Can There Be Too Much Automation?
Within the VDC Industrial Automation and Control Practice we are always applauding the ever expanding role that automated processes are finding in new regional and vertical markets. In most cases, responsible/safe automation products and solutions are clearly superior to the manual industrial processes of the past. Outside of industrial plants there is another area where increased automation has paid dividends and that is in modern passenger aircraft. In either case, well tested industrial and aeronautic automation solutions transfers the control of mission critical processes to avoid operator/pilot type errors.
An article I read last week focused on some aircraft situations where the automated systems failed and, it appears the pilots had lost some of their edge with respect to the correct response they had to make. Planes that are in near stall condition need to point the nose down in order to pick up air speed. Pulling the nose up which might be the intuitive thing actually makes the stall condition more likely. Could there be a parallel from these aircraft situations and those that can take place in industry? In my opinion, similar unexpected problems with automated processes could happen at almost any industrial facility. Most systems are designed with redundant and ultra-reliable components calculated to have the appropriate Safety Integrity Level (SIL) for the safety/process risk that is present. On the other hand, recent events have shown that problems have already occurred in situations where combinations of failures and events overwhelm the automated systems. I was thinking about two obvious ones where this could have happened.
- At Fukushima Daiichi, from what I understand, the earthquake severed the facility's connection to the electrical grid and, then, the resultant tsunami almost simultaneously, destroyed the backup power sources. In such a case would it have been better to keep generating power with at least one of the reactors to keep the facility pumps cooling pumps running? I would estimate that the safest conventional and likely automated process thinking would be throwing the control rods in the reactors and isolating the steam pipes but perhaps this was the wrong move because of the other combined factors.
- In the case of the Deepwater Horizon, which was a well automated modern marvel of engineering we again see a chain of miss-interpreted tests, overridden alarms, and dependence on automated systems that could possibly delayed the reactions needed to save the rig by cutting the it loose from the blown out well. There were other factors in place here that I have discussed in earlier blogs but there is case to be made in that the automated safety systems fell short and it is likely that some manual reactions were not made or at least not in the correct timing and sequence.
Before I close, I do want to take some time to laud the efforts of the personnel directly involved with both cited incidents. They were in desperate situations and had to make difficult decisions and actions and, in many cases, lost their lives doing so. This posting was never intended as a slight to these fine individuals.
In closing I think that in many cases, industrial facility owners are increasingly looking for outside expertise to provide complete automated solutions with respect to process control and safety integrated systems. OEMS and equipment specifiers claim that 55% of their revenues are service related while at the same time end users are claiming ~38% of their expenses were for non-hardware related categories. To me that translates into a lower level of expertise at the point of use. It is unlikely that the automated systems that are put in place will fail but, if they do, hopefully the on-site personnel will have the expertise, decision making, and necessary reactions to prevent disasters.