We have a proposal to create a kit that contains techniques for checking data quality. Do you think this a worthwhile endeavor?
What does data quality mean to you - what types of things would you like to monitor?
Have you pursued this, or is anyone from OSIsoft?
I have done some significant work with various dimensions of PI Data Quality. I've attached a summary document, but I actually implemented something about 3 years ago (pre AFSDK) where the client only wanted to show/report quality issues, not get inundated with alerts/notifications.
No we haven't. I had thought the ideal might be a custom DR.
I'm not seeing your summary doc, I don't know if that's an issue on my end.
Yes, i think it would be of benefit. Especially in the Life Sciences this is a common requirement. So a kit or WhitePaper that demonstrates common techniques would make it easier to answer the questions how PI supports Data Quality.
Any preview availanle on the contents? Curious if there are techniques that i find could be beneficial to add.
No concrete plans at the moment. Our AF kit development is not as formularized as the path (i.e. we do not have a formal program, but some of us still develop them in our down times Rafael Maldonado
How is this work going?
In our company, we tried defining data quality metrics couple years ago. It is a broad and deep topic (as OSIsoft representative said on one of the presentation this week at 2018 PIWorld).
Currently, we are doing simple data validation using PI AF by detecting whether the data is flat-lining (standard deviation is zero) and outside lo/hi limits.
In this week 2018 PIWorld keynote, the number one of the top 10 list from OSIsoft customer advisory board is data quality. I am looking forward to see what OSIsoft will be doing to address this. By the way, the keynote video should be available now.
I found a good example for visualization key process parameters (in my case oil industry) with bars on PI Square: https://pisquare.osisoft.com/message/48815#48815
You can make a template and sent notifications when parameters are out of specification.
Here is my template...
For laboratory tests I suggest a sigma analysis.
Retrieving data ...