← Back to Use Cases

Dealing with uncertainty in risk based integrity assessments

Name Affiliation Phone Number Email Address
Ujjwal Bharadwaj TWI Ltd. ujjwal.bharadwaj@twi.co.uk
Industrial Sectors:

Various, particularly hydrocarbon and chemical process plants


This use case presents the types of uncertainty involved at different levels of risk based integrity assessments and some exemplars on how such uncertainty is treated. Risk based integrity assessments are increasingly used to optimise inspection and maintenance and thereby maintain plant and equipment within tolerable levels of risk from failures. API 580 and API 581(1, 2) are two commonly used established standards in industry sectors such as hydrocarbon and chemical process for implementing risk based approaches. In the write-up below, apart from using terms from these standards, reference is also made to other documents that offer guidelines in the implementation risk based approaches or assessment procedures in general. ‘Risk’ here is a combination of the probability of failure and the consequence of such failure. The equipment covered by the API standards include pressure vessels, process piping, storage tanks, rotating machinery, boilers and heaters, heat exchangers, and pressure relief devices. There are three levels of assessment. One approach to differentiate these levels is to think in terms of increasing resolution of assessment from Level 1 to Level 3. Another way is to look at Level 1 as providing a system- wide perspective, with Level 2 providing a component-wide perspective and Level 3 focused at the subcomponent level. Yet another way is to look at these levels as increasing in the quantification of risk from Level 1 to Level 3. These assessments invariably need to contend with uncertainty in the inputs required. Level 1 is a screening process in order to determine areas within entire process units that have high risk relative to other areas. The outcome of such an assessment, in its simplest form, is a risk matrix depicting the likelihood of failure on one axis and the consequence of failure on the other for the various components within the system being assessed (figure 1). The likelihood of failure assessment would include assessing factors such as the damage mechanisms applicable to the components and the severity of operating conditions. The consequence of failure entails assessing the different types of consequences – economic, impact on personnel and environment. Level 2 follows the identification of components as high risk from the Level 1 screening process. It entails focussing on each such component and gathering more information on the sub-components involved – a weld for example, is a sub-component. Non-destructive inspection (NDT) techniques may be used to determine the current state of the various components. The reliability (probability of detection) of the inspection technique may need to be considered in Level 2 or Level 3 assessments, and is often challenging. Level 3 assessments are more detailed and could involve, for example, engineering criticality assessments (ECA) of specific welds. Numerical modelling techniques may need to be used and tests required to provide inputs or confirm results. Level 3 assessments are usually resource-intensive activities. These Levels should not be seen in isolation. They are often a continuum of assessment activities enhancing the understanding of the behaviour of plant and equipment. Although the use case is focussed on the operation and maintenance phase of plants, increasingly such methods are used at the design stage itself to design out risks.

2.1 Process Inputs

Types of inputs to the three levels of assessments A good assessment requires all available relevant information to be used. The types of information include:

  • Design information which is specific. It provides system and component information, functional requirements, expected loadings, etc.
  • Databases capturing operating experience that are generic and represent industry wide or plant wide data.
  • Expert opinion that is specific to the situation under consideration.
  • Data [1] obtained through engineering studies that could be generic or specific. Such studies include the use of damage assessment models, often requiring a probabilistic approach.
  • Inspection data that is specific and can help in updating initial assumptions.

2.2 Propagation

Processing of data to determine damage, remaining life of components, inspection frequency, impact of inspection, cost benefit analyses and updating data The uncertainty in the inputs mentioned above could be a) well-understood and characterized statistically with reasonable confidence, b) not well-understood due to limited knowledge about the underlying behavior, or c) a combination where it may be useful to condition or calibrate an input from one source with input from another; (3) gives an approach based on Bayesian methods to combine information from different sources or to update information in light of new data. Two applications are described below: Application 1: Inspection data relating to a metallic pipe susceptible to pit corrosion is available. To determine the remaining life of this pipe it is crucial to know areas where the pipe has minimum thickness. The data is analyzed using extreme value distribution using guidelines in (4). Figure 2 shows the probability of failure associated with a remaining life calculation. Application 2: The situation here is that without any inspection there is a probability of failure over the operational life of a component based on a generic database – see figure 3. With inspection one gets specific information and there is more confidence in the data; in the case shown, the probability of failure curve based on annual inspection is below the more conservative curve based on generic values. The figure also shows how inspections followed by remedial action keeps the probability of failure within acceptable limits.

2.3 Interpretation and Communication of Results



The combination of developments in ICT (Information, Communication and Technology) and sensor technology means that real-time condition monitoring data can be used for such assessments. However, this raises the challenge of analysing big data sets. Increase in computing power means that analyses that were seen to be time consuming or intractable can now be undertaken. This development enhances the ability to conduct multi modal, multi-component reliability analyses.   Image References [caption id="attachment_229" align="aligncenter" width="455"]Screen Shot 2015-10-27 at 17.29.45 Figure 1. Risk matrix showing the risk profile within a process unit[/caption] [caption id="attachment_228" align="aligncenter" width="457"]Screen Shot 2015-10-27 at 17.29.55 Figure 2: Pipeline assessment[/caption] [caption id="attachment_227" align="aligncenter" width="466"]Screen Shot 2015-10-27 at 17.30.16 Figure 3: Use of generic data and specific inspection data[/caption]  


API RP 581 Risk Based Inspection Technology. USA: API; 2008.

API Recommended Practice 580 Risk Based Inspection. USA: API Publications; 2009.

ASME. Risk Based Methods for Equipment Life Management, CRTD Vol 41. Introduction: Risk Based Methods for Equipment Life Management. New York, USA.: ASME International; 2003. p. 3-4.

Health and Safefy Executive. Guidelines for use of statistics for analysis of sample inspection of corrosion; Prepared by TWI Ltd for HSE. UK. 2002.