Probabilistic risk assessment

The Probabilistic Safety Analysis ( PSA), and Probabilistic Risk Analysis ( PRA) called, examines the risks of industrial plants using probabilistic methods. The three main objects of investigation of the analysis are:

  • What can fail?
  • How likely is it?
  • What are the implications?
  • 5.1 Model uncertainty
  • 5.2 Data Uncertainty / parameter uncertainty
  • 5.3 Uncertainty due to insufficient knowledge
  • 6.1 Evaluation of the absolute risk size
  • 6.2 Comparative risk assessment
  • 6.3 Assessment of risk importances
  • 7.1 Security Checks
  • 7.2 Living PSA
  • 7.3 Risk Informed Regulation
  • 8.1 Nuclear Technology
  • 8.2 Chemical, Oil & Gasindistrie
  • 8.3 Railway Industry

History of the PSA

From the experience of the reliability analysis of the fifties and sixties out the method of quantitative risk analysis developed. It uses so alike the methods of reliability (technology).

A pioneer of quantitative risk assessment was in England F. Reg Farmer, the 1967 with the risk limit curve (also called "Farmer - curve" ) created the methodological basis for quantitative risk assessment of industrial plants. The risk limit curve is derived from the product of the probability and extent of damage of an accident and sets based on the consideration that the greater the extent of damage of an accident, the lower should be the probability of occurrence, and accordingly reversed.

Farmer also pointed out that in the risk assessment (eg a nuclear power plant ), the full range of possible accidents is to be considered and not just a "maximum " accident ( " Biggest credible accident " - GAU, "Maximum Credible Accident " - MCA ), as in nuclear technology was common until then.

In the American reactor safety study " WASH 1400 " 1975 (so-called " Rasmussen study" ) the quantitative risk of accidents of two nuclear power plants ( NPPs) were first analyzed comprehensively. The event tree and fault tree analysis is in this case the predominant analytical instruments of the PSA, in which captured all the possible accident sequences, modeled and quantified. Developed in the aviation and aerospace engineering fault tree analysis (see Fault Tree Analysis Handbook ) was taken in it. The risk model was created in the PSA of the entire system will then consist of a variety of interlinked event and fault trees. In large-scale systems such as a nuclear power plant risk model and thereby accruing large amounts of data can be quantified and quality-assured only by means of a computer program.

The Reactor Safety Study " WASH 1400 " was also a significant development of the methods of PPE:

  • Failure models redundant systems and components with redundant cross- faults ( CCF - Analysis, Common Cause Failure).
  • Human Factor Analysis ( acquisition of experience from the military).
  • Determination of probabilistic sizes of rare events, such as the failure of pipelines, vessels, rupture of the power plant turbine, internal flooding, crashing heavy loads, off-site events such as airplane crashes, earthquakes and floods.

She sat with it - not only in nuclear technology - the standard for all other risk analyzes. In Germany the so-called was. " German Risk Study for Nuclear Power Plants" for the Biblis by the methodological approaches of " WASH 1400 " created and applicable in Germany as a reference analysis for NPPs.

Risk analyzes are today in all areas of industry, such as nuclear, aviation, rail, marine, chemical, petrochemical and dams, for use.

Significant risk variables

Human Factor

With the accident at American nuclear power plant Three Mile Iceland from 1979 various weaknesses in the area of human- machine interface, the Personnel Qualification and accident management were visible. The accident sparked worldwide retrofits of NPP and the intensive development of methods for human factors analysis.

Internal fire plant

The Browns Ferry nuclear power plant in Alabama, USA in 1975 developed a severe system of internal fire that led to the failure of several safety systems. The cause of the fire was a burning candle that had the maintenance personnel to detect an air leak used in the cable channel of the NPP. The risk -sized " system of internal fire " was clear by this event, which was underestimated in this form before. In 1400, the WASH event fire has not yet been considered. An intensive development of probabilistic fire risk analysis has been initiated with it. It is now standard in the PSA.

Safety culture

In the Russian Chernobyl nuclear power plant in 1986, the heaviest accident occurred in the production of nuclear energy. Starting point of the accident was an experimental procedure to determine the security properties of the system, which took place during shutdown of the plant. Shortcomings of the experimental program, unexpected conditions during the experimental procedure and unplanned intervention of the operating personnel resulted in the sum to a "prompt supercritical power excursion " of the reactor and thus to its catastrophic failure. The cause of the accident analysis revealed significant deficiencies in the safety management and supervision of the system. The importance of asset management and safety culture ( Safety Culture ) on the investment risk was obvious.

Accident cause analysis in virtually all other industries, but also in medicine and pharmaceuticals, equally disclosed that influence quantity. In all the safety standards of the various industrial sectors, there are now appropriate requirements for a safety management and risk management.

Methods to assess the risk - size " safety culture " exists so far only in qualitative form.

Implementation of a PSA

The PSA is created according to the following steps:

Probabilistic input data

The probabilistic input variables in the PSA are:

  • Frequencies of accidental triggering events.
  • Failure rates and repair times of the components.
  • Non- availability of the subsystems through preventive maintenance.
  • Error rates common cause failures of redundant components ( CCF failures / common cause failures).
  • Error rates of human actions ( HF data).

The probabilistic input data ( plant-specific data) or other similar systems ( generic data ) taken either from the operating experience of the plant in question. The reliability indices obtained at the same time provide indicators ( safety indicators ) on the failure behavior of an installation, in particular with regard to systematic errors, aging processes. They provide an early indication of deficits and opportunity for corrective action.

Uncertainties and limitations of the analysis

The quantitative results of a probabilistic risk analysis are generally subject to uncertainties. We distinguish between the following types of uncertainties:

Model uncertainty

The picture of the real system in the risk model under the conditions of the accident / casualty situations always provides only a rough approximation of the real processes dar. simulations of accident scenarios and the analysis of accident sequences with similar systems used to improve the modeling.

Uncertainty data / parameter uncertainty

The reliability indices are subject firstly to " statistical error " and on the other hand, the so-called. " Technical scattering ". The " statistical dispersion " can be reduced by as large a sample. The "technical scattering " results from the fact that the figures used for data analysis components are not completely consistent in style and in their performance with the consideration in the rule. The consequent uncertainty is greater than that of the " statistical error " usually.

Uncertainty due to insufficient knowledge

The accident experience teaches that the knowledge about the - in a complex industrial plant - potential accident sequences usually not complete ( see Section Significant risk variables ), ie that the risk model reflects reality is incomplete.

Risk Rating

Assessment of absolute risk size

The overall results of the quantitative risk analysis, consisting of the probability of occurrence and the impact of the investigated accident sequences sheds light on the collective and individual risks of the population in the vicinity of the plant.

The evaluation approach "MEM" ( "Minimum Endogenous Mortality " ) is based on the minimum death rate of a person ( in age from 5 to 15 years) of 2.10 -4/Jahr. The acceptable risk should be well below this value and carried at 1:10 -5/Jahr ( European railway standard EN 50126, 1997).

In aviation have 4754 depending on the severity of the effects of an error, the following probabilities are detected by ARP 4761 and ARP:

Comparative risk assessment

A practiced to assess the quantitative risk results is the comparison with other industrial risks or alternative system concepts and systems.

The GAMAB principle ( au moins aussi bon Globalement - Generally at least as good ): a new system should be at least as safe or low risk, as any already existing similar system (see European railway standard EN 50126, 1997). In the chemical industry this term "best practice" is applied.

The " ALARP " principle ("as low as reasonably practical " ) is derived from the principle of proportionality, which is always - and as far as possible and practicable - risk mitigation measures to be carried out.

Assessment of risk importances

The risk analysis provides - on the quantitative overall result beyond - information about individual risk contributions (risk importances ) of systems engineering and operation and thus approaches to optimize them (referred to as vulnerability assessment ).

Applications of PSA

Security checks

In Germany, nuclear plants must be regularly subjected to a security clearance. It comprises three parts, deterministic safety status analysis, probabilistic safety analysis and deterministic security analysis.

Living PSA

In the risk analysis usually the system to be analyzed state at the time of the analysis is determined, frozen ("a snap shot in time" ). Subsequent changes in the plant technology, or new levels of knowledge about the PSA data and models remain in consideration. The object of the "Living PSA", therefore, is to keep the PSA over the lifetime of a plant to date. It finds use in the safety management of a system, for the evaluation of proposed technical changes as well as for training of operating personnel

Risk Informed Regulation

In the U.S., the licensing and regulatory process for nuclear power plants is based largely on the PSA. With the " Policy Statement of the NRC " of 1995 following objectives are set:

  • Extensive application of PSA in all decision-making processes on reactor safety.
  • Improving risk assessment by recent findings from events in the reactor operation.
  • Optimize the distribution of available resources by using risk analysis.
  • Review of systems engineering changes using risk analysis in compliance with the general safety principles.
  • Improvement in PSA instruments, for example by creating a PSA standards.