• Proceedings of the Society of Korea Industrial and System Engineering Conference
• /
• 2002.05a
• /
• pp.365-370
• /
• 2002

In this study, according to MARMA(1,0) model which was suggested by Seppala, in case of existing autocorrelation in X control chart and EWMA control chart, the standard method and the non-parametric bootstrap method were compared and analysed using the bootstrap method which use the resampling prediction residual.

• Proceedings of the Safety Management and Science Conference
• /
• 2013.11a
• /
• pp.199-207
• /
• 2013

산업의 빠른 발전 속도에 따라 연구 개발도 함께 발전해야 한다. 따라서 현재 제조공정에 대한 품질 특성치의 분석방법으로 공정 모수의 작은 변화도 쉽게 탐지를 할 수 있는 EWMA 관리도와 Shewhart 관리도보다 공정 변화에 민감하게 탐지 가능한 CUSUM 관리도에 관한 연구가 많이 이루어지고 있다. 하지만 식스시그마 공정관리에 맞춘 평균, 불량률, 미세 분산을 동시에 감지할 수 있는 동시 관리 체계 연구는 많이 미흡하다. 본 연구에서는 기존의 CUSUM, EWMA 관리도 기법보다 빠른 이상 감지를 위해서 평균, 불량률, 분산 3가지가 동시에 관리되어질 수 있도록 Zp-s 관리도를 소개한다. Zp-s 관리도는 ARL을 통해 기존 관리도보다 민감함을 확인할 수 있다.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.33 no.2
• /
• pp.9-17
• /
• 2010

Complex products may present more than one type of defects and these defects are not always of equal severity. These defects are classified according to their seriousness and effect on product quality and performance. So, demerit systems are very effective systems to monitor the different types of defects. Recently, Kang et al.(2009) proposed the revised Demerit-CUSUM for the evaluation of the Demerit-CUSUM control chart performance exactly. In this paper, we present an advanced Demerit control chart using the double EWMA technique. The double EWMA technique is very efficient and strong method for process control where defects and nonconformities occur with various defect types. Moreover, we compare exact performance of Demerit-CUSUM, Demerit-EWMA and Demerit-DEWMA control chart according to changing sample size or mean shifts magnitude. By the result, we confirm that the performance of Demerit-DEWMA control chart is more than the performance of the Demerit-CUSUM and Demerit-EWMA control chart.

• Journal of Korean Society for Quality Management
• /
• v.32 no.4
• /
• pp.169-183
• /
• 2004

The synthetic control chart (SCC) proposed by Wu and Spedding (2000) is to detect shifts in the process mean. The performance was re-evaluated by Davis and Woodall (2002), and the steady-state average run length (ARL) performance was shown to be inferior to cumulative sum (CUSUM) or exponentially weighted moving average (EWMA) chart This paper proposes a simple adaptive scheme to improve the performance of the synthetic control chart. That is, once a non-conforming (NC) sample occurs, we investigate the next L-consecutive samples with larger sample sizes and shorter sampling intervals. We employ a Markov chain model to derive the ARL and the average time to s19na1 (ATS). We also propose a statistical design procedure for determining decision variables. Comprehensive comparative study shows that the proposed control chart is uniformly superior to the original SCC or double sampling (DS) Χ chart and comparable to the EWMA chart in ATS performance.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.23 no.57
• /
• pp.123-129
• /
• 2000

The Bootstrap method proposed by Efron is non-parametric method which doesn't depend on the estimation of prior distribution refer to population. A typical statistical process control chart which is generally used is developed under the assumption that observations follow mutually independent and identically distributed within a sample and between samples. However, autocorrelation greatly affect the developed control chart under the assumption that observations are mutually independent. Many researchers showed that the result which was analyzed by using a typical control chart for the observations which has the correlation violated to the independence assumption can not be true. Therefore, we compared the standard method with bootstrap method and then evaluated them for x control chart and EWMA control chart by using bootstrap method which was proposed by Efron in the AR(1) model when the observations have correlation.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.22 no.52
• /
• pp.337-345
• /
• 1999

The long-tailed, positively skewed exponential distribution can be made into an almost symmetric distribution by taking the exponent of the data. In these situations, to use the traditional shewhart control limits on an individuals chart would be impractical and inconvenient. The transformed data, approximately bell-shaped, can be plotted conveniently on the individuals chart and exponentially weighted moving average chart. In this paper, using modifying statistics with transformed exponential of the data, we give a method for constructing control charts. Selecting method of exponent for individual chart is evaluated. And consider that smaller weight being assigned to the older data as time process and properties and taking method of exponent($\theta$), weighting factor($\alpha$) are suggested. Our recommendation, on the basis result of simulation, is practical method for EWMA chart.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.16 no.28
• /
• pp.87-92
• /
• 1993

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.16 no.28
• /
• pp.93-101
• /
• 1993

In this paper, two adaptive exponentially weighted moving avenge control chart schemes which available for real-time are proposed. The weighting coefficient is estimated using a recursive kalman filter algorithm. Simulated average run lengths indicate the proposed schemes are sensitive to process shifts And their performance is comparable to CUSUM control chart and customary EWMA control chart.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.35 no.4
• /
• pp.118-125
• /
• 2012

When monitoring an instrumental process, one often collects a host of data such as characteristic signals sent by a sensor in short time intervals. Characteristic data of short time intervals tend to be autocorrelated. In the instrumental processes often the practice of adjusting the setting value simply based on the previous one, so-called 'adjacent point operation', becomes more critical, since in the short run the deviations are harder to detect and in the long run they have amplified consequences. Stochastic modelling using ARIMA or AR models are not readily usable here. Due to the difficulty of dealing with autocorrelated data conventional practice is resorting to choosing the time interval where autocorrelation is weak enough then to using I-MR control chart to judge the process stability. In the autocorrelated instrumental processes it appears that using the Shewhart chart and the time interval data where autocorrelation is relatively not existent turns out to be a rather convenient and very useful practice to determine the process stability. However in the autocorrelated instrumental processes we intend to show that one would presumably do better using the EWMA control chart rather than just using the Shewhart chart along with some arbitrarily intervalled data, since the former is more sensitive to shifts given appropriate weights.

• The Korean Journal of Applied Statistics
• /
• v.9 no.1
• /
• pp.91-109
• /
• 1996

Exponentially weighted moving averae(EWMA) control chart has been used widely for process monitoring and process adjustment recently, but there has not been many studies about the selection of the parameters. Design of the control chart can be classified into the statistical design and the economic design. The purpose of the economic design is to minimize the cost function in which all the possible costs occurring during the process are probability given the Type I error probability. In this paper the optimal parameters of the EWMA chart are selected for the economic design as well as for the statistical design. The optimal parameters for the economic design show significantly different from those of the statistical design, and especially the weight is always larger than that used in the statistical design. In the economic design, we divide the model into the single assignable cause model and the multiple assignable causes model caacording to number of which is used as the average context of the multiple assignable causes, it shows that the selection of the parameters may be misleading when the multiple assignable causes exist in practice.