The Power of BI tools
I have taken a little break from blogging with the holidays happening. I knew many of you were on vacation and trying to wrap up your year. I have appreciated all of you reading this blog in 2017 and look forward to writing more in 2018.
One thing I have been spending a MAJOR amount of time with in the last 6 months is learning some of the Business Intelligence (BI) tools. I have focused on both Tableau and Microsoft's Power BI. I learned Tableau first, then I learned Power BI. I think each has its strengths and weaknesses, but both are just incredible tools. They are both in the Gartner "Magic Quadrant" and for good reason.
However, I think many people are missing an opportunity to use these tools with time series data and data historian type applications. I have been most impressed with being able to use it with PI Event Frames. I literally have examples of analyzing 80,000 and 155,000 event frames at a time and these tools perform extremely well when doing so. I also have examples where I have taken time series data and essentially built out event-type analyses with them. I also have examples where I have event frame data, weather data, and time series data all on the same page! The ability to merge multiple data sources and sets of data is just incredible.
In most cases, I have been using PI OLEDB Enterprise to push event frame data into an excel spreadsheet and PI Datalink for time-series data for most of my proof of concept work. Although recently, I have had a chance to use the BI Integrator for Business Analytics from OSIsoft and I am really impressed so far. I will likely write a more detailed post about it at a later date. The great thing I see about it is that you can link PI data (time series or event based) directly to the BI tool of choice without the need of an intermediary tool like Excel. I love being able to link plant floor data directly to the BI tools without an intermediate tool.
The ability to churn through massive amounts of data to gain actionable insights is just incredible. You flat can't do much of that with excel, much less as easily. You can build interactive dashboards for users that are very easy to follow, and dashboards that have numerous data sources, which is next to impossible in Excel.
If you want to see some examples, you can look below for a live report from my Google Analytics account. You can see what my most popular pages are, my most popular blogs are, and the states where most of my website visitors have been from. You can click through these below.
Next, below is a sheet break re-thread tracking tool in a paper mill. The mill was interested in looking at 4 different stages of the re-thread after a machine break. They can now click on the donut chart and see how each crew is performing. The line graphs change based on which crew they click on. They can also analyze these stages by grade of paper as well. The data was generated using PI Event frames and PI Analytics.
Another example of how these tools can be used effectively is in the below Power BI report. I am using the new Power KPI graphic that released in December of 2017 and this visual allows one to color code Sparklines. I have plotted the target values for 3 KPI's in gray, along with actual values on a Sparkline trend. The Sparkline changes colors based on whether the actual KPI is outside of its control limits. One can see that if the line is blue, that the KPI is below the limit, red is above the limit, and green is within limits. This graphic is built with PI Event Frames, PI Analytics, the PI BI Integrator for Business Analytics and Power BI.
Below is a chart with 80,000 PI event frames looking at steam consumption on 10 pieces of equipment by month versus the actual high and low temperature in the geographic location of the plant. This shows over 3 years' worth of data. We were trying to explain why steam usage across the plant had increased over the last 3 years and this showed us that weather could be a factor. I also had several other sheets in this Power BI report that compared these values with 5 minute time series data for nearly 4 years.
Below are several scatterplots in Tableau where I was looking at grade downtime versus "normalized" breaks in a paper mill. In this Tableau workbook, I built a calculation column to normalize breaks by day - for instance, a grade run that lasted 12 hours and had 2 breaks would have 4 normalized breaks (i.e. if it had run a full day with the same break rate) versus another grade run that had 2 breaks over an 8 hour period, or 6 normalized breaks. I had also done a 30 day moving average of normalized break count for analyzing long term trends. This is another incredibly powerful feature of BI tools, which is to do calculations and logic on the data to enhance your analyses.
I also had built a PI Coresight (now PI Vision) display to track KPI's for the machine. As you can see in the second scatterplot below, I allow the user to either click on or copy and paste the Coresight link into their browser for the grade run in question to do further analysis. That way, the user can look at the details over time of the grade run in question very quickly by clicking the link or pasting the link into their browser.
Finally, also using Tableau, I turned trend lines red or green based on a quality parameter that we were looking for a correlation on. If the parameter was three or less, the lines are green, Above 3, they turn various shades of red. I did this for 300 plus variables in a very short time.
As you can see, the types of analyses that you can do, the speed at which you can do them, and how much data that you can evaluate is just incredible. If you aren't using these tools for your time-series data, you are really missing out.