Robot Shift

Experience is important.  Experience ensures the right technology is applied in the right way.  But how do you measure experience and what experience is important?  No one says “This is completely new to us, but please give us a chance to learn on your dime.”  On the flipside, you also want an integrator that isn’t afraid of some healthy stretching.   Too many times I’ve seen integrators who are afraid to step outside of their comfort zone and continue to implement antiquated technologies that sacrifice the final system’s performance, and maintainability.

Experience falls into two categories:

1.  Technology

2.  Application  

Technology Experience:

The integrator may not have solved your exact problem before, but you want to see examples (and references) of how they’ve applied their core technologies in new and innovative ways.  You want to see solutions that were innovative and cutting edge, while still being maintainable by an electrician or millwright at 2:00 AM (no science fair projects).  They may not have solved your exact application before, but if they’ve stretched to solve other problems of similar complexity, then it shows a track record of success.  You can gauge your application’s relative difficulty to what they’ve done.  Talk to their references to make sure it is real.

Application Experience:

Technology only gets you so far.  The integrator needs to have application experience in your industry to put together a complete solution.  Integration is about applying technology to solve a business problem.  If an integrator doesn’t understand your business and its key drivers, how can they apply the technology correctly?  If you’re looking to automate raw food handling, don’t use an integrator that specializes in robotic welding and expect it to be designed to the AMI Meat Safety Standards.  I’m surprised how often this happens.  Ask the integrator to show you completed projects within your industry.  Ask them about key drivers for your industry (product quality, sanitary design, washdown, heat transfer/high temperature, validation, documentation, etc.).  If they don’t know this stuff for your industry, that’s a big red flag.

Finally, I’ll bring it back to my #10 sign an integrator is the real deal.  Ask them where they don’t fit.  What technologies, applications and industries are outside their wheelhouse?  Anyone who claims to do everything, is really a generalist that is great at nothing.

I’m starting to lag behind since my last Top 10 Signs Post.  Things have been busy, but I promise to get the next one up shortly!

I wanted to take a moment to comment on some of the feedback I’ve gotten thus far.  It’s been interesting putting these criteria up here.  It’s sparked lots of great conversations, debates and emails.

I’ve had a couple of comments these criteria are good, but only really apply to high-dollar or technically challenging projects.  For commodity applications, or small low-dollar projects they don’t necessarily apply.  I prefer to think of these as tools to select an integrator use when you (end-user) can’t afford to be wrong.

If you can afford to be wrong on your project (i.e. late project delivery, not meet the desired OEE, incur some extra costs, excessive downtime to install/integrate, etc.) then you should go with the lowest cost provider.  Why wouldn’t you?  There are lots of projects that aren’t mission critical to production where a hiccup doesn’t sink the ship.  For those projects, these criteria don’t apply.

Project size isn’t a good gauge either.  One of our engineers that was embedded at Toyota once said to me “This work [small continuous improvement projects] is some of the simplest, no-glory work, but at the same time the most stressful.  We can make a small change to the line over the weekend and if everything goes well, no one knows about it Monday morning.  It’s business as usual.  If it doesn’t go well, everyone knows”.  They can’t afford to be wrong.

More to come…

You would think every integrator would have solid, proven standards.  It’s logical, and engineers are logical, right?  The problem is engineers love to create.  This often results in re-creating the wheel on a project to project basis, a lot of times when the previous wheel worked just fine.

Great integrators don’t do this.  Great integrators have a culture of consistency coupled with continuous improvement.  They understand the best approach is most often the consistent approach.  This doesn’t mean things stay stagnant.  On the contrary, they channel the engineers’ need to create into improving the company’s collective standards and best practices.

 This culture means that:

• Good designs are evolved to spectacular designs.
  • Bugs have been found in code.
  • Fatigue points have been found in mechanical systems.
• There’s an increased efficiency by the integrator and therefore reduced cost to the end user.
• Better support – since everyone knows the standards, everyone can support it.  This means interchangeability of resources.  When you call at 2AM on a Tuesday and the person who did your project is on vacation, you can get support from someone else within the company, because they understand the standards that were used.

Best practices and standards don’t just apply to technology.  This culture pervades every aspect of the business including billing, project management (costs, scope, schedule), HR and customer service.

An environment where people have to think brings with it wisdom, and this wisdom brings with it kaizen [continuous improvement].  -Teruyuki Minoura

I saw a commercial for the new Apple iPad .  In it they said “You already know how to use it”.  That really resonated with me.  That’s the way great automation systems are – the operators already know how to use them.

What makes them simple to use?  I may be getting pretty nitty gritty here, but it’s the way they are programmed.  Most people tend to think, and therefore program, based on a series of steps or sequences.  The problem is, when a robot gets to step 78 of a 123 step process and something out of the ordinary happens, it doesn’t know what to do next.  Think of how frustrating it is for you when your Windows computer locks up.  That’s the same feeling an operator gets on a daily basis with a glitchy automation system that runs on step or sequence based logic.  This ties into people’s perceptions, beliefs and acceptance of the equipment in your plant.

Great automation systems don’t follow a series of steps, they are programmed with rules or priorities.  These priorities allow the robot(s) to make decisions about what is the most important thing to do based on the state of the equipment within the cell.  This means it handles the the what-if situations that inevitably come up with ease.  This makes it simple and intuitive for an operator to:

• Recover after an E-Stop is pressed
• Recover after a power outage in the plant
• Recover when parts are out of sequence
• Recover when the process is stopped and parts removed
• Start up after a changeover
• Start up on a Monday morning

Not only does it handle these fault scenarios better, but it also runs better and makes more widgets.  Because the system has an understanding of the priorities, it can continually adapt its cycle to optimize the process to keep the highest priority equipment running at the highest capacity possible.

Every integrator is going to tell you that their systems are easy to operate, but how do you really know?  Ask them to talk you through the recovery procedures for the fault scenarios I’ve listed above.  Or ask to see their operator manuals and review these procedures.  Ask their references how they recover from these situations.  It shouldn’t take an operator more than 1 or 2 steps to get the system running after one of these occur.  If they have a 12-step procedure to recover – it’s too complicated.

“Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius — and a lot of courage — to move in the opposite direction”  – Einstein

Automation is a risky business.  You’re designing machines to do the job of people and that can be a tough task (see my post How To Automate A Complex Manual Process).

The main challenges or risks with any automation project fall into three categories: 

1)  Can you go fast enough to keep up? (Throughput/Speed)

 2)  Can you find the parts, pick them up, and place them accurately all while not dropping or damaging them? (Locating and Gripping Strategies)

3)  Can you ensure quality will remain in the process without people there? (Quality Checks)

The devil is in the details.  I’ve seen countless projects go sideways because integrators made assumptions about these variables and they came back to bite them.  You don’t want a science fair project on your floor.

Great integrators are up front and realistic about the risks of projects and have tools/strategies to mitigate these risks.  Kinematic simulation software can be used to determine robot move times very accurately.  Discrete-Event simulation software can plug these robot speeds in with the entire process to get an understanding of the total process throughput/flow (i.e. fork-truck traffic, CNC cycletime, etc.).  Prototyping of grippers or gripping strategies ensures you can grab the parts and handle as required.  Prove out the quality checks.  What tools/technologies will be used to solve this and how well will they work?

Have them show you similar projects they’ve completed successfully using similar strategies/technologies.

…more to come.

Most manufacturers that are new to automation don’t know what they don’t know.  They often select the first or second vendor they talk to, and end up disappointed.  I was reminded of this this week when a new client was introduced to us with the hopes we could fix a number of their outstanding issues that another integrator didn’t/couldn’t finish.  From an integration side, it’s frustrating to compete with integrators that over promise and under deliver.  They often over-simplify and underestimate complexities and risks.

I thought I’d put together my Letterman-style Top 10 List of criteria to select an integrator.  These can be used as a checklist to make sure you’ve got someone who’s the real-deal.  Here’s #10…

#10.  They Know Their Value Proposition 

Another way of saying this is What are they better at than anyone else?  Every company has strengths and weaknesses. If an integrator can’t articulate their value proposition to you – they don’t understand themselves.  Average companies claim to be good at everything.  Great companies are focussed and know where they fit and where they don’t.   You don’t want average.

…more to come.

The natural thinking, when you automate something, is to take a robot and try to have it mimick the tasks a person does when they do the process.  The challenge is, a person has a pretty complex controls system. A robot doesn’t. 

Think about the tools and processes a person uses unconsciously when they perform a task:

• As a person moves to pickup a part, they adjust the speed and position of their hands using their eyesight based on where the part is and where their hands are  (think of catching a baseball).  This is light-years beyond what a robot can do.
• A person has thousands of nerve-endings (sensors) and two hands that serve as phenomonal grippers.
• A person uses force-feedback to ‘feel’ what they are doing.  A person can sense where an object’s centre of mass is and if they’ve grabbed it correctly.
• Finally, a person has intuition.  A person unconsciously processes all this data and gets a gut feel that something is wrong when something doesn’t feel, look, or sound right.

All these abilities, that a person inherently has, make it easy for a person and very challenging for automation to handle complex processes (think of parts that are floppy (i.e. plastic bags, or bags of chips), or parts that are complex shapes and can get tangled together/interlocked, or food products that are soft or jelly-like, etc.)

With robotics, the secret is:  Don’t try to do it all at once like a person does.  Break it down.

Take bin-picking as an example.  A lot of people try to solve it straight-up, the same way a person does.  Sometimes you can, with basic parts.  But sometimes you can’t because of the complexity of the parts. 

A person looks at the bin and, in a split second, decides which part makes most sense to pickup.  An automation system can solve the same problem by breaking the process down into a series of manageable steps.

Step #1:  Get part out of bin – Keep it simple, use brute-force when you can!

• Dump the bin of parts?
• If they’re metal, use a magnet to grab a bunch at a time?
• Use a bowlfeeder to feed them?

Step #2:  Get part singulated – How can I get them separated so I can grab just one?

• Drop them onto a table where they’ll sit flat?
• Set them on a vibratory table where you can get some separation?
• Use a series of belts of increasing speed to create gaps between parts?

Step #3:  Get part located – Accurately locate the one I want

• Use hard tooling to get individual parts into a known location?
• Use simple vision to locate the part in 2D space and pick it up?

You’ve now solved the application with simple technology.  Don’t get me wrong, there’s a place for 3D vision-guided robots and force-feedback systems and sometimes they make the most sense.  BUT, sometimes a series of brute force, simple steps is a beautiful thing!

As engineers, we are technically driven.  We put a strong emphasis on the technical aspects of a solution and often neglect the softer, political and people issues that come with change.

Humans don’t like change.  People are afraid of new technology.   These paradigms are all built on people’s beliefs about their own skills and abilities (or lack thereof), the complexity of technology, and the perceptions of the company’s underlying motives.  It doesn’t matter how well an automation  system is designed and built – it will fail if you don’t address these beliefs and win over the hearts and minds of the people.

Here’s how.

Step #1:  Paint the Vision

Why are you putting robotics?  People need to know.  If they don’t know, they’ll make-up a reason and it’s usually not favorable.

Be up front and explain the reasoning straight-up.  They watch CNN and know the realities of today’s global economy and the need for North America to be innovative to compete.   What is the vision for your company?  Where do you need to be in 5, 10 years?  If you paint a clear vision of where you are going, people will be much more tolerant of change if they believe it is as a logical step to get there.

Step #2:  Identify 3 Champions (a person per shift)

You know who these people are.  They are the guys (and girls) on the floor that are the early adopters of change and ring-leaders amongst the troops.  They are the most respected people on the floor and the most capable.  If they’re involved and engaged early – their ownership is much higher.  Most importantly, it means you respect them.

Step #3:  Collaboration – Engage the Champions

Engage your champions into the design team.  Make it a priority to have them part of the planning from the initial concept, all the way through design, build and runoff of the equipment.  The more they are involved, the more it becomes the group’s solution and everyone has a stake in its success.  This isn’t just lip-service either.   These guys (and girls) know the process better than anyone.  Some of the most discerning process observations and innovative ideas I’ve seen, have come from the people on the floor.

Step #4:  Education

The people on-the-floor that are new to robotics often have the paradigm that robotics/automation will be beyond their skill level , it will highlight their weaknesses and skill gaps and they’ll be embarrassed or found out.

Education is about breaking down this paradigm.  When done right, automation is not complicated.  It’s intuitive, robust and simple to operate.  In this day and age of X-Boxes, Ipod’s and Facebook, people are generally capable of operating a robotic cell.  The sooner they realize it’s simpler than they thought, the sooner the fear and resistance goes away.

Good education or training needs to come in chunks.  Don’t dump it on people all at once – they’ll never retain it.  Build a training and education program that gives people bite-sized chunks and touchpoints throughout the course of the project.  Each touchpoint reinforces the last and helps make the training stick.

Step #5:  Follow-Through, Follow-Up

Just like forming a good habit takes time, you need to keep the champions and the people floor engaged after the equipment is installed.  They appreciate the follow-up and they’ll also help you diagnose any problems or intermittent issues that come up.  You’ve now also got your best people on the lookout for the next process improvement or project that can further improve productivity.  It’s a win-win!

Industrial automation systems are a lot like coin sorting machines.  You dump your coins into the sorting machine and it uses simple rules (tooling) to separate and organize your coins into quarters, dimes, nickels and pennies.  Automation is the same.  Bowlfeeders, part crowding, sieves, robotic vision-guidance, etc. are all really just fancy ways of doing the same thing as a coin sorting machine.  They take random widgets and organize them using tools (hardware or software).  The key is they do it without thinking.

That’s really the question you need to ask yourself when looking at completely automating a process – “Can this process be broken down into a series of steps such that a set of rules can complete it without thinking?”

As much as we think automation and robots are “smart”,  they’re really not.  A robot can’t improvise or handle a new situation it was never designed or programmed to handle.  As technology marches forward, it continues to raise the bar higher and higher by providing more powerful tools, but the overall strategy remains the same.

This strategy means some processes can’t be automated, since they require a person to figure out the anomalies that inevitably come up.  But what if you could semi-automate them?  What if you could remove the laborious, time consuming, non-value added tasks and leave the people to do what they do best, the thinking part?

I’d like to see a new generation of safe, interactive robots that can work with people to perform these tasks. 

A good example where this could fit is putting a gas tank into a car on an automotive assembly line.  The majority of the process, moving the tank from the rack beside the line to the car, is time-consuming, simple and adds no-value to the car itself.  It’s the very final part, where the tank is actually mated with the car, that the person’s brain starts to get used.  That part, is always slightly different and requires thinking. The operator moves fuel lines and wire harnesses out of the way, jiggles the tank into place in order to get the bolts through the holes and threads started, and connects connectors and plugs by wiggling them into place.  It’s this part that makes the process too difficult to completely automated.

If you could introduce a safe, interactive robot into this scenario, the non-value added labor could be eliminated from the process.   A robot could automatically get the tank from the rack, bring it to the car, raise it close to the insertion point, and then track along with the moving car waiting for the operator to help it with the next steps.  An operator, who was working on another value-added task on the car, could then grab a hold of the gripper or the robot arm itself and move the robot and gas tank into place.  Just like two people who work together to lift a heavy object, the robot and person would work together to insert the gas tank.  The difference would be, in this case the robot does most of the lifting and the operator provides enough force to direct and guide it.  Safety scanners or safety mats could be used to determine when the robot and operator are in the same area, and would put the robot into “interactive-safe-mode”.  In this mode, the robot’s torque would be limited such that at the robot could not exert enough force to hurt someone.  The servo motors would provide just enough torque to hold the robot itself up, along with its load.  Any additional force input (from the operator) would be used to guide the robot.

There are other possibilities as well.  In traditional robotic cells, instead of teaching the robot with a pendant, the robot could moved into desired positions by manipulating the arm itself.  This would be easier and more intuitive for people with little robotic experience to work with.  Paths could be taught this way, allowing the operator to move the robot through its the sequence of operations and teach the path points along the way.  Even fault recovery could be easier.  Instead of having to jog a faulted robot out of a tight spot using the pendant (i.e. inside a CNC), the the operator could physically guide the robot arm out.

Currently, in the medical field robots are used to interact with and touch people.  They augment human performance making the doctor’s hand steady or more precise.  They do this in a safe manner such that the patient, doctor and nurses are not endangered.  If this technology exists today in the medical field, then why not apply it in an industrial setting?

I’m often surprised how much time up front is spent talking about the nuts and bolts and specifications of the equipment being purchased, and how little time is spent really defining what success looks like.  The equipment (automation, robots, PLC’s, conveyors, whatever) are all just means to achieve a business outcome.  At the end of the day, what you really care about are reduced production costs, higher product quality, or greater production capacity.

That is what should be measured, that is what an integrator should be held accountable to.  Define the contract such that the integrator needs to deliver this – regardless of the bits and bytes of what they put into it.  If they missed or under estimated something – it’s their responsibility to do what needs to be done to achieve the business outcome that was agreed upon – period.

OEE (Overall Equipment Effectiveness) is a quantitative method used measure performance and takes into account the three major things in an automation system #1 How fast does it run, #2 The system uptime, and #3 The product quality coming out of it.

Here’s how you calculate it.

OEE = Performance Efficiency * Availability * Yield

Performance Efficiency = How fast does it run

Example – the system is designed to run 100 pieces per minute, the final system as it is installed runs 98 pieces per minute.

PE = 98 ppm/100 ppm * 100% = 98%

Availability = How much time the system runs of the total available time (uptime)

Example – In a week’s worth of production, over 2 shifts (4800 available minutes), the system has 5 minutes of downtime and 20 minutes of changeover.

Availability = (4800 minutes – 5 minutes – 40 minutes) /4800 minutes * 100 % = 99.06%

Yield = Percentage of good widgets made (quality)

Example – 2 defective pieces out of 1000 pieces made

Yield = (1000 pieces – 2 pieces)/1000 pieces * 100% = 99.8%

OEE = PE * Availability * Yield

OEE = 98% * 99.06% * 99.8% = 96.9%

Figure this out up front.  Work with the integrator to agree on what this number needs to be.  Agree on the inputs that are required to achieve this (i.e. the system needs to have good product going into it, if it is going to have good product coming out).   Spending the extra time up front to define and agree on this metric will ensure you get what you really want at the end of the project.

In some upcoming blog posts, I’ll walk through some examples of how to put this together.