Predictive maintenance can also be divided into Condition based maintenance (CBM) where you predict the maintenance based of the health status from the data you collect from senors, and Remaining useful life (Rul) where you predict the remaining life of the machine at each point. Either way for PdM the most important thing is to collect lots of failure data in order to train your model predicting failure.
Is Remaining-useful-Life based maintenance not just a software extension of condition based maintenance? Since you need data from the latter to calculate remaining life.
Why would someone wants to live forever? We are already that there just too many of us and the earth can't handle us anymore. Only rich would like to live forever.
Predictive maintenance is the future but I think we need a new profession in the engineering fraternity that will just be responsible for precise data collection.
Yes and also realize that you can have all the data in the world, and then the best you can do is bring the thing down controllably, because: 1. Spare part is too expensive / complex, 2. The lead time before warning is way too short, 3. The actual failure distribution of the data is very Poisson like -> meaning, shoots up super quick just before failure and never earlier… Machine Learning will not help you change the statistical distribution of these failures. So still a design challenge, which can be assisted by AI possibly to take care of all the requirements better.
sounds good as long as I don't have to collect the data myself or device manufacturers are willing to collect, analyse, and validate the data and make them available for its customers. Otherwise predictive maintenance is a severely limited solution that MAY or MAY NOT work even for the organisation that has the capacity to collect the data themselves.
Corrective may have a part/sensor merely out of parameters yet not catastrophically failed. I see corrective maintenance as a site specific instagation of predictive maintenance, catching the part/sensor just as it starts its final slump into failure.
Hi Nchimunya, Thanks for your comment, and great to hear your motivation! With RealPars you will automatically receive free certificates of completion for each completed course series. You will find those when you click on "My Account" followed by "Certifications". Feel free to sign up for our course library via the following link, and start your PLC programming journey! learn.realpars.com/bundles/pro
As a hobby I happen to be working on an oil refinery video game and I intend to have equipment maintenance be a mechanic in the game. If anyone has ideas on how to gamify predictive maintenance, I’m open to suggestions :)
I work in that industry - you need to list assets and components across production lines, and schedule maintenance events based on machine work cycles and asset criticalities. The point is to reduce production line downtime. Yes, there is an asset deterioration line like in the video, but it can be extended with preventative tasks such as appropriate lubrication and replacement of small parts - you need both preventative (ex. lubrication) and predictive (ex. oil analysis) tasks, grouped in weekly/monthly routes.
Taking the human physiology as an example, we can analyse the symptoms and can perform a predictive maintenance but what about a 4-20mA transmitter? I have not yet seen them throw any symptoms, they just simply stop. I would really like to know in detail about the predictive maintenance before I conclude it's just a buzzword
A 4-20ma transmitter is a fairly simple device that either works or it doesn't. The "live zero" feature is useful to detect a broken wire or loss of 24VDC. Preventative maintenance for these devices can be employed by trending the results of calibrations over time. If the "as left" deviations are trending higher, then a prediction can be made concerning the life of the transmitter. Visual inspections (terminals beginning to corrode, etc.) can also be used to predict sensor life. Predictive Maintenance is used most often in terms of equipment performance, but it can be applied to sensors as well.
The manufacturer might be acquiring good data to enhance their "mean time to failure" metric. They may share that to end users, who then follow the recommended predictive replacement. However they may also be getting an unwanted feedback loop of sites changing sensors according to predictions that are themselves being used to time the replacement of the sensor. That would cause negative creep in the metric, leading to shorter & shorter MTTF.
How to get actual diagnostic parameter from transmitter? we may consider to use HART/FF/PROFIBUS signal communication with Instrument Asset Management System (IAMS) and then integrate with PLC, PLC need to build up general rule & online detect specific parameter... Normally temperature transmitter/pressure transmitter/DP type flow transmitter may follow maintenance engineer's experience to build up intelligent decision. However other specific transmitter such as radar, coriolis flow, ultrasonic transmitter, valve positioner usually end user just call vendor to solve at plant. Vendor site engineer sometimes bring their specific software to know other specific trend / signal ...etc. However those specific software also unable to integrate with IAMS...(whatever DTM already installed), Furthermore how is PLC can integrate with instrument for other diagnostic signal??? The video mentioned very good concept such as BN system 1 decision support but many type of instrument & different root cause related instrument life time... For me, 100% agree video concept but I still don't image how to go in to detail & make it reliable..
I would like to see this question answered. Integration is a huge topic indeed, and having asset health diagnosis automated can reduce the subjective and the costly nature of workers (perhaps). But, it tends to be workers that collect that data from SCADA, combine MES data, plus some verbal remarks from mechanics/oilers, put it in Excel and create Work orders in CMMS. I'd say that's a 30ft overview of maintenance in many plants. But, many OEMs start to integrate condition monitoring modules in their products. I'm not aware of many transmitters self diagnosing, perhaps you can tell it from data they output, but that would be software side - on your end to program.
Great questions! In order to use diagnostics from sensors for predictive and preventative maintenance, we need to be able to extract data from the transmitters. With older sensors (4-20ma), this may not be possible. We can use calibration trends to make decisions, but that is not very "rich" data. Use of ProfiBus PA, FFB, and IO-Link opens up a wealth of diagnostic data from the sensors, but still, is only useful if it is read, stored, and actions taken. PLC and DCS systems can certainly be used to read this diagnostic data but may or may not be able to store this data long term, although data historians . AMS systems and other PC-based software is designed to analyze the data and provide actionable items, but it takes time, a deep understanding of the process systems involved, and a commitment to review and update of these systems. This is not an easy task. If the PLC can "see" the diagnostic register(s), it should be able to retrieve and store the data. There may be "local diagnostic" data that is accessible only through a special interface, in which case, the vendor needs to be pressed to provide this data to your system. This scenario is becoming more rare, since it is to the vendor's benefit to be able for the customer to be able to provide this data to them remotely. All of these diagnostics opportunities need to evaluated for how they can be integrated into the plant's systems. It is likely you will never have 100% diagnostic coverage.
What solutions do you use for PDM, please give me some information. I'm going to have plan to research Senseye solutions (Siemens bussines) for this case.
Eventhough it may be called preventive maintenance, many technicians or maintenance personnels do not change their spare parts based on the recommended intervals in vendors' manuals but according to fail history within their plant or based on info of similar operations as failures are based on various factors i.e running hours, operating and environmental conditions, liquid physical and chemical properties, assuming that there is no installation mistake or human error. Hence, it is safe to say that predictive maintenance have long been in practice. 😊
It is true that predictive Maintenance is being used by specialists for a long time. But, by combining the use of computers with human intelligence I believe we can make this decision a lot more accurate.
Also during preventative maintenance it takes more total time for tool to be down. Just because it is preventative maintenance does not mean that the equipment does not have to go down, the difference is that during preventative maintenance you can choose least busiest time to do PM.
@@soheil5710 Same stuff I can say about actual part failing. If I am knowing estimate life-spam of the part, let it fail and without panics, diagnostics replace it. If part is not expensive you should have replacement in stock already. If you have more than one equipment with the same part, you could have a little bit more parts in stock. Your comment does not add any value to the total discussion of the subject. Some parts have life span and sitting in the storage units reduces their life-span. Which reduces their failing time estimates.
@@СтепанМакухов failure in the middle of a production run is much much worse than planned downtime. Just because you have a box of spare sensors doesn't mean the product hasn't been subjected to out of parameters production, probably ruining the batch if not damaging other parts/equipment. Don't be naive.
The point is most of the industrial equipment like that is running 24/7 and is not easy to schedule preventing maintenance and most cases the companies are not willing to do it, resulting in more down time when it breaks costing more thousands of dollars than a simply PM time could takes...
Yes. Pushing time between replacement to near to it's maximum is important, but a small saving for a big risk just isn't a good gamble. Having a redundant sensor/part is my preferred method. Not always possible, but definitely bolsters the collection of failure data.
To answer your question, any translation or modification of our video courses are against RealPars' and TH-cam's copy right policy. You can share our video as long as it remains unmodified, tagged and credited back to us. But any modification or translation is not allowed. You can email us an SRL document with the translated subtitles, and we will happily add those to the specific video course. Thanks for your understanding.
I don't agree with you regarding the information collected by everyone to create a database because many parameters are not the same. For example, a transmitter installed in Africa under specific conditions, such as high temperatures and humidity, is not the same as in Europe where temperatures can drop below zero.
You've raised an interesting point. If I were to create a database based on your question, I'd gather data from the region or area where I'm located for meaningful comparisons. Wishing you a rewarding learning experience with RealPars.
Yeah, "predictive", I'm better off paying a gypsy 5 bucks and her "predict" the fate of devices. All these hofus pocus can't beat good 'ol maintenace work
Predictive maintenance can also be divided into Condition based maintenance (CBM) where you predict the maintenance based of the health status from the data you collect from senors, and Remaining useful life (Rul) where you predict the remaining life of the machine at each point. Either way for PdM the most important thing is to collect lots of failure data in order to train your model predicting failure.
Is Remaining-useful-Life based maintenance not just a software extension of condition based maintenance?
Since you need data from the latter to calculate remaining life.
@@soheil5710 it's an extension of course. But a separate thing according to literature
Could you explain more!
Why would someone wants to live forever? We are already that there just too many of us and the earth can't handle us anymore. Only rich would like to live forever.
Predictive maintenance is the future but I think we need a new profession in the engineering fraternity that will just be responsible for precise data collection.
Yes and also realize that you can have all the data in the world, and then the best you can do is bring the thing down controllably, because: 1. Spare part is too expensive / complex, 2. The lead time before warning is way too short, 3. The actual failure distribution of the data is very Poisson like -> meaning, shoots up super quick just before failure and never earlier…
Machine Learning will not help you change the statistical distribution of these failures. So still a design challenge, which can be assisted by AI possibly to take care of all the requirements better.
I wasn't ready for RealPars to start asking me about immortality, but I'm all here for it
Simple and very instructive video. Predictive maintenance in plants will require huge amounts of data to be analyzed.
Glad it was helpful!
Next time I forget to PM my equipment I’ll just show my boss this video and tell him I’m saving the company money.
Thank RealPars very much
You're very welcome!
Great video, from Brazil thank very much.
Glad you liked it!
Thankyou for Teaching my friend
It's our pleasure!
sounds good as long as I don't have to collect the data myself or device manufacturers are willing to collect, analyse, and validate the data and make them available for its customers. Otherwise predictive maintenance is a severely limited solution that MAY or MAY NOT work even for the organisation that has the capacity to collect the data themselves.
Hello
Sir, Good explains tions
Yes sir we must be careful
Great to hear that!
IS REACTIVE MAINTENANCE SAME AS CORRECTIVE MAINTENANCE?
Yes. They are two descriptions of the same approach to maintenance: wait for a fault to occur, then correct and/or repair.
Corrective may have a part/sensor merely out of parameters yet not catastrophically failed. I see corrective maintenance as a site specific instagation of predictive maintenance, catching the part/sensor just as it starts its final slump into failure.
I'm an electrical engineer from Zambia looking to get into electrical reliability, any suggestions on which sites offer cheap online certification?
Hi Nchimunya,
Thanks for your comment, and great to hear your motivation!
With RealPars you will automatically receive free certificates of completion for each completed course series.
You will find those when you click on "My Account" followed by "Certifications".
Feel free to sign up for our course library via the following link, and start your PLC programming journey! learn.realpars.com/bundles/pro
Thank you, I'm watching
Thanks for your support!
I’m watching this channel since 2015
That's amazing! Thank you for your endless support!
As a hobby I happen to be working on an oil refinery video game and I intend to have equipment maintenance be a mechanic in the game. If anyone has ideas on how to gamify predictive maintenance, I’m open to suggestions :)
I work in that industry - you need to list assets and components across production lines, and schedule maintenance events based on machine work cycles and asset criticalities. The point is to reduce production line downtime. Yes, there is an asset deterioration line like in the video, but it can be extended with preventative tasks such as appropriate lubrication and replacement of small parts - you need both preventative (ex. lubrication) and predictive (ex. oil analysis) tasks, grouped in weekly/monthly routes.
How far are you into developing the game?
Taking the human physiology as an example, we can analyse the symptoms and can perform a predictive maintenance but what about a 4-20mA transmitter? I have not yet seen them throw any symptoms, they just simply stop. I would really like to know in detail about the predictive maintenance before I conclude it's just a buzzword
Good question. Similarly other electronic devices that just fail all of a sudden.
A 4-20ma transmitter is a fairly simple device that either works or it doesn't. The "live zero" feature is useful to detect a broken wire or loss of 24VDC. Preventative maintenance for these devices can be employed by trending the results of calibrations over time. If the "as left" deviations are trending higher, then a prediction can be made concerning the life of the transmitter. Visual inspections (terminals beginning to corrode, etc.) can also be used to predict sensor life. Predictive Maintenance is used most often in terms of equipment performance, but it can be applied to sensors as well.
The manufacturer might be acquiring good data to enhance their "mean time to failure" metric. They may share that to end users, who then follow the recommended predictive replacement.
However they may also be getting an unwanted feedback loop of sites changing sensors according to predictions that are themselves being used to time the replacement of the sensor. That would cause negative creep in the metric, leading to shorter & shorter MTTF.
Thank you very much!
You're very welcome!
@@realpars 🥰
How to get actual diagnostic parameter from transmitter?
we may consider to use HART/FF/PROFIBUS signal communication with Instrument Asset Management System (IAMS) and then integrate with PLC, PLC need to build up general rule & online detect specific parameter...
Normally temperature transmitter/pressure transmitter/DP type flow transmitter may follow maintenance engineer's experience to build up intelligent decision.
However other specific transmitter such as radar, coriolis flow, ultrasonic transmitter, valve positioner usually end user just call vendor to solve at plant. Vendor site engineer sometimes bring their specific software to know other specific trend / signal ...etc.
However those specific software also unable to integrate with IAMS...(whatever DTM already installed), Furthermore how is PLC can integrate with instrument for other diagnostic signal???
The video mentioned very good concept such as BN system 1 decision support but many type of instrument & different root cause related instrument life time...
For me, 100% agree video concept but I still don't image how to go in to detail & make it reliable..
I would like to see this question answered. Integration is a huge topic indeed, and having asset health diagnosis automated can reduce the subjective and the costly nature of workers (perhaps). But, it tends to be workers that collect that data from SCADA, combine MES data, plus some verbal remarks from mechanics/oilers, put it in Excel and create Work orders in CMMS. I'd say that's a 30ft overview of maintenance in many plants. But, many OEMs start to integrate condition monitoring modules in their products. I'm not aware of many transmitters self diagnosing, perhaps you can tell it from data they output, but that would be software side - on your end to program.
@@armelchiza3771 thank your reply... my previous statement still unable to full present how difficult for this topic.. 😅 😑 🙃
Great questions! In order to use diagnostics from sensors for predictive and preventative maintenance, we need to be able to extract data from the transmitters. With older sensors (4-20ma), this may not be possible. We can use calibration trends to make decisions, but that is not very "rich" data. Use of ProfiBus PA, FFB, and IO-Link opens up a wealth of diagnostic data from the sensors, but still, is only useful if it is read, stored, and actions taken. PLC and DCS systems can certainly be used to read this diagnostic data but may or may not be able to store this data long term, although data historians . AMS systems and other PC-based software is designed to analyze the data and provide actionable items, but it takes time, a deep understanding of the process systems involved, and a commitment to review and update of these systems. This is not an easy task. If the PLC can "see" the diagnostic register(s), it should be able to retrieve and store the data. There may be "local diagnostic" data that is accessible only through a special interface, in which case, the vendor needs to be pressed to provide this data to your system. This scenario is becoming more rare, since it is to the vendor's benefit to be able for the customer to be able to provide this data to them remotely. All of these diagnostics opportunities need to evaluated for how they can be integrated into the plant's systems. It is likely you will never have 100% diagnostic coverage.
Sin duda el mantenimiento predictivo es lo mejor en cuanto a costos y confiabilidad
Gracias, Pedro!
Thank you.😊
You're welcome!
Great video
Thank you, Ahmad!
Thanks ❤️
You're welcome!
Very informative video....
Glad you liked it!
What solutions do you use for PDM, please give me some information. I'm going to have plan to research Senseye solutions (Siemens bussines) for this case.
Thanks for your suggestion! I will happily pass this on to our team, hopefully we can create a full video course on this.
Thank you i learned a lot
You're welcome!
Preventive is not necessarily replacing. Like change oil in cars, check breaks, etc
Not Forgetting prescriptive maintenance! All this (office) maintenance doesn't help much when non technical people are running it!
I liked great job
Thank you, Derin!
Eventhough it may be called preventive maintenance, many technicians or maintenance personnels do not change their spare parts based on the recommended intervals in vendors' manuals but according to fail history within their plant or based on info of similar operations as failures are based on various factors i.e running hours, operating and environmental conditions, liquid physical and chemical properties, assuming that there is no installation mistake or human error. Hence, it is safe to say that predictive maintenance have long been in practice. 😊
It is true that predictive Maintenance is being used by specialists for a long time. But, by combining the use of computers with human intelligence I believe we can make this decision a lot more accurate.
Also during preventative maintenance it takes more total time for tool to be down. Just because it is preventative maintenance does not mean that the equipment does not have to go down, the difference is that during preventative maintenance you can choose least busiest time to do PM.
Actually it may take even less time, since you know exactly what parts/tools you need beforehand.
No more panic diagnosing or waiting for part orders.
@@soheil5710 Same stuff I can say about actual part failing. If I am knowing estimate life-spam of the part, let it fail and without panics, diagnostics replace it. If part is not expensive you should have replacement in stock already. If you have more than one equipment with the same part, you could have a little bit more parts in stock.
Your comment does not add any value to the total discussion of the subject.
Some parts have life span and sitting in the storage units reduces their life-span. Which reduces their failing time estimates.
@@СтепанМакухов failure in the middle of a production run is much much worse than planned downtime. Just because you have a box of spare sensors doesn't mean the product hasn't been subjected to out of parameters production, probably ruining the batch if not damaging other parts/equipment. Don't be naive.
The point is most of the industrial equipment like that is running 24/7 and is not easy to schedule preventing maintenance and most cases the companies are not willing to do it, resulting in more down time when it breaks costing more thousands of dollars than a simply PM time could takes...
Thanks for adding that, Julio
Yes. Pushing time between replacement to near to it's maximum is important, but a small saving for a big risk just isn't a good gamble.
Having a redundant sensor/part is my preferred method. Not always possible, but definitely bolsters the collection of failure data.
That's great
Can you translate your videos into Arabic?
To answer your question, any translation or modification of our video courses are against RealPars' and TH-cam's copy right policy.
You can share our video as long as it remains unmodified, tagged and credited back to us. But any modification or translation is not allowed.
You can email us an SRL document with the translated subtitles, and we will happily add those to the specific video course.
Thanks for your understanding.
I don't agree with you regarding the information collected by everyone to create a database because many parameters are not the same. For example, a transmitter installed in Africa under specific conditions, such as high temperatures and humidity, is not the same as in Europe where temperatures can drop below zero.
You've raised an interesting point. If I were to create a database based on your question, I'd gather data from the region or area where I'm located for meaningful comparisons. Wishing you a rewarding learning experience with RealPars.
Yeah, "predictive", I'm better off paying a gypsy 5 bucks and her "predict" the fate of devices. All these hofus pocus can't beat good 'ol maintenace work
You have to compare likenesses. What is the environment ? hostile or passive?
Thanks for the feedback, Wayne! Will go ahead and pass this on to our course developers.