• Proceedings of the Korean Society for Quality Management Conference
• /
• 2009.10a
• /
• pp.23-29
• /
• 2009

직교성은 실험계획에서 중요한 개념이다. 실험계획에서 실험점의 개수보다 인자의 개수가 많은 상황에서 우리는 초포화계획을 사용한다. 이러한 초포화계획은 직교성을 만족하지 못하게 되는 데 얼마나 직교성을 만족하는 지를 평가하는 데 우리는 주로 수치적인 측도들을 사용한다. 우리는 초포화계획의 직교성의 정도를 평가하는 또 다른 탐색적 방법으로서 그래픽방법을 사용할 수 있다. 또한 초포화계획의 예측 능력을 평가하는 방법으로서 우리는 그래픽 방법을 사용할 수 있다.

• Journal of Korean Society for Quality Management
• /
• v.47 no.3
• /
• pp.523-535
• /
• 2019

Purpose: We research on the efficient response surface methodology(RSM) design to develop a design space in Quality by Design(QbD). We propose practical designs for the successful construction of the design space in QbD by allowing different number of replicates at the box points, star points, and the center point in the rotatable central composite design(CCD). Methods: The fraction of design space(FDS) plot is used to compare designs efficiency. The FDS plot shows the fraction of the design space over which the relative standard error of predicted mean response lies below a given value. We search for practical designs whose minimal half-width of the tolerance interval per a standard deviation is less than 4.5 at 0.8 fraction of the design space. Results: The practical designs for the number of factors between two and five are listed. One of the designs in the list could be chosen depending on the experimental budget restriction. Conclusion: The designs with box points replications are more efficient than those with the star points replication. The sequential method to establish a design space is illustrated with the simulated data based on the two examples in RSM.

• Journal of Korean Society for Quality Management
• /
• v.48 no.2
• /
• pp.269-282
• /
• 2020

Purpose: The practical design for experiments with mixtures of q components is consisted in the four types of design points, vertex, center of edge, axial, and center points in a (q-1)-dimensional simplex space. We propose a sequential method for the successful construction of the design space in Quality by Design (QbD) by allowing the different number of replicates at the four types of design points in the practical design when the quadratic canonical polynomial model is assumed. Methods: To compare the mixture designs efficiency, fraction of design space (FDS) plot is used. We search for the practical mixture designs whose the minimal half-width of the tolerance interval per a standard deviation, which is denoted as d₂, is less than 4.5 at 0.8 fraction of the design space. They are found by adding the different number of replicates at the four types of the design points in the practical design. Results: The practical efficient mixture designs for the number of components between three and five are listed. The sequential method to establish a design space is illustrated with the two examples based on the simulated data. Conclusion: The designs with the center of edge points replications are more efficient than those with the vertex points replication. We propose the sample size of at least 23 for three components, 28 for four components, and 33 for the five components based on the list of efficient mixture designs.