Failure statistics

In many cases, a two-parameter Weibull distribution is assumed, where Po ( -1) t e simplified as follows 

4- point bending (L =inner span, L =outer span) 

a plot of the left side of Eq. (9.3) as a function of the natural logarithm of strength should yield a straight line. In such a procedure, a failure probability is needed for each test specimen. This is usually estimated using 

An important feature of the above strength analysis is that a large number of test specimens need to be broken before the Weibull parameters can be known with any accuracy. For example, one often needs to break at least 30 specimens before m is known within 20% and r within 5%. It is, therefore, not unusual to require 30 to 50 specimens in a strength testing program. 

To calculate the relationship between the strength (a ) for one loading geometry L f) and specimen size (F,) and another (tr, L p, V2), Eq. (9.2) can be written for fixed failure probability as 

---

Annual Reports of Gumulative System and Gomponent Reliability for Period from July 1, 1974, through December 31, 1982,serves as a source of engineering and failure statistics for the nuclear industry. It contains data for most components used in nuclear power plants. 

The objective is to estimate, numerically, the probability that a system composed of many components will fail. The obvious question is, "Why don t you just estimate the failure rate of the system from operating experience " There are three reasons IJ the system may not exist, so new data are not available, 2) the injuries and fatalities from the developmental learning experience are unacceptable - the risk must be known ahead of time, and 3) by designing redundancy, the probability of the system failing can be made acceptably remote in which case system failure data caimot be collected directly. The only practical way uses part failure statistics in a system model to estimate the system s reliability. 

Virolainen, R 1984, On Common Cause Failures Statistical Dependence and Calculation of Uncertainty Disagreement in Interpretations of Data, Nuc. Eng. and E 77 pp 103-108. 

Pressure Vessel Failure Statistics and Probabilities Nuclear 4 tables containing failure data for vessels Primarily concerned with boiler failures 113. 

Transmission and Gathering Lines—1970 through June 1984. Pressure Vessel Failure Statistics and Probabilities. 4.7-21... 

At the PC level, the freeze-drying process and all secondary processes are visualized, controlled and documented. Such secondary processes include condenser defrosting, CIP/SIP, plant tests, failure statistics, calibration values, counter of operating hours, plant schematic, table of valves, actual temperature and pressure values, recipes, start parameters and machine parameters. 

FIGURE 11.15 DLC Weibull overpressure failure statistic during a 2.85 VDC voltage solicitation at 70°C. Power factor p = 4.5 and mean lifetime HX0 = 440 h. 

At the production line in question, the profiles surface quality and cross-section geometry deviation was already analyzed and recorded by online optical quality inspection systems. These camera based systems usually mark defective profile sections visibly, to separate them out for the operating personnel [910]. In some cases, this data was analyzed to quantify the amount and kind of failures statistically to control the rate of production faults. For an integrated management of process, user interaction, and fault information, and quantitative quality data, the records of these inspection systems also had to be integrated. 

R, Danzer, P. Supancic, and T. Lube, "Failure Statistics Beyond the Weibull Behavior",... 

It is useful to introduce failure statistics into the time dependence of strength and this can be accomplished by stating the initial inert strength in a Weibull form,... 

Stress-probability-time (SPT) diagrams incorporate the time dependence of strength into failure statistics. They give lifetime predictions. An illustration of the use of SPT diagrams is in bioceramrcs. 

Two things that should be remembered about all statistics, and especially failure statistics, are statistical significance and applicability. The more data we have, the more certain we are that the statistics have meaning. That measure has been fully developed and is called statistical significance. With little data, one cannot be sure about the accuracy of the result. 

That is why the following approach is chosen. The information of the data bank of the Federal Institute for the Environment ZEMA enables one to derive an expected value of 6.4 x 10 per plant and year for an accident with harm to persons outside the premises of plant. This value is arrived at using Bayesian zero failure statistics (vid. Example 9.6). It is about one order of magnimde smaller than that for a fatal occupational accident in the chemical industry, for which about 2.2 X 10 a is obtained (vid. Eig. 1.1). 

Bloch, H.P. 2002. Looking for RCFA databases Consider failure statistics. Hydrocarbon Processing, Vol. 81, No. 1 35-37. 

It is the destiny of HRA specialist that the results of his analysis are considered being more subjective than elaboration of other, objective PSA data (hardware failures statistics). As soon as the PSA results indicate high level of operational risk, there may be conclusion that such unfavorable result is not reflection of reality, but it has been caused by too high level o f conservatism of HRA expert. As a consequence, HRA expert may need to justify his results much more in detail than other members of PSA team. 

The MTBF estimation based on the results of field failure statistic carried out for the fleet consisting of 468 vehicles during more than 3 years, has demonstrated that the requested MTBF values for the significant und major failure categories are satisfied. 

Weibull [42] proposed a failure statistics of a uniformly stressed fiber sample of length L of ccmstant diameter. He represented this by an empirical equation associated with a unimodal distribution. 

There is the prospect that loss of coolant and hence core meltdown can be made incredible and that eventually designs will be released from these serious limitations. However, in the current work, the aim has been to determine how seriously the design is handicapped even if you allow for the unbelievable. Public acceptability of the first large, fast reactors may well require designs that meet the consequences of the meltdown accident. In this context, then, we must strive to understand the reactivity implications of voiding in sodium-cooled reactors. The first reactor of this class will establish failure statistics of control drives, demonstrate independence of different sources of failure, and thereby permit subsequent assertion that failure to scram is impossible. 

Table 5.3 displays the determined failure data of the above-mentioned electronic boards. The statistical confidence area was determined with a degree of safety amounting to 95%. The majority of the failures was found in those electronic boards which were used in the input and monitoring areas of the control system. Boards of the input level are those electronic boards marked contact conversion in Table 5.3. At the monitoring level, the board types nonequivalence monitoring, signalling card, and display module" are used. The following failure statistics are examples of the types of failures found in boards from one manufacturer ...

The failure statistics consider 20 different operators or utilities. As a result of this experiment, it may be safely said that the majority of the boards had been destroyed as a consequence of careless test work and by external influences. Thus the board manufacturers have developed short-circuit-proof electronic boards to replace the old types. 

Clech, J-E, Noctor, D. M., Manock, J. C, Lynott, G. W., and Bader, F. E., Surface Mount Assembly Failure Statistics and Failure FTeeTime, Electronic Components and Technology Conference, 1994, pp. 487 97. 

Glass is the dominant substrate material in the display industry. The brittle failure behavior of glass is exploited in production processes, and taken into account in applications. Knowledge of the failure behavior, and the mechanisms leading to failure, is quite important in the production of reliable products. An appropriate combination of stress distributions, failure statistics and crack extension models is used in reliability models for brittle materials. 

Figure 6.2 Failure-statistics of large chemical process plant in Germany (a), and in the United States (b).

In early operational years of a wind farm, accurate annual failure rate of the components is not available yet and therefore, data from a generic reliability database, like Reliawind database, will be used. After first years of operation, the annual failure rates of the components ean be updated using the actual failure statistics in the farm. 

---