We need good reliability data to take decisions that will help improve safety, environmental, and cost performance. Such data is not readily available, so we tend to use generic data sources. Some of these data sources have well-defined taxonomies and good control of data quality. Even in these cases, there can be a large spread between the maximum and minimum values. While the operating context and maintenance effectiveness of the items in the data sets will be broadly similar, there will be differences that cause this large spread. In most cases, we end up using mean values, which may be significantly different from those applicable in our own operating context. The solution is to use our own data to develop a reliability database.

We had well-defined maintenance strategies and procedures in the company. With the help of a Computerized Maintenance Management System (CMMS), we issued Preventive maintenance (PM), testing, calibration and condition monitoring (CM) work orders, along with relevant procedures. Technicians recorded repair history in the CMMS and test results in calibration sheets.

We removed Pressure Relief Valves (PRVs) periodically for re-certification, and tested them before and after repairs or adjustments. The pre-overhaul data was recorded in relief valve data sheets. Wellhead valve and sub-surface safety valve test...