Application of the Taguchi approach to improve performance of the alignment-layer printing process in liquid crystal displays

Abstract

The uniformity of the alignment-layer thickness on indium tin oxide glass plays an important role in producing liquid crystal display panels of quality images. However, thickness non-uniformity results in producing liquid crystal display panel of poor color image, which increases rework cost and scrap. The Taguchi method is utilized in the define–measure–analyze–improve–control approach to improve the capability of this process. In the define phase, process mapping and quality characteristic definition are discussed. The x̄ and R control charts and process capability analysis are employed in the measure phase. The L₁₈ array and signal-to-noise ratio, followed by analysis of variance, are utilized in the analyze phase. In the improve phase, two-step optimization is conducted to reduce thickness variations and adjust process mean on target. The control charts for the thickness are regularly constructed in the control phase. Initially, the estimates of the capability indices for the alignment-layer printing process are 0.731 and 0.69, respectively, which indicate that this process is incapable. Utilizing the Taguchi method, confirmation experiments showed that the potential and actual capability indices are respectively improved to 3.48 and 3.47. Hence, the process becomes highly capable. This proves the effectiveness of the Taguchi method in improving the performance of the alignment-layer printing process.

Keywords

Define–measure–analyze–improve–control Taguchi method process capability control charts

Introduction

In an information-dominated age, the display of information is becoming increasingly essential in many aspects of daily life. Owing to several important characteristics, such as lightweight, low radiation, and low power consumption, the liquid crystal display (LCD) has become one of the most important systems for high performance information displays.¹ The LCD panel has created a new reality in the electronic information displays industry; such as, television, mobile phones, laptop computers, digital cameras, and other consumer-related communications hardware.

Basically, the LCD consists of a thin-layer liquid crystal sandwiched between a pair of polarizers. To control the optical transmission of the display element (pixel; picture element), the liquid crystal layer is placed between transparent electrodes, which are usually indium tin oxide (ITO) owing to its higher conductivity. In a liquid crystal cell without any external field, the ordering of the molecules is usually determined by the anisotropy of the boundary. Thus, the surfaces of the electrodes or ITO are usually coated with an alignment layer (e.g. polyimide) evenly. The uniformity of the alignment-layer thickness on ITO glass plays an important role in producing panel displays of quality images. In practice, thickness non-uniformity results in color aberration, causes difficulties in assembly process, and increases rework cost scrap. As a result, manufacturing costs of the LCD panel are increased. Therefore, in this research, process engineers aim at improving the capability of the alignment-layer printing process.

Six sigma² is a project-driven management approach to improve the organization’s products, services, and processes by continually reducing defects in the organization. Six sigma focuses on improving customer requirements’ understanding, business profitability, and the effectiveness and efficiency of all operations to meet or exceed customer’s needs and expectation.³ A widely accepted six sigma approach is the define–measure–analyze–improve–control (DMAIC) approach,⁴ which has been adopted to improve the performance and quality for numerous manufacturing processes.^5,6 The DMAIC approach adopts several tools for quality improvement,⁷ including process mapping in the define phase, the capability analysis in measure phase, analysis of variance (ANOVA) in the analyze phase, the robust design in the improve phase, and the control charts in the control phase. This research, therefore, employs the DMAIC approach to enhance the performance of the alignment-layer printing process, and then summarizes.

The define phase

This research is conducted on the cell process for the thin film transistor LCD (TFT-LCD) panel. The TFT-LCD panel is simply composed as shown in Figure 1(a) of two glass substrates sandwiching a layer of liquid crystal. The color filter gives each pixel its own color and is fitted with the front glass substrate. The transistors are fabricated on the back glass substrate. A light source is located at the back of the LCD panel. When a voltage is applied to the transistor, the liquid crystal is bent allowing light to pass through to form a pixel. Typically, a TFT-array substrate is composed, as seen in Figure 1(b), of a matrix of pixels and ITO region each with a TFT switching device, which functions to turn each individual pixel on or off. To create the display image, thousands or millions of pixels are needed.

Figure 1.

Illustration of the LCD structure. (a) The basic structure of LCD. (b) The sketch of TFT substrate.

Two tasks are conducted in the define phase.

Mapping the alignment-layer printing process.

Defining the key quality characteristic (QCH).

Mapping the alignment-layer printing process

The production line of TFT-LCD panels is depicted in Figure 2. As seen in this figure, three main processes are performed in this line; i.e. the array, cell, and module processes. The array process involves cleaning glass substrates, photo-resist coating, and exposure and developing, repeated several times. In the cell process, which is a special step in the LCD manufacturing, the color filter and TFT-array substrate pass through cleaning, alignment-layer printing, and rubbing. Finally, in the module process, the color filter and TFT panel are aligned, combined and fixed together using ultraviolet (UV)-hardened polymer spots. The liquid crystal is then injected between the two ITO layers and the LCD product is completed by adding the back-light unit, intergrated circuit (IC), and other components.

Figure 2.

The main processes of LCD line.

Research on the TFT-LCD panel manufacturing process is comparatively incomplete. Lin et al.⁸ proposed an analytical framework for TFT-LCD production chain planning and scheduling. Ishihara⁹ discussed some mechanisms for the liquid crystal alignment and reviewed various alignment techniques and methods to investigate the interaction of the liquid crystal molecules with substrates. Chen et al.¹⁰ improved the photo-resist coating, which is the main process in the array process, of the TFT-LCD panel using a capability assessment model and the measure–analyze–improve–control approach. However, this research investigates the cell process and mainly focuses on the alignment-layer printing process.

The alignment-layer printing process is displayed in Figure 2, which can be described as follows. A doctor blade distributes a layer of the alignment-layer liquid (polyimide) on a carry-roll, which is then transferred to the printing roll. The printing roll transfers a desired pattern of photosensitive resin (PR) plate gravures filled with a volume of polyimide liquid on ITO glass evenly. The three-dimensional (3D) and two-dimensional (2D) images for the surface of PR plate taken using the classic measuring system are depicted in Figure 3(a) and (b), respectively.

Figure 3.

The 3D and 2D images for PR plate.

In Figure 3, the main problem is summarized by the excessive variations of the PR plate; i.e. the variations contributed by the diameter (d), height (h), and taper angle (θ), which result in non-uniformity of the alignment-layer thickness. In practice, the non-uniformity of the alignment-layer thickness results in color aberrations in the LCD’s image; i.e. color saturation, shades, brightness, and contrast. In addition, it causes difficulties in the assembly process. As a result, the main objective of this research is to enhance the capability of the alignment-layer printing process by reducing the variations contributed by d, h, and θ.

Identifying the critical QCH

For the alignment-layer printing process, the key measurable and continuous QCH is the alignment-layer thickness, which is the nominal-the-best type. The desired target (T) of thickness is 700 Å, while the upper and lower specification limits, USL and LSL, are 800 Å and 600 Å, respectively. Practically, a layer thickness above the USL results in insufficient LCD color saturation, or pixel brightness, while below the LSL limit causes oversaturated color, or pixel darkness. As thousands of pixels create the LCD image, color shades may occur. Such quality defects, in reality, cannot be reworked and, hence, the LCD would be scrapped.

The measure phase

In this phase, the $\bar{x}$ and R control charts, followed by the process capability analysis, are adopted to access the performance of the alignment-layer printing process at the initial stage.

The x̄ and R control charts

The $\bar{x}$ and R control charts have been widely applied for online monitoring of the mean and variability for many manufacturing processes. To construct these two control charts, two ITO glass samples, each of size 300 mm × 400 mm, are randomly selected from the daily production for eight working days. The alignment-layer thickness is then measured by the contour measurement instrument at five different places on the surface of ITO glass; i.e. left, right, upper, lower, and middle. Sixteen samples, each with a sample size of five thickness measurements, would result in 80 observations. Using the collected data, the $\bar{x}$ and R control charts are constructed and depicted in Figure 4.

Figure 4.

Control charts at initial and improvement stages.

The lower control limit (LCL), center line (CL), and upper control limit (UCL) for the $\bar{x}$ chart are calculated as 633.9, 696.0, and 758.1 Å, respectively; while for the R chart are estimated as 0.0, 107.6, and 227.6 Å, respectively. In Figure 4, it is observed that there are no points detected outside the control limits and no patterns or runs are observed within the control limits of the $\bar{x}$ and R control charts. Hence, the alignment-layer printing process is concluded in the control at the initial stage.

Process capability analysis

Process capability analysis is a vital part of an overall quality-improvement program, by which the capability of a manufacturing process can be measured and assessed. The potential and actual process capability indices, $C_{p}$ and $C_{pk}$ , respectively, are the cornerstones of the process capability analysis and the most popular of all the indices. The $C_{p}$ is defined as¹¹

C_{p} = \frac{USL - LSL}{6 σ}

(1)

In practice, the process standard deviation (σ) is unknown and frequently estimated by

\overset{⌢}{σ} = \bar{R} / d_{2}

(2)

where $\bar{R}$ is the CL value of the R chart and d₂ is a constant related to sample size. The estimator of $C_{p}$ , ${\overset{⌢}{C}}_{p}$ , is obtained by replacing σ̑ instead of σ in equation (1). Contrary to $C_{p}$ , the $C_{pk}$ attempts to take the target (T) into account.¹² The estimator of $C_{pk}$ , ${C ⌢}_{pk}$ , is calculated as

{\overset{⌢}{C}}_{p k} = \min {\frac{U S L - \bar{\bar{x}}}{3 \overset{⌢}{σ}}, \frac{\bar{\bar{x}} - L S L}{3 \overset{⌢}{σ}}}

(3)

where $\bar{\bar{x}}$ is the process mean estimated by CL of the $\bar{x}$ chart. Juran and Gryna¹³ recommend a minimum $C_{p}$ value of 1.33 for an ongoing process to indicate that a process has the potential to produce within specification and uses only 75% of the specification range if the process is perfectly centered at the specification midpoint. Similarly, a value of 1.33 is considered the standard minimum boundary for $C_{pk}$ (Kotz and Lovelace)¹⁴ to conclude that a process is actually performing at a level where at least 75% of the specification range is being used. It can be said that the $C_{p}$ provides an upper bound on the $C_{pk}$ value, while, when the process is perfectly centered at the specification midpoint, the values for $C_{p}$ and $C_{pk}$ are the same. Evidently, the $C_{p}$ and $C_{pk}$ , when used together, provide a very good indication of how well a process is now performing and how well a process could perform if it was centered on target. Therefore, both $C_{p}$ and $C_{pk}$ will be adopted for assessing the capability of the alignment-layer printing process, where this process will be judged as a capable process if the values of both indices exceed 1.33.

In Figure 4, the $\bar{R}$ value is 107.6 Å. The d₂ value for a sample size of five thickness measurements is 2.326. Substituting these two values in equation (2), the σ̑ value is calculated as 46.26. The C̑_p value is then equal to 0.731. To estimate the C̑_pk, the $\bar{\bar{x}}$ value, as shown Figure 4, is equal to 696 Å. Using equation (3), the C̑_pk is calculated as 0.69. Since the C̑_p and C̑_pk values are both far below l.33, it is concluded that the alignment-layer printing process is incapable for providing thickness variations within the allowable specification range at the initial stage.

The analyze phase and improve phase

The Taguchi method^15–16 is widely used in quality engineering for analyzing and improving quality and yield. The main goal of this method is finding the optimal setting of control factors involved in a system to make its performance insensitive to noise, and thus it has been utilized in many business applications; such as, chip-on-board technology,¹⁷ surface mounting technology,¹⁸ and the SUS304 wire drawing process.¹⁹ The Taguchi method is utilized in the analyze phase and improve phase as described in the following subsections.

The analyze phase

Three tasks are carried out for analyzing the experimental data of the alignment-layer thickness. These tasks are described as follows.

The orthogonal array experiments

The cause and effect diagram for thickness non-uniformity is shown in Figure 5. Notice that the variations contributed by the diameter, height, and taper angle of the PR plate are the main causes for thickness non-uniformity. These variations can be controlled by four cell process factors, including: arrangement method of the alignment-layer molecules on the ITO surface (x₁), molecule diameter (x₂), UV exposure energy (x₃), and UV exposure time (x₄).

Figure 5.

Cause and effect diagram for thickness non-uniformity.

Based on process knowledge, factor A is assigned at the two levels as illustrated in Figure 6, whereas factors B to D are assigned each at three levels. The level values for these factors are displayed in Table 1. The alignment-layer printing process is performed at the initial factor settings A₁B₁C₁D₁.

Figure 6.

Illustration of the arrangement method.

Table 1.

The control factors and their levels.

Process factor	Factor level^*
	Level 1	Level 2	Level 3
A. Arrangement method	Square	Triangular
B. Molecule diameter (µm)	35	40	45
C. Exposure energy (mW/cm²)	35	40	45
D. Exposure time (seconds)	100	110	120

Initial levels are in bold.

An efficient way to study the effect of several control factors simultaneously is to plan matrix experiments using orthogonal arrays (OAs). The appropriate OA to investigate a two-level factor and three three-level factors is the L₁₈ (2¹ × 3⁷) array shown in Table 2. In this array, factor A is assigned to column 1; since it has two levels. Factors B, C, and D are assigned to the next three three-level columns, respectively. The other four columns are left empty for evaluating error term.

Table 2.

The experimental results using L₁₈ array.

Experiment number	Factor^*								Average QCH	Standard deviation	S/N ratio
	A	B	C	D	e	e	e	e	${\bar{y}}_{i}$ (Å)	s_i	η_i (dB)
1	1	1	1	1	1	1	1	1	713.40	54.76	22.30
2	1	1	2	2	2	2	2	2	744.10	30.56	27.73
3	1	1	3	3	3	3	3	3	708.60	30.20	27.41
4	1	2	1	1	2	2	3	3	687.00	51.19	22.55
5	1	2	2	2	3	3	1	1	724.40	33.13	26.80
6	1	2	3	3	1	1	2	2	694.70	35.38	25.86
7	1	3	1	2	1	3	2	3	663.40	42.24	23.92
8	1	3	2	3	2	1	3	1	661.75	36.25	25.23
9	1	3	3	1	3	2	1	2	653.15	42.41	23.75
10	2	1	1	3	3	2	2	1	713.95	23.44	29.68
11	2	1	2	1	1	3	3	2	726.50	10.89	36.48
12	2	1	3	2	2	1	1	3	719.70	16.69	32.69
13	2	2	1	2	3	1	3	2	676.50	13.61	33.93
14	2	2	2	3	1	2	1	3	681.40	11.36	35.56
15	2	2	3	1	2	3	2	1	685.25	17.06	32.08
16	2	3	1	3	2	3	1	2	671.20	11.91	35.02
17	2	3	2	1	3	1	2	3	675.75	8.13	38.39
18	2	3	3	2	1	2	3	1	660.35	9.58	36.77

e denotes columns for error.

QCH: quality characteristic; S/N: signal-to-noise.

Preliminary experiments are conducted to ensure the measuring instruments are repeatable and capable. Next, the 18 experiments in the L₁₈ (2¹ × 3⁷) array are performed by varying the factor levels in a systematic manner. Four repetitions are conducted at the same factor levels of each experiment. The thickness is then measured at five different places on the surface of the ITO sample.

Signal-to-noise ratio analysis

The signal-to-noise (S/N) ratio (η) is always used in the Taguchi method to interpret experimental data and measure performance. For the nominal-the-best type QCH, η _i is calculated for each experiment i using

η_{I} = 10 lo g_{10} {(\frac{{\bar{y}}_{i}}{s_{i}})}^{2} i = 1, \dots, 18

(4)

where ${\bar{y}}_{i}$ and s_i are the mean and standard deviation of the measured observations for experiment i, respectively. For explanation, in Table 2 the calculated ${\bar{y}}_{1}$ and s₁ for the first experiment are 713.40 and 54.76, respectively. The corresponding S/N ratio ( $η_{1}$ ) is estimated as

η_{1} = 10 lo g_{10} {(\frac{713.40}{54.76})}^{2} = 22.30 dB

The $η_{i}$ values for the other 17 experiments are obtained in a similar manner. Further, ANOVA is usually conducted to determine the process factor with significant effects. In ANOVA, the sum of squares (SS_j) is contributed by factor j of the four factors. The percentage contribution (ρ _j ) by factor j in total sum of squares (SS_T) is then adopted to evaluate its importance on the quality characteristic of interest, which can be expressed as²⁰

ρ_{j} = S S_{j} / S S_{T} \times 100 %

(5)

The results of ANOVA for the S/N ratio are summarized in Table 3.

Table 3.

(a) ANOVA for S/N ratio.

Factor (j)	SS_j	df_j	ρ _j (%) (SS_j/SS_T)	MS_j (SS _j /df_j)	F ratio (MS_j/MS_e)	F(0.05, df_j, 10)	p value
A	401.96	1	81.43	401.96	99.86	4.96	0.000
B	4.78	2	0.97	2.39	0.59	4.10	0.570
C	43.31	2	8.78	21.66	5.38	4.10	0.026
D	3.29	2	0.67	1.65	0.41	4.10	0.410
Error	40.25	10		4.025
Total	493.60	17

Table 3.

(b) ANOVA for S/N ratio (after pooling insignificant factors).

Factor (j)	SS_j*	df_j	MS_j (SS _j /df_j)	F ratio (MS_j/V_e)	F(0.05, df_j, 14)	Decision	SS′_j	ρ′_j (%) (SS′_j/SS_T)
A	401.96	1	401.96	116.46	4.60	Significant	398.51	80.74
B	4.78	2	2.39
C	43.31	2	21.66	6.28	3.74	Significant	18.21	3.69
D	3.29	2	1.65
Pooled error	(48.32)	14	(3.45)					15.57
Total	493.60	17						100%

* Pooled sum of squares (SS) are identified in bold.

It is noticed in Table 3(a) that the ρ_A (= 81.43%) and ρ_C (= 8.78%) contribute more than 5% to total variation, whereas ρ_B (= 0.97%) and ρ_D (= 0.67%) contribute less than 5%. The F ratio is calculated for each factor then compared with the F value at 5% significance level. That is, the F ratio for factor A (= 99.87) is compared with F_0.05,1,10 (= 4.96), whereas the F ratios correspond to factors B to D are compared with F_0.05,1,10 (= 4.10). If the F ratio of a factor is larger than the F value at 5% significant level, this factor is judged significant; otherwise, it is considered insignificant. Based on F testing, the effects of factors A and C are considered significant. However, the effects of factors B and D are judged insignificant, and hence they will be pooled into the error term. The ANOVA results for the S/N ratio after pooling factors B and D are shown in Table 3(b). In this table, the degrees of freedom associated with pooled error are increased to 14, which is calculated as the degree of freedom of error (= 10), plus the degrees of freedom corresponding to pooled factors B and D. Consequently, the F ratios of factors A and C are calculated by dividing the mean square of each factor by the pooled error mean square (V_e). The F ratios are then compared with F_0.05,1,14 (= 4.60) and F_0.05,2,14 (= 3.74) for factors A and D, respectively. These factors are found still significant. Let ρ′_j denote the pure percentage contribution by factor j calculated as

{ρ'}_{j} = S {S'}_{j} / S S_{T} \times 100 %

(6)

where SS’_j is the pure sum of squares contributed by factor j with degrees of freedom (df_j), which can be estimated using

SS'_{j} = S S_{j} - d f_{j} \times Ve

(7)

For example, the ρ′_A is calculated as follows. The df_A and V_e are equal to 1.00 and 6.90, respectively. Substituting these values in equation (7), the SS’_A is then

SS'_{A} = 401.96 - 1 \times 3.54 = 398.51

Using equation (6), the ρ′_A is estimated as

{ρ'}_{A} = 398.51 / 493.60 \times 100 % = 80.74 %

The ρ′_A value indicates that 80.74% of the total variation is contributed by the arrangement method of the surface molecules (factor A). Therefore, factor A is considered the most influential factor on variation reduction. The ρ′_C (= 3.69%) is calculated in a similar manner.

The results of ANOVA for the S/N ratio are summarized in Table 4, where the ρ_A and ρ_D values are calculated 0.72% and 2.51%, respectively, which are less than 5%. While, the molecule’s diameter (factor B) has the largest contribution (ρ′_B = 78.62%) to the total variations followed by contribution of the exposure energy (factor C), ρ′_C, of 7.37%. The F ratios of factors B and C of 36.47 and 3.42, respectively, are compared with the F values at 5% significance level, F_0.05,2,13 (= 3.81). Obviously, only factor B significantly affects the thickness mean, while it does not affect the S/N ratio. Hence, this factor will be used to adjust process mean on target.

Table 4.

ANOVA for thickness.

Factor (j)	SS_j *	df_j	ρ _j (%) (SS_j/SS_T)	MS_j (SS _j /df_j)	F ratio (MS_j/V_e)	F(0.05, df_j, 13)	Decision
A	88.44	1	0.72	88.44
B	9675.15	2	78.62	4837.58	36.47	3.81	Significant
C	907.15	2	7.37	453.58	3.42	3.81	Insignificant
D	309.33	2	2.51	154.67
Error	1326.74	10
Pooled error	(1724.52)	13		(132.66)
Total	12306.82	17

Pooled sum of squares (SS) are identified in bold.

The improve phase

Taguchi’s two-step optimization is conducted in this phase to improve the alignment-layer printing process. The first optimization step aims at reducing variation, whereas the second one considers adjusting the process mean on the desired target (= 700 Å).

Variation reduction

Reducing variation is accomplished by choosing the factor levels that maximize the S/N ratio. Factors A and C are found to significantly affect variability. The S/N ratio averages for each factor levels are summarized in Table 5 and then depicted in Figure 7. It is observed that the combination of optimal factor levels that reduces variability is A₂C₂.

Table 5.

S/N ratio averages for each factor level.

Control factor	Factor level^*
	Level 1	Level 2	Level 3
Arrangement method	25.06	34.51
Molecule diameter (µm)	29.38	29.46	30.51
Exposure energy (mw/cm²)	27.90	31.70	29.76
Exposure time (seconds)	29.26	30.31	29.79

Initial levels are in bold.

Figure 7.

Average S/N ratio plot.

To confirm the improvement gained by setting factor A and C levels A₂C₂, two ITO samples, each of size 300 mm × 400 mm, are randomly selected per day for eight working days, which results in 16 ITO samples. The alignment-layer thickness is measured using the contour measuring instrument at five different places on the surface of each ITO sample. The $\bar{x}$ and R control charts for thickness are constructed, as shown in Figure 4. At improvement stage, the LCL, CL, and UCL values for the $\bar{x}$ chart are estimated 689.4, 702.2, and 715.1 Å, respectively. For the R chart, the LCL, CL, and UCL values are calculated 0.0, 22.31, and 47.18 Å, respectively. In Figure 4, it is obvious that the x̄ and R control limits are both in control and tighter than the corresponding limits at the initial stage for each chart. The σ̑ value at improvement stage is calculated as 9.59. Moreover, the C̑_p and C̑_pk are enhanced to 3.48 and 3.40, respectively, which indicates that the alignment-layer printing process is highly capable. The anticipated improvement in the alignment-layer thickness owing to setting factor A and C at optimal level combination A₂C₂ is estimated and found equal to 15.43 dB; which is calculated as the sum of the S/N ratio averages at optimal factor levels minus the sum at initial factor levels. Meanwhile, the process mean (= 702.2 Å) is still off target and hence it should be adjusted on the target (= 700 Å). To adjust the process mean on target, the second optimization step is conducted in the following subsection.

Adjustment of the process mean on target

Process mean can be adjusted on the target by selecting the factors that have significant effects on the mean but practically no effect on η. Previously, it is concluded that the molecule diameter (factor B) is only factor for this purpose. Therefore, factor B is used to adjust the process mean on target. Table 6 displays the thickness averages for each factor level.

Table 6.

QCH averages for each factor level.

Control factor	Factor level
	Level 1	Level 2	Level 3
Arrangement method	694.50	690.07
Molecule diameter (µm)	721.04	691.54	664.27
Exposure energy (mw/cm²)	687.58	702.32	686.96
Exposure time (seconds)	690.18	698.08	688.60

In Table 6, the thickness average decreases as the level value of factor B increases. In order to reduce the process mean from 702.2 to 700 Å, the best level value of factor B is calculated by interpolation and found equal to 35.37 µm, which is calculated as the value of level 1 (= 35) plus the difference between thickness averages at level 1 and 2 divided by 2.2. Confirmation experiments are finally conducted at the level value of 35.37 µm for factors B and A₂C₂. Confirmation results show that the process mean is 700.20 Å, while the σ̑ value is not affected. The $C_{p} and$ values are then calculated and found to equal 3.48 and 3.47, respectively. These values indicate that, due to setting process factors at optimal levels, the alignment-layer printing process becomes highly capable.

The control phase

In this phase, the $\bar{x}$ and R control charts are maintained for monitoring thickness of alignment-layer on ITO glass, while the capability of its printing process is regularly assessed. Moreover, the molecule’s diameter and exposure energy, which are important for reducing thickness variation and adjusting process mean on target, respectively, are regularly kept monitored and controlled.

Conclusions

Image color-quality has been recognized as one of the top considerations in the display manufacturing industry and has become a benchmark that influences the consumers’ purchasing decision. At the initial stage, the $\bar{x}$ and R control charts reveal that the printing process is in control. However, the C̑_p and C̑_pk indicate that the alignment-layer printing process is incapable. The DMAIC approach is implemented to improve the capability of the alignment-layer printing process. The optimal values of factor levels are decided as (1), the arrangement method is triangular (A₂), the molecule’s diameter is 35.37 µm, the UV exposure energy is 40 mw/cm², and the exposure time is 110 s. In ANOVA, the arrangement method (factor A) and the molecule’s diameter (factor B) are the key control factors for reducing variations and adjusting the mean of the alignment-layer printing process, respectively. Confirmation experiments are conducted at these level values followed by constructing the $\bar{x}$ and R control charts and process capability analysis. Confirmation results show that the σ̑ value is reduced from 42.26 to 9.59, hence, the obtained C̑_p and C̑_pk indicate that the alignment-layer is highly capable. The process mean is adjusted almost on target. This shows that the DMAIC, including Taguchi method, is a powerful approach for improving the performance of the alignment-layer printing process.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Yeh

. Optics of liquid crystal displays. John Wiley & Sons, Inc., 1999.

Kwak

Anbari

. Benefits, obstacles, and future of six sigma approach. Technovation 2006; 26: 708–715.

Antony

Banuelas

. A strategy for survival. Mfg Eng 2001; 80(3): 119–121.

Goh

. A strategic assessment of six sigma. Quality and Reliability Engng Int 2002; 18: 403–410.

MHC

Al-Refaie

. Improving wooden parts quality by adopting DMAIC procedure. Quality and Reliability Engng Int 2008; 24: 351–360.

M-H

Al-Refaie

Yang

. DMAIC approach to improve the capability of SMT solder printing process. IEEE Trans Electronics Packaging Mfg 2008; 31(2): 126–133.

Tang

Goh

Lam

. Fortification of six sigma: expanding the DMAIC toolset. Quality and Reliability Engng Int 2007; 23: 3–18.

Lin

Chen

Huang

. A hierarchy planning model for TFT–LCD production chain. Int J Electronic Business Mgmt 2004; 2(1): 59–68.

Ishihara

. How far has the molecular alignment of liquid crystals been elucidated? IEEE/OSA J Display Technol 2005; 1(1): 30–40.

10.

Chen

Wang

Chen

. A MAIC approach to TFT–LCD panel quality improvement. Microelectronics Reliability 2006; 46(7): 1189–1198.

11.

Spring

. A unifying approach to process capability indices. J Quality Technol 1997; 29(1): 49–58.

12.

Kane

. Process capability indices. J Quality Technol 1986; 18: 41–52.

13.

Juran

Gryna

. Quality planning and analysis. 2nd ed. New York: McGraw-Hill, 1986.

14.

Kotz

Lovelace

. Process capability indices in theory and practice. London: Arnold, 1998.

15.

Phadke

. Quality engineering using robust design. Englewood Cliffs, NJ: Prentice-Hall, 1989.

16.

Al-Refaie

. Grey-data envelopment analysis approach for solving the multi-response problem in the Taguchi method. Proc IMechE, Part B: J Engineering Manufacture 2010; 224(B1): 147–158.

17.

Hong

. Optimal parameter design for chip-on-board technology using the Taguchi method. Int J Adv Mfg Technol 2005; 25: 145–153.

18.

Al-Refaie

. Optimizing SMT performance using comparisons of Efficiency between different systems technique in DEA. IEEE Trans Electronics Packaging Mfg 2009; 32(4): 256–264.

19.

Al-Refaie

MHC

Tai

. Optimizing SUS 304 wire drawing process by grey analysis utilizing Taguchi method. J Uni Sci Technol Beijing, Mineral, Metall, Mater 2009; 15(6): 714–722.

20.

Belavendram

. Quality by design-Taguchi techniques for industrial experimentation. Prentice Hall International, 1995.