Machinery and Digital Transformation

It is estimated that 70% of industrial companies have digital transformation initiatives underway and will collectively spend an astounding $2T in 2023 toward that end.  A new member of the C-suite has even emerged – the Chief Digital Officer – underscoring the emphasis on digital transformation and industry 4.0.  Yet, with all this emphasis, a paltry 16% of such initiatives to-date have met internal expectations, placing enormous pressure on CDOs and digital transformation business leaders to “deliver the goods” with projects that meet or even exceed expectations.  Here, we explore what CDOs are thinking about based on a recent interview with one such transformation leader.

 

Bently Nevada had a recent opportunity to go deep inside the mind of a digital transformation leader with a series of probing interview questions on how projects are selected, what type of ROI is expected, where machinery fits in the grand scheme of improvement initiatives, the role of Artificial Intelligence (AI) as well as so-called Machine Learning (ML) in such initiatives, and more.  This individual was an ideal source of insights because he possessed a hands-on engineering background covering rotating machinery and reliability coupled now with a role leading digital transformation initiatives of all types – whether machinery-related or not.  His employer is a multi-national industrial petrochemical business with more than 30 manufacturing facilities in 20+ countries across every global pole.  His insights thus reflect a deep understanding of machinery, a deep understanding of success criteria for digital transformation projects, and a deep understanding of needs that transcend any particular plant, country, or pole.

 

Q1: What level of expected ROI is required to even get to pilot phase from the “sea of good ideas” brought to you?

A1: Any initiative that cannot deliver an ROI of at least 20% is immediately dismissed from consideration.  But in today’s environment, the bar is often set even higher and projects that can’t return 30% or better rarely make the cut. 

 

Q2: What keeps projects from moving forward even if they make the cut?

A2: To even be considered, there is not only the ROI hurdle but also the scalability hurdle.  We really aren’t interested in any initiative that we can’t roll out on an enterprise-wide basis because that’s where the big returns come from.  No matter how big an impact I might get at a single plant, it isn’t a digital transformation initiative if it can’t be applied at most or all of my facilities because it can then be compounded 30+ times.  There have to be economies of scale present, too.

 

 

Q3: Can you elaborate a bit more on economies of scale?

A3: When I’m going to deploy a solution across the enterprise, I expect it to get increasingly easier with each incremental site.  For example, we rolled out some maintenance software a few years ago and deployment at the first site took 6 months.  At sites 2-4 it took 3 months, and at all the remaining sites it took only 6 weeks.  I really don’t want to pursue something that is just as difficult to deploy on the last site as on the first site.  

 

Q4: What is the role of the supplier in that first deployment or first few deployments?

A4: Quite frankly, I won’t even consider rolling something out as entirely DIY.  I don’t really even know how you can roll something out without the intimate involvement of the supplier.  They are the expert on how to do it right, to know fully whether it is working properly or not and to be able to get factory support when problems are encountered. I actually go even further in insisting upon supplier involvement with the first few deployments: they actually have to be literally embedded with my team on-site so that we work through the process together.  If a supplier can’t do this, it really becomes a non-starter for me.  I’d say its mandatory – not optional.

 

Q5: Do things get stuck in pilot phase  and if so, why?

A5: Well yeah – of course some projects get stuck. It’s usually because they either fail to meet the ROI or they prove to be unscalable.  

 

Q6: So, the pilot phase1 is about validating both ROI and scalability?  Do you ever forego the POV¹?

A6: Yes – it is about validation.  I have to validate these things for myself and I rarely if ever forego the POV because no matter how wonderful a testimonial might be from someone inside or outside my industry, I have to show that it works with the particulars of my plants and my business conditions.

 

 

Q7: When you think about machinery, where are the problems still occurring?

A7: We have more than 100,000 rotating machines across our 30+ facilities that aren’t so-called “critical”.  We aren’t having problems with those critical machines anymore.  Bently Nevada has been a large part of making those the most reliable and best-instrumented.  The real issues are with that population of 100,000+ machines.

 

Q8: What are those issues?  

A8: It’s a numbers game.  The machines may or may not have a production impact.  In fact, many of them are spared.  The reasoning has been that because they are spared, we don’t need to be as concerned about them.  However, when you consider the costs of maintenance on 100,000 assets, it adds up.  For example, many of these assets are garden-variety, ANSI centrifugal pumps driven by an electric motor.  We’d like to take the MTBF on these machines from something like 18 months to something more like 28 months.  When you do the math across many thousands of assets, the savings to the business are tens of millions of dollars.  Reduction of maintenance costs is the driver here – not primarily downtime.

 

Q9: Don’t you already address such assets with some type of condition monitoring?

A9: To some extent, but we can’t get any economies of scale by simply expanding a program based on portable data collection.  It’s too expensive if I want more frequent data or want to include more assets in the program.

 

Q10: So, what exactly are you envisioning – if anything?

A10: Right now, I’m actively involved in a POV of IIoT sensors on these machines and then using AI to analyze the data. Inexpensive, continuous data acquisition coupled with AI means I am not relying on an army of people to collect and then review the data.  I can automate those activities and that is at the heart of a digital transformation initiative: what can I automate that I’m currently not doing at all or doing manually and thus inefficiently.

 

Q11: What is important to you in this POV?

A11: First, sensor cost is enormously important.  If I’m going to put multiple sensors on these machines, they have to be inexpensive².  One of the problems with AI is that of “not enough data” and it means I have to find a way to get enough data into the system – cost-effectively – that my AI is no longer hampered by that.  This means sensors in the hundreds of dollars – not thousands.  And it obviously means wireless communications not wired.
 

 

Q12: AI and ML has a poor reputation with many customers.  Have you experienced that and if so, why are you giving AI a “second chance”?

A12: I have extensive experience with so-called “generic” AI using large systems relying on pure data science.  The main reason those systems are problematic is partly because they require so much data, but also because you have to build models.  Out of the box, they know nothing at all about machinery.  I do think there is a place for such systems and the truth is, our company is investing heavily in them – we even published a press release about the work we’re doing along those lines – but we understand that we have to create the models and many of those are focused on things we know a lot about and might be proprietary to us – such as our chemical and manufacturing processes, so that’s OK and expected.  But when you try to apply those systems to machines, you usually fail.  That has certainly been my own experience and I can tell you that I have spent many hours building models only to get non-sensical outputs because I didn’t have enough sensors.  However, purpose-built AI is different.  The models are already in the system and it understands pumps, motors, sensors, failure modes, etc. So, I don’t have to spend time reinventing the wheel and if the sensors are suitably inexpensive, I can easily get enough data. Our very early investigations are promising and in 2023, we will be doing a full-scale pilot with wireless IIoT sensing and AI for those machines I describes – primarily pumps.  But, I’m optimistic because purpose-build AI does not require the training that generic AI requires.

 

Q13: Does this come out of a capital expenditure budget or an operating expenditure budget?

A13: Well, that will vary by company.  I have seen CapEx thresholds as low as $10,000 and as high as $130,000.  So, a pilot might perhaps be within an OpEx budget, but usually not.  I will say that whenever we can do something under an OpEx budget, it is preferred.  CapEx ends up being amortized and goes into annual reports.  There is a lot less scrutiny on OpEx spends.  But this is another reason why inexpensive sensors are important².  If I need to add more machines or more points on already-monitored machines, it is much easier when it can be done within an OpEx budget².

 

Q14: Is data ownership important to you?

A14: Data ownership is a huge issue.  We won’t consider any solution where we don’t own the data.  This is because we need to be able to use the data as we see fit and may want to use it in ways not envisioned by the supplier, but highly valuable to us.

 

Q15: What are you ultimately hoping to get out of the system we have been discussing?

A15: The ultimately goal is to increase the Mean Time Between Failure (MTBF) for these machines so that we are not repairing so many, so often.  This is ultimately about the reduction of maintenance costs because the machines run longer and the outages are planned rather than unplanned.  There are secondary benefits, like safety, but the primary driver here is maintenance costs because there is a huge compounding effect when you have 100,000 such assets.

 

1. Many customers refer to a pilot phase as “Proof of Value” or POV.  POV is used throughout this piece as it better captures the essential purpose of a pilot phase.
2. Bently Nevada’s solution for this class of machines is offered as a service.  Although the wireless sensors used in this solution have a very modest cost, this is not of concern to customers because the pricing structure is on a per-machine basis, not a per-sensor basis.  Bently Nevada is responsible to provide a sufficient complement of sensors in order to fully monitor the machine and detect malfunctions.  Additional machines represent incremental, modest pricing as part of the subscription.  Your Bently Nevada sales professional can provide further information on subscription pricing and granularity.