2005-automation-skilled-trades-hero

Automation and skilled trades in the future

May 13, 2020
The demand for a skilled workforce across all trades is on the rise, and the role of the human decision maker has never been more crucial.

According to a number of academics, the skilled trades industry is being replaced by automation and equipment is self-healing. Marketing for IoT companies claim that you can eliminate your skills gap with their AI or self-learning systems. Media and government have been stating that we are purely a service economy, as computers and automation have replaced labor and manufacturing has primarily moved abroad.

For those of us in industry and those of us who are experiencing automation and data analytic reliability, the hype is not matching reality. In fact, once we separate from vendor-authored articles and delve into real data, such as that produced by the U.S. Government Accountability Office (GAO) in their March 2019 “Workforce Automation” report, the facts appear quite different.

What they discovered is that whether a company has automated or not, a majority have neither seen a change in headcount nor have had layoffs, and hiring numbers did not change significantly. In companies that claimed an increase in automation, the data do not actually show the impact through the U.S. Department of Labor. So far, data have been used to present estimates on the impact, and include significant gaps.

The demand for a skilled workforce across all trades is on the rise due both to a large retiring, or retired, baby boomer labor force and commercial/industrial growth. Several of the needs that have been in demand for this workforce since the 1990s have been critical thinking and an understanding of evolving technology. Even more so, now, is the growing need for these trades to have an understanding of such foreign concepts as smart devices and cybersecurity, especially in cloud-based and internet systems.

The prevailing thought of no manufacturing and a “service economy” still resides in politics, media, and academia. This mindset has been prevalent enough that, during the coronavirus pandemic crisis, the suggestion that people should just work from home was a bit of an eye-opener. During a discussion with a decision-maker, they only saw things in terms of service professionals and hourly/minimum wage.

Can the age of phones and computers responding with human voices really have had that great an impact on our perceptions? Do true artificial intelligence or learning systems exist that do not require human interaction? Won’t the installation of these devices replace the problems we have with a skills gap issue with solutions?

One of the things we know about automation and computer control is that any existing problems will be found, and found quickly, sometimes with disastrous results. The problem resides in the difference between marketing promises, programming, and implementation. For instance, product marketing copy will claim that the software can solve everything including that the system will do all the work for you. In reality, existing software is linear in nature and can only provide results based on how it is programmed. The systems really do not think for themselves and, if they did, the thought processes would be so foreign to us that we would not understand the internal processes or actions taken.

The perception of intelligence comes from the ability of modern computational systems to rapidly process an increasing amount of data and compare to tables in much smaller devices than in the past. These actions can be presented in almost human-like responses or repetitive responses. Where things go wrong is when an event or action occurs that is a variance that was not planned or designed into the system, or when a sensor fails and provides incorrect information. The benefits related to the repetitive motions and basic decision-making can have huge beneficial impacts such as increased productivity and mostly repeatable results, but human interaction is needed for deviations or errors.

On August 14, 2003, the Northeast and parts of Canada suffered a significant blackout. The problem started with poor tree maintenance combined with high heat of the day. A system was in place to identify the problem, but the software had a glitch and, once the problem started, a large number of power plants went off-line within 3 minutes. The glitch caused an alert to operators not to function and as the system was overloaded the problem progressed. The impact of the minor glitch, an error in a line of code, was the loss of power and essential resources to over 50 million people.

On a smaller scale, another facility had placed automated monitoring devices on large, critical equipment and was taking in over 38 million points of data per day. Two subsequent failures occurred that required months to repair, failures that should have been detected by the instrumentation in advance. An investigation identified that the automated equipment was not set up correctly, so the failures were not detected in advance. It was also identified that although massive amounts of data were collected and analyzed, providing significant operational benefits and the detection of some defects, there were massive holes in the system for which standard testing practices actually provided the data necessary to take action on critical machines.

Much of the data provides information to investigate failures and provide indicators, but experience with IoT device failures identified that human interaction is paramount. At this location, manpower requirements have also not decreased and, instead, have increased to include specialists and data scientists. The combination has provided overall benefits in understanding the complex system but has not accomplished a reduction in trade requirements.

What is particularly worrying about many of the automation approaches, even including CMMS and enterprise systems, is that they should be viewed as an expense. This concept even comes from some of the developers who suggest that the increased expenses are a cost of doing business. For the most part, many business managers will consider that the application of these types of automation systems will automatically result in cost avoidance and/or cost benefits.

In fact, the successful initial implementation of most business intelligence and enterprise systems has been reported as low as 8%. A quick Google search of enterprise system implementation will show “lessons learned” from failures of many large corporations with losses that small companies would not be able to bear. Even at this level of automation we see challenges and, in many cases, problems with measurable ROI. If there is no return in dollars, efficiency, or reputation, then there is no need for such a system. In reality, when implemented properly, the system will and must have value in terms of effectiveness and real costs.

Automation and business information systems can have huge benefits for corporations of all sizes when they are implemented correctly. Each time my team and I have performed a root cause analysis on failed equipment or overall system that could have been avoided, we observe some very specific patterns, enough so that we have to make sure to take a step back and avoid assumptions on each RCA.

The result can be viewed as seven basic considerations:

  1. Determine a need. Programs, processes, and automation should not be implemented because they are the latest fad. Read into, or follow up on, success articles related to automation and other systems and also who has written them. You may discover a need you were unaware of, or you may discover that the article and hype is actually marketing.
  2. Develop a strategy around your business and business context. Beware of cookie-cutter implementation techniques or anyone recommending that all stakeholders should not be involved. We have run into situations where IoT systems were about to be implemented for maintenance based upon what vendors had convinced executives they needed that did not match the needs of reliability engineers nor the types of issues seen in their facilities.
  3. Don’t start overwhelming yourself with data and then figure out how it fits. Put your program together first. With maintenance systems, implement your processes whether that is reliability-centered maintenance, or some other program development task, in order to determine what information and actions are required. For other parts of the business, get your processes in place and ensure they are successful. Once you apply whatever your system is into automation, things will happen quickly and you will probably not have time to correct any errors prior to what will be a series of expensive disasters.
  4. Planning is critical. Make sure everything is in place and that communications with vendors and stakeholders are working. This would include all information, any special planning meetings, etc. Beware of vendors that keep everything they do as “trade secrets,” or are neither responsive nor willing to work with your third party or internal quality assurance. This includes vendors that are unwilling to conform with your specifications, or that find reasons to want to change them once work is in progress.
  5. Have a third party external or internal resource performing quality assurance and verification (commissioning) that the equipment is set up correctly and doing what was specified or sold. Ensure that the commissioning internal or external personnel understand the system that the solution is being applied to and expectations. We’ve been brought into multiple instances where a consultant who was brought in was unfamiliar with the system, and items that should have been identified were not. Eventually significant issues were drawn out until warranty periods expired.
  6. Cybersecurity and updates. There must be a plan in place for system updates and eventual retirement. Is the system in place meant to be isolated, or will there be any connection to wireless, cloud, or other vulnerabilities? It is important to remember that if you can see the data in some form of connection, then there is access to the system.
  7. If they tell you that their automation system, AI, machine learning, or enterprise system requires no human interaction to verify its operation: They. Are. Lying.

Based upon present GAO reports since 2018, the information being used to project workforce reduction are primarily assumptive because separate surveys show very different results. The problem, as we have seen with corporate implementation of automation, is that incorrect and incomplete information is providing erroneous conclusions. The result has been a change in direction related to skilled trades through Continuing and Technical Education (CTE) and related education even though federal, state, and local legislation has been passed in order to fill in the widening skill gap.

The Bureau of Labor Statistics should alter their surveys in order to obtain improved data in order to better understand workforce impact conditions. Proper implementation of automation should not have a significant impact on the workforce with the exception of an increase in skills and critical thinking by the skilled workforce. We will see an improvement in throughput and efficiency when systems are installed correctly, or a greater need for additional workforce or third parties to assist when there are deviations.

About the Author: Howard W. Penrose
Howard W. Penrose, Ph.D., CMRP is the president of MotorDoc® LLC and a past chair of the Society for Maintenance and Reliability Professionals. He currently serves as the Cybersecurity and Infrastructure chair of the SMRP Government Relations committee. MotorDoc provides industrial and electrical reliability consulting, electric machinery design and RCFA support, and is heavily involved in commercial, industrial, and federal testing and reliability and wind generation and storage system reliability and testing through Electrical Signature Analysis. He can be contacted at [email protected].