The viability of a risk model is dependent on the availability and quality of the input parameters. The data gathering effort can either push the risk model (i.e., the model is built around the available data) or be pulled by the risk model (the model dictates the data needs). The former option is not available to US pipeline gas operators that must now adhere to a list of data items prescribed in 45 CFR § 192.917(b). This work produced data quality diagnostic tools, Key Performance Indicators (KPI), and an approach to incorporate data uncertainty into the different types of risk models. A Data Quality Score was developed to allow the internal stakeholders to assess the suitability of the input database before running risk, which is also useful to demonstrate the progress with the data acquisition effort. Data quality KPIs can be evaluated in many dimensions; their development was based on a review of data quality systems for scientific and engineering processes, which had many coincidences with the parameters in the guideline in API Bulletin 1178. The multidimensionality of the data uncertainty makes the definition of the associated meta data a complex task and a possible issue for the database definition. The sensitivity analysis can be leveraged to assess the data importance and minimize the amount of meta data stored. Finally, a guideline for modifying the risk model to compensate for data with high uncertainty. For probabilistic models, the distribution of the input parameter with high uncertainty needs to be modified depending on the nature of source of uncertainty.
The sensitivity analysis can be leveraged to assess the data importance and minimize the amount of meta data stored. Finally, a guideline for modifying the risk model to compensate for data with high uncertainty: For probabilistic models, the distribution of the input parameter with high uncertainty needs to be modified depending on the nature of source of uncertainty.
Available after the event



