This is going to be a continuing series of posts that will try to analyze the current state of automation and attempt to predict where we are going to wind up by the end of this decade.
First, I find it absolutely instructive and maybe even enlightening to take a look at what we actually do when we automate a process or a manufacturing cell.
A hundred years ago, we had excellent control of both continuous process, batch process and discrete manufacturing processes. Most people do not understand this. We actually had better control of some processes than we have had for decades afterward.
How could this be?
Simply put, each control loop had a human in the loop. The human was trained and experienced, and could see the process by observing its behavior, and made changes based on tuned intuition.
Unfortunately, corporate managements began to see this type of control as too expensive. They also claimed that it was not as accurate or repeatable as some sort of automatic control, and that quality suffered. There is no real proof that this was true in most cases.
But we started to attempt to replace the human in the loop with controllers, first pneumatic and then electrical, and then electronic. The controllers were designed to react to the behavior of one variable and make changes to a final control element, exactly as the human had done.
Of course, it never worked right. It was okay, and we played games with tuning algorithms like PID to more closely approximate the behavior of a human in the loop.
When we moved to distributed control, we had the microprocessor power to be able to do complex mathematical models that again attempted to mimic the human in the loop. This Advanced Process Control gave good results as long as the model was accurate and kept up, but again, it did not do as well as a trained operator could do, when the operator was fresh, not distracted, and fully experienced.
So what we have, in continuous process, in batch process, and in discrete manufacturing, is an overarching model that what we are doing is replacing the trained human in the loop.
Does anybody see anything interesting about the fact that this is what we’ve been doing for nearly a hundred years?
We have about run out of string. For the past ten or twenty years, we have been getting less and less improvement, regardless of the expense we push into our control systems. It is time, and far past time, for somebody to come up with a new model– a new, overarching philosophical method for doing more than replacing the human in the loop.
Your comments are invited.
First, I find it absolutely instructive and maybe even enlightening to take a look at what we actually do when we automate a process or a manufacturing cell.
A hundred years ago, we had excellent control of both continuous process, batch process and discrete manufacturing processes. Most people do not understand this. We actually had better control of some processes than we have had for decades afterward.
How could this be?
Simply put, each control loop had a human in the loop. The human was trained and experienced, and could see the process by observing its behavior, and made changes based on tuned intuition.
Unfortunately, corporate managements began to see this type of control as too expensive. They also claimed that it was not as accurate or repeatable as some sort of automatic control, and that quality suffered. There is no real proof that this was true in most cases.
But we started to attempt to replace the human in the loop with controllers, first pneumatic and then electrical, and then electronic. The controllers were designed to react to the behavior of one variable and make changes to a final control element, exactly as the human had done.
Of course, it never worked right. It was okay, and we played games with tuning algorithms like PID to more closely approximate the behavior of a human in the loop.
When we moved to distributed control, we had the microprocessor power to be able to do complex mathematical models that again attempted to mimic the human in the loop. This Advanced Process Control gave good results as long as the model was accurate and kept up, but again, it did not do as well as a trained operator could do, when the operator was fresh, not distracted, and fully experienced.
So what we have, in continuous process, in batch process, and in discrete manufacturing, is an overarching model that what we are doing is replacing the trained human in the loop.
Does anybody see anything interesting about the fact that this is what we’ve been doing for nearly a hundred years?
We have about run out of string. For the past ten or twenty years, we have been getting less and less improvement, regardless of the expense we push into our control systems. It is time, and far past time, for somebody to come up with a new model– a new, overarching philosophical method for doing more than replacing the human in the loop.
Your comments are invited.