3 Steps to Improving

Robot Accuracy
Robot process isn't performing as expected?
We all know that robots are a great technology for carrying out repetitive tasks over and over.
However, robots are increasingly being used in challenging applications that tend to expose
their weaknesses rather than leveraging their strengths. The use of robots is often based on
several assumptions, which are worth considering in a little more detail.

High Repeatability ≠ High Accuracy

As a robot can repeatably return to the same position time after time, it may be assumed that
it is globally accurate – but this is rarely true. Repeatability for an industrial serial axis robot
may be around 50µm, but if the same robot were programmed to move nominally one metre,
it would likely have inaccuracies at the millimetre level.

High Payload ≠ High Stiffness.

Another assumption is the interplay between robot payload and stiffness. The payload rating
is the robot’s ability to lift and manipulate mass, whereas the stiffness is concerned with the
robot’s rigidity under load, usually manifesting as end-effector stability during an operation.
We have encountered numerous examples where larger payload robots have been down-
selected with the assumption that the robot will be more rigid when exposed to external
forces.

However, the stiffness is more than just a function of the lifting power of the motors. Stiffness
of the robot end-effector can be heavily influenced by load orientation, backlash in gear
boxes, and robot linkage stiffness. These elements are not inherently improved just by using a
larger robot. Poor stiffness typically manifests as positional inaccuracy and process variation.

High Flexibility ≠ In-process Adaptability.

Lastly, robots are highly flexible both in terms of how you deploy them and how they can
manipulate an end-effector in space. It is understandable that for a continually changing set
of tasks or parts that robot flexibility is seen as an advantage – however, given that robots
have poor global accuracy, their ability to adapt in-process to absorb variation is also poor.

Introducing a robot process without

recognising these limitations can result
in a sub-optimal application and a loss of
productivity, ultimately increasing costs.
How can a robot's performance be improved?
Understanding and accepting the limitations of industrial robots should not preclude them
from adoption in difficult applications, but it should change our approach as to how best to
deploy them.

MEASURE
Measuring your current process is a logical first step and yet is rarely
carried out in practice as it is difficult to do accurately. We need to
establish what is actually happening during the manufacturing
operation. This is subtly different from a more common approach, which
is to measure the process outcome, i.e. the manufactured part or
assembly, from which conclusions are inferred.
Measurement of the produced outcome can only ever give part of the story. If the
part/assembly/process is good – that’s great! – but if you need to dive deeper then measuring
of the outcome will have many conflated signals as to where the error may be arising. For
example, your problem may be fixture/part location relative to the robot, but this could be
mistaken for robot positioning inaccuracy.

Another approach is to measure the robot path as a “dry run”, without any external loads. This
is not always easy as measurement systems capable of such tasks often require specialist
resource, expensive software, and even modification of the process to ensure targeting and
line of sight are maintained. Although a useful diagnostic step, this kind of measurement can
provide a false sense of security as it is often the case that payloads and external forces have a
significant impact on performance. Want to measure in-process? Read on!

https://insphereltd.com
 TEACH
The ability to teach a robot is a very useful function, updating a position
from the pendant and relying on good repeatability to ensure the robot
returns to the ‘taught’ position. This can be used to generate a program
from scratch or to ‘touch-up’ an existing program to achieve the original
intent. This typically occurs when an offline program (OLP) is being
brought online, but there are of course other times you would touch-up
a program, perhaps to negate some drift, or accommodate a new
fixture.

Traditionally, touching-up a robot program is carried out by a skilled operator relying on

visual markers. Touching-up a program can be a good way to wash-out some of the tolerance
stacks in your manufacturing process, for example inaccuracies in the tool centre point (TCP)
offsets or positioning of the fixture relative to the robot. However, it is a lengthy process and
only valid for the one-operation. This can be extremely arduous in complex robotic
applications and can become impractical or even impossible for dynamic paths. Could
leveraging robot repeatability unlock you process?

Another approach is to measure the robot path as a “dry run”, without any external loads. This
is not always easy as measurement systems capable of such tasks often require specialist
resource, expensive software, and even modification of the process to ensure targeting and
line of sight are maintained. Although a useful diagnostic step, this kind of measurement can
provide a false sense of security as it is often the case that payloads and external forces have a
significant impact on performance. Want to measure in-process? Read on!

https://insphereltd.com
MONITOR

If the purpose of the process measurement and robot teach was to

ensure a baseline performance could be achieved, then the next question
to ask should be: is the process performance maintained over time?

Although robots have good repeatability from one run to the next, they are still mechanical
systems and as such, they have limitations when faced with:
Variation in the fixture/tooling/part location.
Temperature variation.
Variation in the static and dynamic loads.
Wear and tear of gearboxes/moving parts.

The same measurement principles should be applied as before – can I measure my process as
a whole, whilst it is being carried out?
Monitoring the process provides confidence that a robot process is consistent, and your
process is controlled. This can – as confidence grows – provide a proxy for process control,
reducing the part inspection requirement. Additionally, monitoring the process can work as
an early warning system for maintenance teams, catching and scheduling down-time before
scrap is produced.

What are the challenges with taking this

approach?
Carrying out these three steps is straightforward in principle, however, in practice, there are a
number of hurdles to overcome. Until now, measurement and sensors systems have not been
designed to allow operators to capture this data easily.
Portable metrology equipment typically requires a specialist to operate the equipment and
software.
Even with specialist resource on hand, inherent limitations with the metrology system
remain, typically:
x Only able to measure a single point of interest.
x Only able to measure in 3D, without angular information.
x Must maintain a limited line of sight.
x Requires a stable mount.
x Not built to remain in-situ within the robot environment.
x Requires third party software.

https://insphereltd.com
What's unique about IONA System?
IONA is the culmination of INSPHERE’s experience in tackling these challenging
applications. The hardware and software have been developed by us from the ground-
up to monitor and improve robot processes. Enabling simple in-process measurement
of the end-effectors and fixtures, we have created a system for production engineers,
not metrologists.

As consequence, IONA:
Is designed to remain in-situ, in industrial environments and continually monitor
24/7/365.
Can simultaneously measure multiple points of interest (in 6D).
Overcomes line-of-sight issues by virtue of numerous vantage points and built-in
redundancy.
Does not require stable mounting.
Has a non-specialist, easy-to-use software for collecting data and improving robot
accuracy.
Our measure module can provide true position of robot end-effectors relative to the
cell datum, fixture datum or any other reference (all of which can move!). Data is
readily available in real time to import back into your OLP environment or integrate
directly into a digital twin. The engine works best when combined with teach, our
module for improving robot performance. Analogous to the manual method of
touching up programs (but much quicker and simpler) teach takes the measured data
and updates the robot program to reflect the design intent. It negates operator-to-
operator variation, reduces the potential for mistakes, and provides an external
reference to buy-off. In most applications, this enables a cell to be commissioned
without the requirement for parts. Once the program is optimised, the software
switches over to monitor, giving continuous assurance that in-process accuracy is
maintained.

https://insphereltd.com
The ability to monitor a process 24/7 is often not given the value it deserves, simply
because it has not previously been viable with available technologies. The costs of bad
parts and process interruptions are known to be very high. IONA can shift the mind-set
of manufacturing from one that tolerates scrap, waste and stoppages, to one that is
constantly vigilant, is able to learn from historic data, and therefore delivers far greater
productivity than has previously been possible. This shift is a key part of the move
towards Smart Manufacturing Industry 4.0 philosophies, and makes possible a major
uplift in manufacturing quality and productivity for a wide range of complex
automation processes.

IONA directly improves the accuracy of robotic applications, giving you:

More process confidence
Right-first-time programs
Significant time and cost savings
Improved part quality
Early detection of process drift (reducing scrap!)
Increased uptime and better planning of preventive maintenance
Higher productivity

Find out more about the IONA System:

https://insphereltd.com/iona