Like many of my colleagues, I became an engineer because I enjoy problem solving: stripping a challenge to its core elements and experimenting with different solutions. We engineers need to understand the root cause before proposing the fix.
With data science entering the Operations and Maintenance domain, the possibility that Artificial Intelligence replaces human intelligence is cause for concern for traditionally trained engineers.
In this article, I explore the evolving role of data science and provide a primer for engineers that are now required to deploy Industrial Analytics or Maintenance 4.0 solutions without the benefit of a formal data science education.
Let’s start with some of the justifiable concerns about Machine Learning on the part of O&M employees. The valley of despair.
In 1999, two social psychologists, Justin Kruger and David Dunning published a study suggesting that people of low ability have an illusionary view of their own superiority and lack the self-awareness of their own incompetence. Another phenomenon relates to people with higher ability and self-awareness who recognize their lack of knowledge. This is captured in the graph below as the “valley of despair.”
How is this relevant to a reliability engineer? The adoption of Artificial Intelligence to the maintenance arena requires new ways of analyzing data generated from industrial machinery. In the past, most of this data, even though stored, was not made accessible for analytics.
Many of the topics relating to Machine Learning are new and were not covered in the academic curriculum of engineering programs.
As a result, there is concern on the part of engineers that plants’ data science competencies is the weak link to deploying Machine Learning based solutions. This was verified as part of the Emory University / Presenso Maintenance 4.0 Research study. O&M employees were asked to rank six statements related to Maintenance 4.0 readiness. The lowest ranked statement was “we have sufficient staff of Data Scientists to deploy Maintenance 4.0,” which ranked 4.6 out of 9.0 (Strongly Agree). Furthermore, in face to face interviews, some OEMs expressed concerns that they would need to assume responsibility for certain Machine Learning related tasks.
Although engineers’ perspectives on Machine Learning cannot be fully plotted on the Dunning-Kruger graph, the O&M employees’ concern that they lack expertise in this area could lead to the valley of despair.
This may seem like an exaggeration to some. However, Gartner uses the term “Citizen Data Scientist” to describe someone who is not formally qualified as a data scientist but is a “power user” of Machine Learning analytical tasks. Furthermore, tech giants such as Amazon and Google are releasing Machine Learning tools to help non-data scientist apply data science to their roles. Although these tools are fairly generic today, in the future, they may be targeted to plant employees tasked with Predictive Maintenance.
In reality, Maintenance 4.0 industrial analytics can only scale with solutions that do not require Machine Learning expertise at a plant level.
The data science discipline is undergoing its own revolution as Machine Learning algorithms are cannibalizing labor-intensive tasks such as data pre-processing and algorithm selection. As robotics and automation replace human labor, Automated Machine Learning is similarly changing the role of the data scientist and eliminating laborious work.
In conclusion, let me address the question: how much of the data science discipline does a Reliability Engineer need to know? In practical terms, very little.
For that purpose, Presenso’s AI driven Industrial Analytics solution extracts data to the cloud and provides plant technicians with warnings of evolving asset breakdown and Root Cause Failure Analysis. The Machine Learning heavy lifting is performed by our algorithms without O&M input. This is AutoML.
No Citizen Data Scientists need apply!