Reliability Model for a Dental Glass-Ceramic with Multiple Flaw Populations

Dental ceramics suffer from intrinsic, processing, and service flaws. It is desirable to predict the lifetimes of all-ceramic prostheses when multiple flaw types present. Objective: To determine a statistical model for fitting fatigue lifetime data with multiple flaw populations. Method: Rectangular beam specimens of a fluorapatite glass-ceramic (IPS e.max ZirPress, Ivoclar-Vivadent) were fabricated by pressing. The polished specimens (15 μm diamond) were subjected to cyclic loading between 4.5 and 45 MPa at 2 Hz and 10 Hz (N=30 each). Tests were performed using a fully articulating four-point flexure fixture in deionized water at 37˚C. Fractographic techniques were used to identify initial flaw types. The lifetime data were fit to a partially concurrent model (Johnson, 1979). Weibull++ software was used to estimate the parameters of the mixed Weibull model. A power analysis was performed by Monte Carlo simulations using the Simumatic tool of Weibull++ to estimate the appropriate sample size. Results: Fractographic examination revealed that failure originated from surface porosity and grinding flaws. The Weibull moduli for surface porosity and grinding flaw populations are denoted by m_p and m_g, respectively, while the characteristic lifetimes of surface porosity and grinding flaw populations are denoted by η_p and η_g, respectively. A sample size of 500 specimens per group was estimated to be necessary for tight confidence intervals for a five-parameter Weibull model. The parameters of the mixed Weibull model are summarized as follows:

	Fatigue 2 Hz		Fatigue 10 Hz
Weibull modulus	mp=0.87	mg=0.72	mp=0.47	mg=1.28
Characteristic lifetime (s)	ηp=114000	ηg=1090	ηp=166000	ηg=274
Proportion	0.32	0.68	0.45	0.55

Conclusion: The fatigue lifetime distribution was dependent on the flaw distribution, and it fit well (KS=0.58%) to a partially concurrent model. The effect of frequency on the model parameters could not be determined due to unfeasible sample size. NIH grants DE013358 and DE017991.