Taguchi based Case study in the automotive industry: nonconformity decreasing with use of Six Sigma methodology

Abstract

In this study, we applied a conceptual Six Sigma/design of experiment hybrid framework that aims to integrate the Taguchi method and Six Sigma for process improvement in a complex industrial environment. In this context, the Six Sigma methodology was employed on a company operating within the automotive industry to improve a manufacturing process which caused a customer complaint within the company. Studies employing the Taguchi experiment design usually focus on a single variable and neglect the effects of adjustments on remaining quality characteristics. In this study, a multi-response Taguchi design of experiment was preferred, and all of the quality characteristics were taken into account. In our study, define, measure, analysis, improve and control phases were used to reduce the nonconformity rate from 23.940 percent (baseline) to 0.049 percent. As a result of implementing Six Sigma, the sigma level increased from 2.21 (baseline) to 4.80.

1. Introduction

Industrial companies require good process management and continuous improvement (CI) to acquire an advantageous position among the competition and achieve customer satisfaction at the same time. To this end, companies employ methods and techniques featuring a series of statistical and non-statistical tools. Among others, Six Sigma (SS) has become a general framework for this purpose. SS has successfully been applied in many fields such as manufacturing, services, R&D, health sector, nuclear facilities, maintenance, repair, and the public sector, following a rapid expansion from the point it was introduced in an industrial environment at Motorola in 1980s [13,19,27,28].

SS defines ‘quality’ as 99.73% of products, services or any other characteristics being in compliance with the required specifications. Companies which apply SS have determined the defect rate of a product, service or any other characteristic to be less than 3.4 in a million to achieve better results regarding the overall quality level [5]. SS was designed, built upon statistical thinking and approaches with a particular focus on eliminating variation in processes [17]. Improvement in effectiveness and efficiency constitutes the main core of the concept of SS [10]. SS is considered a methodology that makes tools for improvement of process capabilities available for all businesses [42]. The purpose of SS is to detect and eliminate sources of variation [24].

In recent years, it is evident that the number of SS projects is rapidly increasing. The most important reason for this is various profits brought about by SS projects. Especially, due to benefits such as decreasing the non-conformity rate in the industry, increasing performance, and enhancing customer satisfaction, SS projects are being preferred [22].

As known, there is a strong competition in the automotive industry. Therefore, the company in this study did not want its name to be used. In this study, the company is referred to as the rubber manufacturer. The reason for the choice for the rubber manufacturing company was that the company does not just apply widely accepted industrial standards but is also compliant with the specific quality standards and procedures of customer quality systems. The company is manufacturing rubber and plastic-derivative products for the automotive industry where the competition is fierce and counts as one of the top actors in the sector. The aim of the company is to drive the quality of its products to the utmost level in line with the CI mentality. The rubber manufacturer as the main supplier of the globally leading ABC car brand lately receives customer complaints about the flexibility of the door glass seals it produces.

This study has been done as an SS project in the automotive industry which has severe competition and a strict quality perspective. One of the aims of this industry is to solve problems with a fast and low-cost approach. In this study, our aim is to improve the factors causing customer complaints by using SS. Using the SS methodology, the factors triggering customer complaints will be detected and improved. In more detail, this study analyses the root cause and identify the actions to optimise front door glass seal flexibility problems, using the SS define, measure, analysis, improve, control (DMAIC) method. The firm did not want to share the number of customer complaints. Therefore, the number of customer complaints from the ABC brand is not specified. 270 experiments were performed instead of 7290 experiments with the use of MRSN Taguchi DoE. As a result of this project, the sigma level increased from 2.21 to 4.80.

The remaining part of this paper is structured as follows. The DMAIC methodology is provided in Section 2. Section 3 presents SS applications which are related to the automotive industry. In Section 4, we conclude the study with the results and further research.

2. Methodology

In this study, the multi-response Taguchi design of experiment (MRSN) was utilised along with the DMAIC methodology. This was because, in this method, decisions are made based on data analysis. It is a method based on the continuity of the results obtained. Moreover, in contrast with many studies utilising the Taguchi experiment design, where the goal is to optimise a single variable, simultaneous optimisation of multiple variables was aimed. In this regard, the study distinguishes itself and contributes to the literature by setting an example for an alternative use of the method. In this study, MRSN was preferred, and all quality characteristics were taken into account.

DMAIC that allows structural flexibility provides a roadmap for the SS team. The DMAIC roadmap provides a problem-solving process that had not existed before [37]. DMAIC projects can be developed using a top-down or bottom-up approach. The SS DMAIC method focuses on a specific area of interest. However, SS projects with a large scope may concentrate on multiple areas, and it may take three to eight months to complete these projects [23].

These DMAIC stages are designed to lead a team through a process improvement project from start to finish [35,36,41]. The problem and customer requirements are defined at the definition phase [16]. The potential causes of the problems are defined at the measurement phase by using various statistical tools. The analysis phase focuses on analysis of the product/process metrics and their comparison [31]. The root causes of the problems in the process are determined by SS teams [24,37]. The improvement phase aims to take the causes found at the analysis phase into account, and it also aims to choose and target solutions to eradicate such causes [3]. The control phase seeks to implement continuous measures and actions to preserve the improvement by tracking, standardising, documenting and integrating the new process on a daily basis [30].

The use of DMAIC is not limited to solving quality problems but includes all aspects of business development, such as time savings, cost savings and service improvement [39]. The detailed composition and progression of utilisation of DMAIC tools could diverge from one organisation to another and from project to project. However, their application and logical flow are what facilitate the total impact of SS [14]. DMAIC focuses on elimination of inefficient business processes, application and development of new metrics and the use of technology to ensure improvement [20]. Companies preferring the SS DMAIC methodology utilise this method to achieve real improvements and results [31].

In this article, the Minitab 17 statistical software was used to conduct the process capability analysis.

3. Application

In the modern competitive market, the chances of survival are rather low for a company making mistakes. Therefore, companies aim to minimise possible sources of defect by taking the production process under control. To achieve this goal, various statistical and non-statistical techniques and tools are employed. The SS team preferred the SS DMAIC methodology as the application method. At various phases of DMAIC, many statistical techniques such as Taguchi design of experiment, Gage R&R and SPC are used. This way, products of higher quality are manufactured, costs are lowered, and efficiency and customer satisfaction are enhanced.

3.1. Define phase

The first step of the SS project is the definition phase of the DMAIC problem-solving method. In this step, the indications, goals and limitations of the project are defined in a way apprehensible by every individual, and the effect on the customer is determined. Since the definition phase constitutes the initial stage of the project start, establishing the correct metrics is of importance to be able to reach suitable solutions in the following phases.

As the first step of this SS project, the customer complaint of the ABC car manufacturer concerning the front door glass jamming and the glass coming loose off the seal housing was defined. It was determined that most of the customer complaints were about the glass jamming at times and therefore not being able to complete the upwards/downwards movement during the opening and closing of the electrically actuated front door window or the glass coming loose off the seal housing. The glass actuator automatically locks itself out and becomes inactive to minimise problems caused by jamming. The ABC car manufacturer detected that the problem was caused by the front door glass seals. Therefore, the scope of the SS project was the door glass seal with the part number only MTA2034, manufactured for ABC Cars. The seal is manufactured only on the machine with the code number X15. Front door glass seals manufactured for other models of this firm were excluded from the scope of this project. According to the SS DMAIC methodology, the schedule for each planned step was prepared. SS designated a period of 12 months for implementation of the project. The rubber manufacturing company formed an SS team to solve the problem triggering the customer complaints of the ABC car manufacturer and prevent a recurrence. The SS team was formed among competent employees, working at relevant departments or cooperating actively in addition to that on a voluntary basis. A ‘CI Manager’ with a black belt was selected as the team leader. The university supported the statistical knowledge to the company within the collaboration of academia and industry.

The SS team evaluated the customer complaint from the ABC automotive company regarding the front door glass jamming and glass coming loose off the seal housing. There had been an evident increase in the number of ABC customer complaints regarding the door glass seal supplied by the rubber manufacturing company. The occurrence of the error on the car depends on several factors like the frequency of use of the vehicle and the environment in which the vehicle is used. Therefore, the number of errors could be unstable throughout months.

In this study, the flexibility of the door glass seal was determined as the quality characteristic that triggered the customer complaint. The front door glass should move in the seal with a certain degree of flexibility. The technical specification provided by ABC to the rubber supplier company stated the limits as 9 ± 2 Newton. The shape and location of the door glass seal are shown in Figure 1.

Figure 1.

Location of the front door glass seal on a vehicle.

Glass seals ensure that the moving glasses of the car operate silently and with the minimum friction possible in the housing. Additionally, these seals assist water, dust, heat and sound insulation, therefore contributing to the cleanliness of the windows.

3.2. Measure phase

The total variation consists of the measurement system variation and the process variation. To determine the actual process variation in SS projects, principally, the measurement system needs to be confirmed [2,6,10,15,30,34]. The aim of Gage R&R studies is to ensure that the measurement system is statistically reliable. A Gage R&R study evaluates how much of the total variance is due to the measurement system [21]. Measurements are critical components of any quality system. Measurement is an inseparable constituent of the DMAIC problem-solving process. An ineffective measurement system may affect work performance adversely due to the possibility of incorrect decision-making [26].

Flexibility leads to continuous data. The measurement system was analysed based on the flexibility data measured by a rubber elasticity metre. Measurements were conducted randomly on unformed five single parts with three repetitions and three appraisers, manufactured at different times within six workdays and three shifts. In every shift, one operator mans the machine with the code no. X15. Only MTA2034 glass seals for the ABC Company are manufactured with this machine. To achieve randomness in part selection, the operators do not know which box contains the parts of which profile and the order of parts. The results of the measurement system analysis are shown in Table 1. It is accepted in AIAG that a repeatability value between 10 and 30% is marginal. So, it may be seen that 25.25% of the total variance was due to Gage R&R, and the number of distinct categories was 5. Thus, the capability of this measurement system was reliable [1,4].

Table 1. Measurement system analysis of front door seal flexibility.

		Study Var	%Study Var	%Tolerance
Source	Std. Dev (SD)	(5.15 × SD)	(%SV)	(SV/Toler)
Total Gage R&R	0.0089028	0.045849	25.25	1.15
Repeatability	0.0066667	0.034333	18.91	0.86
Reproducibility	0.0059004	0.030387	16.74	0.76
Operators	0.0000000	0.000000	0.00	0.00
Operators*Parts	0.0059004	0.030387	16.74	0.76
Part-To-Part	0.0341131	0.175682	96.76	4.39
Total Variation	0.0352557	0.181567	100.00	4.54

Number of Distinct Categories = 5

Due to the finding that the variation caused by the measurement system turned out to be acceptable, the factors triggering the variation stemming from the process were analysed at the analysis phase of the project.

3.3. Analysis phase

According to Pyzdek [33], the current status of the process and significant inputs and their levels causing process variance are analysed at this phase. The SS team collected data from the production line to check whether the process was under control and receive early warnings for process variation. As the measurements were conducted with destructive testing, the data were collected by five measurements for each shift, in a ten-day period. 150 measurements in total were conducted and gathered in 30 subgroups. In this study, S1, S2 and S3 indicate the day, evening and night shifts respectively. Firstly, the SS team conducted an Anderson–Darling normality test. The test of normality using the Anderson–Darling test statistic (A-squared = .37) resulted in a large p-value (0.422) indicating that the data could be considered to have a normal distribution. Thus, an $\overline{X}-R$ control chart was drawn up to check for the stability of the process and whether the production was in control. This chart indicates that the production is in control when no observation is out of the control limits. The team investigated how much the process changed from the customer specification bounds by applying a process capability analysis as a statistical criterion [25]. The SS team examined the $C_{p_k}$ value since the nominal (target) value did not coincide with the process mean ( $\overline{\overline{x}}$). The value of $C_{p_k}$ was obtained as 0.24 < 1.33, which indicated that the process did not comply with the specification limits. Since the data followed a normal distribution, the ‘Expected Overall’ performance value given in Figure 2 was examined to see the value of the nonconforming products in PPM. The $PP{M}_{\text{Total}}$ value was calculated as 239,396, which indicated that 239,396 of every million MTA2034 parts produced with the X15 machine did not comply with the flexibility limits stated in the contract requirements. Appendix 1 shows that the quality level corresponding to the 239,396 nonconformities is $2.21\sigma$. This shows that the company was capable of producing on the quality level of $2.21\sigma$.

Figure 2.

Process capability analysis.

The SS team tried to determine the cause/s of the deviation with the brainstorming method and a cause and effect matrix. After a detailed examination and evaluation of every single factor obtained through brainstorming and an Ishikawa diagram, six factors and three levels for each factor having an effect on the flexibility quality characteristic were determined. The levels of these factors were determined by the R&D department. These factors and the levels thereof affecting the flexibility quality characteristic are shown in Table 2.

Table 2. Factors affecting the flexibility of the glass seal.

Factor Knowledge			Factor Knowledge
Code	Control factor	Unit	1	2	3
A	Cooling time	Sec.	5	10	15
B	Injection pressure	Bar	250	450	650
C	Injection speed	cm/s	5	10	15
D	Melt Temperature	°C	470	480	490
E	Mold Temperature	°C	150	155	160
F	Wall Thickness	mm	2.10	2.15	2.20

In the present situation, factors and the levels thereof causing the process variation were defined, which were $A_2{B}_2{C}_1{D}_2{E}_1{F}_1$ and highlighted with blue colour in Table 2. According to Chen and Brahma [7], ‘the most important part of the analysis phase is to identify the potential factors (X) that affect the project index (Y)’.

3.4. Improvement phase

The levels of critical factors leading to process changes are determined and verified in this stage. At the improvement phase, the goal is to narrow down the gap between the current state of the process and the target value. Project management and other planning and administrative tools are utilised to introduce and enforce the new approach. Statistical techniques are implemented to confirm the enhancements [33]. At this stage; the differences in the process are reviewed and which factors contribute significantly to the outcomes are determined. The factors and their levels affecting the flexibility of the door glass seal in the current condition were $A_2{B}_2{C}_1{D}_2{E}_1{F}_1$. The SS team aimed to determine the optimal factors and their levels to eliminate the flexibility problem of the door glass seal. Factoring and fractional factoring tests may require a large number of test units. It may, therefore, be reasonable to use one or more of the methods for controlling haphazard change [8]. In this practice, a full factorial experiment design would require $3^6=729$ experiments. Since experiments are to be conducted with ten repetitions each for the mean and the variance to be calculated effectively, there would be a total of 7290 different experiments to perform. This enormous amount of experiments would increase the cost and cause time loss. Therefore, the SS team decided to apply the Taguchi experiment design which is widely used in phase to improve process or product. Taguchi [38] takes concern in experiments generally to estimate the main effects, trying the significance level based on the experiences of the experimenter. These experiments are not based on the potential trial combinations as a whole but rather on examination of a fraction by using vertical columns, triangular tables and linear plots [9]. ‘The goal of the Taguchi method is to find control factor settings that generate acceptable responses despite natural environmental and process variability’ [32]. In this study, a vertical column was formed in the $3^k$ system by using Graeco-Latin Squares stemming from the Latin Squares method. It was determined by using the Graeco-Latin Squares method that the most suitable orthogonal array for six factors and three levels was $L27$. Therefore, 27 individual experiments along with ten repetitions each, constituting a total of 270 experiments, were performed. As a result of the Taguchi experiment design, the total amount of experiments to be conducted went down by 7,020 (96.29%) in comparison to the full factorial experiment design. The $L27$ Orthogonal experiment array obtained through the Graeco-Latin Squares method is shown in Table 3.

Table 3. Suitable orthogonal arrays OA (27, 3⁶).

	FACTORS
Experiment No	A	B	C	D	E	F	Cooling Time (A)	Injection Pressure (B)	Injection Speed (C)	Melt Temperature (D)	Mold Temperature (E)	Wall Thickness (F)
1	1	1	1	1	1	1	5	250	5	470	150	2.10
2	1	1	1	1	2	2	5	250	5	470	155	2.15
3	1	1	1	1	3	3	5	250	5	470	160	2.20
4	1	2	2	2	1	1	5	450	10	480	150	2.10
5	1	2	2	2	2	2	5	450	10	480	155	2.15
6	1	2	2	2	3	3	5	450	10	480	160	2.20
7	1	3	3	3	1	1	5	650	15	490	150	2.10
8	1	3	3	3	2	2	5	650	15	490	155	2.15
9	1	3	3	3	3	3	5	650	15	490	160	2.20
10	2	1	2	3	1	2	10	250	10	490	150	2.15
11	2	1	2	3	2	3	10	250	10	490	155	2.20
12	2	1	2	3	3	1	10	250	10	490	160	2.10
13	2	2	3	1	1	2	10	450	15	470	150	2.15
14	2	2	3	1	2	3	10	450	15	470	155	2.20
15	2	2	3	1	3	1	10	450	15	470	160	2.10
16	2	3	1	2	1	2	10	650	5	480	150	2.15
17	2	3	1	2	2	3	10	650	5	480	155	2.20
18	2	3	1	2	3	1	10	650	5	480	160	2.10
19	3	1	3	2	1	3	15	250	15	480	150	2.20
20	3	1	3	2	2	1	15	250	15	480	155	2.10
21	3	1	3	2	3	2	15	250	15	480	160	2.15
22	3	2	1	3	1	3	15	450	5	490	150	2.20
23	3	2	1	3	2	1	15	450	5	490	155	2.10
24	3	2	1	3	3	2	15	450	5	490	160	2.15
25	3	3	2	1	1	3	15	650	10	470	150	2.20
26	3	3	2	1	2	1	15	650	10	470	155	2.10
27	3	3	2	1	3	2	15	650	10	470	160	2.15

Adjustments on the factors affecting the flexibility, which triggered the customer complaints, and their levels might also affect the following quality characteristics of the front door glass seal appearance, weight and hardness. The ABC car manufacturer considers all of the quality characteristics, i.e. response variables, equally important. Since all quality characteristics are ‘nominal is best’. Equations (1)–(6), which are given below, were used to calculate the signal to noise ratio, $S/N_{A_1}$, given in equation (7).

$$L_{ij}=k\left(\frac{{s_{ij}}^2}{{\overline{y}}_{ij}^2}\right)$$ (1)

$L_{ij}$ = Quality loss for the $i$th response at the $j$th trial; $y_{ijk}$ = Observed data for the $i$th response at the $j$th trial, $k$th repetition; $n_i$ = Replications for the $i$th response.

$$\begin{array}{rl}{\overline{y}}_{ij}& =\frac{1}{n_i}\sum _{k=1}^{n_i}(y_{ijk})\end{array}$$ (2)

$$\begin{array}{rl}{s}_{ij}^2& =\frac{1}{n_i-1}\sum _{k=1}^{n_i}(y_{ijk}-{\overline{y}}_{ij}{)}^2\end{array}$$ (3)

$k=$ Quality loss coefficient

The four responses (in order of importance) are:

• Flexibility, in which the target value is 9 ± 2 Newton,

• Appearance, in which the target value is 1,

• Weight, in which the target value is 321 ± 10 gr,

• Hardness, in which the target value is 65–70 shore A.

In order to determine the multi-response signal to noise (MRSN) ratio, reducing the variability needs normalising the scale of the quality loss for each characteristic. The ratio of multi response S/N can be calculated by the total normalised quality loss (TNQL). These three phases are [40]:

• Normalise the quality loss of each trial for each response.

  $$C_{ij}=\frac{L_{ij}}{L_i^{\ast }}$$ (4)

  where $L_i^{\ast }=\text{max}(L_{i1},L_{i2},\dots ,L_{ij})$.
• Compute the total normalised quality loss (TNQL) of each trial.

  $$\text{TNQ}{\text{L}}_l=\sum _{i=1}^n{w}_i{C}_{ij}$$ (5)

Where $w_i$ = the weight of the $i$th normalised response ( $i=1,2,\dots ,n$).

• Determine the optimal MRSN ratio ( $\eta$) for each trial:

  $$\text{MRS}{\text{N}}_j=-10\text{log}(\text{TNQL})$$ (6)

The Taguchi experiment results were analysed by taking the response variables ‘flexibility, appearance, weight and hardness’ quality characteristics into account. The TNQL and MRSN ratios related to these trials are given in Appendix 2. Each factor and the effects of their respective levels were computed with the following calculation of the MRSN values. An example calculation method and results of $S/N_{A_1}$ are as follows [29]:

$$S/N_{A_1}=\frac{{\eta }_1+{\eta }_2+{\eta }_3+{\eta }_4+{\eta }_5+{\eta }_6+{\eta }_7+{\eta }_8+{\eta }_9}{9}$$

In a similar manner, all factors and their respective effect values were calculated, and the results are demonstrated in Table 4. At the same time; all calculations and results are shown in Appendix 3. The optimal level combination was determined as A₃B₂C₁D₂E₁F_2.

Table 4. S/N ratio value by factor level.

Level	A	B	C	D	E	F
1	0.041	0.124	1.459	0.419	1.338	−1.735
2	1.015	1.209	0.315	1.003	0.611	3.726
3	1.305	1.027	0.587	0.938	0.411	0.369
Delta (Δ)	1.264	1.085	1.144	0.584	0.927	5.461
Rank	2	4	3	6	5	1

ANOVA table was calculated as shown in the equations below:

$$S{S}_T=\sum _{i=1}^n({\eta }_i-{\eta }_m{)}^2=180.922$$

${\eta }_m$: The total mean S/N ratio; ${\eta }_i$: The mean S/N ratio for the ith experiment; $n$: The number of experiments in the orthogonal array.

$${\eta }_m=\sum _{i=1}^n\frac{{\eta }_i}{n}=0.78690$$

Sum of squares for the factor ‘A’ is calculated as shown in the equation below [29]. The sums of squares for the remaining factors are given in Table 5.

$$S{S}_A=9(m_{A_1}-m{)}^2+9(m_{A_2}-m{)}^2+9(m_{A_3}-m{)}^2=7.898$$

As it may be seen in Table 5, the most influential factor on the door glass seal flexibility was the ‘glass seal walls thickness factor’, ‘F’ with a rate of 75.483%. The optimum level combination had been determined as ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$.

Table 5. ANOVA results for response variables.

	Degree of	Sum of	Mean of		Percentage
Source of Variation	Freedom	Square	Square	F-ratio	Contribution (P)
Cooling Time [A]	2	7.898	3.95	3.10487	4.365%
Injection Pressure [B]	2	6.084	3.04	2.39181	3.362%
Injection Speed [C]	2	6.439	3.22	2.53133	3.558%
Melt Temperature [D]	2	1.842	0.92	0.72407	1.018%
Mold Temperature [E]	2	4.285	2.14	1.68433	2.368%
Wall Thickness [F]	2	136.567	68.28	53.68642	75.483%
Error	14	17.807	1.27		9.842%
Total	26	180.922			100.0%

The optimum level combination ‘ ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$’ was required to be verified. Verification was performed with the test of the optimum combination condition obtained through the Taguchi experiment design. For this purpose, data pertaining to the combination ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$ were collected. Since the orthogonal array $L27$ was used in the project, and the experiments were performed with ten repetitions, 270 data sets were present. In order to verify the optimum combination ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$ obtained through the MRSN Taguchi experiment design, 30 verification experiments were performed. Defining statistical data regarding the standard deviation, variation and range of the initial condition ( ${\text{A}}_{\text{2}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{1}}$), ten best conditions and the optimum combination ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$ obtained through the Taguchi experiment design of the four quality characteristics are given in Table 6.

Table 6. Comparison of confirmation data to starting conditions, best 10 conditions and optimum conditions.

Response		Starting condition		Optimum condition
Variables	Statistics	(A2B2C1D2E1F1)	Best 10	(A3B2C1D2E1F2)
Flexibility	Average	8.768	9.133	9.045
	Std. Dev.	1.322	0.555	0.072
	Variance	1.748	0.308	0.005
	Range	3.760	1.800	0.240
Appearance	Average	0.981	1.000	1.000
	Std. Dev.	0.135	0.000	0.000
	Variance	0.018	0.000	0.000
	Range	1.000	0.000	0.000
Weight	Average	320.331	320.668	320.956
	Std. Dev.	3.673	1.732	0.185
	Variance	13.492	2.999	0.034
	Range	11.900	7.370	1.060
Hardness	Average	67.171	67.235	67.508
	Std. Dev.	0.735	0.279	0.046
	Variance	0.540	0.078	0.002
	Range	2.370	1.190	0.180

As the mean, standard deviation, variation and range data that are given in Table 6 are examined, it may be clearly seen that there was an improvement in flexibility, appearance, weight and hardness quality characteristics. Therefore, it may be assumed that the combination ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$ was suitable. Following the experiment, the mean values were drawn closer to the nominal values.

By using the data pertaining to the initial condition, the best ten conditions and the optimal condition of the quality characteristics normalised quality costs ( $C_{ij}$) were calculated, and the results are presented in Table 7.

Table 7. TNQL and MRSN values according to the normalized costs ( $C_{ij}$).

	Flexibility	Appearance	Weight	Hardness	Flexibility	Appearance	Weight	Hardness	1.0; 1.0; 1.0; 1.0
Combination	L1j	L2j	L3j	L4j	C1j	C2j	C3j	C4j	$\text{TNQ}{\text{L}}_j$	$\text{MRS}{\text{N}}_j$	Improvement (dB)
Starting Condition (A2B2C1D2E1F1)	0.0001879	0.0320073	0.0000023	0.0000011	1	1	1	1	4	−6.021
Best 10 Condition	0.0000969	0.0000000	0.0000011	0.0000008	0.5156998	0	0.4782609	0.7272727	1.7212334	−2.358	3.663
Optimum Condition (A3B2C1D2E1F2)	0.0000630	0.0000000	0.0000003	0.0000005	0.3352847	0	0.1304348	0.4545455	0.9202650	0.361	6.382

The achieved improvements were calculated by including the normalised costs ( $C_{ij}$) as shown below:

$$\text{Best }10\text{ conditions}\wedge \text{the initial condition}=-2.358-(-6.021)=3.663dB$$

$$\text{The optimal}\wedge \text{the initial condition}=0.361-(-6.021)=6.382dB$$

These values obtained in decibel demonstrate the improvement rates. This verification outcome attests to the success of the optimal condition ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$. While examining the verification results, it is seen that significant improvements in flexibility, weight and hardness quality characteristics were achieved.

As a result of the confirmation experiment, the suitability of the optimum combination ${\text{A}}_{\text{3}}{\text{B}}_{\text{2}}{\text{C}}_{\text{1}}{\text{D}}_{\text{2}}{\text{E}}_{\text{1}}{\text{F}}_{\text{2}}$ obtained through the Taguchi experiment design was confirmed. Response variables flexibility, appearance, weight and hardness values were drawn closer to the nominal levels. It was seen that especially the quality characteristic that triggered the customer complaint, flexibility, approached the nominal value quite significantly.

3.5. Control phase

As pointed out by Hsieh, Lin and Manuca [18], the aim of this control phase is not only applying already accepted solutions but also guaranteeing to implement CI at all times. It is the phase in which the required control and monitoring operations are performed for the process to continue under the improved conditions. As shown in Figure 3, the SS team performed a before-after analysis to standardise the process and control the process capability.

Figure 3.

Control phase.

The optimum factor levels and the values of these levels verified following the improvement phase are shown in Table 8.

Table 8. Before-after comparison.

Factor Knowledge
			Before	After
Code	Control factor	Unit	(A2B2C1D2E1F1)	(A3B2C1D2E1F2)
A	Cooling time	Sec.	10	15
B	Injection pressure	Bar	450	450
C	Injection speed	cm/s	5	5
D	Melt Temperature	°C	480	480
E	Mold Temperature	°C	150	150
F	Wall Thickness	mm	2.10	2.15

The ABC car manufacturer wanted to reduce the customer complaints concerning the front door glass seals to a minimum. Therefore, the contract was renewed in consultation with the rubber manufacturer to include the revised specification value for the glass seal flexibility as 9 ± 0.5 Newton. A process capability analysis was conducted by using the new specification limits. Relevant data were collected by five samples from each shift, in a 10-day period. 150 measurements were performed in total. A normality test was performed on the measurements, and the obtained $P-\text{Value}=0.547$ indicated that the data followed a normal distribution. Then, an $\overline{X}-R$ control chart was drawn up to check for the stability of the process and whether the production was in control. When the $\overline{X}-R$ control chart result was examined, no special cause was observed, and thus, the observations were in the vicinity of the mean values. Under these circumstances, it may be deduced that the process was stable. A process capability analysis was performed to evaluate the capability of the process, and the results are shown in Figure 4.

Figure 4.

Process capability index.

The calculation delivered the following outcome: $C_{p_k}=1.13$. This value proved the improvement in the process capability consequent upon the SS project. The process mean was brought reasonably close to the nominal value with 9.09193 Newton. Improvements were documented and entered into the system. Personnel were trained in the changes and certified. Standardised work instructions were created and made available along with the production order to the production area. Following the SS project, the variation in the process decreased, and the process mean converged to the nominal value. Statistical improvement results obtained through the SS project conducted to eliminate the flexibility problem of the door glass seal are given below:

• Process means increased from 7.0951 Newton to 9.0919 Newton and approached the nominal value.

• The sigma level of the process increased from 2.21σ to 4.80σ.

• The standard deviation of the process decreased from 0.140589 to 0.123759.

The results acquired at the control phase showed that the glass seal flexibility value was drawn near the nominal value. The SS team arranged a closing meeting following the successful execution of the project. A presentation was prepared to outline the improvement to the management achieved with the project. The success of the project was approved by the management.

4. Conclusion

In today's global economy, especially the automotive industry, which is a leading industry, faces fierce competition. To remain competitive, manufacturer companies must be therefore continuously efficient and improve. For this reason, companies are producing their products or services to satisfy their customers. In order to do this, companies are using highly sophisticated statistical tools and techniques. These tools and techniques play an important role in identifying and controlling the processes during production. There are several managerial and scientific ways to use these techniques. These managerial approaches are usually the resultant methodologies of the evolution of previous approaches. One of these new management approaches known as SS all around the world is quite effective. SS may be created with the right corporate infrastructure in all sectors.

It plays an important role to make SS successful because the problem is focused on not only according to customers’ needs but also based on the production process of the front door window seal. First of all, the problem was identified, and then, the measurement system was verified. In order to verify the system, a process productivity analysis was performed, and the causes of the problem were identified. CI was achieved by eliminating factors caused by process insufficiency and by eliminating their effects. For this, the improvement phase of the SS methodology used the Multi Response Taguchi Design of Experiment method. As well-known, Taguchi experimental design employed in the improvement phase of many SS projects might neglect other quality characteristics, while focusing on a single one. This circumstance may have a negative influence on the financial state of the company, product/service quality and customer satisfaction. In this study, the possibility of an adjustment made on cooling time, injection pressure, injection speed, melting temperature, mould temperature and wall thickness factors and levels that affected the flexibility of the front door glass seal along with the appearance, weight and hardness quality characteristics was evaluated by applying the ‘multi-response signal noise’ Taguchi experiment design. Consequently, it was ensured that the remaining quality characteristics of importance to the customer were taken equally into consideration. Therefore, CI was satisfied not only by having a response variable but also by identifying the quality characteristics of the product. By using MRSN Taguchi design of experiment, the CI of quality characteristics was identified not only on the flexibility of front door window value but also on appearance, weight and hardness.

In terms of avoiding any customer complaints with the front door window, the statistical values/results from the Taguchi test helped prove the success of this project. To be successful in SS projects, having and using a statistical technique will be helpful not only in the automotive industry but also in the aerospace, defence, chemistry, health, service, R&D and public industries. For this reason, employees, who would have a statistical background, would be beneficial for SS projects.

In tandem with the advancements of the industrial age, the area of application for SS continues to expand. Industry 4.0 may be roughly described as innovation projects including robots taking over production completely, development of artificial intelligence, shift of production from factories to homes via 3D printers, analysis and evaluation of vast amounts of information through data analysis methods, etc. [11,12]. Up until today, SS applications were conducted at the third phase (computing and automation) of the industrial age. However, changes and improvements on SS applications will be inevitable with the Industry 4.0 era. Therefore, the usage of SS will adapt to these innovations and keep on spreading. The integration of SS with innovation has become inevitable in the context of Industry 4.0. The scope of application of SS is destined to expand much more broadly and rapidly during Industry 4.0 than ever.

Disclosure statement

No potential conflict of interest was reported by the author(s).