted as necessary.

4. Results and Discussion

In producing components using layer-by-layer methods, the uniformity of the created layers plays a crucial role in determining the precision of the final piece. Any deviation from uniformity can result in distortion or dimensional changes in the component. Therefore, this study investigates the uniformity of weld bead height, employing the effective parameters of the GMAW process.

The research explores the impacts of welding voltage, wire feed speed, nozzle travel speed, and nozzle angle, considering three levels for each parameter through a factorial experimental design. Subsequently, the smoothness of the weld bead top line is analyzed using image processing and wavelet transform techniques. The aim is to identify the optimal parameters that yield a desirable smoothness in the weld bead top line for producing components using additive manufacturing via arc welding with shielding gas.

To attain a uniform height for the weld bead line as per the experimental design, 81 experiments were conducted. Figure 12 shows the weld beads according to the experimental design.

Following this, the algorithm described in the second part of this article was used in connection with image processing to extract the lines indicating the uniformity of the height of the weld bead. The algorithm described in the third section of this article was employed in conjunction with image processing to extract the lines representing the weld bead’s height uniformity. Figure 13 illustrates the stages from the weld bead image to the top-line image. Given that the objective is to obtain an image of the top edge of the weld for uniformity assessment, the camera angle and the specific line selected as the weld bead top line are not of paramount importance; only the parameter of line smoothness is under consideration.

Following the extraction of weld line coordinates using image processing techniques, the weld line is transformed into a one-dimensional waveform through one-dimensional wavelet transform, decomposing it into details and approximations. Utilizing these details allows for distinguishing between a smooth and non-smooth weld bead. Based on the results obtained from image processing and wavelet transform, the smoothest weld bead, or, in other words, the weld with the most uniform height, is selected for the production of components.

To evaluate the effectiveness of the proposed method, a specific geometric component, in this case a turbine blade geometry, was chosen. This particular component featured complex geometry, necessitating the fabrication of a three -axis CNC table. The precision table enabled movements with an accuracy of 0.05 units in the z, y, and x directions, facilitating the manufacturing process.

The selected model consisted of multiple layers, making it suitable for assessing the performance of the proposed method. Based on the weld bead height, it was determined that 63 layers were needed to construct the component. A Lisp-based program within the AutoCAD software environment generated the required motion codes for creating each layer. It should be noted that the thickness between the layers was obtained by creating a wall of the best welding beads according to the experiments.

Welding was conducted according to specific parameters, including a voltage of 27 V, a wire travel speed of 252 cm/min, a nozzle angle of 90 degrees, and a nozzle travel speed of 315 mm per minute, based on the results of the conducted experiments. These parameters yielded the smoothest weld bead line. Welding was executed along predetermined paths, resulting in a fabricated component that exhibited satisfactory performance based on the proposed method.

The model was reprinted using this method to conduct a more detailed examination and re-evaluation. An analysis of the heights of the produced components confirmed the precision of this method in component fabrication. The data collected for several points in this model are presented in Figure 14 and

Table 3. Height of points measured for several points in experiments.

Experiment Number	Height (mm)
	h1	Error (%)	h2	Error (%)	h3	Error (%)	h4	Error (%)
Model	240	-	240	-	240	-	240	-
Experiment 1	236	1.66	235	2.08	234	2.50	236	1.66
Experiment 2	238	0.83	236	1.66	237	1.25	237	1.25

. As a second example illustrating the effectiveness of this method, a polishing process lasting nearly one hour was employed to produce a final usable component. Figure 15 shows the height values for the considered points of the obtained model and the first and second experimental tests.

Figure 16 depicts the two manufactured components. As evident in Figure 16B, the uniformity in the height of the weld bead ensured that the final layer’s uniformity was maintained after stacking 63 layers on top of each other.

5. Conclusions

Today, the use of additive manufacturing techniques for producing metal components has found a progressively diverse range of applications. Among these techniques, one that stands out for its exceptional ability to fabricate components swiftly and cost-effectively is WAAM. This research delves into one of the primary challenges associated with WAAM, which is the uniformity of weld bead height, and seeks to investigate the influential parameters affecting this geometric parameter comprehensively. A total of 81 experiments were meticulously conducted, encompassing varying levels of welding speed, nozzle angle, wire speed, and voltage parameters.

To accurately ascertain the weld line’s top edge, sophisticated image processing techniques were harnessed. These techniques enabled the precise determination of the weld line’s top edge, representing the uniformity of weld bead height. This study scrutinized the acquired line profiles by applying wavelet transform. This analytical approach allowed for a thorough assessment of the uniformity of weld bead height. Key findings regarding height uniformity according to experimental results are listed as follows:

• The smoothest welding line was obtained at the highest voltage value in the selected range.
• The most uniform welding was achieved by setting the nozzle in a 90° position to the welding path.
• Increasing the wire feed rate improved the quality of the welding line in terms of height uniformity.

This rigorous experimentation and analysis revealed an optimal configuration for manufacturing a sample component. This configuration proved instrumental in producing a component that met satisfactory quality standards. Furthermore, significant emphasis was placed on assessing the repeatability of the component production process, ensuring consistency and reliability in the manufacturing process.