A data base system has been developed to analyze root causes of failures and unplanned outages in combined-cycle power plants and related equipments. Raw data in the form of plant work orders and outage reports are provided by thirteen utilities. Data encompasses both scheduled and unscheduled maintenance activities involving all plant equipments of mostly combined cycle installations. Primary objectives are to evaluate combined-cycle plant maintenance records to determine root causes of equipment failures, to analyze maintenance data to identify key areas for reliability, availability, and maintainability (RAM) improvement, and to develop an automated data processing system for feedback to project participants.
ERAS (EPRI Reliability Assessment System) data were analyzed to determine failure rates and mean downtimes of critical equipment used in combined-cycle plants and integrated gasification combined-cycle plants of the future. A primary ERAS data requirement is to document plant equipment failures involving planned outages, unplanned outages, and noncurtailing maintenance. Maintenance data are mailed directly from the plant on a monthly or weekly basis. In order to maintain significant coordination and feedback, documentation is also mailed by ARINC Research to manufacturers and EPRI project personnel.
ERAS is foremost a repository, based on dBASE II software, of raw data records which address combined-cycle plant equipment scheduled and unscheduled maintenance activity. By using dBASE II commands, data contained in “DAILY”, “MONTHLY”, “PROBLEMS”, and “FIREHRS” data base files (DBFs) are analyzed in many ways. Data may be separated by manufacturer or utility. RAM statistics may be analyzed according to six data hierarchies, grouping data by common plant design, system, subsystem, or components. Special report forms can be developed depending on specific data processing needs.
During 1982 a total of 1744 maintenance records for the thirteen participating plants were processed. Seventy percent of the events were noncurtailing, of which 612 events involved component failures. The data base included 968 failure events to over 200 different combined-cycle plant components each described by a three digit code. Component repairs accounted for over 50 percent of these failures. Detailed descriptions of these failures identifying causes and failed piece parts can be listed by the computer from the “PROBLEMS” DBF using “FRM’s SYSCOMP” (failures grouped by components) or “SYSPLANT” (failures grouped by plant).
It is concluded that work orders do provide insight into possible root cause and can assist engineering in followup failure investigations, although discussions with power plant personnel are often required for additional insight into root cause. The extent of successful root cause determination in the ERAS data base is difficult to quantify due to different root cause interpretations. However, project efforts have successfully identified the failed piece part in 86 percent of equipment failures. An analysis of plant reliability problems showed that failures of combustion turbine and heat recovery boiler panel controls and remote sensing equipment are the most frequent cause of combined-cycle plant unavailability. Each plant appears to have specific problem areas generic to these control system failures. Generic problems were observed in drum level set points, control valves, panel cards and timers, computers and data links, flame scanners, and thermo-couples. Many of the control related problems documented in ERAS records identify the failed part number of the control circuit affected. It is expected that these documented control system failures will begin to yield better insight into root cause as the number of records increase, and cause and effect relationships are established.
Modeling Control System Failures and Attacks – The Waterloo Campaign to Oil Pipelines