At present, 3D printing techniques are thriving. They have been extensively used to produce cultural and creative individualized goods, according to Bogue [1], and products with complex shapes that traditional production technologies cannot achieve. However, the study is still restricted to small quantities of customized products. 3D printing parameters include the printing material, fill thickness, wall thickness, temperature, and speed. Chen et al. [2] discussed the effect of material infill density, fill thickness, extruder head temperature, and printing speed parameters on the tensile strength of a product. Sanatgar et al. [3] printed PolyAmide (PA) fibre (nylon) and PolyLactic Acid (PLA) materials on a PA fabric surface. They indicated the impact of platform temperature, extruder head temperature, and printing speed on the adhesion force. However, the studies above only considered single-quality characteristics and did not consider all of the parameters of 3D printing that may limit the print quality and stability.

Related documents have discussed the correlation among specification stipulations, cost, and pricing practices in automobile, computer, and medical biotechnology product development and manufacturing processes [4– 6]. There are no current product specifications for 3D printing manufacturing technology or standard specifications for judging product quality or identifying products that cannot meet requirements in the operational phase. In addition, previous research mostly discussed the application of 3D printing to the cultural and creative, medical treatment, and food areas. However, although the electronics manufacturing industry has the highest output value in the present economic development structure, few scholars have used 3D printing in the electronics manufacturing industry.

This study aimed to extend the 3D printing application area into the electronics manufacturing industry, taking production equipment components as examples to expand the application benefit of 3D printing techniques. The purposes are described below:

I. The tolerance determination of the Taguchi quality loss method was used to create the product specifications of 3D printing manufacturing technology for components of the electronics manufacturing industry so that products could meet requirements in the operational phase.

II. The effect of material selection and printing parameters on product quality characteristics was discussed using a Taguchi parameter design experiment, and the Principal Components Analysis (PCA) method was developed to measure multiple quality characteristic indicators. An optimum parameter level combination was proposed to confirm that the products would meet the specification above requirement.

III. Finally, the printing cost was discussed, and the benefit of using 3D printing techniques to produce the component was evaluated.

This study used a 3D printing technique to produce tools used for electronics manufacturing. First, the uses and characteristics of the components in the case were analyzed to sort out the overall external structure and assembly and determine the essential quality characteristics of the printed end product. Secondly, the tolerance determination of the Taguchi quality loss method was used to calculate the production specifications of various quality characteristics. The Taguchi experimental design was used to optimize the printing material and printing parameters. Multiple quality characteristics were considered in this study. Therefore, the PCA method was combined with various measurement indexes to propose the optimum printing parameter combination. Afterwards, the confirmation experiment was conducted, i.e., the optimum parameter combination was used for printing to judge whether the printing result was in the Confidence Interval (CI), determine the additivity of various factorial effects, and whether the print quality conformed with the tolerance specifications. Without error, the printed parts would be assembled and mounted on the machine for testing, ensuring stable operation. Finally, the benefit was assessed.

The clip body of the parts shall be assembled with a spring and movable plate after printing. Four key quality characteristics were established, as described below.

I. Inside diameter of the left and right holes (Fig. 5. a): the left and right holes of the clip body hold the spring, and the elastic force causes the clip body to clamp onto the capacitor pins. The spring cannot be inserted into the hole if the inside diameter is too small, and the spring will be loosened if the inside diameter is too large.

II. Outside diameter of the locating point (Fig. 5. b): the front and back sides of the clip body have two locating points placed in the movable plate's circular holes fixed to the clip body. Meanwhile, the movable plate can move side to side so that the capacitor pin can enter the sharp middle corner of the body and be clamped. The movable plate cannot be mounted if the outside diameter of the locating point is too large, and the movable plate will be loosened if the outside diameter of the locating point is too small.

III. Height of locating point (Fig. 5.c): the movable plate cannot be fastened if it is too low; if it is too high and exceeds the body plane, there will be interference with other components, and the inserter will fail.

IV. Middle recess thickness (Fig. 5.d): the middle recess of the clip body holds the movable plate, which is fixed by the locating above point. The movable plate cannot be mounted if the middle recess is too thick, and the movable plate will be loosened if the middle recess is too thin.

a. b.

c. d.

outside diameter of locating point; c. height of locating point; d. intermediate thickness.

To begin with, this study evaluated two printing bottom contact surfaces, the parts clip body backside down and body side down, to contact the printing platform, as shown in Fig. 6. a and Fig. 6. b In addition, if the printed object feature had a hanging design, the support structure would be built with the slicing software to stabilize the object during printing to prevent it from sagging or dropping. There were two scenarios: one with support and one without support. Afterwards, when printing the object's bottom layer, a peripheral frame was built to enhance the bottom strength and the bendability between the print object and the printing platform. There were two scenarios with and without a frame. The factor combinations considered in the pre-experiment are shown in Table 2.

a. b.

Bottom contact surface Back side down Side down Support structure No Yes Peripheral frame No Yes

The results indicated that if the parts clip body were printed using combination 2, the print object body would be damaged when removing the support material and the peripheral frame. In addition, during side-down printing, if the body locating point was too low, there would be a dimensional error, and the walls of the left and right holes would be incomplete, as shown in Fig. 7. Therefore, the subsequent printing experiment used the conditions of combination 1, i.e., a bottom contact surface with a back side down, no support structure, and no peripheral frame.

This experiment discussed the material characteristics and printing parameters that could influence the quality described above characteristics. The experimental control factors and levels are shown in Table 3, as described below:

I. Material: FDM process type 3D printing usually uses polylactide PLA material and Acrylonitrile Butadiene Styrene (ABS) resin. ABS resin material is stronger than PLA material but releases toxic gases during heating. High-Impact PolyStyrene (HIPS) and ABS resin materials have similar mechanical strengths. Therefore, this study evaluated PLA and HIPS materials.

II. Fill thickness: fill thickness refers to the thickness of each material layer used when printing the object. The smaller the fill thickness is, the finer the printed object surface will be; however, the process takes time. This study considered fill thicknesses of 0.16mm and 0.3mm.

III. Wall thickness: wall thickness refers to the thickness of the outer shell material while printing the object. It is generally set as a multiple of the nozzle bore. A thicker object wall contributes to enhancing the object surface's firmness. The nozzle bore was 0.4mm in this study, and the printed wall thicknesses were 0.8mm and 1.2mm.

IV. Bottom/top thickness: bottom/top thickness refers to the thickness of the topmost and bottommost stacked material. It is generally set as a multiple of the fill thickness. This study considered bottom/top thicknesses of 0.4mm and 1.2mm.

V. Infill density refers to the solid ratio of the material filling the print object. High infill density enhances the object's internal strength, but the process takes time and consumes material. An infill density higher than 20% approximates the entity. This study considered infill densities of 20% and 50%.

VI. Printing speed: slow printing may enhance the object's surface fineness, but the process takes time. This study considered printing speeds of 20 mm/s and 30 mm/s.

VII. Extruder head temperature: the extruder head temperature must be appropriate for the melting point of the printing material. If the temperature is too low, the material cannot adhere to the printing platform smoothly. The material's shaping quality will be affected if the temperature is too high. The material manufacturer suggested that the appropriate extruder head temperature ranges for HIPS and PLA material are 190~230℃ and 195~250℃, respectively. This study considered extruder head temperatures of 195℃ and 220℃.

This study built an L8 orthogonal array for the experiment. The experimental configuration is shown in Table 3, including the seven control above factors with two levels. Each combination was printed twice, and two observations were made. The experimental order was random.

The experimental results are shown in Table 3. As the four quality characteristics were the nominal-the-best characteristic of Taguchi quality, the Signal-to-Noise ratio (SN) of various parameter combinations could be calculated by Eq. (4), in which y ̅ is the average of the observations, and s is the standard deviation of the observations. The response table and factor effect diagram of various quality characteristics is shown in Table 4 and Fig. 8, respectively, to show the influence of multiple factors on the quality characteristic (SN).

η = 10 × log10 ( ). (4)

NTB 2

As the ideal parameter combinations of the four quality characteristics were inconsistent, the mutually independent linear combinations were transformed for the probability correlated quality characteristics using Principal Component Analysis (PCA) to determine the optimum parameter combination [13, 14]. This approach aligns with established methodologies where PCA simplifies the complexity in multivariate data by reducing its dimensions, thus facilitating a more straightforward identification of the optimal settings [15]. PCA's effectiveness in optimizing process parameters is further supported by its ability to handle data intercorrelations that may affect the output quality [16], thus increasing interpretability and preserving the essential characteristics of the dataset [17]. The technique has been detailed in various studies, showcasing its utility in enhancing data analysis outcomes across different research fields [18]. It is renowned for its capacity to analyze data tables where observations are described by several inter-correlated quantitative variables [17]. Furthermore, PCA is instrumental in reducing the number of space dimensions while retaining the variance present in the dataset, which is crucial for efficient data analysis [19]. The analysis steps are described below. Step 1: the original SN ratio matrix is standardized M quality characteristics evaluate l parameter combinations. The original matrix D is built, where SN1, SN2, SN3, and SN4 are the inside diameter of the left and right holes, the outside diameter of the locating point, the height of the locating point, and the middle recess thickness, respectively. The results are shown in Table 5. The SN ratio of various quality characteristics is standardized by Eq. (5), and the matrix [Sij] is obtained, i.e., the standardized value of the SN ratio of the No. j quality characteristic in the No. i parameter combination, as shown in Table 6.

, (i = 1,2, … , l; j = 1,2, … , m) (5) Step 2: the matrix of the correlation coefficient, eigenvalue, coefficient of determination and eigenvector are calculated.

The correlation coefficient of the standardized SN ratio is calculated by Eq. (6), and the correlation matrix [R] is obtained, as shown in Table 7. The matrix of the correlation coefficient, eigenvalue and eigenvector relation are expressed as Eq. (7), and the matrix of the correlation coefficient, eigenvalue and eigenvector can be obtained. The coefficient of determination of various principal components is obtained by Eq. (8), and the results are shown in Table 8 and Table 9.

∑ (S − S )(S ′ − S )

i=1 ij j ij j′

[γ ′] = .

jj (6)

√ l l ∑ (S − S ) ∑ (S ′ − S )

i=1 ij j i=1 ij j′

Where is the coefficient of the correlation between the No. j quality characteristic and the No. j' quality

characteristic, is the average of the standardized SN ratios of the experimental groups of the No. j quality characteristic, and is the average of the standardized SN ratios of the experimental groups of the No. j' quality characteristic.

Where is the eigenvalue of the No. p principal component, is the m×m unit matrix, is the eigenvector corresponding to of the No. p principal component, is the coefficient of determination of the No. p principal component, and T is the total eigenvalue of various principal components.

Step 3: Calculate the principal component score and multiple performance characteristic index (MPCI)

The principal component scores and MPCI Ωi are calculated by Eq. (9) and Eq. (10), respectively, where is the score of the No. p principal component in the No. i experimental group, and is the eigenvector corresponding to the No. p principal component. The results are shown in Table 10.

Ωi = Cp × Φip. (10)

Step 4: determine the factor level combination of the optimum parameters The total score response table of various parameters at different levels is calculated. The results indicated the material should be PLA, a fill thickness of 0.16mm, a wall thickness of 0.8mm, a bottom/top thickness of 1.2mm, an infill density of 50%, a printing speed of 20mm/s, and an extruder head temperature of 220℃. The optimum printing parameter combination was A1 B2 C2 D1 E1 F2 G2, as shown in Table 11.

Huang et al |Manag. Anal. Soc. Insights. 1(2) (2024) 154-170

First, the SN ratio under the optimum printing parameter combination was predicted using Eq. (11), in which SN is the overall average SN of all experimental groups and A1, B2, C2, D1, E1, F2, G2 are the average SN ratio of various control factors under the optimum level condition, respectively. The 95% CI of the predicted SN ratio of various response values was calculated using Eq. (12). The results are shown in Table 12. The measurement results of different quality characteristics fell into the CI, meaning the additive model of relevancy of various control factors was tenable, and there was no significant interaction among parameters.

SN = T ̅ + (̅ A̅̅ − T ̅) + (̅ B̅̅ − T ̅) + (̅ C̅̅ − T ̅) + (D ̅̅̅̅ − T ̅) + (̅ E̅̅ − T ̅) + (̅ F̅̅ − T ̅) + (̅ G̅̅ − T ̅).

1 2 2 1 1 2 2

CI = √F × V × [ + ]. (11)

α;1;V2 e

neff r

n = . (12)

eff ∗

1 + Df Where α is the significance level (α=0.05), v is the DOF of the pooled error variance, Ve is the pooled error variance, neff is the number of effective observations, L is the total instances of processing (L=8), Df* is used to estimate the total DOF of factors of the mean value, and r is the sample number for the confirmation test (r=4).

The inside diameter of the left and right holes 46.894 43.024~50.764 44.519 Outside diameter of locating point 49.903 44.832~54.973 47.424 Height of locating point 40.076 34.401~45.751 36.686 Middle recess thickness 39.52 29.724~49.316 33.559

The size feature of the clip body 3D model built by the drawing software was the target value, and whether the size feature of the four quality characteristics of the printed samples in the confirmation test was within the specification tolerance was judged. The results are shown in Table 13. The four quality characteristics of the printed samples were within the specification tolerance.

This study used a 3D printing technique to produce components for the electronics manufacturing industry to evaluate the benefits. The cost of production was estimated, including the equipment cost, labour cost, and material cost. The equipment cost consisted of the purchase cost of the 3D printing machine and the depreciation and amortization cost during production. The labour cost included creating the 3D model file, operating the printing machine object, the subsequent surface treatment, and component assembly. The material cost covered the materials for the print object and the support structure. The benefit of using 3D printing to produce components for the electronics manufacturing industry is expressed as Eq. 13, in which Vp is the component value (price), Tp is the total production hours, tp is the time needed to print the component, C3D is the acquisition cost of the 3D printing machine, rd is the depreciation rate per unit time of the 3D printing machine, ∑(tp×ro) is the sum of the products of the component printing, subsequent surface treatment and assembly time and the corresponding unit salary of the operators, Cm is the material cost for printing the component, tm is the time needed for component size measurement and modelling, and re is the salary per unit time of the modelling engineer.

Tp Tp

Benefit=(Vp × ) − {[(C3D × rd × tp) + ∑(tp × ro) + Cm] × + (tm × re)}. (13)

tp tp

C.-Y. Huang: Methodology, Software, Validation, Formal analysis, and investigation. Y.-W. Chung: Writing, creating the initial design, conceptualization, and editing.

All authors have read and agreed to the published version of the manuscript.

All the data are available in this paper.

The authors declare no conflict of interest.