In this first part of my blog series on Intelligent Maintenance, I introduce some basic maintenance concepts and look at how more sophisticated maintenance strategies powered by AI – whether Predictive (PdM) or Prescriptive (RxM) approaches – can create substantial value for asset-intensive businesses.
Yet, we often – narrow-mindedly – tend to restrict the AI use case in maintenance for these more sophisticated strategies, while there are other ways AI can benefit maintenance objectives even under more basic maintenance models (reactive and preventive). Hence, in the second part on this topic, I will review the constraints of the advanced maintenance strategies and highlight some pragmatic ways in which traditional maintenance models can still benefit from the use of AI (and digital). In the third part of this blog series, I will review some common problem definitions and ML techniques used in advanced PdM, and give some tips on how to best get started with AI.
Managing Director, Prosperitree Consulting
June 2020
Predictive Maintenance (PdM) is one of the most obvious AI application areas for a range of industries with large asset pools, critical and continuous operations, and – as a result – a high-impact maintenance function. In industries, such as Oil & Gas, Mining, Chemicals, Utilities, Manufacturing, Construction or Infrastructure management, well-run maintenance operation costs at least between 2-5% p.a. of the asset replacement value, and Overall Equipment Effectiveness (OEE) losses due to unplanned maintenance range from 3 to 5 percent. Studies show that unplanned downtime is costing industrial manufacturers an estimated $50 billion each year. 1
So, for businesses running at high operating leverage and operating expensive asset pools, increasing maintenance efficiency (labor intensity and productivity, spare parts inventory level, optimal frequency of maintenance activities, efficiency of material usage) and improving asset reliability (uptime, plannability, yield level) can easily represent a double digit per centage uplift opportunity in profit margin. Better maintenance practices also lead to safer operating and maintenance environments (fewer field visits resulting in reduced exposure to hazards, assets run safer, improved compliance with safe operating and maintenance protocols).
So, the stakes are high.
Portfolio approach
Organizations can select from different types of maintenance approaches to increase their asset reliability – the most common ones being reactive, preventive, and predictive maintenance.
Predictive maintenance uses condition-based indicators and alerts to surface maintenance needs only when machines are at risk of breaking down — optimizing maintenance cadence and maximizing machine availability. It uses data from various sources like historical maintenance records, sensor data from machines, and weather data to determine when a machine will need to be serviced. Leveraging real-time asset data plus historical data, operators can make more informed decisions about when a machine will need a repair.
Predictive maintenance takes massive amounts of data and through the use of AI and predictive maintenance software, translates that data into meaningful insights and data points. A key element in this process is the Internet of Things (IoT). IoT allows for different assets and systems to connect, work together, and share, analyze and action data.
IoT relies on predictive maintenance sensors to capture information, make sense of it and identify any areas that need attention. Some examples of using predictive maintenance and predictive maintenance sensors include vibration analysis, oil analysis, thermal imaging, and equipment observation.
Ideally, predictive maintenance allows the maintenance frequency to be as low as possible to prevent unplanned reactive maintenance, without incurring costs associated with doing too much preventive maintenance.
These three types of maintenance each have unique benefits and disadvantages depending on which asset is being monitored, at what stage of reliability the plant is, and the impact of downtime on the business. Since there are many downsides to reactive maintenance, organizations tend to move from reactive to preventive maintenance.
However, this does not mean that reactive maintenance can be completely eliminated. Situations will always arise that demand some sort of reactive maintenance. In spite of an organization’s best efforts and planning, the potential for equipment to break down or malfunction still exists. Your operations are not immune to external forces beyond your control (like impact from extreme weather), requiring ad-hoc reactive maintenance. But you may decide to wait with the replacement of an equipment despite its age and condition (so called “planned” reactive maintenance approach), where the asset is not part of a critical system, or where the cost of routine servicing may not be worth the effort and man-hours in the long term.4
This is where reliability-centered maintenance (RCM) comes into play. RCM is a highly involved process that aims to analyze all the possible failures for each piece of equipment and customize a maintenance plan for each piece. With this strategy, all three types of maintenance are implemented in a way that is best for the equipment being monitored. According to RCM, less than 10 percent of maintenance should be reactive, 25 to 30 percent preventive, and 45 to 55 percent predictive. Enterprise asset management (EAM) or computerized maintenance management system (CMMS) tools can even help you identify the best mix of maintenance strategies for your site.5
The advent of Advanced Predictive Maintenance
Advanced predictive maintenance (PdM), enabled by extensive sensor integration and machine-learning techniques, is one of the most widely-heralded benefits of the fourth industrial revolution. Advanced predictive maintenance has been seen as a killer app for Industry 4.0. The approach combines many of the technologies that underpin the new wave of industrial digitization, such as networked sensors, big data, advanced analytics, and machine learning.6
Advanced analytics algorithms based on information like historical sensor data, maintenance records, or failure mode analyses help define thresholds per asset or component that act as decision criteria in day-to-day monitoring. Asset and component conditions are then continuously monitored remotely. If a threshold is about to be surpassed, an intervention is carefully planned and scheduled. Ideally, the remote monitoring is done in an automated way and integrated within a company’s ERP and maintenance workflow systems. This way, manual monitoring can be avoided, and planning, scheduling, and spare-parts management can be automated according to the assets’ conditions.7
Many strategies for transitioning to predictive maintenance leverage the Industrial Internet of Things (IIoT). However, organizations should first take advantage of existing operations’ data sources before investing in IIoT-enabled sensors or software. Capitalizing on the billions of data points already being generated by Supervisory control and data acquisition (SCADA) and automation systems augments maintenance records front-line workers use daily.
Data from SCADA systems is an untapped gold mine of a facility’s inner workings. The systems gather data that can mirror asset health, like Cycle counts, Gas detection, Runtime hours, Valve status, Electrical readings, etc.8
Also, spot readings at regular intervals using portable instruments by field maintenance or by operators as part of autonomous maintenance can provide valuable datapoints for ML models, without the need to expand IIoT-enabled sensor networks across the entire plant or asset base.
Decisions on when to intervene and perform a maintenance activity based on the condition of the equipment can be rule-based or AI-powered, latter typically denoted as advanced PdM. Rule-based predictive maintenance has traditionally been done using SCADA systems set up with human-coded thresholds, alert rules , and configurations. SCADA systems are an excellent tool for managing operations and resources. However, many SCADA systems were deployed before condition-based maintenance technologies were available commercially.
In contrast, ML algorithms are fed OT data (from the production floor: sensors, PLCs, historians, SCADA), IT data (contextual data: ERP, quality, MES, etc.), and manufacturing process information describing the synchronicity between the machines, and the rate of production flow.
In industrial AI, the process known as “training”, enables the ML algorithms to detect anomalies and test correlations while searching for patterns across the various data feeds.9
When predictive maintenance is working effectively as a maintenance strategy, maintenance is only performed on machines when it is required. That is, just before failure is likely to occur. This brings several bottom-line impacts:
- Reduces the chances of collateral damage to the system
- Reduces the cost of asset failures
- Improves equipment reliability
- Optimizes maintenance intervals (more optimal than manufacturer recommendations)
- Minimizing the time the equipment is being maintained
- Minimizing the production hours lost to maintenance
- Minimizing the cost of spare parts and supplies
- Minimizes overtime costs by scheduling the activities
- Minimizes requirement for emergency spare parts
- Improves worker safety
Predictive maintenance programs have been shown to lead to a tenfold increase in ROI, a 25%-30% reduction in maintenance costs, a 70%-75% decrease of breakdowns and a 35%-45% reduction in downtime.
A CXP Group report says that 90% of manufacturers who implemented Predictive Maintenance in their work noticed reductions in repair time and unplanned downtime, while 80% saw that their old industrial infrastructure was improved.
According to McKinsey, the recent advancements in AI-based quality assurance promise productivity increases of up to 50%. Detection accuracy of defects increases while simultaneously flexibility is enhanced and deployment times decrease. Improvements of up to 90% in defect detection as compared to human inspection are feasible using deep-learning-based systems.11
Real-life examples
A major pharmaceutical and health products company used connective solutions to schedule preventive maintenance on actual runtime, as recorded by their automation system. The results were astounding. On average, preventive maintenance was reduced by 30% and mean time to repair was reduced by 20% through earlier notification of potential faults. All of this translated to almost a 50% increase in asset availability.8
Highly automated operations, such as aerospace or automotive manufacturing, have cut response time to equipment outages by as much as 70% simply by routing alarms to front-line maintenance professionals. Some even achieved a full return on investment within six months through a combination of cost savings and increased efficiency. In one instance, a major aerospace engine manufacturer caught an emerging fault on a critical motor with IIoT-enabled vibration sensors. This gave them three months to plan an effective replacement and avoid unplanned downtime.8
Its core, prescriptive maintenance allows thyssenkrupp Elevator to focus on servicing elevators in the most efficient ways possible to increase uptime. The company’s predictive model can already predict five days in advance when an elevator will shut down because of a door problem. This early warning has proved to be highly accurate with no false positives, making it therefore invaluable to thyssenkrupp customers.
Even before a service technician arrives on site, the expert system advises the technician on the four most likely causes of the problem, based on the data, with 90% accuracy. This means thyssenkrupp technicians can fix an issue on the first visit more than 90% of the time.12
The future: Prescriptive Maintenance
Prescriptive maintenance, abbreviated as RxM, is a maintenance concept that collects and analyzes data about an equipment’s condition to come up with specialized recommendations and corresponding outcomes to reduce operational risks. It takes Predictive Maintenance a notch higher by not only predicting failure events, but also recommending actions to take.
Apart from taking data from existing assets, and providing real-time asset health along with potential defects and prescriptions, prescriptive maintenance algorithms arbitrate multiple prediction instances to rank decisions based on extraneous factors such as cost, criticality, and customer-specific data. By way of such arbitrage, prescriptive maintenance enables predictions to transform into decisions and tangible actions.13
For example, given an equipment running with varying bearing temperature, predictive concepts will tell you when the equipment is likely to fail given its temperature profile. Prescriptive methods, on the other hand, tells you that reducing the equipment speed by a certain amount can double the time before it is likely to fail.
Though RxM is still in its infancy, many thought leaders are considering its potential to become the next level of reliability and maintenance best practice.
Part 2 of this blog series looks at the constraints of the advanced maintenance strategies and highlights some pragmatic ways in which traditional maintenance models can still benefit from the use of AI (and digital).
Part 3 of this blog series reviews some common problem definitions and ML techniques used in advanced PdM, and highlights some tips on how to best get started with AI.
- https://partners.wsj.com/emerson/unlocking-performance/how-manufacturers-can-achieve-top-quartile-performance/
- https://www.proaxion.io/corrective-maintenance/
- https://www.uptake.com/blog/how-ai-is-making-predictive-maintenance-a-reality-for-the-industrial-iot
- https://www.mromagazine.com/features/the-role-of-reactive-maintenance-in-your-overall-maintenance-strategy/
- https://www.reliableplant.com/Read/31679/reactive-maintenance-strategy
- https://www.mckinsey.com/business-functions/operations/our-insights/digitally-enabled-reliability-beyond-predictive-maintenance
- https://www.mckinsey.com/business-functions/operations/our-insights/the-future-of-maintenance-for-distributed-fixed-assets
- https://www.controleng.com/articles/leverage-scada-data-in-maintenance-workflows/
- https://towardsdatascience.com/how-to-implement-machine-learning-for-predictive-maintenance-4633cdbe4860
- https://www.seebo.com/predictive-maintenance/
- https://www.mckinsey.com/~/media/McKinsey/Industries/Semiconductors/Our%20Insights/Smartening%20up%20with%20artificial%20intelligence/Smartening-up-with-artificial-intelligence.ashx
- https://www.plantservices.com/articles/2017/rxm-what-is-prescriptive-maintenance-and-how-soon-will-you-need-it/
- https://assets.ext.hpe.com/is/content/hpedam/documents/a00091000-1999/a00091769/a00091769enw.pdf
About the Author
Dr. Adam Flesch is the Managing Director of Prosperitree Consulting, and a former McKinsey Jr. Partner. He has been in the management consulting space for more than 15 years serving Clients across a wide range of industries on strategy, risk management, and operation effectiveness topics. He is a strong advocate for the wider use of AI in business, from decision-making to front-line automation. He advised an integrated international Oil and Gas company on designing and implementing a lean transformation program across its refinery operations, including the entire maintenance function.
About Prosperitree
Prosperitree Consulting is a boutique strategy consultancy focusing on AI-centered business solutions and traditional management consulting. It helps businesses build future-proof strategies and establish smarter day-to-day decision-making routines while turning their data assets into actionable insights and enhancing their respective capabilities.
0 Comments