The step-change from reactive to predictive condition monitoring
Condition monitoring has gained prominence over the years as equipment owners continually prioritise asset utilisation and productivity. For condition-based maintenance to be possible, it is essential to have knowledge of the machine condition and its rate of change with time. Thus the main function of condition monitoring is to provide this knowledge, Munesu Shoko.
The crudest way of operating machines is to run them until they fail, and then try to repair them in order to make them fit for further service. This method of operation can be very costly in terms of lost output and machine destruction, and in addition can involve hazards to personnel. It is now well recognised that, particularly in the case of large and expensive plant, it is more economical and operationally satisfactory to carry out regular maintenance.
Just how important is condition monitoring in optimising asset utilisation and productivity? The main savings which can be made by the application of condition monitoring to industrial machinery arise by avoiding losses of output due to the breakdown of machinery, and by reducing the costs of maintenance.
Lourens Zeelie, online condition monitoring specialist at SKF Group, says the importance of condition monitoring has gone from an optional, nice to have, to a requirement on new plants. “Asset utilisation and productivity can be significantly increased by knowing the condition of the asset. Preventative measures can be taken to prevent failures. Similarly, assets can be utilised for much longer by knowing the exact condition and at what point it is expected to fail,” says Zeelie.
Arveen Gobind, asset reliability executive at Martec, says condition monitoring is fundamental in assisting asset owners to determine the condition of their assets and improve reliability and availability. “Condition monitoring provides information specific to an asset, which is used as input to predict potential failures and a proactive approach to maintenance,” says Gobind.
James Cowling, CM product advisor at Martec, says assets that remain in good condition will ensure that the designed process that they are serving is in an optimised state. “Knowing the condition and the risk status of your assets enables better production decisions. For example, defects on a pump that are identified by using ultrasound and vibration can help the production team to make an informed decision to avoid further damage on the main pump instead,” says Cowling.
Gobind adds that condition monitoring systems further give advice on improvement areas, which allows maintenance/repair time to be allocated accordingly and with a more defined scope.
Available industry figures show that the global use of predictive maintenance has risen from 47% to 51% in the past five years, while running equipment to the point of failure has dropped from 61% to 57% during the same period. What is the state of affairs locally? Zeelie says there is definitely an increased interest in predictive maintenance systems. He reasons that on some new plants it has become a requirement in the design phase that assets be equipped with the necessary monitoring equipment.
“Existing plants also show an increased interest due to tight budgets and high production requirements which do not allow for unexpected downtime or maintenance. Despite the rise in interest, some existing plants are not yet willing to commit to the capital costs involved in installing a monitoring system,” says Zeelie.
Cowling says over the last couple of years, the traditional view that condition monitoring is primarily applicable to rotating devices (bearings, gearboxes pumps and fans, among others) has changed. “With the continued improvement in technology, sensors and condition monitoring applications, we see significant growth in electrical condition monitoring on MV switchgear, cables and transformers,” says Cowling. “These are critical assets to many of our clients and just as strategic as the large mechanical equipment assets. Plant owners are thus now in a position to improve overall equipment reliability, resulting in sustainable plant integrity.”
Gobind adds that the rapid expansion of condition monitoring technologies and the integration with IIoT applications is further evidence that companies see the benefit of condition monitoring as a primary technology to optimise asset utilisation and productivity. “Constant economic pressure forces asset-intensive organisations to work smarter to avoid costly breakdowns that hamper production and service delivery and ultimately reduce the useful life of their assets. By predicting failure modes and applying corrective actions, asset owners reduce the risk of failures and high repair costs while extending reliable asset life,” says Gobind.
Towards smart management
Machine condition monitoring has made its evolutionary turn towards smart asset management. Cowling says historically the primary condition monitoring tools were vibration, tribology (oil sampling) and in some cases thermography (infra-red technology).
“This was traditionally performed as a service or periodically by using handheld equipment or in some instances installing permanent sensors on critical or strategic assets. However, the condition monitoring data was not always accurately captured or immediately made available for the plant engineer to utilise,” says Cowling.
According to Gobind, with the onset of IIoT, the significant technological changes are being able to connect condition monitoring sensors (vibration, temperature, DGA, moisture, debris, energy, impact, among others) to web-based agnostic edge processing devices that allow the client to see the condition of their asset in real time and online.
“At Martec, we launched a new service called the Reliability Nerve Centre (RNC). The RNC is an online monitoring solution where key parameters are selected and monitored, and which provides insight into the assets condition and risk, related to defect development and potential asset failure,” says Gobind. The data obtained from the monitored parameters is turned into useful information, utilising engineered algorithms. The algorithms even provide automated digital notifications with recommended actions based on failure modes to improve reliability and availability.
“We also perform advanced diagnostic analysis on critical alarm levels – being significant deviations or breached thresholds. This is a huge benefit as more than one parameter is usually monitored on a critical asset, which requires an expert’s understanding of when the combined result of these parameters require the asset to be pulled for maintenance. Such informed interventions are vital in the prevention of unplanned outages that could result in significant production and financial losses for the organisation,” says Gobind.
Zeelie says SKF is involved in some large projects aimed at smart asset management. Inputs from multiple systems are combined at a central point for analysing. With the available information a very good understanding of the asset condition can be achieved.
“With this information available, preventative maintenance strategies can be put in place. In the long term it can be possible to determine whether it is maybe more feasible to replace the asset rather than doing maintenance. The aim of this is to have the most efficient plant running at the lowest, most feasible, cost,” says Zeelie.
The rise of AI
Artificial intelligence (AI) in machine condition monitoring is said to be gaining ground. Zeelie says AI is much spoken about and definitely being explored. He is of the view that it is not yet at a stage where it can be implemented without human intervention. “Currently available systems rely on algorithms and mathematics to simplify the diagnostic process for humans. Condition monitoring, especially Vibration Analysis, is a complex field that still requires human interpretation. Existing systems can however alert humans to possible defects,” says Zeelie.
Cowling says AI and Machine Learning are going to play a significant role in the asset management realm in the future. Currently, the bulk of analytical data processing, diagnostics and reporting in many cases are still assessed by subject matter experts. However, there are already some of the basic fundamental condition monitoring signatures or thresholds, which with engineered algorithms have been developed to provide automated failure mode identification, resulting in the appropriate corrective action being recommended.
The analytical models and algorithms behind AI, in some cases, can require a significant amount of condition monitoring data to be developed and fine-tuned. With the deployment of more online solutions and IIoT based CM applications, the speed at which this data is being generated and analysed to speed up the research and development processes will significantly reduce the timeframe to implement AI in machine condition monitoring applications.
Over time the implementation of AI models and algorithms may progressively replace some of the more basic analytical work currently being performed by condition monitoring practitioners. “However, during this initial development, we predict that the subject matter experts will stay in demand to address consultative issues,” says Cowling.
“At Martec, our RNC currently has automated algorithms running to provide warnings and alarms to clients via SMS or email and as we continue to gather more data as part of our research and development programme within the larger Pragma group, we will look to introduce our own AI platforms and provide these to our clients as part of a larger reliability centred asset management solution,” adds Gobind.
The condition monitoring equipment market is segmented by product type, for example, vibration monitoring equipment and lubricating oil analysis equipment, among others. According to Zeelie, the most prominent type of condition monitoring is Vibration Analysis. Another important method is Lubrication Analysis.
“When monitoring an asset, defects will first be picked up by Lubrication Analysis. Many plants do Lubrication Analysis based on a schedule and manual sampling. However, online lubrication analysis systems are also available. SKF offers online Lubrication and Vibration Monitoring systems, but commonly only receive enquiries for online Vibration Monitoring systems. It must be noted that online lubrication analysis can only be done with oil, which is only used in special applications where grease cannot be used,” explains Zeelie.
SKF has a full range of online condition monitoring systems, including a protection system for critical assets such as turbines. There are also different sensors for the different applications ranging from Accelerometers to high accuracy Eddy Current probes.
Gobind says historically in local markets, the bulk of condition monitoring technology in the industrial and mining sectors has always been dominated by mechanical applications, focusing on mechanical systems due to their higher failure rates, hence vibration, tribology and thermography are still very popular.
“In the last few years, the introduction of ultrasound as a new complementary technology to existing condition monitoring programmes has been growing and used for different applications and not just purely mechanical or rotating equipment and applications ( bearing lubrication control, compresses air leaks, MV electrical substation, hydraulics, pneumatics and valve applications),” says Gobind.
Cowling adds that condition monitoring of medium and high voltage electrical assets in the power generation, transmission, utilities and mining sectors has been growing at a steady pace. Asset owners are looking to monitor their electrical networks and assets due to the adverse effects of failure resulting in significant financial loss and personnel safety. “Condition monitoring applications related to electrical assets such as transformers, switchgear, cables, generators and motors, among others, has continued to expand and grow,” says Cowling.
“IIoT will continue to drive condition monitoring applications and technologies, and as more companies look to digitise and implement IIoT as part of their Smart Asset optimisation drive, more equipment and assets will have condition monitoring sensors or systems installed,” reasons Cowling.
“At Martec, our primary focus is to assist our clients to improve the reliability and availability of their assets. We provide CM sensors, monitors, products, services and in-time online monitoring solutions for nearly all asset and plant types,” concludes Gobind.
First published in the Capital Equipment News.