Weibull Analysis Method Applied to Optical Fiber Breaking Stress Data

This paper details application of a 2-parameter Weibull maximum likelihood estimation (MLE) method to optical fiber breaking stress data. Optical fiber is used in a broad range of telecommunications applications and its associated performance in fabricated component assemblies is of critical importance to the proper functioning of telecom networks. Fiber optic components incorporating stripped optical fiber include optical couplers, optical splitters, WDM devices, connectors, and mechanical and fusion splices. The strength of optical fiber in component assemblies is dependent upon numerous aspects of the fabrication process. These include the fiber coating stripping process, fiber cleaning after stripping, any manual handling of the fiber after stripping, and the time period after fiber stripping until the fiber is inserted and potted into a component or subjected to mechanical or fusion splice processes. The ambient moisture content of the air is a critical parameter affecting the resultant strength of stripped fiber. The fabrication and assembly processes utilized to produce component assemblies can cause damage to the fiber and the resultant creation of macro defects. From a reliability and analysis perspective being able to separate the effects of fabrication-induced macro defects from those due to intrinsic micro flaws present in the unstripped optical fiber is critical. The Weibull statistical analysis method permits identification of distinct fiber breaking stress regions associated with intrinsic fiber strength vs. macro-defects created by stripping and fabrication processes. The Weibull analysis is most effectively applied to fiber breaking stress data in the lower 25% quartile since this stress region will be of most interest from a reliability perspective. Another significant benefit derived from the Weibull analysis is the ability to predict cumulative failure rates for any selected values of fiber breaking stress. The predictive capability of the Weibull model provides more useful information regarding the lower breaking stress region than conventional statistical analysis methods such as analysis of variance (ANOVA). The Weibull MLE data analysis method provides more conservative and robust results relative to other Weibull analysis methods such as Median Rank Regression (MRR) [1]. The Weibull MLE analysis method was applied to individual and cumulative fiber breaking stress data sets generated from 10 new fiber strippers of a selected specific type. The objective of these analyses was to provide trustworthy predictions of Weibull unreliability for the selected stripper type in the lower modal breaking stress region of interest for component end-use performance. This paper details the following: • methodology for producing the initial Weibull data plots, • identification of modal regions associated with fiber macro and intrinsic defects, • selection of a representative breaking stress region for application of the Weibull analysis, • generation and interpretation of the two Weibull parameters (β and η), and • prediction of cumulative failure rate or unreliability confidence intervals (C.I.’s) for selected fiber breaking stress values. This study showed that combining the data from all 10 fiber strippers into a cumulative data set and then performing the Weibull MLE analysis is the preferred approach. This method provides the best and most comprehensive definition of the lower fiber breaking stress modal region and more trustworthy predictions of Weibull unreliability at critical selected independent variable values of 50 and 100 kpsi.