Home Health Case Studies in Maintenance and Reliability: A Wealth of Best Practices
This large volume of data presented an opportunity to compute reliability parameters for equipment using our own data. The numbers, based on our operating context and maintenance history, would be far more relevant and applicable in managing risks that we faced in our operations. Hitherto, reliability data from generic sources were used for such efforts as quantitative risk analysis and Reliability Centered Maintenance (RCM) studies.
Many of the items under consideration were subject to hidden failures, i.e., failures that would not be known to the operator under normal conditions. For example, failures of PRVs, ESD valves, firewater deluge valves, gas detectors, etc., would not be known to the operator. If they failed during a test, they would be replaced with a new item or previously shop-repaired item (to as good as new or AGAN) immediately. An applicable reliability parameter to use in these cases is the mean time to failure (MTTF).
The operator can know about evident failures, because they will result in some local effect such as low flow or pressure, a pool of product on the floor, high current, etc. Once the operator knows of the defect, corrective action can be initiated, and the item would be repaired. Such items are termed repairable, while items subject to hidden failures are termed non-repairable. When items are repaired, it could be to AGAN standards as before and in such a case the parameter to use is MTTF. In most cases, however, the repair is to a lower standard than that of a new item. There are many reasons for not being able to achieve AGAN standards in the field. These include, e.g., unavailability of special tools, measuring instruments, skills, controlled environment, etc. The reliability of the repaired item is not 100%, as is the case with AGAN repairs. The applicable parameter for such repairable items is mean time between failures (MTBF).
The formula for calculating MTTF and MTBF is identical. In both cases, we divide the cumulative time in operation by the number of failures in that period. In this chapter, we will use the terms interchangeably, knowing that they are not the same and with apologies to the purists.
Cumulative time in operation = Sum of operating time of all items in set.
Number of failures is the sum of all the failures in the time period under review.
PRVs are bench-tested before cleaning and inspection. Data recorded in the calibration sheets include
a. Leakage, if any, below 90% of cold set pressure
b. Pressure at which valve lifted
c. Simmering or chattering between 90% and 100% of set pressure
d. General internal condition, including e.g., fouling
e. Date of installation
f. Date of removal from service
We collected all the PRV data sheets (archived over the years) from storage cartons, and entered the data into a spreadsheet. The first item (a), namely leakage below 90%, is useful for calculating the mean time between failures (MTBF) for the failure mode—internal leakage. In cases where item (b) is more than 110% of set pressure, these are marked as 'failed to lift on demand.' A count of these failures allows us to compute the MTBF for the failure mode—failed to lift at set pressure. The number of (calendar) days in service is computed as the difference between items (f) and (e).
As far as the operating conditions are concerned, we divided the PRVs into broad groups, such as oil, gas, produced water, and air. We divided them further in pressure ranges e.g., 0-100 psig, 100-500 psig, 500-1000 psig, 1000-2000 psig, over 2000 psig. In selecting the mechanical design features, instead of using the make, model, and size, we used the valve type, e.g., conventional spring-loaded, balanced-bellows, or pilot-operated. While we recorded the manufacturer's name, model number, and size, we did not use this for sorting the data.
For each set of PRVs, sorted by service, pressure rating, and valve type, we computed the cumulative days in service by adding up the days in service for each valve in the set. We added up the number of failure events in two separate lots, namely those that leak below 90% and those that do not lift above 110%.
We divided these MTBF values by 365 to give their value in years.
|< Prev||CONTENTS||Next >|