Where do ‘failure fingerprints’ and ‘machine inoculation’ fit into plant maintenance? Well everywhere, as we shall soon explore.
ARC Research Group asserts that more than 80 per cent of plant failures cannot be detected by traditional preventive age, usage, and wear-based maintenance practices; assets continue to fail on a time-random basis, no matter how frequently they are inspected or serviced.
Many leading manufacturers are turning to predictive and prescriptive analytics to unearth the truth at the confluence of operational and maintenance data to derive deeper insights into the real causes of asset failure.
Today, industries work relentlessly to maximise product yield and quality, lower operational costs by improving plant reliability, increase personnel safety, and strive for ‘greener’ operations. The rise of the Industrial Internet of Things (IIoT) and Machine Learning technologies has encouraged advances in asset performance management (APM). Plant managers who ignore these trends do so at their own peril.
Industry pundits label the grouping of such technologies as APM 2.0 and advise firms to adopt a plant strategy in line with APM 2.0 to overcome operational difficulties, such as:
- Asset reliability and product quality issues, which hinder the ability to drive manufacturing processes harder;
- (Unknowingly) operating assets outside of design and safety limits that causes more than 80% of all equipment breakdowns;
- Insufficient early warnings about asset degradation and failures;
- Legacy maintenance routines that incur unnecessary costs;
- Ensuring that plant and people are protected as the processes are driven to the limit;
- Providing shared responsibilities between operations and maintenance when it comes to asset management.
Such challenges are complex. Despite companies spending millions of dollars on preventive maintenance and risk-based inspections, process disruptions and unplanned downtime continue. ARC cautioned, “The global process industry loses US$20 billion annually from unplanned plant downtime.” The dynamic nature of production processes makes a viable solution elusive. Pushing plants to the limit with thousands of simultaneous and sequential process variations means human-based predictions of patterns leading to failure, are virtually impossible.
Enter APM 2.0, a new portfolio of software innovation, including advanced analytics, ensemble modelling and machine learning technology. The newfound proactive approach allows capital-intensive companies in industries such as energy, chemicals, mining and transportation, to predict events and prescribe accurate remedial guidance. The results help to avoid failures or mitigate the consequences, resulting in greater reliability and availability during a much longer asset lifecycle.
To understand APM 2.0 it is necessary to explore the current information sources – process and asset. Data is generated by real-time process measurements such as temperature, pressure, flow rates, pH, whereas asset measurements are linked to asset status such as drive motor vibration, impeller cavitation, liquid carryover into compressors, valve wear-and-tear, mean time before failure (MTBF) monitoring, intermediate product storage, to name but a few. Using much of this information, and information generated by control systems, SCADA systems, advanced process control software, maintenance regime tools, and the likes, it is possible to derive preventative maintenance routines that attempt to improve asset reliability.
But these are insufficient to prevent time-random asset disruptions as plants are driven to the limit to produce more, higher quality output, at reduced costs, without jeopardising safety.
Sifting through the legions of data generated in the plant, and extracting only that which is meaningful, has proved to be a monumental challenge. Additionally, the volume of data that is available for analysis is growing exponentially, aided by the popular acceptance of Industry 4.0, including the use of the IIoT (Industrial Internet of Things).
Tools are required to mine and analyse this data and predictively ensure greater asset reliability, by prescriptively advising how the assets need to be managed.
Asset Performance Management 2.0 integrates classical descriptive and diagnostic analytical tools, combining them with other asset information, to improve reliability, use machine learning to predict asset failures, and apply advanced analytics to deliver prescriptive actions to eliminate random plant disruptions.
Machine learning is integral to APM 2.0. Distilling volumes of data down to several key latent variables provides powerful predictive and prescriptive models suitable for a variety of uses including both continuous and batch monitoring, analysis, and control. Machine learning automatically mines and discovers the ‘failure fingerprints’ that represent the early indications of an impending failure.
Autonomous in nature, and requiring minimal human resources, this advanced technology constantly learns and adapts to new signal patterns when operating conditions change. Failure signatures learned on one machine ‘inoculates’ that specific machine, so the same condition does not recur. Learned signatures readily transfer to similar machines, preventing the same degradation conditions from affecting them.
Once the failure fingerprints have been identified the Maintenance Engineer can add prescriptive guidance based advanced analytics to avoid potential future failures.
In closing, ARC suggests, “With a good APM strategy, operations and maintenance groups become more collaborative, exchanging information manage critical issues and operational constraints, while improving overall performance. Combining the information from the traditionally separate operations and maintenance solutions improves the effectiveness of both areas, and offers new opportunities for managing risk and optimizing performance.”
By Michael Brooks, Senior Director, APM Business Consulting and Anand Gadgil, Senior Sales Account Manager, AspenTech