|dc.description.abstract||This research defined, investigated, and developed a performance reliability concept, which uses on-line performance information and is capable of working in the computer aided manufacturing environment. Furthermore, the conditional performance reliability structure developed is capable of real-time, look-ahead projections of a failure-free "next" cycle. The conditional tool performance reliability structure enables one to maximize system through-put and product quality as well as resources.
In the performance domain, physical performance is a measure that represents some degree of system, subsystem, component or device success in a continuous sense, as opposed to a classical binomial sense (success or failure). If applicable sensing and monitoring means exist, physical performance can be observed over time, along with explanatory variables or covariables. Performance reliability represents the probability that performance will remain satisfactory over a finite period of time or usage cycles in the future. An empirical physical performance function is constructed to incorporate the explanatory variables, operating, and environmental conditions over a time or usage dimension. This function enables one to model device performance and the associated classical reliability measures simultaneously, in the performance domain, when a performance critical limit (which represents an appropriate definition of failure in terms of performance) is set at a fixed level, based on application requirements.
After the performance reliability theory was developed, it was demonstrated through a carbide-tipped HSS drilling tool example. This example was based on cutting forces (thrust) generated while drilling Duralcan aluminum composites. The development included the capability for online, real-time conditional tool reliability prediction as well as traditional classical reliability measures.
In the case of inadequate knowledge of the failure mechanics, this empirical modeling concept along with performance degradation knowledge can serve as an important analysis tool in reliability work in product and process improvement. This research provides an innovative linkage between actuarial and physical based reliability work. Results are expressed in a non-parametric form, with minimal assumptions. The parametric case is demonstrated using the normal distribution to represent performance measurements. Traditional regression analysis and response surface techniques can be used to develop performance function models. The resulting failure density, cumulative failure density, reliability and failure hazard functions are empirical in nature and follow the time or usage dimension of the performance data.||