When to Automate?

Automation may seem like a natural progression for manufacturers looking to improve their production, but the choice of when and if to automate depends on many factors. Automation can be a great way to improve quality, speed, and safety, however, it is not a magic wand to solve all problems.

Where Automation Excels

Automating a process is a good choice when the process is well defined and repetitive. This means that a product being assembled has reached design maturity and that toleranced drawings are available. The input material and parts are consistent which allow the automation system to behave the same way every time to produce the same output product. In other words, the problem well bounded.

Calvary Robotics has produced many custom automation systems for our customers to achieve unique and complicated results, and often these systems are constructed of stations which perform one very simple task at a time. A pressing station may be dialed-in to press to a single depth or to press with a single force. A feeding system will take in one type of component and orient them all in the same direction and present them in a consistent way to a pick-and-place station. A welding station will produce the same weld bead between the same mating components. A vision inspection station can confirm the presence or absence of known features in an expected area.

Does That Mean There’s No Room for Input Error?

No, automation can still handle some variation. Manual or automatic change-over tooling can enable a system to work with different, similar product lines with slightly different components. Modern vision-guided pick-and-place systems can identify the orientation and location of known components on a platform and pick them and present them in a single, consistent way to the next stage of the process. Many types of in-process checks can be designed such that out-of-tolerance input material or a failed operation can be identified. Many times, an automated reject process can quickly eject bad input or failed output without stopping the system. Notification is then made to the operators who can investigate the cause at an appropriate time.

Automation also is a good choice when the process requires working with material that could be dangerous to humans or in cases where the risk of contaminating the product would be very high. Human workers applying adhesives with hazardous fumes, manipulating small, sharp objects or large, heavy objects can be avoided. Medical devices can be assembled and packaged in a clean and sterile environment.

Where Automation Struggles

Automation is not a solution for general problem solving. The question is often asked “Why can’t we just have a robot arm do that?” Usually the problem is not that a robot arm couldn’t move the weight or that it couldn’t achieve the motion, but it is that the task cannot be properly defined such that the robot can be told what it is it has to do.

Take, for instance, the task of changing wheels and tires on a car which is on a lift. For most cars this is a reasonably defined problem. Wheels are round. Inside of the round area there typically 4 to 6 lug nuts that must be removed. The tires are round and rubber and can be gripped with plenty of force to remove them without damaging them. A robotic arm system is designed with some grippers, a vision system, and a nut-driver with a tool-changer area to adjust the bit as needed. A car rolls onto the lift, the driver waits in the waiting area while the car rises and the robot goes to work removing the wheels and replacing them with ones from a rack behind it. The job is over in a few seconds.

The process works so well, management now wants to put this robot to use for everything. A quick study shows that the arm should be able to move under the car and reach all points of the exhaust system so there is a push to now have the robot perform muffler changes. The assumption is that since we have a vision system we should be able to see what type of muffler is present, swap out some tooling for cutting and welding, and perform the job, reducing the need for skilled labor.

The problem is that this task is both very poorly defined and significantly more complex. Unlike the round wheels, the exhaust system changes drastically from car to car. Even the seemingly simple task of identifying where the muffler is becomes much more difficult because it is not a basic shape. There is a significant chance of contamination under the vehicle from dirt, mud, snow, and even intentional things like under spray. The muffler may not even be completely whole if a customer is looking for a new one. With all the challenges, the robot programmer must still find a way to describe a motion path. And this motion will need to be even more accurate than in our wheel problem, because with our wheels it is likely that we can align the tool to the wheel center then over-travel until hit the wheel with compliant tooling. Under the car we will have an unknown and complex surface with nothing for our robot to “feel”.

Because of these challenges there is a significant risk of damage to our customer’s vehicle. The design safeguards required could not be practically designed with current technology. This task now falls out of the area of automation and into the area of true general AI.

Is Automation Stealing my Job?

Chronic health problems resulting from repetitive tasks are a real concern. Humans were not meant to be put in boxes and perform only a single task for hours on end. We are the ultimate general problem solvers and should use our skills to carry out complex tasks that machines cannot do. In our hands alone we have a complex set of tactile sensors, temperature sensors, and dozens of actuators all controlled by our brain. With a single sentence you could explain to most people how to fold a basket of laundry, carry a group of objects up stairs, or pull weeds out of a garden, yet these are still incredibly difficult challenges for automation.

Automation can and should be used for tasks which are dangerous and unrewarding. When implemented properly, automation can exist alongside a human workforce. Automation systems should perform a scripted series of tasks. Humans should solve problems, create, be adaptable, compassionate, human.