Robot Safety Training

Where you are located on this big, blue marble is going to be relevant to the answers that you get. In the US, there are RIA standards, which they are going to want to sell you and ANSI, as well. There could be some state requirements but not likely. Europe has their own that they follow, I think.

https://www.automate.org/a3-content/robot-safety-and-risk-assessment-training

A3 (formerly RIA) is the industry body that writes most of the robotic safety standards, so they're probably the place to start. Lots of 3rd parties offer training as well. And each robot brand has their own implementation of safety, which you would have to learn separately.

There's also more than one way to do safety. Many robots have enough brute force to punch right through a fence and hit someone, so we used to use adjustable hard stops mounted on the joints to limit the robot's range of motion. More often, these days, you might buy the safety option for the robot that uses digital realtime measurement of the robot's position and prevents the robot from entering or leaving certain areas.

At minimum, a robot must have an E-Stop and a Safety Gate signal. E-Stop acts as on a CNC, and Safety Gate operates rather like opening the door on a CNC -- it disables Auto/high-speed mode, but allows the programmer to still use Teach mode (which relies on a deadman switch built into the teach/jog pendant). These signals can be hard-wired, or can be network signals from a safety-rated PLC -- which you need to use will influence which options you buy on the robot. And it's possible to have many more, with more complex operations, but everything has to be safety-rated.

It should go without saying that the Safety Gate signal need to be tied to an actual physical gate, which needs to be part of an actual fence, that prevents people from getting close enough to the robot to get hurt.

Then there's Cobots (Collaborative Robots), which are (supposedly) inherently safe -- basically, they're so slow and weak they are incapable of injuring someone, and as such have no need of fencing or gates. Right up until someone mounts a sharp tool to one of them....

This subject can get extremely complicated and brand-specific very quickly, and every end user has their own spin on implementing the general safety standards. But the basic rule of thumb I start with is: If someone can get hurt without making a deliberate effort to circumvent safety, it's probably not safe enough.

Obviously, there's wiggle room there. The RIA standard rates hazards on a multi-tier matrix that includes: degree of harm, ease of avoidance, and probability of happening. It's impossible to make robots completely non-hazardous, but we have to try to block anything that makes it easy to get hurt.