IoT GUIDE

Smart asset management: Another case of “over-promise, under-deliver?”

With technology, vision often comes way before realization does. In other words, the hype reaches you before the application does. This is exactly how the evolution of smart asset maintenance has charted its path as well. So why exactly has adoption been so painfully slow? In today’s IIoT era, MROs need to identify aging machine parts and predict their lifespan. For instance, does a part need to be replaced, or can it be repaired? If repaired, how long will it last, and how will it affect the aircraft’s estimated lifespan? These examples, which seem simple enough, are a far cry from the much more complex questions that harass MROs. According to The Manufacturer, a leading automotive manufacturer reported that for its employee base of almost 18000, 85-95% of their time was spent in [maintenance] crisis work.

Considering the vast adoption of IoT and the already snowballing data being gathered via zillions of sensors, how can companies take advantage of the predictive approaches to maintenance?

The drudgery of the evolution - Reactive versus Predictive

To begin with, most industrial sectors used corrective maintenance – or the practice of fixing assets after they’ve suffered a breakdown. Then, we had preventive maintenance, which involved the fixing/ replacing of parts at predetermined intervals (a common practice in the context of this strategy was to run maintenance schedules at the same frequency at which the most heavily used machines were prone to breakdown). This approach led to three issues – one, asset performance is inherently variable. That is, the same equipment need not always be the weakest link. Second, the breakdown frequency of an asset is also prone to variation. Third, the maintenance schedule planned based on the behavior of one asset need not hold good for another asset. These shortcomings led to the birth of the next level of maintenance – predictive maintenance. This approach involves, as the term implies, identifying asset failure before it happens.

Predictive maintenance requires the execution of system checks at predetermined intervals to analyze system health. These controls are usually in the form of continual data collection (regarding temperature, light, pressure, sound/ vibration, etc.) from equipment via sensors. The results of these checks are used to determine whether or not maintenance activities are required.

The evolution of each of these stages has followed a painstakingly arduous path spread across years, mainly because of the cost barrier imposed by technology – that is, the cost-benefit ratio in each of these cases has been too low to justify the investment. Initially, the cost of sensors prevented adoption. This was overcome with the advent of digitization. Next, it was the cost of connectivity – but wireless communication took care of that. Finally, the worry about the cost of data science/machine learning (which was high due to the scarcity of sufficient skilled expertise) has been addressed through smart, self-learning machines - teaching machines to do machine learning. What this means is that we’ve now reached a pinnacle of sorts, where we can converge cloud-based technologies, IIoT, and data science to get those insights about machine health.

Who wouldn’t like to play God? From Sensor to Insights

The array of assets making up the IIoT is extensive – pumps, conveyors/vehicles, transformers, unending pipes, and cables – the risk of failure is that much higher. It goes without saying that a fail proof maintenance strategy is of prime importance for the industry. All the insights amassed through predictive maintenance can offer little help unless something is actually done to predict issues. This phase of maintenance, called cognitive predictive maintenance, will, therefore, involve not just advance flagging of potential issues but also the suggestion of viable “what-if” solutions that can change the outcome.

For instance, Rockwell Automation uses the prescriptive approach to churn health and diagnostics analytics from asset performance data. The insights thus gained help minimize productivity losses resulting from unscheduled downtime. “That same information will also help the application make prescriptive recommendations,” says Mike Pantaleano, Global Business Manager of device/edge analytics. “This way, manufacturers can improve equipment uptime and lower maintenance costs.”

Another company that has adopted the prescriptive approach is ABB, a leading technology provider for the infrastructure, transportation, and utility industries. Its grid automation business leverages the prescriptive approach to facilitate a “stronger, smarter, and greener grid.” In fact, the company uses customized models to help companies improve asset performance and reliability by shifting from descriptive to prescriptive analytics. This, in turn, facilitates risk-based investment optimization.

Perfecting Predictive Maintenance - From Insight to Outcomes

We’re now at a point where we can leverage all the information and insights offered by the cloud computing technologies, IIoT platforms, and cyber-physical, smart systems ushered in by Industry 4.0 to drive intelligent maintenance. Moreover, sensors and communication networks are a lot cheaper than they used to be, and communication is easier than ever before. However, only 20% of the sensor data is conspicuous and clear while the remaining whooping 80% still remains disorganized, ambiguous and fuzzy. So how do organizations get into tackling this dark data smog which is hovering their industrial machines?

To get to the bottom of this mine, a manual approach to machine learning would only result in a myopic scope of data analysis (where few parameters are considered at any point in time). The question is, how do companies then take all the parameters into account by incorporating all relevant algorithms to generate the accurate output? How do they go about ‘identifying the real triggers’ of these failures by digging deep into the unknown?

Dan Miklovic, a Principal Analyst at LNS Research and one of the pioneering thinkers in the field of prescriptive maintenance says “No longer will you need an ensemble of experts to tell you how and when to maintain your assets, as the assets themselves will tell you what they need if they are unable to fix themselves.” So how do companies tap into this potential of digital transformation, something that they have been struggling with for long.

Final Frontier of Smart Asset Maintenance 'Cognitive Predictive Maintenance'

Capturing the true potential of Industry 4.0 often requires a company-wide transformation to re-invent itself and its core capabilities at a much more accelerated pace than before. With cognitive predictive maintenance, machines will start answering maintenance questions you didn’t know existed.

Getting the orchestra to perform

As asset management needs are evolving, organizations need to realize the importance of adopting a digital approach to managing asset life cycles to stay competitive. This approach should ideally involve the following steps:

 Getting a hang of big data analytics

There’s no escaping the fact that you should have a basic understanding of data – it’s going to be pouring in from every conceivable direction, and it’ll be as unstructured as it comes. While the analysis and subsequent recommendations will be taken care of by machine learning, you’ll still need to know if they’re heading in the direction you want them to and tune them if not. Note that simply monitoring asset performance data and using it to schedule preventive maintenance doesn’t qualify as predictive maintenance. Cognitive Predictive maintenance involves the application of big data thinking to structured, unstructured, readily available, and hard-to-find dark data to redefine operational excellence for connected factories.

Setting up a solid operational architecture

A robust operational architecture is essential to ensure the successful implementation of cognitive predictive maintenance for asset-intensive organizations. This helps clearly distinguish between strategy and tactics, which in turn helps understand current capabilities and what they’re capable of. This exercise provides the roadmap to advance from how things are to how they should be.

Adoption of industry best practices

Since predictive maintenance is an ensemble of efforts, all stakeholders, processes, and technologies should be aligned with industry best practices to ensure seamless implementation. Moreover, maintenance as a complete function should be aligned with all other business functions, while being sufficiently flexible to be configured to achieve specific business goals.

Reiteration of business objectives

This step is what starts the transformation from asset performance the way it stands currently to the ultimate goal of where it should be to facilitate business objectives. This is also the crucial turning point wherein maintenance shifts from being a cost center to a value generation point.

To Conclude... 

A paradigm shift in maintenance approaches

The shift to cognitive predictive maintenance will radically alter the way industrial businesses function, rendering many a business model obsolete. Companies will have to start focusing on defining optimization objectives for businesses and on how to take advantage smart tools to achieve these objectives. These smart tools will then themselves let you know when they need fixing, and how best that fixing can be done.

 “Time is money,” goes the age-old adage – intelligent asset maintenance is exactly what we need to save both.

This article was originally posted on LinkedIn.