The digitalization of oil and gas facilities is becoming a new technical arena. Effective solutions can be used to convert data into information and knowledge, which can then be used to improve maintenance operations. This paper discusses several aspects of this process, ranging from a discussion of maintenance strategies to the opportunities presented by extracting new information from big data.
Fieldbuses, device diagnostics, and advanced management-and-control systems collect large amounts of data, but acquiring and applying methods of exploiting the data have lagged. End users have expressed doubts that they are realizing value from these solutions and have wondered whether a return to simpler systems is needed. In this paper, the authors conclude that condition-based maintenance can reduce the cost of maintenance operations significantly and that further potential in predictive-maintenance regimes exists as experience with base data is gained.
Reactive, Scheduled, and Condition-Based Maintenance Strategies
A reactive, or break-and-fix, type of maintenance strategy will often present the lowest maintenance operation cost, seen in isolation. But also implied is the cost of production unavailability, the safety risk posed by very hazardous events, and the risk of high repair costs after catastrophic equipment failure.
Scheduled or periodic maintenance requires one to estimate failure modes and consequences for all equipment in the plant; then, on the basis of the equipment’s expected lifetime, one calculates inspection intervals and replacement cycles. Because this is often impractical to perform for each individual piece of equipment, equipment is divided into classes depending on type and is given maintenance that is based on that type and its criticality. This approach can have some undesirable effects, including the following:
The 80:20 rule may apply; it states that 80% of maintenance intervention is caused by 20% of equipment.
A conservative approach that prioritizes avoiding operational failures will lead to short inspection intervals and high maintenance costs.
Many sources conclude that human error is the source of as much as 50% of all maintenance issues. Thus, an excessive maintenance intensity will by itself lead to additional failures and the need for even more maintenance.
Thus, a transition to condition-based maintenance is desirable. For this form of maintenance to be possible, several conditions must be satisfied:
The fault condition must be detectable (i.e., one must have a fault progression that allows detection with sufficient time to allow corrective action).
The equipment must have sufficient instrumentation to allow parameters to be observed.
A model of the equipment must exist that can collect information about fault progression and measured data and form a sufficiently accurate “digital twin” to assess equipment condition.
A maintenance strategy and procedure must be implemented that can take advantage of this information.