ARC Advisory Group developed its Asset Performance Management (APM) concept and model almost a decade ago to provide industrial organisations with a framework to analyse their asset management needs, develop effective strategies, and ultimately optimise their asset availability and utilisation.
Asset Performance Management 2.0 now incorporates emerging industrial IoT (IIoT) approaches and technologies and new analytics solutions. It also uses information from production management and control systems in asset management applications to provide new opportunities for optimising asset performance.
Reliability studies show that, on average, 82% of assets have a random failure pattern. Traditional preventive maintenance assumes that the probability of equipment failure increases with use, which applies to just 18% of assets. IIoT and analytics, using engineered algorithms and/or machine learning techniques, provide a new means to predict failures. This can enable organisations to drive down unscheduled downtime to near zero through predictive and, ultimately, prescriptive methods. This can have a positive effect on a broad range of KPIs.
Consumer electronics, including more than a billion mobile phones, drove the creation of infrastructure and provided economies of scale for the sensors, networking and cloud computing used in IIoT. APM 2.0 leverages this new capability to enable new business processes and business models in industrial organisations.
APM 2.0 incorporates IIoT, analytics, and other predictive and prescriptive technologies to bring performance to a higher level. It provides a means to systematically improve key metrics like uptime, mean time to repair (MTTR), asset longevity, cost, quality/yield and safety for maintenance; as well as on-time shipment, quality, and inventory for operations. This optimisation crosses functions to reduce or even eliminate traditional inefficiencies, waste, and dysfunction.
Aligning production management with maintenance management
Asset performance management helps align production management (making the product) with asset management (ensuring the capability to produce).
This enables increased visibility, collaboration, and communication for higher productivity, reduced risk, and improved return on assets (ROA). Goals and objectives become more clearly communicated and shared.
An APM 2.0 strategy includes information sharing and application integration among enterprise asset management (EAM), manufacturing operations management (MOM), plant asset management (PAM), and other solutions to provide a comprehensive view of production and asset performance.
It incorporates IIoT, advanced analytics, and other predictive and prescriptive technologies to reveal new opportunities to optimise asset availability and production. It includes an improved understanding of risk, with fact-based risk assessment. Integrating information from production and asset management applications provides new opportunities for organisations to balance operational constraints (capacity, inventory, labour, etc.) and improve ROA.
Impact of industrial IoT
Industrial IoT provides the enabler for next generation APM 2.0. The wealth of data provided through IIoT, combined with analytics, opens new opportunities for improving asset performance.
Industrial IoT essentials
IIoT solutions typically include three major components:
• Data acquisition from various systems, equipment, devices or sensors. For the owner-operator, the data usually comes from sensors attached to the control system and passes into an historian or other time-stamped database. For an OEM monitoring its installed equipment, the sensor is usually part of an intelligent device with a processor, memory, and small software applications. In both cases, the data can be process values (pressure, temperature, flow, etc.) or asset health information.
• Communications, networking and security: Usually, information is transferred in a hierarchal manner from sensor to control system to cloud applications. However, peer-to-peer communications among machines offers interesting possibilities for energy management or process coordination that have yet to be applied broadly.
• Cloud applications: Currently, the dominant use of the data involves analytics to predict equipment failure so that repairs can be made prior to a fault resulting in unscheduled downtime. Some applications identify production or operating issues that require attention. When conditions warrant, an alert goes to operations or maintenance. Ideally, business process automation initiates an appropriate workflow in the applications used by these functions.
New class of PaaS applications with analytics and Big Data
Historian and analytics software solutions have been around for at least three decades. In that time, the effect of Moore’s Law has vastly improved data storage and computing capacity. Also, the number of intelligent sensors and devices feeding data has exploded. This confluence of Big Data, low-cost cloud computing, and advanced analytics enables a new class of applications.
Analytics have become pervasive for finding value that was “hidden” in the data. Standardised platforms - like IBM’s BlueMix, PTC’s ThingWorx, GE’s Predix, and Microsoft’s Azure (plus many other more focused IoT applications) – reduce the engineering and programming investments needed to implement a solution. Implementation costs and ease-of-use for these PaaS (platform as a service) solutions have significantly improved, increasing acceptance.
Avoiding unplanned downtime
A time-stamped data repository like a historian brings pervasive access to operating data for IIoT applications. Most industrial organisations have critical equipment for which unplanned downtime would disrupt operations. A small undetected problem could cascade into a much bigger issue – much like how failure to spend $25 for engine oil could lead to your car’s engine seizing, requiring a $5000 repair. IIoT enables new techniques for condition monitoring and predictive maintenance (PdM) to enable organisations to identify issues that could otherwise lead to costly unplanned downtime, negatively impacting KPIs and executive metrics. Using process data to alarm and drive maintenance work (like the oil pressure light on a car’s dashboard) is a well-recognised concept. PdM – which is often the “low hanging fruit” for IIoT applications - is a proven approach for improving uptime while reducing maintenance costs.
Role of analytics
The software to analyse the data provided by IIoT is segmented into two general categories: engineered algorithms and machine learning.
Engineered algorithms
Many types of assets (like large power transmission transformers) have well-researched and well-understood failure patterns and an extensive knowledge base of the attributes related to failures. For specific classes of equipment, this research has been used to create algorithms to predict failures. Parameters collected from an asset go through engineered algorithms (i.e., predetermined formulas, Boolean logic, rules and/or decision trees) to determine the health of an asset.
Packaged software algorithms can often provide a viable approach for developing, executing, and managing specific types of assets. This software helps select process data, configure diagnostic algorithms, and create alerts that drive the appropriate actions by maintenance or operations. The process data typically come from the plant historian and/or other systems, usually via OPC-based communications.
Machine learning
An emerging technology for analytics involves advanced pattern-recognition and other types of machine learning. Open source code for various types of machine learning approaches has speeded adoption. This technology generates an empirical model by ‘learning’ from an asset’s unique operating history during various stable and dynamic process conditions. This model becomes the baseline profile for normal operations for a specific piece of installed equipment or a broader processing unit. The learning system automatically compares an asset’s model with real-time operating data to detect subtle changes. These changes provide early warning signs of impending equipment failure before they reach alarm levels and possibly an unplanned shutdown.
Though machine learning can operate with a limited number of existing sensors, IIoT offers a richer set of process data (variety, volume, and velocity) for a higher fidelity model that can improve condition monitoring and asset reliability. The combination of these two emerging technologies offers an opportunity to take prescriptive maintenance to a new level. Case stories indicate that combining machine learning with a large number of IIoT sensors provides more advanced notice of a pending failure than traditional, single-variable condition monitoring systems.
Positioning engineered algorithm and machine learning
With many of the same asset and well-understood operating parameters, the engineered algorithm usually provides a good fit. A machine manufacturer with standard products has the economies of scale to create software for monitoring its equipment via the Internet and sell aftermarket services for predictive maintenance and reliability. In some cases, an end user has many of the same type of asset, providing similar economies of scale.
Machine learning provides a good fit for unique equipment or processes. A common approach has the machine learning application reading data from the historian. Initially, there will be many false-positive alerts. The software must be “trained” to improve the proportion of positive alerts. This can take from as little as a few days to as much as six months, depending on the sophistication of the machine learning software and the scope of the process being monitored. Critical factors for success include:
• Authorised trainers with deep expertise that can provide accurate guidance to the machine learning algorithm.
• Prescriptive information when an alert is generated to help humans assess priority and diagnose the issue (otherwise, they will ignore it).
On the path to prescriptive maintenance
As organisations move along the maturing maintenance curve from run to failure towards preventive, predictive, and prescriptive maintenance; they will almost certainly achieve improvements in the core KPIs for asset management and maintenance: uptime, asset longevity, cost control, yield/quality, and safety. Note that some industries, such as refining, rank safety higher. These KPIs relate directly to executive metrics for the C-suite – hence their importance.
Reduced maintenance costs
Predictive maintenance allows maintenance personnel to anticipate failures, schedule work orders, and prevent failures. A study by a major petroleum company showed that, compared to calendar-based preventive maintenance, a predictive approach reduces maintenance costs by up to 50 percent. The specific benefits reported include:
• Maintenance costs reduced by 50%.
• Unexpected failures reduced by 55%.
• Mean time between failures (MTBF) increased by 30%.
• Machinery availability increased by 30%.
Preventive maintenance fits 18% of assets
Preventive maintenance assumes that the probability of equipment failure increases with use and thus scheduled maintenance based on calendar time, run time, or cycle count. However, data on failure patterns from four different studies show that (on average) only 18% of assets have an age-related failure pattern (reference the next chart). As a result, preventive maintenance benefits just 18% of assets.
Predictive and prescriptive maintenance fits the other 82%
Doing preventive maintenance on the other 82% of assets could well cause failures by placing some assets at the beginning of the Type B curve for early life failures. APM 2.0 with predictive/prescriptive strategies using IIoT and analytics (to identify randomly occurring failures) provides an appropriate maintenance strategy for the other 82% of assets.
Where to go from here
APM 2.0 incorporates the potential of IIoT and provides a strategy for systematically improving key metrics like uptime, mean time to repair (MTTR), asset longevity, cost, quality/yield, and safety. Rather than accepting waste among APM functions, ARC recommends that industrial organisations develop a disciplined approach for improvements:
• Start predictive and prescriptive maintenance projects using IIoT and analytics. Consider targeting those critical assets that have a random failure pattern – for which preventive maintenance is ineffective and can be counterproductive. Focus on the ‘small data’ for a specific type of equipment.
• Use smaller pilot projects to build confidence within the organisation and get the sceptics on board. Avoid going directly to a large, expensive, and high-risk Big Data initiative that will take too long, resulting in decaying executive support.
• Start with an IIoT platform and architecture that can grow so that you can continue past the first project completion and build on success.
• Wherever possible, eliminate paper-based business processes – particularly those involving data collection by technicians and operators – to assure data integrity and confidence in the associated APM systems. Adopt mobile devices for the mobile technicians to use to collect data.
• Automate the business process to connect the alerts generated by the predictive maintenance application with the EAM system so that issues get attention. Avoid dependence on ad hoc communications among functional groups.
• As part of the supplier selection project, include a review of the vendor’s IIoT strategy. Those with a solid strategy will have a first-mover advantage, and a more sustainable business.
ARC continues to explore the potential benefits that IIoT-enabled technologies such as smart sensors, Big Data, and predictive analytics can provide to help improve both operations and maintenance in manufacturing plants and other industrial facilities. To join in this conversation, readers are encouraged to visit ARC’s Industrial IoT and Industrie 4.0 Viewpoints blog (https://industrial-iot.com).
Ralph Rio
Ralph Rio, ARC Advisory Group, vice president, focuses on Enterprise Asset Management (EAM), Field Service Management (FSM), Global Service Providers (GSP), and 3D scanning systems & software. Ralph has more than 40 years’ experience with manufacturing and industrial applications and has been with ARC since 2000. He holds a BS in Mechanical Engineering and an MS in Management Science from Rensselaer Polytechnic Institute, Troy, NY.
For more information contact Paul Miller, ARC Advisory Group, +1 781 471 1141, [email protected], www.arcweb.com
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version