IoT Agenda

Jun 17 2019   11:35AM GMT

Unraveling the predictive maintenance conundrum in the IoT era

Rajesh Devnani Profile: Rajesh Devnani

Tags:
ai
augmented reality
Enterprise IoT
Internet of Things
iot
IoT analytics
IoT data
IoT strategy
Predictive maintenance

One of the pioneering applications fueling the exponential growth of IoT in the digital era has been the rise of predictive maintenance as the poster child of IoT applications. Estimates from analysts vary widely, but the overall predictive maintenance market will grow at a rapid clip compared to any other IoT use case. With global assets under operation amounting to roughly 2.5 times the world GDP, the economic impact from effective maintenance practices can be truly transformative and extend to trillions of dollars.

In the past, maintenance was traditionally relegated the status of a support function within the realm of manufacturing and was treated as a pure-play cost center. Predictive maintenance flips this status over and elevates the maintenance function from a cost-centric role to a prime strategic role within the organization.

With the potential that predictive maintenance promises, there is an excessive hype surrounding the topic and enough content peddling the power of algorithms as a magic wand to realize maintenance nirvana.

The reality, unfortunately, is more nuanced. Predictive maintenance is not a pure-play technology gig. Predictive maintenance can be a distinct competitive differentiator — whether you are competing on cost, customer service or innovation — and can truly bring immense business value to your organization. It’s imperative, however, to peep under the hood to gain perspective on the practical application of predictive maintenance and resolve the conundrum:

  1. It’s all about the metrics.
  2. Busting the lone ranger attitude.
  3. There is no “one size fits all.”
  4. You may have a hammer, but everything is not a nail.
  5. The great horizontal versus vertical divide.
  6. It’s not just in the asset.
  7. Predictive is not the end-all.
  8. Measurement is key.
  9. Deployment and sustainability is the end game.
  10. The future beckons, so embrace it

1. It’s all about the metrics
The risk that the technology works just fine but fails to deliver the goods is quite high in the case of a predictive maintenance application. It’s easy to get enamored by a new shiny toy and dive straight in. Instead, it is imperative to begin with a clear understanding of the business goals. It is imperative to have a clear understanding of the specific KPIs that you wish to impact and more specifically by how much — for example, overall equipment effectiveness, mean time to repair, mean time between failures, on time in full, maintenance effectiveness and so forth. An overall maintenance function audit is a wise strategy in this startup phase.

2. Busting the Lone Ranger attitude
Functional maintenance specialists have a tendency to be enamored with the fruits of their toil and believe that they can alone deliver the results. This induces a siloed mentality that is counterproductive. It is imperative to bust the silos and have integrated thinking when considering a predictive maintenance initiative. Predictive maintenance is a team sport. Maintenance should work as an interdependent function and balance considerations related to other functions including production, inventory, human capital and customer service to optimize overall performance.

3. There is no “one size fits all”
Do not paint all your assets and asset classes with the same brush. Before diving deep, it is imperative to classify your assets into distinct classes, each with its own maintenance strategy suited to maximize business value. Some situations need reliability-centered maintenance for critical high-value assets, but a reactive approach may be sufficient for some simple noncritical assets. The segmentation should enable classifying assets into distinct classes, each with its own strategy but ultimately delivering the best business outcomes.

4. You may have a hammer but everything is not a nail
It is important to have an in-depth understanding of the physics of each asset and its potential failure modes and root causes. Vibration monitoring may be a great technique, but it may not be the right strategy for what you are trying to monitor, for example loose electrical connections. An effective understanding of failure modes and how to proactively preempt them through measuring what characteristics is essential. Knowing where to apply what sensing technique — thermal imaging, ultrasound, infrared, spectral analysis, vibration analysis and so forth — needs good domain expertise.

5. The great horizontal versus vertical divide
One school of thought contends that maintenance is a horizontal function and given enough historic data — hopefully labeled — smart algorithms can figure out all underlying patterns and correlations, delivering near-perfect insights without an iota of understanding on the vertical or asset class. The other school believes that an in-depth understanding of the asset, its constituent components and its functioning modes is essential to a good maintenance strategy. The reality lies somewhere in between. Having domain expertise on the pertinent asset classes and the contextual environment is definitely useful, but there is truly a bit of magic behind the data science algorithms. While they do uncover interesting and often counterintuitive insights that are humanly impossible in the end, the objective should be to balance the two perspectives to derive the most optimal value from your implementation.

6. It’s not just in the asset
Birds of the same feather may flock together, but assets of the same make and model don’t always perform similarly. An inordinate focus on just the asset data — based on sensors — can trip us up. The sensor data needs to blend with the context data — such as ambient conditions, operational environment, asset operation mode, general asset upkeep, etc. — to deliver the right insights. Context is really the king.

7. Predictive is not the end-all
According to Gartner, predictive maintenance is a strategy on a continuum from reactive to financially optimized. In that sense, predictive maintenance is a cog in the wheel that powers higher end objectives. The insights gleaned through predictive maintenance should be actionable in a sustainable way for translating them into real tangible value for the organization. The end goal is clearly either financial optimization or driving new innovative business models. Maintenance should work as an interdependent function and balance considerations related to other functions including production, inventory and human capital to optimize overall financial performance.

8. Measurement is key
Very often, strategic transformation initiatives like predictive maintenance get a bad rap for not delivering the goods. A major issue is often the lack of effective communications on the benefits achieved — e.g., by averting major breakdowns through predictive maintenance, avoiding truck-rolls, etc. — and quantifying the financial impacts. An effective baselining and measurement strategy is key to ensure that the benefits realized through predictive maintenance are adequately measured. Communicating the positive results to gain a wider traction within the organization and keep the momentum going is an imperative.

9. Deployment and sustainability is the end game
Creating standalone algorithmic models and proving their efficacy may get all the attention, but is not the end game. The real value is derived once those models have been deployed in a production context and are integrated into the business applications landscape. It is imperative to refresh models periodically to avoid model fatigue and ensure that the model incorporates new contextual information to sustain high levels of accuracy.

10. The future beckons so embrace it
Predictive maintenance is a fast-evolving function and greatly benefits from its confluence with technologies like AI and augmented reality. McKinsey forecasted productivity increases of up to 20% and reduction of maintenance costs greater than 10% through application of AI in predictive maintenance. Computer vision advancements will further enhance traditional predictive maintenance applications. Augmented reality applications will enhance maintenance worker productivity through wearables for applications like guided repairs. The era of self-healing machines is also not too distant. The field will continue to make rapid strides and it is imperative to stay plugged in and start piloting with new emerging technologies on an experimental basis.

Delivering true benefits from the implementation of a predictive maintenance program calls for a holistic perspective and requires the right blend of domain, consulting, technology and analytic skills. In summary, it’s the recipe that matters more than any one of the above single ingredients.

All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.

 Comment on this Post

 
There was an error processing your information. Please try again later.
Thanks. We'll let you know when a new response is added.
Send me notifications when other members comment.

Forgot Password

No problem! Submit your e-mail address below. We'll send you an e-mail containing your password.

Your password has been sent to:

Share this item with your network: