The chemical processing industry incorporates a broad range of machinery, both rotating and reciprocating, which are monitored using various types of sensors and other advanced technologies that measure equipment performance/health. The determining factor as to what must be monitored and how to do so is the criticality of the equipment, based on its importance to the process and/or operation.
Machinery typically falls into two broad categories, with some gray area in-between:
Centrifugal critical machinery that uses steam or gas turbine drives generally turn at speeds in the 6,000 rpm to 12,000 rpm range, with some expanders turning at 20,000 rpm. These units incorporate hydrodynamic bearings to support the rotor, with extremely close clearances between the shaft and bearings in a nominal range of 6 to 12 thousandths of an inch.
Axial positioning of the rotors relative to the non-rotating parts, likewise, require very tight clearances. Units operating with such tight clearances can respond very negatively to changes, such as rotor balance, alignment, internal shifts, etc., and process changes. Should these changes result in the loss of the clearances, the damage can be substantial – up to and including a catastrophic failure.
These changes almost always manifest themselves as changes in vibration. And because of the potential suddenness at which the changes can occur, particularly with respect to the axial position measurement, they require continuous monitoring systems. Fortunately, most changes are subtle within the vibration characteristics. However, a continuous monitoring system is necessary to detect and alarm on the smallest changes, giving an analyst adequate time to perform early diagnostics and remediation.
Due to the critical nature of these machines, industry has developed several standards, written and used by end user groups, as part of the American Petroleum Institute (API) standards program. API-670 is the standard that outlines requirements for Continuous Monitoring Systems used on these critical machines. So, for those machines considered critical to the operation of a process, providing early detection of potential problems is of utmost importance, and if automatic shutdowns are incorporated, they can mitigate the extent of a serious problem.
Just as critical to some processes are reciprocating machinery, which typically run at low speed in the less than 600 rpm range. Unlike a centrifugal machine having one primary moving part, reciprocating machines have many more moving parts. Loss of these machines can have just as much impact on a plant’s operation as a high-speed centrifugal machine.
These machines often incorporate a greater number and type of sensors, as the design is more complex. Continuous monitoring of the dynamics of such machines can also provide early detection of changes, allowing for early analytics and intervention. In a protection mode, these systems provide an automatic shutdown based on predetermined alarm/shut-down levels. The potential destructive nature of a reciprocating unit is equally as severe as with a centrifugal machine.
Well over half the machines in a processing unit are considered “not critical,” in that complete loss of a machine will have no, or very minimal, impact on the continuous operation of the plant. And virtually all of these BOP machines have spare backups available. Thus, they have not historically been continuously monitored. However, because of their great number, significant maintenance costs can be incurred. Most of them are electric-motor driven with speeds of typically 3,600 rpm or 1,800 rpm. At these speeds, rolling element bearings are almost exclusively used and can wear out quickly.
Instead of continuous monitoring, data on many BOP machines are still gathered via a route-based program. In this approach, an analyst takes periodic samples – typically, vibration readings – via a battery power unit with the data processed by a dedicated server. The concept is based on detecting trends of increased vibration due to bearing degradation, lubrication issues, alignment issues, process issues, etc.
Though generally considered successful, there are several drawbacks associated with this methodology. Of increased concern is the safety aspect of having personnel in or around the operating unit. Several major end users are taking steps to minimize the exposure of operators in the operating unit by locating control rooms remote to the processing unit. Another issue is the duration between data samples, which is typically monthly, with some even quarterly. Much can happen between data samples, limiting the effectiveness of detection.
Considering the two main issues associated with route-based programs – personnel safety and possibly missed data – users are starting to employ various types of wireless or online monitoring systems. For critical machines, the cost of a continuous monitoring system is about equal between the hardware and the infrastructure. With the lower cost of BOP monitoring hardware, the infrastructure cost on a comparable basis is still considered hard to justify. Therefore, the use of wireless sensors is being deployed, as it significantly reduces the infrastructure costs.
While there is a significant reduction in the overall cost of wireless systems, there are some compromises to be considered. Since wireless sensors are battery powered, battery replacement or complete sensor replacement, depending on the hardware selected, are issues to be managed. Battery life is typically managed by keeping the polling frequency at a reasonable rate, but the sampling is not continuous.
At times, there are also issues with “line-of-sight” between the wireless sensors and the gateway, but with a comprehensive survey, these issues can be easily mitigated. So, while wireless solutions are not continuously monitoring the asset, they are significantly better than typical route-based approaches, and there are installation savings to be considered.
When continuous monitoring is necessary, there is another method that employs sensors hardwired to local receivers requiring short cabling runs, subsequently incurring minimal installation costs. The receivers still require a wired connection to a local server; this can typically be done through a fiber optic connection, which has become a very common part of plant infrastructure. The advantage here is in the continuous monitoring over periodic data samples, but not all assets will be critical enough to plant operations to justify the installation costs over wireless approaches.
Software for all three solutions – route-based, wireless, and continuous monitoring – are primarily based on monitoring the various frequencies in the dynamic data, which reflect specific aspects of the machine, e.g. an inner race defect on a rolling element bearing.
Sophisticated systems are being deployed in conjunction with the continuous monitoring/protection systems providing descriptors that further tie the anomaly to a specific problem – cavitation, alignment, inboard bearing, etc. Hence, minimal analysis time is needed for identifying faults.
Today, with properly designed and technology leveraged wireless approaches, the cost can often be less than that of operating a route-based program, and the frequency and quality of data collection is significantly improved. Several companies, including Allied Reliability, are starting to incorporate artificial intelligence (AI) and machine learning (ML) into their technology to further enhance the quality and accuracy of the software performing the first pass analysis of equipment data.
Ernest McCurdy is the Vice Present of Engineers at T.F. Hudgins.
