• International Journal of Reliability and Applications
• /
• v.5 no.4
• /
• pp.129-145
• /
• 2004

We use a novel, forced censoring technique that closer fits the lower tails of strenth distributions to better estimate extremly smaller percentiles for measuring progress in continuous improvement initiatives. These percentiles are of greater interest for companies, government oversight organizations, and consumers concerned with safely and preventing accidents for many products in general, but specifically for medium density fiberboard (MDF). The international industrial standard for MDF for measuring highest quality is internal bond (IB, also called tensile strengh) and its smaller percentiles are crucial, especially the first percentile and lower ones. We induce censoring at a value just above the median to weight lower observations more. Using this approach, we have better fits in the lower tails of the distribution, where these samller percentiles are impacted most. Finally, bootstrap estimates of the small percentiles are used to demonstrate improved intervals by our forced censoring approach and the fitted model. There was evidence from the study to suggest that MDF has potentially different failure modes for early failures. Overall, our approach is parsimonious and is suitable for real time manufacturing settings. The approach works for either strengths distributions or lifetime distributions.

• International Journal of Reliability and Applications
• /
• v.7 no.1
• /
• pp.27-39
• /
• 2006

Many real world cases in material failure analysis do not follow perfectly the normal distribution. Forcing of the normality assumption may lead to inaccurate predictions and poor product quality. We examine the failure process of the internal bond (IB or tensile strength) of medium density fiberboard (MDF). We propose a forced censoring technique that closer fits the lower tails of strength distributions and better estimates extremely smaller percentiles, which may be valuable to continuous quality improvement initiatives. Further analyses are performed to build an accelerated common-shaped Weibull model for different product types using the $JMP^{(R)}$ Survival and Reliability platform. In this paper, a forced censoring technique is implemented for the first time as a software module, using $JMP^{(R)}$ Scripting Language (JSL) to expedite data processing, which is crucial for real-time manufacturing settings. Also, we use JSL to automate the task of fitting an accelerated Weibull model and testing model homogeneity in the shape parameter. Finally, a package script is written to readily provide field engineers customized reporting for model visualization, parameter estimation, and percentile forecasting. Our approach may be more accurate for product conformance evaluation, plus help reduce the cost of destructive testing and data management due to reduced frequency of testing. It may also be valuable for preventing field failure and improved product safety even when destructive testing is not reduced by yielding higher precision intervals at the same confidence level.