It’s well established that unplanned downtime is a headache. Software vendors have advertised for years that analytics are supposed to cure this headache. Instead, many companies are finding that analytics bring headaches of their own. Why haven’t analytics magically solved the downtime problem?
You may have heard of the Pareto Principle; 20% of your effort produces 80% of your results. This sounds great, but the tricky part is figuring out which part of the effort is the valuable 20%.
There are many ways to embark on an analytics journey, many of which will not be time well spent. So, Atonix Digital has identified 4 areas of focus to ensure your effort is delivering results.
The data science process is well established – if you google it, you’ll get something like the below diagram.
As applied to Operational Intelligence, the traditional application of this process looks something like this:
The flaw in this approach is that you’re working through a single problem at a time. You identify a single failure mode, you work with operations experts to decide what inputs to use, you comb through the data to identify “good data” and “bad data”, you look at correlations of individual inputs to decide which inputs to keep, you experiment with different math algorithms, you determine how sensitive the model should be, and you deploy it into production. At the end of this process, you have a good model (singular). Repeat this process 1000 times, and you’ll have a single plant covered.
However, there is a way to significantly reduce the amount of time spent building models as most of these steps are automatable. And when you start thinking of the automated approach, you stop thinking about an individual problem and you start thinking about categories of problems across your entire asset-base. You leverage templates to determine the right inputs, you use rules to clean the data, you use input elimination algorithms to determine which inputs to keep, you use scoring algorithms to pick which algorithm to use, and you use relative metrics to determine how sensitive the model should be.
The automated approach results in 80% asset coverage in 20% (or less) effort.
We’ve seen companies struggle with identifying what type of resource to use on their analytics team.
On one hand, some companies favor using Data Scientists or Analytics Specialists in the building of models. Data Scientists are clearly experts at data and mathematics but have limited experience with the details of equipment or operations. Data Scientists build high-quality predictive models as they follow the traditional data science approach noted above. They build models from scratch and focus on the details of every individual model. So, you get great models but it’s a time-consuming approach.
On the other hand, Asset Experts (think engineers and operations) understand the details of the operation, but have less experience with Data and Mathematics. Since they understand the process, they understand the process data – which tags are related and which are independent. Since they have a wide view of operations, Asset Experts tend to want as much asset/process coverage as possible as quickly as possible.
Our suggestion is to have Data Scientists focus on new and complicated problems, while you enable asset experts with tools, like AtonixOI, that make the modeling and evaluation easy.
Each company is at their own stage in the analytics journey, and each company only has their own experience to leverage as they progress. This leads many companies to start from scratch – coming up with unique ways to move and clean data, testing many different algorithms, defining their priority failure modes and what data is needed to detect the failures.
Many people say “Don’t reinvent the Wheel.” This statement is sometimes rebutted by saying “Reinventing the wheel is how we innovate.” Here are some thoughts on when to leverage existing wheels and when to innovate.
Operational Intelligence solutions are widely deployed, and Atonix Digital has done much of the reinventing for you. AtonixOI contains thousands of model templates, preselected algorithms, and automated input selection that dramatically reduce your time to value. This experience is built into the software, so there is no need for you to reinvent the OI wheel
The People, Process, and Technology framework has been around since the 1960s, but the industry tends to talk a lot about data and digitalization and analytics and AI.
All of these topics fall into the Technology category and ignore the other two circles.
If you picture yourself on a journey from data to results, the entire data science process noted above fits in the first 20% of the journey. It’s important to think about what happens after the analytics are deployed. Many companies do think about the process, but it feels like forced organizational change to adopt a technology rather than adopting a technology that anticipates (or even drives) a process. Analytics are important, but the right technology sees analytics as the first step in a detect, diagnose, and resolve process that drives results.
We can all agree that unplanned downtime is a headache. The right analytics will cure the headache; the wrong analytics will compound it. Use AtonixOI to automate the modeling, enable your asset experts, bring built-in experience, and to drive a process that delivers results.
See what Atonix can do for you with a free demo. Contact us below.