Safety-instrumented systems (SISs) are among the most important and effective safety barriers in reducing the likelihood of hazardous events and/or mitigate their consequences to assets (humans, environment, and material assets). This PhD thesis focuses on the reliability of SISs.
The overall objective of this PhD thesis has been to develop new methods and new concepts for reliability assessment of safety-instrumented systems. With the knowledge generated in this PhD project, the decision-makers are able to make more rational decisions related to SIS reliability in design, technology qualification, implementation, and operation, hence to achieve a better strategy for major risk prevention.
This PhD thesis has been a theoretical exercise with the functional safety standards (IEC 61508, IEC 61511, etc.), probability theory, and system reliability theory as bases. SISs in the process (mainly oil and gas) industry have been extensively used as examples and cases, but the reliability assessment methods and models developed during this PhD are applicable to all industry sectors.
This PhD thesis investigates several important issues in SIS reliability assessment, and significant achievements have been made to obtain better SIS reliability assessment results. The main contributions of this PhD project are documented in the form of ten articles, among which, four articles have been published in relevant international journals, two are currently under review and the other four have been presented in peer reviewed international conferences and published in the conference proceedings. In addition to the articles, the results from this PhD thesis are also partly implemented in the 2013 version of the PDS method handbook.
Simplified formulas are the preferred approach for SIS reliability assessment among practitioners, but the current formulas from IEC 61508 and PDS method fail to account for some important aspects such as dangerous detected (DD) failures, non-perfect proof tests, and partial tests. In this PhD thesis, several extensions are proposed such that the new formulas are able to treat the DD-failures, non-perfect proof tests, and partial tests properly, such that the applicability of the simplified formulas is extended. For complex SISs, advanced methods are needed to study their reliability. This thesis points to the Markov methods and Petri nets as promising candidates. These two methods are investigated in depth in relation to SIS reliability assessment and their advantages are demonstrated.
Common cause failures (CCFs) have significant influences on the SIS reliability. Despite the efforts made in the past decades, there are still inconsistency between different CCF definitions and a commonly accepted definitionis missing. In SIS reliability assessment, CCFs are usually modeled by the beta-factor model and the multiple beta-factor (MBF) model without the adequacy of these models being checked. This PhD thesis proposes to define CCF on component and system level separately to harmonize the differences between the current CCF definitions. Based on the new definitions, the adequacy of the beta-factor model and multiple beta-factor (MBF) model are verified with respect to several assumptions, and conservative models are identified for different system configuration.
Human and organizational factors (HOFs) influence SIS reliability, but they have not been systematically studied in the context of SIS. This PhD thesis investigates the HOF influence on the component failure rate by extending the failure rate model in MIL-HDBK-217F such that HOFs are considered. A Bayesian approach is proposed to integrate field data and expert opinion to quantify the HOF influences on failure rate. By using the proposed approach, the company and local influences are considered and better SIS reliability assessment are achieved.
Process demands are threats to the systems safety, at the same time, they also reveal the state of a SIS. Using demands as tests and taking credits from demands in SIS reliability assessment have been controversial topics. The industry wants to use the information about the state of the SIS from an actual demand to support decisions but fears of the possible accidents due to the demand. This PhD thesis systematically investigates this issue, and provides a thorough discussion of the pros and cons of using such a “testing strategy”, and highlights cautions, challenges, and conditions of use. With the material from this PhD thesis, the decision-makers can have a broader and better picture of using demands as tests, and can decide whether and how to use the information from demands in SIS-related decisions without failing to maintain the due safety level.
The functional safety standards classify SISs into low-demand, high-demand, and continuous modes of operation based on the demand frequency, and use different measures to quantify the reliability of SISs working in different modes. The classification and use of reliability measures are, however, lacking of scientific basis, and the practitioners are sometimes confused. This thesis provides a thorough discussion of this issue, and suggests a common approach to integrate demand frequency into SIS reliability assessment with Markov methods so that the demand frequency is considered in the assessment and no classification is needed. This thesis also proposes a common reliability measure, that is applicable to all demand frequencies, to be used together with the common approach.
The standpoint of this PhD thesis is that all reliability and risk analyses are merely tools to provide inputs for better and more rational decision-making, if there is no decision to make, a reliability or risk analysis should never be initiated. Uncertainty plays an important role in SIS-related decisions. Without knowing the uncertainty level of the reliability assessment results, erroneous decisions may be made and an unacceptable risk level may result. This PhD thesis adopts the uncertainty classification from the quantitative risk analysis in nuclear industry, and provides a thorough discussion of each uncertainty category in relation to SIS. It is concluded that the completeness uncertainty is the most important to address in decisions under uncertainty, followed by model uncertainty and parameter uncertainty. To consider the uncertainties in decision-making, this PhD thesis proposes a simple and practical approach to quantify the uncertainty, and hence help to reach more rational decisions.