Equipment maintenance in the cognitive era
I recall my days as a young production supervisor (I am certainly aging myself). I would need to track down the maintenance crew every time a machine went down unscheduled and be at their mercy to get production running again. Over the last 30 years we have seen a slow evolution as we have moved away from purely reactive maintenance to a point where the standard practice is preventive maintenance. Through the 1980s and 1990s, the growth of programmable logic controllers (PLCs), distributed control systems (DCSs), digital communications, and the proliferation of statistical analysis tools drove the change.
The age of IoT and digital disruption
Enter the Internet of Things (IoT). IoT has ushered in a new era of rapid change – digital disruption enabled by embedded low cost sensors, pervasive wireless connectivity, cloud computing, and the mainstreaming of data science and analytical engines. There has been a sharp decrease in infrastructure costs over the last 10 years.
This has, in turn, contributed to rapid advances in maintenance management.
We are now at the threshold of cognitive maintenance management – combining human intelligence with machine learning and Artificial Intelligence (AI) to unlock the full value of IoT.
What is cognitive maintenance management?
Cognitive maintenance management is the science of accessing data from connected equipment and products, combining it with existing service knowledge and other sources of external relevant data to understand what it reveals, and using this understanding to improve equipment quality, uptime, service time, and the overall service experience. Cognitive maintenance management can:
- Connect to what you have, where you have it
- Prevent issues before they happen
- Provide proactive step-by-step guidance on doing field repairs once, right, and at lowest possible cost and disruption
- Optimize the use of systems, individual machines, and people
- Apply reasoning and learning systems to become continually smarter and more optimized over time
The challenge, however, is that the data we need is largely distributed data, full of inconsistencies and difficult to combine.
Data from billions of interactions between machines, devices and people is massive, complex and variable. We can fix one machine, but not understand how one machine’s complications might drive another machine’s failure.
The challenge of data analysis
Pre-defined programs aren’t able to analyze all the data holistically. Traditional systems cannot make sense if all the IoT data combined with unstructured data. We have data rich machine and technician logs but parsing them is difficult and extracting value is difficult. We can predict and gain insight on part/machine breakage or failure but have not made it easy for a technician to troubleshoot the problem; many repairs replace fully functional parts.
Traditional data analysis using operational reports and dashboards relies on humans to identify patterns and glean insights. The deluge of IoT data just cannot be handled effectively by humans. The risk of coming to the wrong conclusions and decisions exponentially increases.
A cognitive system makes sense of all types of data – it works across data sources and decides which patterns and relationships matter. It uses machine learning and advanced analytics to organize the data and generate insights. It evolves and improves through learned self-correction and adaptation and can:
- Address ambiguous defects
- Provide answers with a margin of error and choices
- Handle information without explicitly knowing semantics
- Interact in natural language with humans and machines
So, you may ask – how does machine learning in the context of cognitive maintenance provide more data insight than what you can learn from traditional predictive maintenance?
Predictive maintenance does not (and cannot easily) look at the differentiating environmental factors impacting a piece of equipment. Nor can it create a learning environment by leveraging the knowledge base of expert technicians and their log books. Training for a marathon at sea level is vastly different from training at altitude. How environmental variables affect you determines how you train. Similarly, with maintenance, how do different operating environments (temperature, humidity, dust, pollen) affect performance?
Imagine the power of being able to leverage the “tribal” knowledge that expert technicians hold in their heads and personal logs to fix things faster? All those things should be factored in holistically when looking at the data and coming up with accurate predictions of when and how to service and replace parts. Cognitive maintenance can provide this insight on the combinations of factors that will lead to equipment failure along with how to best service the product and at the lowest possible cost.