“Prediction is hard, particularly about the future,” so it was said by either Voltaire or Yogi Berra. Despite the warning from these august individuals — and my own failed predictions of a paperless office, Beta as the format of the future and the certainty of Netscape — I’m going to try again to assess the impact of technological changes on process plant operations and staffing.

Early human factors (1940s) focused on the allocation of tasks between people and machines. What did people do well, e.g., handle missing or ambiguous data, and what do machines do well, e.g., repetitive actions and exerting large forces. Science fiction of the late 20th century predicted that eventually robots could replace humans entirely. What was once fantasy is predicted to become fact soon. What impact does this have on process plant operation in general, and the operators in particular?

Many tasks performed by operators are rapidly becoming obsolete. Any sensory-related task done by field operators can be automated (feeling for vibration, seeing valve position, smelling for leaks, hearing pump cavitation). The notion that the field operator is the “eyes and ears” for the console operator is losing its meaning. Physical tasks that were the domain of field operators are now handled by machines (motors and remote control can open and close valves or start/stop pumps). Control room operators once had valve handles sticking through the control room wall, which they used to adjust the process. PID controllers long ago eliminated that, so control room operators are more monitors of the control system, ensuring the process hits its target. Multivariable control, a relatively simple type of AI, now ensures targets are met, with the operator relegated to handling events if things go awry.

Other industries have already faced these problems to some extent. Automotive repair has had to deal with sophisticated diagnostics, changing the role of mechanics. Assembly line production has had to deal with the use of robots, removing many of the physical tasks previously done by people. The selection and training for these positions have changed as the jobs have changed.

Process plant operator tasks can be grouped into three major categories: 1. operational (making adjustments to the process), 2. reliability (including the entire maintenance cycle from diagnosis through repair, and 3. quality (lab analysis ensuring the products meet the necessary specifications). These tasks are typically done by a single operator working alone. This will change.

Looking Ahead

The operator of the future will engage in more activities and fewer tasks, if at all. An activity is not a solitary exercise; it requires collaboration and cooperation between multiple groups (operations, technical, maintenance, lab) and entities (AI). These activities will fall into the same three major categories that defined the tasks. Operational activities will bring together board operators, field operators, engineering (process and control), and instrumentation to ensure the process operates at maximum efficiency. Reliability activities will bring together operators, maintenance, rotating equipment and reliability engineers to keep plants running with minimal equipment breakdown. Quality activities will bring together operators, chemists and environmental engineers to ensure that monitoring of products and effluents is accurate and appropriate.

The operator of tomorrow will need to understand the nature of the work and the purpose for which it is done. Collaboration is essential and requires that people understand what others are doing and how their work fits with the work of others. It requires a higher level of interdependence for success in the work than simply performing tasks does. Collaboration is a win-win situation because all those involved have a common goal.

Most plants currently staff the field by assigning operators areas of responsibility based upon geography. Perhaps that will no longer be the best way. Instead, operators could be assigned to activities for the entire unit. One operator might inspect the rotating equipment throughout the unit. A different operator could be looking after the instrumentation. A third might specialize in the analyzers used to establish product specifications.

The output of these activities will essentially be judgments and decisions. No amount of data from AI or any other source can answer the question. “Is it a good idea to ….?”.[1] For example, the unit is approaching a limit of some type, quality, mechanical, whatever. AI cannot decide whether it is a good idea to continue, cut rates or shut down. It can help the operator answer the question by providing the past distribution of outcomes, but it does not provide the answer. You cannot derive an “ought” from an “is”. This barrier between what is and what ought to be was established by philosopher David Hume in the 18th century. Whether something is a “good” idea is a value judgment that nothing in the past can answer, only the operator in the now. That is the human factor.

AI, robots and automation will impact process plants. Operators will be doing activities rather than tasks. They must be trained to understand the goals of the activities and selected to work in this collaborative environment. They cannot blindly follow instructions. They will need to be able to say, “I don’t think this is a good idea.” But they will be needed.

[1] Megill, Kenneth A, Thinking for a Living: the Coming Age of Knowledge Work. De Gruyter Saur, 2004