Overview of Reliability in Structural Ceramics

The failure prediction requirements and the pertinent accept/reject criteria for structural ceramics are derived, and the available failure prediction techniques are examined, vis-a-vis the failure prediction relations , in order to highlight the capabilities and limitations of each technique. The need for additional techniques is thereby demonstrated. The capabilities of the ultrasonic technique are extensively evaluated in order to determine its ability to satisfy the deficiencies in the existing failure prediction repertoire. The prospects are shown to be very encouraging, but the results of several key studies must be awaited before defining the ulti~~ate role of ultrasonic failu·re prediction techniques. I don't intend to emphasize particular failure prediction techniques--that is the intent of the Poster Session and the subsequent talks--but I wish to indicate at the outset that there are several ways of predicting failure in ceramics. It is convenient to separate these into two groups. One group consists of direct techniques, where the defect is detected directly and the fracture mechanics analysis is applied to the defect to predict failure; these include, ultrasonics, microfocus x-radiography, microwaves (see Poster Sess ion) and, of course, dye penetrants. The other group consists of indirect techniques that are particularly pertinent to ceramics. These include overload proof testing (see Poster Session), flaw statistics, and intriguingly, ultrasonic a ttenua ti on. In attempting quantitative failure prediction, we recognize that the defects are irregular in shape and that the defect size range of concern is 10 to 100 microns. What defect features need to be characterized to enable us to predict failure effectively? The first parameter is the defect size, especially the maximum dimension of that defect. The realization that the size distribution of defects in most materials generally has an extreme value form can also be used to good effect. The second defect characteristic is its orientation; normally, in ceramics, the distribution of orientations is essentially random. The third parameter is the aspect ratio of the defect; for which a normal distribution seems reasonable. The fourth feature of great importance, perhaps even more important than the aspect ratio and the orientation, concerns the physical properties of that defect, e .g., the elastic properties, and the thermal expansion coefficient. The sequence involved in predicting failure is in three parts: firstly, we would like to characterize the defect, then the crack evolution from the defect, and finally, the slow crack growth. I shall describe an approach based on a fairly quantitative inspection technique and a more qualitative technique. Starting with a series of samples or actual components containing defects of known size and geometry, we can relate the actual defect size to the interpreted defect size throu~h a specific model and signal analysis procedure (Fig. 1). Of course, this isn't just a line through the origin; because, for different types of defect the inferred size can be different for a specified actual size. Also, there is a lower limit for each type of defect and i nspection technique. ACTUAL OEfta SIZE, •. Figure 1. A plot illustrating a probable relation between the actua l and detected defect sizes for a quantitative ultrasonic flaw detection ~thod. For a more typical technique (e.g., microfocus x-rays or a conventional ultrasonic method}, certain types of defects will never be detected, regardless of their size, and there is a wide range of inferred defect sizes (Fig. 2). For illustration, some data obtained on a silicon nitride material, containing several types of defect, using high frequency ultrasonics and using microfocus x-radiography, are summarized in Table I.