During the past month, I interviewed several engineers for an article on automatic screwdriving. Along the way, I learned a bit of wisdom that is applicable no matter what process you might want to automate: You can learn a lot about your parts when automating a process that was previously done manually.

Automation is not a cure-all. It can improve quality, boost output and lower labor costs, but it can’t fix parts or designs that are flawed from the start. The most flexible assembly machines ever invented—also known as people—can easily cope with design shortcomings or process inconsistencies. Machines cannot. Accuracy, consistency and control are the watchwords of automation.

To illustrate the point, Jarrod Neff, marketing manager at screwdriver supplier Visumatic Industrial Products, shared a couple of examples.

The first involved a medical device manufacturer that wanted to automate the process of installing screws in a plastic part. The application called for sharply pointed screws to be driven into recesses several inches deep. The operation was relatively easy for an operator using a driver with a magnetic bit. He could feel whether the point of the fastener was seated in the hole before triggering the driver.

A robot has no such sense. It can’t tell if a screw is in the hole or lodged in the side of the recess. “If the screw wasn’t perfectly aligned, it would bore into the side of the recess instead of the hole,” says Neff. “Switching to a fastener with a blunt point took care of the problem.”

Another example involved a manufacturer of large-screen televisions. The manufacturer wanted a Cartesian robot to install screws around the perimeter of the television’s bezel (the plastic frame that surrounds the screen). The assembly was delivered to the robot on a conveyor. The application worked great, but would experience problems at seemingly random periods of time. The robot wasn’t driving the screws successfully.

After much head scratching, a clever engineer realized the holes weren’t where they were supposed to be. On very hot days, the temperature inside the assembly plant would get quite warm. So warm, in fact, that freshly molded bezels did not have enough time to fully cool before they reached the robot. As a result, the parts were ever so slightly larger than they were supposed to be.

Temperature-related expansion and contraction might not matter on a small part. But on a 65-inch television frame, it could mean a difference of 0.2 millimeter or more—enough to throw off a screwdriving robot. In the end, a low-tech solution solved the problem. The manufacturer installed a fan on the assembly line and gave the bezels an extra 10 to 15 seconds to cool off and contract to their specified dimensions.

 Lesson learned. So before you dial your favorite automation vendor, ask yourself this: How well do you know your parts?