Failure databases

Several industry failure databases exist. Analysts gather failure records, make estimates of time in operation and calculate failure rates. The resulting information is published in a book in various forms or provided in a computer database. The main advantage of such documents is that they provide actual field failure based information. 

Several problems exist with this method of getting failure rate data. Often needed information about a failure is not collected. This includes total time in operation, failure confirmation, technology class, failure cause and stress conditions. The results are usually a significantly higher failure rate than the number needed for probabilistic SIF verification. This is due to  

Lack of distinction between random failures and wear out failures, 

Lack of distinction between systematic failures and random failures, 

When total time in operation is not recorded, failures due to wear out cannot be distinguished from random failures during the useful life. If these failures are grouped together, the data analyst cannot distinguish between them and will typically assume that all failures are random. The resulting failure rate number is too high. In addition, the opportunity to establish the useful life period is also lost. 

---

Frequency Phase 3 Use Branch Point Estimates to Develop a Ere-quency Estimate for the Accident Scenarios. The analysis team may choose to assign frequency values for initiating events and probability values for the branch points of the event trees without drawing fault tree models. These estimates are based on discussions with operating personnel, review of industrial equipment failure databases, and review of human reliability studies. This allows the team to provide initial estimates of scenario frequency and avoids the effort of the detailed analysis (Frequency Phase 4). In many cases, characterizing a few dominant accident scenarios in a layer of protection analysis will provide adequate frequency information. 

Eide, S. A. et al., 1990b, Generic Component Failure Database for Light Water and Sodium... 

For the appropriate development of covariate distribution models, the pharmaceutical industry has huge amount of data in their clinical databases. In addition, there are also public databases available which can be used, like the Congestive Heart Failure Database (http //www.physionet.org/) derived from patients undergoing cardiac catheterization at Duke Medical Centre during 1990-1996 (about 4000 patients, data on demographics, risk factors histories, cardiac catheterization, EKG, cardiac scores, follow-up data). 

Journaling is the writing of all before and after images of modifications of the database to a journal file as well as to the database file. The journal device should be a device other than that used to store the database in case of failure. Database Operator utilities (DBO) are provided to specify the after image journal device (DBO/ AFTER-JOURNAL), make backup copies of the database (DBO/BACKUP), restore the corrupted database with the backup (DBO/RESTORE), and reapply all changes since the last backup from the after-image journal to the backup database (DBO/RECOVER). 

Other industry failure database sources include ... 

Signalling Telecommunications equipment failure database Her Majesty s Railway Inspectorate database/annual reports Generic failure and reliability data sources. 

As already discussed in Chapter I in connection with risk matrix, in qualitative analysis, likelihood is estimated and categorized based on experience and judgment applicable for the project. Also these risk categorizations may he done on a quantitative basis as already discussed (say once in a year, etc.). In quantitative analysis the same is done based on previous records or a failure database for which quantitative PHA may be helpful. Failure occurrence data from other plants within or outside the company could be a good source of data. 

In the USA, the nuclear power utilities have a common database (NPRDS), which is only a failure database. It does not include information on planned or unplanned unavailability, which is also necessary for probabilistic safety evaluations. In die US, true reliability and availability data has been collected sporadically by the NRC and the utilities to dicilitate probabilistic evaluation of specific issues. The USNRC is considering rulemaking for owners of nuclear power plants to routinely collect and report reliability data that is suitable for use in PSAs. 

Wierman, T., Rasmuson, DM. Mosleh, A. 2007. Common-cause failure database and analysis system event data collection, classification, and coding. Division of Risk Assessment and Special Projects, Office of Nuclear Regulatory Research, US Nuclear Regulatory Commission. 

USNRC. 1998b. NUREG/CR-6268 Common-cause failure database and analysis system. 

NUREG/CR-6268 (2007) Common-Cause Failure Databases and Analysis System Event Data Collection, Classification, and Coding, Washington DC U.S. Nuclear Regulatory Commission. 

Wang, T, Xuan, W., Wang, X., Ren, K., Overview of oil and gas pipeline failure database. Proceedings of the International Conference on Pipelines and Trenchless Technology, 2013, pp. 1161-1167. 

Potential resources include Naval component databases. Reactor Safety organization knowledge, and the following commercial or government nuclear and non-nuclear component failure databases and reliability resources ... 

The hardware and software used to implement LIMS systems must be vahdated. Computers and networks need to be examined for potential impact of component failure on LIMS data. Security concerns regarding control of access to LIMS information must be addressed. Software, operating systems, and database management systems used in the implementation of LIMS systems must be vahdated to protect against data cormption and loss. Mechanisms for fault-tolerant operation and LIMS data backup and restoration should be documented and tested. One approach to vahdation of LIMS hardware and software is to choose vendors whose products are precertified however, the ultimate responsibihty for vahdation remains with the user. Vahdating the LIMS system s operation involves a substantial amount of work, and an adequate vahdation infrastmcture is a prerequisite for the constmction of a dependable and flexible LIMS system. 

A DBMS performs what is called transaction management. This process allows multiple users to access and store data in the database without cormption. The abiUty to do this is particularly important when data are being written to the DBMS, because power intermptions or hardware failure can cause database transactions to be incompletely processed. Transaction managers use the "all or nothing" principle all the data is written to the DBMS, ie, the transaction is completed, or none of it is written. 

In the simplest terms, a fault-tree for risk analysis requires the following information probabiUty of detection of a particular anomaly for an NDE system, repair or replacement decision for an item judged defective, probabiUty of failure of the anomaly, cost of failure, cost of inspection, and cost of repair. Implementation of a risk-based inspection system should lead to an overall improvement in the inspection costs as well as in the safety in operation for a plant, component, or a system. Unless the database is well estabUshed, however, costs may fluctuate considerably. 

Reliability. There has been a significant rise in interest among pump users in the 1990s to improve equipment reflabiUty and increase mean time between failures. Quantifiable solutions to such problems are being sought (61). Statistical databases (qv) have grown, improved by continuous contributions of both pump manufacturers and users. Users have also learned to compile and interpret these data. Moreover, sophisticated instmmentation has become available. Examples are vibration analysis and pump diagnostics. 

Once the fault tree is constructed, quantitative failure rate and probability data must be obtained for all basic causes. A number of equipment failure rate databases are available for general use. However, specific equipment failure rate data is generally lacking and. 

The presence of errors within the underlying database fudher degrades the accuracy and precision of the parameter e.stimate. If the database contains bias, this will translate into bias in the parameter estimates. In the flash example referenced above, including reasonable database uncertainty in the phase equilibria increases me 95 percent confidence interval to 14. As the database uncertainty increases, the uncertainty in the resultant parameter estimate increases as shown by the trend line represented in Fig. 30-24. Failure to account for the database uncertainty results in poor extrapolations to other operating conditions. 

Duplicate submissions not identified. Facilities sometimes send multiple copies of the same Form R report to insure that EPA received a copy. Duplicate submissions must be identified by printing the word DUPLICATE" in red Ink on page one in the box marked "THIS SPACE FOR YOUR OPTIONAL USE". Failure to clearly identity a duplicate report may result in the duplicate appearance of the data in the database and the appearance of increased emissions from the facility. 

A continually updated database including submitted descriptions of inherently safer technology successes and failures. 

A failure modes and effects analysis delineates components, their interaction.s ith each other, and the effects of their failures on their system. A key element of fault tree analysis is the identification of related fault events that can contribute to the top event. For a quantitative evaluation, the failure modes must be clearly defined and related to a numerical database. Component failure modes should be realistically and consistently postulated within the context of system operational requirements and environmental factors. 

Systems analyses are like formulas, they have little usefulness until the variables are assigned probabilistic numbers from nuclear or chemical data bases. These data concern the probability of failing vessels, pipes, valves, instruments and controls. The primary difference between chemical and nuclear data is that the former may operate in a more chemically active environment, while the later operate in radiation. This chapter addresses both, but most of the data were gathered for nuclear systems. It covers 1) failure rate databases, 2) incident databases, 3) how to prepare failure rates from incidents, and 4) human factors for nuclear and chemical analyses. 

Failure rates for generic E E equipment -IEEE 500 (1977) r hi method to database devcloj... 

Appendix HI, of WASH-1400 presents a database from 52 references that were used in the study. It includes raw data, notes on test and maintenance time and frequency, human-reliability estimates, aircraft-crash probabilities, frequency of initiating events, and information on common-cause failures. Using this information, it assesses the range for each failure rate. 

Commonly, there are components that are not in any database of failure rates, or the data do not apply for the environment or test and maintenance at your plant. In addition, site specific data may be needed for regulatory purposes or for making the plant run safer and better. For both cases there is a need for calculating failure rate data from incident data, and the mechanics of database preparation and processing. 

The mode of failure in a test is examined carefully before the failure is included in the database. In the diesel-generator example, unsatisfactory performance may have been reported because of a trip on a low oil pressure signal, high oil temperature, or both. 

Table 4.1-5 shows in these records, there has been 3 control rod drive failures. Assume 100 plants in the U.S. with an average of 30 control rods/plant and 10.7 years of experience in this database. Estimate, the mode, 90% and 10% confidence limits for the failure rate. 

The nuclear equipment failure rate database has not changed markedly since the RSS and chemical process data contains information for non-chemical process equipment in a more benign environment. Uncertainty in the database results from the statistical sample, heterogeneity, incompleteness, and unrepresentative environment, operation, and maintenance. Some PSA.s use extensive studies of plant-specific data to augment the generic database by Bayesian methods and others do not. No standard guidance is available for when to use which and the improvement in accuracy that is achieved thereby. Improvements in the database and in the treatment of data requires, uhstaiui.il indu.sinal support but it is expensive. 

This folder contains three files bnlgener.xls, bnlgener.wql, and referen.txt. The database, bnlgener is an eclectic collection of 1,311 failure rates from 31 references. The database is provided as a spread sheet in the Microsoft Excel 4 format (xls) and in Corel Quatro-Pro format (wql). Tbis information is not provided in ASCII because of format... 

---