Latest / Red Talks / Episode 21: AI & Reliability
Transcript
- 0:04SPEAKER_00: How will AI transform reliability over the next five years?
- 0:10SPEAKER_01: Okay.
- 0:11SPEAKER_01: So I think we are at the cusp of one of the biggest transformations in the maintenance and reliability industry.
- 0:19SPEAKER_01: For decades we have progressed from reactive maintenance to predict to preventive maintenance, preventive maintenance to predictive maintenance.
- 0:27SPEAKER_01: And now the next journey will be predictive maintenance to prescriptive maintenance and autonomous reliability.
- 0:35SPEAKER_01: So when I say prescriptive maintenance, so predictive maintenance with AI will power the prescriptive maintenance and increasingly autonomous reliability will be the uh trend and the transformation in the next five years.
- 0:52SPEAKER_01: We are now monitoring our assets on the live basis, and uh where I see AI gonna help transform the reliability uh in next five years is uh is helping the asset owners to make sure how their performance of assets will be uh uh health-wise, how good their asset gonna uh how good they can track their ROIs.
- 1:18SPEAKER_01: AI has you know
- 1:21SPEAKER_02: made its uh you know you can say significant role now in the industrial analytics to take it to the next level, right?
- 1:28SPEAKER_02: So as far as the oil analysis is concerned, it's you know we are already using uh AA-based recommendations, you know.
- 1:36SPEAKER_02: We have uh, in fact, actually, you know, we have what we made uh you know a bit forward in this particular domain.
- 1:44SPEAKER_02: We have developed an integrated condition monitoring software, you know, it's an application where every condition monitoring you know technology, be it a vibration analysis or oil analysis or maybe ultrasound analysis, MCSCA, or maybe you know the thermography, right?
- 2:01SPEAKER_02: Every technology now you know is into a bits and pieces.
- 2:05SPEAKER_02: You know, the users basically use it like every technology.
- 2:08SPEAKER_02: When a report comes in, they take an action, you know, just based on that report, right?
- 2:12SPEAKER_02: It is true for even oil analysis.
- 2:14SPEAKER_02: The moment oil analysis, when we give a report, they always do a filtration or they always do some sort of a topping up, you know, this is what the major activity.
- 2:22SPEAKER_02: But how we transform that into a unified uh, you know, you can say recommendation, we have integrated AI with our lubrication management as well with the other condition monitoring practices.
- 2:36SPEAKER_02: So, along with you know the machine context data.
- 2:39SPEAKER_02: So we are collecting the data of the machine, you know, the environmental conditions, operating conditions, you know, the practical operational parameters, along with whatever the lubrication related maintenance and then the load factors, all that stuff, you know, which is going on with the machine, and we are integrating the you know condition monitoring reports as well.
- 2:58SPEAKER_02: So, with all this, what is happening, you're going to get the you know, AI is going to basically look at the root cost.
- 3:06SPEAKER_02: You know, already you can say how this is different from the AI, what is already doing with the vibration and other things.
- 3:13SPEAKER_02: It is independently it is checking that report.
- 3:16SPEAKER_02: Because you are asking over the next five years, but uh I can see you know a lot of things are happening, a lot of improvements in the reliability area are already happening uh because of the AI.
- 3:28SPEAKER_02: Uh it's uh uh because many times you know there's a lot of data basically, you know, the uh reliability is a data-driven actually.
- 3:38SPEAKER_02: A lot of data comes from you know sensors and many other you know uh instruments, instrumentation which is there in the refineries and you know major petrochemical complexes.
- 3:50SPEAKER_02: So it's all the data.
- 3:53SPEAKER_02: So you can everything is you want to take any decision regarding the reliability, it's the data only.
- 3:59SPEAKER_02: So the data which is coming and the large amount of data, large volumes of data, it's really
- 4:04SPEAKER_03: impossible for a human being to you know handle such large volumes of data uh and to take a you know uh a good decision or you know um which really you know improves the reliability.
- 4:19SPEAKER_03: So and we a human being, you know, is literally has got limitations, you know, uh analyzing such large volumes of data, but when this AI, what it is really giving us is in the you know ready-made you know analytical views and the uh various amounts of data in a few minutes it gives you know filters the data or and uh it um uh analyzes the data and it gives you know quick decisions, you know.
- 4:50SPEAKER_03: Uh probably the decision given by the AI at this stage may not be very matured actually, but over the five years when the tools are you know evolving and when the knowledge is uh these tools are you know continuously being trained with the you know human knowledge because they have got experience of 30 years, 40 years in the industry, what will happen is these tools will evolve over the period of definitely the next four or five years, which will be trained you know to take more uh reliable and dependable decisions.
- 5:25SPEAKER_03: But I feel because a lot of work is being done on AI, so they are fears also, but we can uh see
- 5:36SPEAKER_02: in our lives also how it is
- 5:38SPEAKER_03: touching.
- 5:40SPEAKER_02: Today uh we put sensors
- 5:42SPEAKER_03: on various
- 5:43SPEAKER_02: machines,
- 5:44SPEAKER_03: various rotating
- 5:45SPEAKER_02: parts or
- 5:46SPEAKER_03: various uh insulating parts, and we monitor that uh and then the data
- 5:53SPEAKER_02: goes somewhere, people analyze and give up uh give their suggestions.
- 6:00SPEAKER_02: But with AI, what I feel slowly this will not be only for predicting what is going to happen because in AI you are able to link the whole life of the machine.
- 6:18SPEAKER_02: You have got data also that is also integrated for the AI to give us a quality result, top quality result, our primary data, whatever we are feeding to AI should be the quality one.
- 6:36SPEAKER_02: So if we are feeding uh raw data or some uh local sensors or uh you can say uh unhealthy data, then your results should be getting that kind of quality only.
- 6:51SPEAKER_02: So AI will surely be the provide us a reliability solutions absolutely perfect in next five years, it will be a total change uh in maintenance industry or CBM industry.
- 7:05SPEAKER_02: But the main concern is the quality of data should be the top quality.
- 7:10SPEAKER_04: Some of the software's uh for demonstration purpose we have already used it, but so far we have not used in full fledged, but we are hoping to use it uh within short within a few months or so.
- 7:23SPEAKER_04: Maybe uh next generation AI we will our maintenance system will be totally AI-based.
- 7:28SPEAKER_04: So I hope so.
- 7:29SPEAKER_04: AI will we will have a very big impact in improving the reliability of the system as well as machines.
- 7:36SPEAKER_04: See, uh reliability consists of machines and data.
- 7:41SPEAKER_04: So uh it exists for quite a long time.
- 7:45SPEAKER_04: That AI has made that you know that data linkage with machines in a much faster way, much precise way.
- 7:53SPEAKER_04: And uh, there are pros and cons of any new technology.
- 7:56SPEAKER_04: Initially, there will be resistance, or you know, people vote about people think about the cons but not the pros.
- 8:02SPEAKER_04: But I am sure it is going to work, the life will be much easier of the shop floor people, my maintenance colleagues, as well as you know the reliability procedure.
- 8:13SPEAKER_04: Uh reliability engineer will get an AI co-pilot.
- 8:17SPEAKER_04: When I say AI co-pilot, he he will be assisted with the uniform actionable data to take the uh chatboard-based approach or uh uh app-based approach to do a faster FMECA, faster root cause analysis.
- 8:37SPEAKER_04: So FMECA will not be a static Excel sheet, it will be a living app system.
- 8:44SPEAKER_04: Then uh RCA root cause analysis will be much faster with the introduction of AI.
- 8:50SPEAKER_04: Digital twins will become the reliability twins.
- 8:54SPEAKER_04: So that is another uh it is work in progress presently.
- 8:57SPEAKER_04: It is not that it is not there, many industries are adopting slowly, but will more become an adopted and accepted uh methodology in the times to come.
- 9:08SPEAKER_04: We are moving to
- 9:08unknown: an age whereby data-based decision making will become more and more important as competition becomes fierce, nobody wants to stop an equipment.
- 9:18unknown: So if you don't want to stop an equipment, you need to know real-time health of the equipment.
- 9:22unknown: If real-time health of the equipment is supposed to be known, you need to collect that data, you need to know what has happened, you need to know what you did
- 9:30SPEAKER_02: in past, what could lead to those failures.
- 9:33SPEAKER_02: But yes, only one thing I see when we look at uh industries which are already in place.
- 9:40SPEAKER_02: That there's a lot of investment that people need to make in terms of instrumentation, in terms of having a digital twin whereby this data can be collected.
- 9:49SPEAKER_02: But yes, with whatever work available, AI can give excellent output at a very fast pace.
- 9:55SPEAKER_04: AI is now moving very fast and it is uh coming up in almost all the fields.
- 10:01SPEAKER_04: Initially, it was uh more of uh software development and uh IT and this field, but now we are seeing it is choking over the all the aspects of life.
- 10:11SPEAKER_04: It could be industries, it could be my manufacturing, my logistics, my healthcare, my resource management, it would be anything.
- 10:18SPEAKER_04: So it is going very fast.
- 10:20SPEAKER_04: And then when we talk uh specifically about reliability, then it is uh uh we have seen systems like uh operative maintenance or condition monitoring systems.
- 10:31SPEAKER_04: So these were these some basic systems which were already being used for the reliability measurement or increasing the overall life cycle of the systems or the products.
- 10:40SPEAKER_04: But what I feel when we integrate AI into this, integrated with along with these products, then we can have very accurate prediction about the future, about the performance, about the life cycle, about the issues which might come in the future.
- 10:55SPEAKER_04: So this is one thing, and then what I feel is going forward, uh right now it uh we still don't know what are the areas and how much would be the impact of the AI in the overall system.
- 11:08SPEAKER_04: So, but then what I feel we should make more use of it in uh we are as of now looking at it in very selected fields.
- 11:16SPEAKER_04: As I said uh so far, it was mostly into IT and coding, and but uh I feel in the uh resource management, this is one area for critical resources, natural resources, especially.
- 11:28SPEAKER_04: For I'll take a very simple example, like water.
- 11:31SPEAKER_04: So, water is a very important resource, a very scarce resource, although it is present abundantly, but then there is also a very huge shortage of water in various parts of the world.
- 11:43SPEAKER_04: So, if we use AI just to understand very basic things like what is the precipitation level or what is the available of uh availability of water, what is the groundwater levels, and then especially if you see in India in the last few years and still it is going on, we have uh a lot of uh these solar pumps are coming up, solar pumping systems, and they are being installed in like uh in millions of quantities.
- 12:11SPEAKER_04: Now, what is happening is once all these together when they start functioning, they will take the water from the ground level.
- 12:18SPEAKER_04: So the groundwater reserves will deplete.
- 12:21SPEAKER_04: Now, with by using AI, we should be able to understand and we will be able to understand what is the groundwater levels, what is the level it is again being replenished.
- 12:32SPEAKER_04: India has a very good monsoon, but then most of the monsoon water, the rainwater, is going to waste.
- 12:40SPEAKER_04: So we have to again once this I predicts this, and then it enables, it will enable us to see that again we ensure that the groundwater is replenished and we just got just uh pumping up the water.
- 12:53SPEAKER_04: So this is one very small example.
- 12:55SPEAKER_04: Uh, another example from the field I will give you like uh uh just using this condition monitoring system.
- 13:01SPEAKER_04: We had uh one customer of ours where we had installed a very basic uh condition monitoring system, and during the production process, the system predicted we found some jerks in the system.
- 13:13SPEAKER_04: So initially it was thought there's some electrical or a voltage spike or something or a mechanical problem.
- 13:19SPEAKER_04: But when we instigated we investigated that and we went too deep into the problem, we understood all this system was okay, but then the problem was action in the raw material, the thickness of the raw material, there was some problem.
- 13:32SPEAKER_04: So this we could find it out after a lot of time, after a lot of investigation, but then that helped to prevent a problem which would have come later on, and the complete product, the end product were they might have to maybe even scrap that.
- 13:50SPEAKER_04: So if we integrate AI into this along with the existing products and some future products which might come up, then this complete cycle, the entire time duration for this for the investigation and coming up with a solution will be minimized.
- 14:06SPEAKER_02: Some some are able to solve the complex problems very, very easily, and there are some challenges while dealing and interacting with AI.
- 14:16SPEAKER_02: So, what I believe that in future, humans will learn more how to deal and work and interact with AI, and then the AI definitely will impact the decision making.
- 14:29SPEAKER_02: So,
- 14:30SPEAKER_04: in the back end, there will be some one person, maybe he is looking, watching, monitoring the data, what's happening around, and then remaining all the data collection because of IoT, data collection is there, sensors are everywhere, and it's going to increase a lot.
- 14:46SPEAKER_04: There are a lot of things we still we are not monitoring monitoring.
- 14:50SPEAKER_04: So there will be more sensors, people will monitor them also, and then every all the data will come to the AI, and AI will
- 14:57SPEAKER_02: definitely uh do the things which normally it would took a lot of time for others.
- 15:03SPEAKER_02: One more important aspect AI
- 15:05SPEAKER_04: will
- 15:05SPEAKER_02: create a greater impact on the
- 15:06SPEAKER_04: robots
- 15:07SPEAKER_02: also.
- 15:08SPEAKER_02: So AI-enabled
- 15:09SPEAKER_04: robots will be able to work like a human in the factory, that will really be a variable transformation.
- 15:16SPEAKER_04: AI is going to stay, it's going to evolve, and it's going to definitely have its impact on
- 15:24SPEAKER_05: reliability also.
- 15:26SPEAKER_05: For example, the first thing which comes to my mind is regarding RAM simulation, reliability, availability, and maintainability simulations.
- 15:35SPEAKER_05: Today, when we try to build the reliability block diagram for an RAM engineer for a big plant, it might take 2 to 3 days to complete the RVD.
- 15:49SPEAKER_05: But if you have digital twins already available for the plants and the whole facility, and if you put that as an input to the RVD module of an RAM simulation software, perhaps the RVDs can be developed in less than an hour.
- 16:10SPEAKER_05: So that's productivity.
- 16:12SPEAKER_05: Now coming to how it can help reliability, pattern recognition.
- 16:18SPEAKER_05: There are systems available, but they are not widely implemented because they are mostly server-based.
- 16:26SPEAKER_05: Some cloud-based systems are there.
- 16:29SPEAKER_05: But with AI coming into the picture, they can make things happen much faster and they can extract more information from that data than ever before.
- 16:45SPEAKER_04: You know, uh this post COVID, especially the post COVID, uh the remote condition monitoring gained uh more momentum.
- 16:54SPEAKER_04: Uh okay, basically, you know, uh to avoid dependency on the you know on the people And uh that's how the evaluation gained more, you know, uh, you know, relevance or the importance uh to teach your sensors, okay, uh a kind of machine learning to their sensors, put some AI over there so that self-predictions, self-predictions can be done through the sensors, and you can get to know about the condition of your machines well in advance before they can fail or become inefficient.
- 17:33SPEAKER_04: So, yes, so uh in next uh five years, uh I see
- 17:40SPEAKER_03: uh this technology to the next level.
- 17:43SPEAKER_03: I think it's already taken uh in this one year, and it is very essential because it's not a luxury, it is a need now.
- 17:55SPEAKER_03: It is need now for all because I have experience about the maintenance and all these things in all over India
- 18:07SPEAKER_02: and nearby at hundred clubs.
- 18:09SPEAKER_02: I was looking the maintenance of that.
- 18:12SPEAKER_02: And in there are the lots of problems was there, and we solve manually.
- 18:21SPEAKER_02: So it takes it takes the money, lots of money, lots of time.
- 18:28SPEAKER_02: Sometimes sometimes it is not solved, it was not accurate.
- 18:35unknown: And that is
- 18:36SPEAKER_03: the that is where is the biggest problem.
- 18:40SPEAKER_03: But now the reliability is there.
- 18:44SPEAKER_03: Um in these five years, it will be it will be amazing, it will be astonishing.
- 18:53SPEAKER_03: Why?
- 18:54SPEAKER_03: Because for example, one we can give one example.
- 18:58SPEAKER_03: Um if some machine works, that best machine, for example, in uh uh in uh wellness or fitness, we can say the pre-cor machine is also there.
- 19:11SPEAKER_03: Okay, well so basically there is need of oiling some dust that should not come.
- 19:20SPEAKER_03: Okay, so we have to take uh periodically checks about that machines and all these things.
- 19:27SPEAKER_03: But now AI if AI machines there, I think Air machines already uh already coming, so AI machines is there, so they are maintaining that uh they are giving reports to the main office and they will solve at their office about our for example we are sitting in Mumbai and uh for we are sitting in uh Madras and Mumbai is a corporate office.
- 19:57SPEAKER_03: So Madras, if we call them, oh there's no need to call also.
- 20:03SPEAKER_03: All machine will send the report automatically to the uh Mumbai corporate office and it will see uh what is the problem and it will solve automatically.
- 20:15SPEAKER_00: Red Talks features voices driving safety, reliability, and innovation forward in the industrial world.
- 20:22SPEAKER_00: If your work challenges the status quo, we'd like to hear from you.
- 20:26SPEAKER_00: Learn more at redtalk.com slash about.