Data Analytics for Superior Plant Performance and Productivity Outcomes
14/09/2022 Aspen Technology, Inc
By Mike Brooks, Global Director, APM Solutions, AspenTech
A number of dimensions of plant performance are important in substantiating outcomes. However, it’s been pretty clear that equipment failures resulting in unplanned downtime are one of the most extreme consumers of profit margins in the manufacturing industry. Indeed, in 2016 the National Association of Manufacturers suggested worldwide manufacturing is a $14 trillion a year business. Furthermore, it went on to suggest the big problem is that 10% manufacturing losses are unexpected breakdowns costing $1.4T each year. Likely the number is even bigger now. Today, we have technology solutions that can avoid the failures while also maximising equipment use and driving efficiency and being sensitive to sustainability issues such as personnel safety, environmental concerns, and resource issues.
To power the improvements, the new flavors of condition-based monitoring (CbM) have come a long way from simple calculations, logical expressions, engineering and statistical models. State-of-the-art solutions deploy intense pattern recognition technology to recognise changes in equipment behavior from normal, and the best detect the precise patterns that signify degradation is starting and will lead to failure in a specific time if the condition is left unattended. A huge breakthrough comes from solutions that monitor and detect the process-induced behavior that ARC suggests causes around 82% of equipment breakdowns. Such solutions search intensely for those early-stage destructive forces – the root cause – that instigate degradation and impending failures. We often call them Predictive Maintenance (PdM) solutions since they deliver the insights to plan the appropriate action. Solutions that pre-empt degradation and imminent failures with much earlier warnings, in terms of weeks and months’ notice instead of hours and days, provide more time to consider the appropriate action. Often that time can mean minor service intervention is needed to fix a small problem such as a blocked filter before a bearing failure develops. Better still, products providing early insight into the actual root cause and not just the failure mode may allow a slight process adjustment to eliminate the degradation and need for service altogether.
There are even solutions that now deliver prescriptive guidance to advise operators, reliability and maintenance staff instructions on how to proceed; what process changes to make, or likely root cause and failure modes to advise maintenance planners. However, the fundamental truth is that when superior pattern detection provides more time, it becomes an enabler of greater optionality to solve the detected issue. For example, a 45-day warning of a compressor failure can allow planners and schedulers to make different plans for the outage including pre-building intermediate inventory to sustain the outage, or make different feedstock choices, or purchase spot market finished product to fulfill product delivery commitments. There’s the appropriate time to plan and execute a safe, orderly, and environmentally shutdown. The great percentage of accidents happen during sudden transient and unexpected process behavior such as an emergency shutdown. Those conditions also stress the process so that on a refinery for example that’s the time when the flare valves can lift exposing to the problem to the public and during that shutdown, more carbon may be released to the atmosphere than in a whole year.
It's the AI/ML (machine learning) in the CbM product that makes the solution so powerful, but not just the fact that a product uses AI/ML, but ‘how’ the product uses AI/ML. Trimming an estimated model with ML does not produce the same results. But using AI/ML to exactly measure the patterns from data feeds in dozens (and more) data dimensions and time is far more accurate in detecting precise normal conditions in all circumstances including seasonal weather, different operating conditions, even across batches and different operating conditions. When products have more accurate baselines of normal conditions, they are much more accurate with far fewer false positives in detecting deviations and abnormal conditions. Such solutions can also update automatically without expert intrusion and keep up-to-date even as the process changes.
The uptime or availability is not the only concern of the plant management. Of course, the machines must be available when needed, but they must also provide full optimal performance and produce the appropriate yields and quality of products. The data-driven analytics in solutions may assess yield and quality performance over many years of production to identity the multi-dimensional correlations between operating parameters and likely outcomes. Continuously monitoring can alert when conditions go awry, away from conditions that produce good yield and product and offer prescriptive advice on what to change and how, to bring the operation back to a good track.
Additionally, a key factor for any manufacturer is the cost of doing business and continuous assessment to ensure a balance between the cost of decisions, the risks in those decisions, and the probable outcomes. The equipment in a plant operates as a system in concert with other equipment and the process itself. It’s well known that you cannot isolate the machine from the process or the process from the machine, along with other equipment they operate as a defined system. An implementation of a comprehensive data-driven solution can do a couple of things. First, it may remove opinion from a course of action, especially when the opinion can be misguided and not explicitly supported by factual data – that often in the past were unavailable. Second, now we can use data with process analytics to analyse and examine the precise cost and very probable consequences of implementation of various manufacturing decisions to identify equipment and areas that are inefficient and explain how changes can increase productivity and lower cost.
Products can accelerate performance by concentrating on internal work process that guide the user and abstract the engineering and data science, democratising the product to enable regular staff to use it with what already know and without data science or deep engineering skills. People do more with lower skills because the AI/ML helps them. Similar to the Apple iPhone, the smarts are provided on the inside and not by experts on the outside. The efficacy of data-driven insights and product ease-of-use assures accelerated performance and time-to-cash, increased ROI, besides increased reliability and uptime, resulting in a better yield of end product, more sustainable outcomes and ultimately more satisfied customers.
Print this articleFor more information, please contact:
Aspen Technology, Inc
20 Crosby Drive
Bedford
Massachusetts
01730
USA
Tel: +1 781 221 6400
Email: info@aspentech.com
Web: https://www.aspentech.com
Share article:
Process and Control Today are not responsible for the content of submitted or externally produced articles and images. Click here to email us about any errors or omissions contained within this article.