Mission control

With constant pressure to increase productivity, condition monitoring technology is sure to become a vital tool in a plant engineer's portfolio. Mark Venables explains.

With growing integration between the control functions within production facilities and the manufacturing and enterprise systems that monitor them, plant engineers have more ways available with which to keep a check on the condition of machinery.

Monitoring increases overall knowledge of asset condition and allows condition-based maintenance (CBM) to be carried out by scheduling downtime, labour, and materials based on machinery health.

The benefits of a CBM programme are many, and they drive a variety of key performance indicators (KPI) by improving plant metrics, such as return on net assets (RONA), and lower inventory costs and increased overall equipment effectiveness (OEE).

There are many forms of condition monitoring (CM), from sensors for individual items of equipment such as bearings, motors or drives, to fully-integrated, plant-wide systems that monitor the condition of an entire factory or manufacturing process. And its utilisation is growing rapidly.

Electric motors is one area where

CM technology

has become well established. Problems occur in motors for a variety of reasons, ranging from basic design faults and manufacturing quality to application and site conditions. They are most likely to arise from bearing failures — probably the most common cause — with insulation deterioration a close second.

While not trivialising bearing failure's effect on plant operational efficiency, established vibration analysis methods can readily pinpoint the potential for failure and trigger preventive measures.

Electrical problems pose a much more complex task, however, as they require plant engineers to both anticipate insulation failures and determine the cause of motor windings problems after they occur.

Early identification of a weak motor and its replacement or repair during an outage is demonstrably less expensive than plant downtime and rushed repairs.

Fortunately, there are now specific CM techniques available to monitor motor performance, identify early indications of potential trouble and, better still, help to predict and proactively minimise problems by non-instrusive on-line methods.

One novel approach to revealing complex internal problems, developed by

Glasgow Caledonian University

, proposes the use of RF antennae to pick up radio interference generated by the motor, then store, process and analyse resultant data.

There are a number of simple methods — including DC coil resistance test, insulation resistance test, polarisation index test, DC high potential test and surge comparison test — to determine if motor windings are approaching failure. These procedures can reveal whether winding insulation is deteriorating, or if there are localised flaws that will eventually escalate into total failure.

While information from isolated sensors is relatively common, the constant drive to maximise performance from asset investments and automate maintenance processes, means plant operators need accurate data as well as knowledge. Artificial neural networks, in particular those based on novelty detection, can often provide solutions to complex and/or demanding monitoring needs where traditional techniques fail.

Traditionally, the health of industrial plant and assets have been monitored by individual sensors. These have pre-programmed alarm limits based on a prior knowledge or understanding of the changes in condition of the asset.

Artificial neural networks differ in that they are trained with data from multiple sensors. Although this data is known to have some relationship to the condition of the asset, the actual relationship may not be understood or definable. Training allows normal and fault conditions to be recognised. Given the right application, neural network-based solutions can provide excellent results in a relatively short time.

Neural networks learn from experience and are particularly well suited to solutions which are complex or difficult to specify, but where sufficient data can be provided from which a response can be learnt. In addition, a neural network can even be trained to give the right response to data it has not previously seen. This is possible, with expert design, based on its ability to interpolate from a previous learning experience.

'CM applications can provide protective and/or predictive solutions combined with intelligent diagnostics,' said

Oxford BioSignals

industrial business director Paul Nicholls. 'It is likely that potential applications will already have some form of monitoring due to the critical nature of its performance and/or safety requirements. Monitoring the data from a number of disparate sources and measures, such as vibration, temperature and pressure, will allow for fault detection much earlier than single sensor systems.

'For safety systems, a better theoretical framework in the design of controllers for non-linear environments is claimed to improve performance — better specificity, for example. In the process industries there is potential for closer plant surveillance and consequently predictive maintenance, including plant life extension.'

The main difference between using neural networks and pure knowledge-based systems lies in the ability of those networks being able to integrate complex signals at a low level. This provides more information, and allows for more accurate interpolation of parameter settings as well as automation of increasing complex systems.

Another example of an holistic approach comes from

Nottingham

asset management specialist Monition. Responding to the need for increased production efficiency and plant reliability, the company has developed a CM service to meet demanding operational environments.

Monition's approach is multi-technological and includes both on-site vibration monitoring of operational plant and laboratory analysis of machinery lubricants. This not only identifies developing mechanical faults at a very early stage — enabling maintenance to be scheduled well in advance of predicted breakdowns — but also provides organisations with the information to target their maintenance at fault prevention.

Machinery is individually profiled according to application, usage and criticality factors. Monitoring programmes are then developed to ensure that optimum levels of analysis are implemented — but only where and when necessary.

'Far too often we find that condition monitoring has become routine and, in some cases, is totally unnecessary,' said managing director Mike Burrows. 'Monitoring simply for monitoring's sake is costly, and goes completely against the philosophy of CM as an efficient and cost-effective predictive maintenance tool.'

Rapid technological development has widened the potential application of condition monitoring. No longer just a simple indicator of plant health, the powerful prognostic capabilities of CM are the cornerstone in a reliability-focused maintenance approach, aimed not just at predicting failures, but at eliminating them entirely.

Vibration signatures are monitored using the most advanced CSI equipment available. The analysis of this data is refined to such a degree that not only are mechanical faults such as worn bearings and misaligned shafts easily identifiable, but also problems can be detected before they have a chance to contribute to component breakdowns.

With constant pressure to increase productivity, CM technology is sure to become a vital tool and take its place alongside the likes of lean, Six Sigma and 5S.