단기 생산 공정에 활용되는 통계적 공정관리 기법의 비교 연구 = Comparison of the Statistical Process Control Techniques for Short Production Runs

다국어 초록 (Multilingual Abstract)

Short runs are common in modern business environments. Owing to customer and short product life cycles , manufacturing trends have shifted toward a wide variety of mixed products with small batch sizes. It is difficult to apply traditional control charts efficiently and effectively in such environments.
When traditional control charts are applied to short run manufacturing environments, sampling difficulties are inevitable because there is never enough data from a part type to calculate meaningful control limits for control charts, and then different strategies are needed to deal with the situations. Special variable techniques that use mathematically transformed statistics and the corresponding charts can be needed for short runs.
In this thesis, the sensitivity of the Standardized Control Chart, Hillier's Exact Method and Q Chart for variables among many SPC (Statistical Process Control) techniques for short runs have been studied.
The Standardized Control Chart is a representative method among the transformation techniques. The control limits of this chart are calculated as in the traditional way but using the computed values of differences or normalizing ratios between actual measurements and nominal or target values.
Hillier's Exact Method has a two-stage procedure, using the first m subgroups to assess control of subsequent ones, that provide valid control limits for control charts regardless of the number of subgroups. This procedure compensates correctly for the uncertainty involved when computing control limits with small amount of data.
Quesenberry(l991) defined Q statistics that can be computed in a number of situations, assuming a normal process distribution and either one or both parameters unknown. We have considered Shewhart Q chart for the process mean and variance of a normal process. This chart method avoids the Phase- I /Phase-Ⅱ structure and attempt to start meaningful charting almost immediately.
Simulation experiments to compare with performances of three short run X￣ - R control charts are conducted for combinations of subgroups size(n), scales and timing of shift for process mean and standard deviation.
The probability and Average Run Length(ARL) to detect a one-step permanent shift in a process mean and/or a process standard deviation are computed for 5000 simulation runs. The probabilities of signals on the first thirty observations after shift of two parameters or Phase-Ⅱ have been used as the basis of comparison.
Based on simulation results, we observe as followings
(1) As n(subgroup size) and C(shift point) increase, probabilities of signals after parameters' shift increases and ARL decrease.
(2) n and C affect more Hillier's Exact Method than others.
(3) The Standardized Control Chart has the best performance but has the highest false alarm rate at the nominal situations.
(4) The false alarm rates of Hillier's Exact Method are less than others, and probabilities of signals after parameters' shift are better when subgroup size is about 5 and initial subgroup number is sufficient.
(5) In cases of the Standardized Control Chart and Q chart, process means and ranges are no longer plotted in the original scale of measurement.
In results, we may recommend that it will be a better method to apply Hillier's Exact Method to the Standardized or DNOM values.
Future researches are needed to modify the control charts that have greater sensitivity than three methods above. This will include how to best estimate process parameters with little initial data and how to best update the estimates as more data are collected.

목차 (Table of Contents)

• 목차
• Ⅰ. 서론 = 1
• 1. 연구목적 및 필요성 = 1
• 2. 연구방법 및 범위 = 3
• Ⅱ. 단기 생산 공정에 활용되는 SPC 기법 = 6
• 1. 표준화된 관리도 = 6
• 2. Hillier의 방법 = 9
• 3. Q관리도 = 14
• 4. 기법별 특성 비교 = 18
• Ⅲ. Simulation 실험 및 SPC 기법의 적용 = 20
• 1. Simulation 의 설계 = 20
• 2. 실험의 결과 및 비교 고찰 = 24
• 3. 항공기 부품 공정에의 적용 = 35
• Ⅳ. 결론 = 40
• 참고문헌 = 42
• 부록 = 44
• SUMMARY = 53