All Member Open Forum

Are you worried about keeping your skills modern going into the new age of data and analytics-as many of us should be? But relax. You already have, by far, the most difficult skill to come by for staying modern. You are a subject matter expert in maintenance and reliability by virtue of years of experience. In contrast, the skills of data and analytics relevant to you in the new age can be acquired in short order as you put them into play.

If you have a fellow maintenance and reliability SME to guide you, you can learn to work with data in a day. Guided by a fellow SME, in hours you can be brought to know the breadth of ways data can be transformed to newly obtainable insights and embedded in operational processes. Guided by a fellow SME, in hours you can be brought to know the methods to each insight deliverable.

You can see why. As an SME, when you work with data extracted from your CMMS and then join and cleanse the data into super tables, every variable, every row and every cell has meaning to you. As an SME, when you are introduced to the range, power and purpose of data-based insights, you immediately recognize which would matter in your case. While putting the recognized insight analytics into play, it is you as an SME that is the most important input to building and evaluating each insight deliverable.

So what are you going to do about it?

------------------------------
Richard Lamb
Analytics4Strategy.com
Houston TX
832-710-0755
------------------------------

​Really great opening question that needs much discussion.  As much as I embrace technology, we need to proceed with caution.  The caution is to not forget that basic fundamental skills in maintenance must still be taught and understood even when mixing advanced technology with analytics.  The problem I see is blind trust in the technology (and sensors, programming, etc.) and little application understanding by new end users.  This then handicaps troubleshooting and correcting future problems in the field.  My feeling is we are forcing some technology before it or the workforce is ready to deliver on the results.

Here is a somewhat relative example of the technology vs skill creep.  Machine alignment was practiced and taught for years with dial indicators, graph paper, etc.  Skilled labor understood the moves and what they were doing and why.  The invent of laser alignment offered many advantages but no longer requires a depth of understanding the moves (the computer does all that but not all bad).  However, newer crafts people now mostly learn to push buttons and just get the screen results. 

I see errors all the time with forcing the alignment to get the reading and forgetting the fundamentals such as eliminate soft foot, rough alignment, proper shim practices, etc.  We begin to lose skilled labor but have trained button pushers.  When they can't get the laser to give them the correct numbers or repeatable, everyone blames the laser but it is really a lack of fundamentals to alignment.  I see this all the time.  It is even to the point that the older hands have even given way to not thinking about what they are doing.

We haven't heard the result of the Boeing 737 crashes but it sure sounds like some technology went wrong and human intervention failed (probably for several reasons) leading to the crashes.  Hard to replace a skilled, trained human pilot able to adjust to an infinite set of variables which is impossible to program.

We need to remember to let the technology work for us and not us be slaves to the technology.  I personally think the human mind is the most complicated and marvelous creation by the Creator and artificial intelligence while it has its place will never be able to replace the human mind and reason that is well trained.  I believe our future must be a marriage of both technology and skill (white collar and blue collar) not just one to be successful.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

Original Message:
Sent: 03-19-2019 09:49 AM
From: Richard Lamb
Subject: Are you worried about keeping your skills modern going into the new age of data and analytics

Are you worried about keeping your skills modern going into the new age of data and analytics-as many of us should be? But relax. You already have, by far, the most difficult skill to come by for staying modern. You are a subject matter expert in maintenance and reliability by virtue of years of experience. In contrast, the skills of data and analytics relevant to you in the new age can be acquired in short order as you put them into play.

If you have a fellow maintenance and reliability SME to guide you, you can learn to work with data in a day. Guided by a fellow SME, in hours you can be brought to know the breadth of ways data can be transformed to newly obtainable insights and embedded in operational processes. Guided by a fellow SME, in hours you can be brought to know the methods to each insight deliverable.

You can see why. As an SME, when you work with data extracted from your CMMS and then join and cleanse the data into super tables, every variable, every row and every cell has meaning to you. As an SME, when you are introduced to the range, power and purpose of data-based insights, you immediately recognize which would matter in your case. While putting the recognized insight analytics into play, it is you as an SME that is the most important input to building and evaluating each insight deliverable.

So what are you going to do about it?

------------------------------
Richard Lamb
Analytics4Strategy.com
Houston TX
832-710-0755
------------------------------

Original Message------

​Really great opening question that needs much discussion.  As much as I embrace technology, we need to proceed with caution.  The caution is to not forget that basic fundamental skills in maintenance must still be taught and understood even when mixing advanced technology with analytics.  The problem I see is blind trust in the technology (and sensors, programming, etc.) and little application understanding by new end users.  This then handicaps troubleshooting and correcting future problems in the field.  My feeling is we are forcing some technology before it or the workforce is ready to deliver on the results.

Here is a somewhat relative example of the technology vs skill creep.  Machine alignment was practiced and taught for years with dial indicators, graph paper, etc.  Skilled labor understood the moves and what they were doing and why.  The invent of laser alignment offered many advantages but no longer requires a depth of understanding the moves (the computer does all that but not all bad).  However, newer crafts people now mostly learn to push buttons and just get the screen results. 

I see errors all the time with forcing the alignment to get the reading and forgetting the fundamentals such as eliminate soft foot, rough alignment, proper shim practices, etc.  We begin to lose skilled labor but have trained button pushers.  When they can't get the laser to give them the correct numbers or repeatable, everyone blames the laser but it is really a lack of fundamentals to alignment.  I see this all the time.  It is even to the point that the older hands have even given way to not thinking about what they are doing.

We haven't heard the result of the Boeing 737 crashes but it sure sounds like some technology went wrong and human intervention failed (probably for several reasons) leading to the crashes.  Hard to replace a skilled, trained human pilot able to adjust to an infinite set of variables which is impossible to program.

We need to remember to let the technology work for us and not us be slaves to the technology.  I personally think the human mind is the most complicated and marvelous creation by the Creator and artificial intelligence while it has its place will never be able to replace the human mind and reason that is well trained.  I believe our future must be a marriage of both technology and skill (white collar and blue collar) not just one to be successful.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

Good info in the slides; not sure we are talking exactly the same things but certainly related.

As far as "big data" from any source (CMMS, field sensor, etc.), garbage in will lead to garbage out.  So one thing we must do is if anyone leans on the result of an automated AI platform is make sure the data source is working properly.  This means high skill will be required by folks keeping these data sources working properly - calibrated, discipline inputs where human input is required like a CMMS, etc.  So we better get the input sources maintenance and function skill level up by our employees.  Been part of many organizations where these data systems were not maintained for several reasons - cost, skill level, business model, culture, etc.

The next thing​ is we better have a proven algorithm in place that is doing the correct analysis and conclusions.  The "smart" systems I've seen early on have had big flaws.  Predictive maintenance algorithms do the math but the assumptions, wrong inputs and misc variables have yet to be proven consistent and reliable without a skilled analyst intervening. 

Do a better job screening any system that goes in.  One OEM had some of the first "smart" systems several decades ago.  The algorithm was actually pretty good but only had one problem.  It was setup to control parameters to extend or preserve equipment reliability and in doing so it would sacrifice the process in the plant.  The priority is all wrong, plants have to run - reliable or not.   The other related issue is over selling "AI" and "smart" systems.  No one has an algorithm good enough to predict component life on a simple machine down to the day but I've heard it several times.  It would take some at length study with 10s and 100s of variables to model that and then enough sensors to model those variables in the field.  I'm not convinced there is a business model to support that yet.  So get long term buy in before investing time and resources on this type of project.

Industry has been doing condition monitoring and process predicting for years in software platforms so the foundation of this is nothing new.  With new technologies the ability to improve that and grow that capability is certainly better than ever before but we have to remember the rest of industry, culture and skill labor has to keep pace with it to be successful long term to be more than a flash in the pan project. 

Don't let your limbs reach further than your roots are deep.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

Original Message:
Sent: 03-21-2019 10:33 AM
From: Richard Lamb
Subject: Are you worried about keeping your skills modern going into the new age of data and analytics

The big question is what are you going to do about it if you are worried about keeping your skills modern going into the new age of data and analytics? Because the comments made by Randy Riddell are commonly voiced, a large part of the answer is suggested by them. We need to establish among us maintenance and reliability practitioners a clear, implementable understanding of data-drivenness.

The slide set, "First Step to Becoming a Data-Driven Operation," was prepared to begin building a common understanding amongst us. To download the slides, scroll to the download button at the bottom of the webpage https://analytics4strategy.com/train-frststpdtdrvnops.

A "data-driven" maintenance operation is defined as one that harnesses its CMMS data to augment the experience and judgement of operatives, managers, analysts and engineers as they plan, organize, conduct and control the network of processes that decide their plant's maximal uptime, stay abreast of its site and facility deterioration, and match its maintenance capacity to those ends. In fewer words, best practices can only do best when we use our data.

Original Message------

​Really great opening question that needs much discussion.  As much as I embrace technology, we need to proceed with caution.  The caution is to not forget that basic fundamental skills in maintenance must still be taught and understood even when mixing advanced technology with analytics.  The problem I see is blind trust in the technology (and sensors, programming, etc.) and little application understanding by new end users.  This then handicaps troubleshooting and correcting future problems in the field.  My feeling is we are forcing some technology before it or the workforce is ready to deliver on the results.

Here is a somewhat relative example of the technology vs skill creep.  Machine alignment was practiced and taught for years with dial indicators, graph paper, etc.  Skilled labor understood the moves and what they were doing and why.  The invent of laser alignment offered many advantages but no longer requires a depth of understanding the moves (the computer does all that but not all bad).  However, newer crafts people now mostly learn to push buttons and just get the screen results.

I see errors all the time with forcing the alignment to get the reading and forgetting the fundamentals such as eliminate soft foot, rough alignment, proper shim practices, etc.  We begin to lose skilled labor but have trained button pushers.  When they can't get the laser to give them the correct numbers or repeatable, everyone blames the laser but it is really a lack of fundamentals to alignment.  I see this all the time.  It is even to the point that the older hands have even given way to not thinking about what they are doing.

We haven't heard the result of the Boeing 737 crashes but it sure sounds like some technology went wrong and human intervention failed (probably for several reasons) leading to the crashes.  Hard to replace a skilled, trained human pilot able to adjust to an infinite set of variables which is impossible to program.

We need to remember to let the technology work for us and not us be slaves to the technology.  I personally think the human mind is the most complicated and marvelous creation by the Creator and artificial intelligence while it has its place will never be able to replace the human mind and reason that is well trained.  I believe our future must be a marriage of both technology and skill (white collar and blue collar) not just one to be successful.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

Original Message------

Good info in the slides; not sure we are talking exactly the same things but certainly related.

As far as "big data" from any source (CMMS, field sensor, etc.), garbage in will lead to garbage out.  So one thing we must do is if anyone leans on the result of an automated AI platform is make sure the data source is working properly.  This means high skill will be required by folks keeping these data sources working properly - calibrated, discipline inputs where human input is required like a CMMS, etc.  So we better get the input sources maintenance and function skill level up by our employees.  Been part of many organizations where these data systems were not maintained for several reasons - cost, skill level, business model, culture, etc.

The next thing​ is we better have a proven algorithm in place that is doing the correct analysis and conclusions.  The "smart" systems I've seen early on have had big flaws.  Predictive maintenance algorithms do the math but the assumptions, wrong inputs and misc variables have yet to be proven consistent and reliable without a skilled analyst intervening. 

Do a better job screening any system that goes in.  One OEM had some of the first "smart" systems several decades ago.  The algorithm was actually pretty good but only had one problem.  It was setup to control parameters to extend or preserve equipment reliability and in doing so it would sacrifice the process in the plant.  The priority is all wrong, plants have to run - reliable or not.   The other related issue is over selling "AI" and "smart" systems.  No one has an algorithm good enough to predict component life on a simple machine down to the day but I've heard it several times.  It would take some at length study with 10s and 100s of variables to model that and then enough sensors to model those variables in the field.  I'm not convinced there is a business model to support that yet.  So get long term buy in before investing time and resources on this type of project.

Industry has been doing condition monitoring and process predicting for years in software platforms so the foundation of this is nothing new.  With new technologies the ability to improve that and grow that capability is certainly better than ever before but we have to remember the rest of industry, culture and skill labor has to keep pace with it to be successful long term to be more than a flash in the pan project. 

Don't let your limbs reach further than your roots are deep.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

The science using regression model with good data is good stuff and I agree can produce some good information to manage our maint.  In theory, I agree with the possibilities of AI, VR, MR, etc (industry 4.0).  The challenge is getting good, reliable data (from the CMMS in the case you mention) from actual failures, maint plans, etc.  If this is all done then predictive analytics, doing the math can certainly add some value. 

The first challenge with good data is getting discipline in the organization to keep good data.  Many organizations are not good at that and have the wrong PMs in place (too often or not often enough) or do not document actual failures.  Documenting actual failures and types of failures (failure modes) will be critical going forward for industry 4.0 to build confidence in the maintenance/reliability area. 

In the case of industry 4.0, too many companies were too eager to market something they didn't have or could not execute.  Kind of like the news reporter that thought he had a headline story and released the sensational headline without researching the facts.

Similar situation with the implementation several decades ago when predictive maintenance (vibration analysis was new in industry).  If the database has wrong info (bearings, rotor, etc.) then analysis will not be reliable.  The automated vibration algorithms will then apply the known physics (math) to the wrong data and thus wrong results.  Had this happen a lot especially with inexperienced vibration analyst making calls based on software analysis.  To keep this accurate and working properly, the analysist and other reliability professionals must work to keep this up to date. 

This is where I'm not sure the stakeholders understand what is required on the ground to back feed these "smart" systems to make them work correctly.  Maintaining data sources with good data is not free but will take some resources.   Yes the platform can be done on a one time install but keeping it working properly will cost the organization something.

Richard, I appreciate the discussion.  Looks like you have really done your homework on how to make the behind the scenes data analytics work.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------

Original Message:
Sent: 03-22-2019 05:34 PM
From: Richard Lamb
Subject: Are you worried about keeping your skills modern going into the new age of data and analytics

To Randy Riddell's comment, "Good info in the slides; not sure we are talking exactly the same things but certainly related."

Yes and no. The slides explain the framework of data driven operations in which the experience and judgement of SMEs are augmented with a long menu of Insight Deliverables (slides 6 and 58 through 70). In contrast, Randy is talking about a particular sector of insight deliverables. However, the framework is still present behind the sector just as it is for all types of insight deliverables-basic to exotic.

Randy's comment, "No one has an algorithm good enough to predict component life on a simple machine down to the day, but I've heard it several times," points to an important point. Some insight deliverables are way over-promised. Worse, they are causing us to not notice the 98 percent of the range of insight deliverables that are practicable in augmenting us in the conduct of maintenance and reliability-best practices are practiced best with data.

Additionally to Randy's point, we should never forget that machine learning is simply the act of fitting data to a many dimensional line with an algorithm. In turn, AI is using the learned fit to predict that, in the long-run, cases with particular characteristics have a predicted outcome. And when we hang our hat on a prediction (AI), we should always remember that a good AI model is one that gets it right 85 percent or greater-think about that when crossing the street in front of a driverless car.

Randy, Thank you for your comments. Each comment I am privileged to receive through the forum, allows me to learn how to improve the message for reaching to data-driven reliability and maintenance.

Original Message------

Good info in the slides; not sure we are talking exactly the same things but certainly related.

As far as "big data" from any source (CMMS, field sensor, etc.), garbage in will lead to garbage out.  So one thing we must do is if anyone leans on the result of an automated AI platform is make sure the data source is working properly.  This means high skill will be required by folks keeping these data sources working properly - calibrated, discipline inputs where human input is required like a CMMS, etc.  So we better get the input sources maintenance and function skill level up by our employees.  Been part of many organizations where these data systems were not maintained for several reasons - cost, skill level, business model, culture, etc.

The next thing​ is we better have a proven algorithm in place that is doing the correct analysis and conclusions.  The "smart" systems I've seen early on have had big flaws.  Predictive maintenance algorithms do the math but the assumptions, wrong inputs and misc variables have yet to be proven consistent and reliable without a skilled analyst intervening.

Do a better job screening any system that goes in.  One OEM had some of the first "smart" systems several decades ago.  The algorithm was actually pretty good but only had one problem.  It was setup to control parameters to extend or preserve equipment reliability and in doing so it would sacrifice the process in the plant.  The priority is all wrong, plants have to run - reliable or not.   The other related issue is over selling "AI" and "smart" systems.  No one has an algorithm good enough to predict component life on a simple machine down to the day but I've heard it several times.  It would take some at length study with 10s and 100s of variables to model that and then enough sensors to model those variables in the field.  I'm not convinced there is a business model to support that yet.  So get long term buy in before investing time and resources on this type of project.

Industry has been doing condition monitoring and process predicting for years in software platforms so the foundation of this is nothing new.  With new technologies the ability to improve that and grow that capability is certainly better than ever before but we have to remember the rest of industry, culture and skill labor has to keep pace with it to be successful long term to be more than a flash in the pan project.

Don't let your limbs reach further than your roots are deep.

------------------------------
Randy Riddell, CMRP, PSAP, CLS
Reliability Manager
Essity
Cherokee AL
------------------------------