Evolution of industrial preventive maintenance. We interview Diego Galar (part 1)

Trained as a Telecommunications Engineer and Doctor in Industrial Engineering with eight books published to his credit and more than 150 scientific articles published, Diego Galar is one of the great experts in the world of maintenance in Europe. In addition to his extensive professional experience, he is a senior researcher at Tecnalia in the Industry and Transport division and Professor of Condition Monitoring at Luleå tekniska University in Sweden, which is why his every sentence exudes a profound knowledge of the discipline of maintenance and its future in the coming years.

In this extensive interview, from which we have extracted 3 instalments, we take a look at the current and future situation of maintenance.

- How does a Telecommunications Engineer end up in the world of industrial preventive maintenance?

It is true that for a telecommunications engineer to end up in industrial preventive maintenance is something very curious. I started in the world of maintenance through vibration analysis, which was to know the health of the machines, the equivalent of doing electrocardiograms on the machines and see what could happen. At that time, vibration analysis was the germ of the maintenance that was practised in the 70s and 80s. I became interested in both the diagnostic and prognostic processes of seeing "what" was wrong with the machine and seeing "what could happen to it" and so I started to get into the world of maintenance in the 1990s.

- And how has the discipline of maintenance evolved since then?

Maintenance has changed a lot. In the year 2000 and the following 20 years, we have seen the incorporation of electronics, automation and finally the whole ICT part. I would say that maintenance processes were still very classical but they have been empowered by these technologies. Monitoring that was unfeasible in the 90s is not unfeasible now. You don't need complex equipment, now with a laptop it is possible to monitor in real time, which was impossible in those days. And maintenance has also changed a lot as a process in companies.

When I joined, maintenance was seen as a necessary evil, you have to have maintenance because equipment can fail, we have unavailability, we have OE losses etc...; it was a necessary evil that had to be mitigated with preventive maintenance, which was very labour intensive and very intensive in terms of machine downtime, but it helped you to keep your machine in order.

All that has been changing, and maintenance has been evolving. Now you have methodologies such as VDM, Value Driven Maintenance, which in some way configures maintenance that adds value to the organisation. Maintenance is no longer seen as a necessary evil, but as a contribution of value, and the paradigm is changing in that sense.

Also in the 2000s, the concept of asset management began to expand, and now we are talking about "intelligent asset management", i.e. machines are no longer operated or maintained, but are managed as assets; assets that produce wealth, that produce a profit and therefore the maintainer is not only a person who maintains but also manages the asset. Operation and maintenance merge to give rise to asset management, and in 2011 the fourth industrial revolution arrived , bringing more tools to the world of maintenance, of a technological nature; some more successful than others and which has allowed us to take a giant leap forward.

And in this respect, maintenance has been greatly reduced . Companies always play with maintenance mixes. You have a % of preventive, another % of corrective and another % of predictive and the technologies of the fourth industrial revolution have made predictive maintenance more important, it is more reliable and therefore, you can increase that % in the mix. We have been reducing what was invasive maintenance, corrective, preventive and of course reactive maintenance, unplanned maintenance, emergency maintenance, which has been reduced to practically nothing, because with today's technologies we are able to predict the evolution of almost any failure in a system or subsystem.

As a result, reactive maintenance, emergency maintenance is very limited, scheduled corrective maintenance is greatly reduced and preventive maintenance, which is very costly, is also reduced.

- On occasion I have heard you say that industrial preventive maintenance is not an option.

Yes, it is not a question of whether I implement it or not, but when. Companies that don't enter into it will be left out of the game, as they will not be competitive. With a classic maintenance mix, you are not going to be competitive. If we are involved as a link in the global supply chain, we have to be efficient and effective, we have to produce a lot and we have to produce well and reliably. We see it now with the chip crisis, as part of a global supply chain, any small break in the chain is causing a major disaster.

If we have traditional maintenance we will move with KPIs and OEEs from the 80s and 90s and now that is no longer competitive, we don't have the profit margins of the 80s. Now we move with such small profit margins that we can't afford any maintenance failures. It's a question of gaining competitiveness.

Industrial Preventive Maintenance, like other technologies that have entered with the 4.0 is no longer something that is fashionable, it is a question of gaining competitiveness because the KPIs that we move are so tight that Predictive Maintenance will help us not to be an element of rupture in this global model of the supply chain.

- How do we get started in industrial preventive maintenance and what are the big challenges for Industry?

When you decide to monitor something it has to be because the impact you are going to get through the cost of monitoring is something that is going to provide you with a benefit. It is not about monitoring all the potential points of the machine just to give a lot of information from the machine where the failure never occurs or if it does occur it is not critical.

Years ago, if you wanted to monitor a machine in a combined cycle power plant for example, you would say to yourself: as it is so expensive to monitor continuously, what am I going to monitor? the turbine, which is perhaps the most critical thing. And what we did was an analysis of failure modes and their effects, we calculated by criticality and severity and it told you the possible failures. In case of failure, we analysed where it could hurt the most and we put the resources there. And this is something that has never been done in the industry because we have not been very methodological.

We have lacked a lot of methodology in the industry. We have not had the methodology to detect those critical points where we can monitor and we have not known how to take advantage of cheaper technologies, even today in the industry, I do not see IoT as very widespread.

There are companies that still don't trust if there is no wiring, there are wireless sensors that can be used, however they doubt the battery life. In many places what doesn't go through the cable tray and into the PLC is not an option.

So one of the challenges is to apply methodology.

Another challenge is that the maintenance manager has to speak the language of management, he has to be able to transform his maintenance budget not into an expense, but into an investment to make the plant reliable, the maintenance manager is no longer the sharpest technician in the plant.

If we can convince management that asset degradation will be lower and performance higher, that our OEE will be higher and that the lifetime cost of the assets will be much more positive for the company, i.e. if we can keep the assets longer and producing more, this means that we will avoid costly investments, we will avoid costly retro fittings and we will produce a lot.  

If we convince management that we move with the technologies in those parameters to keep the assets in good shape and producing a lot, then we will be able to make that profit.

The third challenge is cultural and organisational change. Industrial preventive maintenance is about change and change is always traumatic.

If we are in a company where maintenance is a mix of corrective and preventive where predictive maintenance is residual, the change will be brutal. It will be brutal because it implies a cultural change but also an organisational change. It is not just a way of doing things but also how to do them. In this sense, it is very difficult because expectations have been sold for many years that have been totally false.        

In other words, there has been a crystal ball of unreliability that has been sold by deciding that one could predict practically everything and this has generated expectations that have not been fulfilled and which has caused many maintenance managers to return to their winter quarters and say I'll stay where I am, as I know I control certain availability because this technology is not sufficiently mature.

Now when we go to companies and meet these same maintenance managers, they are much more demanding when it comes to demanding both accuracy and low uncertainty in what that prediction is.

When it comes to predictive maintenance, the maintenance manager becomes a pure and simple risk manager. He will receive daily reports on the state of the machinery and potential predictions. And it will be up to him to assume the risk or not of scheduling an intervention. The risk will be taken when you have a sufficiently reliable element for that prediction to be good enough.

I understand that maintenance managers, even those who are prone to such a transition, demand that in order to take the risks of such a change, they must be given security, because it can be really catastrophic.

So we have to bear in mind that predictive maintenance will have to coexist, it is not going to totally replace corrective and preventive maintenance. In fact, I give the same example, for car brakes it doesn't make sense to do predictive maintenance because the wear is so linear, it's so homogeneous, and you know that changing it every x amount of time you don't need anything else that it's not worth spending a sensor there, so preventive by calendar or scheduled corrective maintenance is going to be there.

What predictive has to do, is to make room for and mitigate over all the reactive and programmed correctives that are very onerous.

- Once we have the money and we get the green light to do the predictive plan, what do we do?

Of course, we have to analyse the reliability of our plant, look at where the reliability bottlenecks are, i.e. where that failure can cause a quality problem in the final product or can cause a breakage that can have an impact of several hours or whatever. I have to analyse those reliability bottlenecks and that's where I have to put all the money.

We have to do a reliability analysis, an AMP analysis, what failure modes, what symptoms we discover, and that will tell us what kind of sensors we should use to detect, identify, locate and predict the failure. When I see monitoring every minute for failure magnitudes that occur every four months, it makes me sick, it's a brutal waste of resources.

Depending on this analysis, we will know more or less how often the failure may occur, do we have to monitor continuously or do we monitor with a weekly, monthly, or every 6 hours latency, and depending on that, we will size our data architecture to adapt it to what is potentially a failure.  

Once we have that which is the OT part of data collection, we have to take into account the maintenance manager has to play in that OT-IT convergence. And what do I mean by that?

When data scientists come and sit down to work on maintenance data and they see the stream of data coming from the sensor and they see that in that stream of data there comes a moment when there is a stop and the data changes, they have to be told why this event occurs, it is because there is a work order and something has been done on this machine, but of course that is not OT, that is in the CMMS and that is IT.

It is therefore necessary to merge the two data to produce metadata that allows me to explain the reality of maintenance, which is not a continuous flow of data; it is often interrupted by stoppages, it is interrupted by reference changes, it is interrupted by work orders, all these are semantic ontologies and mean that we cannot work only with the data that come from the sensor but we also need what is in the IT systems to reconstruct the reality of maintenance and this is perhaps what maintenance people now find more difficult.  

When I show them sensor data and they see big discontinuities and you tell them that in order for this to make sense, I have to tell the Data Scientist how in that event there's a gash in the signal and if I'm able to put labels on the events, which are disruptive in that signal, I'll be able to explain and I'll be able to predict.

To think that it all boils down to a sensor reading and making a prediction is a simplistic view of maintenance.

- In which sectors is industrial preventive maintenance being adopted?

One of the sectors is the aerospaceThe reason is very simple, failure is not an option. This has then been transferred to the high-speed train sector, which has inherited a lot from the aeronautical sector railway, to high-speed trains, which have inherited a lot from the aeronautical sector, and little by little it has been permeating the rest of the industry. However, there are many sectors where predictive maintenance is still in its infancy and industry, for example, has had a hard time getting into the game.

The power sector is another clear exponent of Predictive Maintenance. energy sector is another clear exponent of Predictive Maintenance and this is because it has tremendously expensive and often neglected assets in remote locations such as wind turbines or substations.

All these industries have had something in common. The large business groups, whether in the aeronautical sector with Airbus or Boeing or the gas sector with the Oreda Group, have all shared knowledge in the field of maintenance and this has helped them to grow a lot, they have grown and have matured maintenance together.

The problem of the industrial sector is that it is much more heterogeneous, we have from cement to the steel industry, to the automotive or pharmaceutical industry. And the industry has not shared that knowledge so much, and if it has shared it, the components that have been shared are very disparate; and I would say that they are still far from 4.0 maintenance or real predictive maintenance.

Progress has been made in some specific assets, but what we see in an oil platform or in an aircraft where the whole system is very reliable in factories is not, they are still disjoint systems where maintenance is, with apologies, "ñapas".

If you were interested in this first part of the interview, we invite you to read the second part where we talk about business models in maintenance here