This paper proposes a set of reliability targets that can be used in the design and assessment of onshore natural gas pipelines. The targets were developed as part of a PRCI-sponsored project that aims to establish reliability-based methods as a viable alternative for pipeline design and assessment. The proposed targets are calibrated to meet risk levels that are considered widely acceptable. The proposed criteria are based on a detailed consideration of both societal and individual risk criteria. Two societal risk criteria were considered: the first based on a fixed expectation of the number of fatalities and the second based on a risk aversion function as characterized by a F/N relationship. Societal risk criteria were calibrated to match or exceed the average safety levels implied by current codes. Individual risk criteria were based on published tolerable levels. The target reliability levels corresponding to the three criteria are presented and a recommended set of targets is presented.

Quantitative analysis approaches based on structural reliability methods are gaining wider acceptance as a basis for assessing pipeline integrity and these methods are ideally suited to managing metal loss corrosion damage as identified through in-line inspection. The essence of this approach is to combine deterministic failure prediction models with in-line inspection data, the physical and operational characteristics of the pipeline, corrosion growth rate projections, and the uncertainties inherent in this information, to estimate the probability of corrosion failure as a function of time. The probability estimates so obtained provide the basis for informed decisions on which defects to repair, when to repair them and when to re-inspect. While much has been written in recent years on these types of analyses, the authors are not aware of any published methods that address all of the factors that can significantly influence the probability estimates obtained from such an analysis. Of particular importance in this context are the uncertainties associated with the reported defect data, the uncertainties associated with the models used to predict failure from this defect data, and the approach used to discriminate between failure by leak and failure by burst. The correct discrimination of failure mode is important because tolerable failure probabilities should depend on the mode of failure, with lower limits being required for burst failures because the consequences of failure are typically orders of magnitude more severe than for leaks. This paper provides an overview of a probabilistic approach to corrosion defect management that addresses the key sources of uncertainty and discriminates between failure modes. This approach can be used to assess corrosion integrity based on in-line inspection data, schedule defect repairs and provide guidance in establishing re-inspection intervals.

The optimal design level for onshore natural gas pipelines was explored through a hypothetical example, whereby the pipe wall thickness was assumed to be the sole design parameter. The probability distributions of the life-cycle costs of various candidate designs for the example pipeline were obtained using Monte-Carlo simulation. The life-cycle cost included the cost of failure due to equipment impact and external corrosion, and the cost of periodic maintenance actions for external corrosion. The cost of failure included both the cost of fatality and injury as well as the cost of property damage and value of lost product. The minimum expected life-cycle cost criterion and stochastic dominance rules were employed to determine the optimal design level. The allowable societal risk level was considered as a constraint in the optimal design selection. It was found that the Canadian Standard Association design leads to the minimum expected life-cycle cost and satisfies the allowable societal risk constraint as well. A set of optimal designs for a risk-averse decision maker was identified using the stochastic dominance rules. Both the ASME and CSA designs belong to the optimal design set and meet the allowable societal risk constraint.

This paper proposes a set of reliability targets that can be used in the design and assessment of onshore natural gas pipelines. The targets were developed as part of a PRCI-sponsored project that aims to establish reliability-based methods as a viable alternative for pipeline design and assessment. The proposed targets are calibrated to meet risk levels that are considered widely acceptable. The proposed criteria are based on a detailed consideration of both societal and individual risk criteria. Two societal risk criteria were considered; the first based on a fixed expectation of the number of fatalities and the second based on a risk aversion function as characterized by an F/N relationship. Societal risk criteria were calibrated to match or exceed the average safety levels implied by current codes. Individual risk criteria were based on published tolerable levels. The target reliability levels corresponding to the three criteria are presented and a recommended set of targets is presented.

Offshore codes do not give sufficient guidance regarding design criteria for loads resulting from combinations of stochastic environmental processes such as wind and waves. To assist design engineers in defining such criteria, a suite of methods that use environmental data to calculate the probability distributions of load effects resulting from combination of stochastic loads were investigated. An approach has been developed for using the results to calculate structure-specific and generalized load combination criteria. Extensive application of this approach in connection with Environment Canada’s wind and wave data bases for the Canadian East Coast region formed the basis for some interesting conclusions regarding the process of estimating combined extreme loads on offshore structures. It was found that data based on actual measurements of wave height and wind speed are preferable to hindcast data, since the latter have artificially high correlations that lead to overly conservative results. External analyses are most reliable when 20 or more years of data are used with analysis methods based on distribution tails. Reasonably good results can be achieved with 10 yr of data. Methods based on the point-in-time data and using mathematically convenient assumptions regarding distribution types and process characteristics can lead to large errors if the assumptions made are not substantiated by appropriate data. Load combination solutions are highly dependent on the geographic location and data base. Therefore, a separate analysis should be carried out for the structure and location being considered if possible. Wind and wave load combination solutions are sensitive to correlations and assumed distribution types; closed-form solutions for independent and Gaussian correlated processes can lead to significant errors. If site-specific analyses are not practical, companion action factors of 0.65, 0.60, and 0.55 for return periods of 20, 100, and 1000 yr, may be used for wind and wave loading on slender offshore structures in the Canadian East coast region. For wide structures in the same region, the suggested companion factors for the same return periods are 0.75, 0.70, and 0.65.