In 1968, long before most of those who do advanced analytics work today were born, Kenny Rogers and the First Edition had a chart hit with «Just Dropped In». For almost as long, the spreadsheet has been around, and it has been the staple of how most engineers today go about advanced analytics. Thankfully, now there's a better way!
Said to reflect the LSD experience, Kenny Roger's song was intended to be a warning about the dangers of using the drug. Fast forward to 2019, and the chorus of the song can be seen as a starting point for how many companies do advanced analytics today.
«...I just dropped in to see what condition my condition was in Yeah, yeah, oh-yeah, what condition my condition was in...»
Most analytics efforts done today are centered around time periods of interests. Conditions define when a state is true, in other words when time periods of interest occur, and each instance of the condition is called a capsule.
You’re an expert. You know something about something
If you’re the responsible for a piece of equipment, you clearly often drop in to see if your wells are on-stream (condition), vibration levels are ok (condition), emissions within expected range (condition) etc.
These are all examples of conditions, and you probably have a display on your PC, or handheld devices, with «traffic lights» to show what condition your condition is in.
But what happens when there is an alert, a warning, and some of the lights go into red? Your condition is no longer where you want it to be. You need to take action. This is what they are paying you for.
So, you start digging in to your time series data to understand what is going on. Sometimes a simple trend will help you understand, but most of the time further investigations are required to get to the root cause.
This is when the fun begins. Do you have all the signals required for your analysis? It’s too late to start collecting them at this point, but hopefully your company already has a data infrastructure for collecting this data in place. Systems like OSIsoft PI System do a good job at that, but you will most likely also need some context from other databases, e.g. maintenance- and alarm/events systems, to get the full picture.
Data quality
Your next challenge will be that a lot of your signals are in need of cleansing. There will be gaps in your data, time periods you want to exclude, there will be fliers and outliers, and there will be timestamps that need to be aligned.
In short, you will need to clean your data before you do your analysis. Doing an average of compressor speed over a time period that includes when the compressor was offline for maintenance, will yield an incorrect result. A diesel usage report that includes an outlier value, will throw your daily consumption calculation way off.
You, as a process engineer, should be able to do this cleansing yourself, and, once it’s been done, it should be available for others to use so they won’t have to do it all over again.
Once your time periods of interest (conditions) have been identified, you will want to quantify and create statistics for said periods.
- Every time we drain the diesel tank, what volume do we drain and how long does it take? Why are certain draining capsules longer/shorter than others?
- When our well is on-stream, what is the average wellhead pressure, and how does that compare to our similar wells?Sometimes we run two generators when power demands dictates only one is needed.
- How much energy are we wasting? How much unnecessary CO2 tax will we have to pay?
Now you are gaining real insights into your process and your assets.
You’re identifying good behaviour and abnormal behaviour. You do your root cause analysis, and you report your findings in a collaborative way with colleges and external partners.
You’re doing real analytics now, and YOU have a hand on the steering wheel.
Seeq applications enable the complete lifecycle of advanced analytics, from first connecting to data to landing insights in improved business and production results.
In the current hype around A.I. and machine learning, good 'ol fashioned knowledge about your equipment and assets, coupled with the right tools, still equates to improved operational outcomes.
Your newfound insights can in turn be used to model expected, and predict future, behavior, which leads you back into monitoring and looking for those pesky boundary excursions.
It’s a lifecycle of continuous improvement, from first connecting to data to landing insights in improved business and production results.
Børre Heggernes, CTO at Amitec
Maybe it’s time you drop in to see what condition your conditions are in?
Seeq let's you search your data, add context, cleanse, model, find patterns, establish boundaries, monitor assets, collaborate in real time, and interact with process data like never before.
