Data-Driven Reliability: Insights from Field to Headquarters

Abstract Objective/Scope The aim of this paper is to demonstrate how collecting and qualifying, in an industrial manner, data reported by field teams in the Computerized Maintenance Management System (CMMS) allows us to improve operational excellence. The application presented in this communication is the construction of reliability databases that are used in risk analyses and Reliability, Availability, Maintainability (RAM) studies to optimize the design and/or the operating philosophy of production and safety systems. Methods, Procedures, Process The international standard ISO14224 and OREDA (Offshore and onshore REliability DAta) consortium provide a set of structuring guidelines for reliability data collection in petroleum, petrochemical and natural gas industries. However, there is a gap between the operational reality on field and the standards, with a heterogeneity of data source implementations and a variety of data entry practices. Thanks to the data-driven workflow we have implemented and a set of quality controls, we are now able to track more than 100,000 pieces of equipment and ensure the correct classification of related events (failure and maintenance) to populate efficiently and relevantly internal reliability databases. Results, Observations, Conclusions The data ecosystem we have implemented has allowed us to save time in data collection. What used to take about 100 man-days can now be done in about 10 days, with the prospect of going even faster with machine learning classification approaches. The time savings allow us to have a more substantial database, updated more regularly, and with a wider variety of equipment items. This enables us to obtain finer and less-conservative failure rates than those of OREDA that is the main reference in our industry (e.g., on safety & environmental critical elements) and more specific to our environments. It is also possible for us to study equipment items not yet covered by OREDA neither by the other data sources available in the literature (e.g., cranes) or to work on specific indicators like "on demand failure rate" for fire & gas detectors. This also allows us to compare different assets from our company and better understand the impact of preventive maintenance on corrective maintenance. By starting using these new failure data in studies, one of the objectives of the company is to streamline better our designs, extend test frequencies, optimize spare part strategies and preventive maintenance programs, etc. Several millions of dollars of savings are forecasted. Novel/Additive Information Improving the design and safety of our facilities through field feedback has concretely demonstrated the value of a data-centric approach. This also gives us better reactivity when we need to integrate "new objects" (e.g., wind turbines, integrated compressors); in our case, a fundamental building block is the use of a standardized taxonomy. Finally, this structured approach provides a better understanding to local operators and maintainers about the extended use of their data entries.