Views: 66

Replies to This Discussion

I like the video. Though I disagree to call every automated algorithm or traditional data mining to be Artificial Intelligence.
Thank you for the thumbs up on the video Chris. You're absolutely correct that every automated algorithm or tradional data mining fall under the category of artificial intelligence.

The crieterion we use for the category of systems that perceives its environment and takes actions that maximize its chances of success. And that is what Falkonry does.

Falkonry examines the source data then makes intelligent decision on how and what signal processing and machine learning techniques and algorithms will yield the greatest chance of "success" — the ability to classify conditions of operation from the signal data while minimizing not false positives.

Crick

HI Crick,

I get the impression from the video that you are assuming that all of the data that are being collected and analyzed are of value.  Just because we have capability of creating vast amounts of data, does not mean that all measures are of value.  What do you do to help a client focus on the minimum set of value-added measures?  And what do you do to ensure that you are not mixing apples and oranges when analyzing the data?  Does Falconry only work with automatically generated data?  What do you do about manually generated, collected, and reported data?  Does Falconry only work in real time?  What do you do with historical data?  What if the historical data are in paper form only?

The Oil & Gas industry is now subject to extra federal regulations regarding "climate change."  In point of fact, Marathon Oil is being required to provide data back to 2006.  And a lot of these data are not easily accessible.  Can Falconry help with this problem?

With Kind Regards,
Henry Schneider

Hi Henry, let me answer your questions:

What do you do to help a client focus on the minimum set of value-added measures?

```You are correct that not all data is of value. That's both true and known. That said, new data analysis techniques enable domain owners to extract value from data that heretofore was not possible. Simple example: the motor current collected from industrial pully drive motors in coal power or mining operations. These motors fail 12% per year. Collecting the data, or simply watching hundreds of current meters doesn't help. Using Falkonry, the subtle changes in the motor curren pattern predict motor failure ~20 miunutes shad of catastrophic failure — there is value in the data.```

And what do you do to ensure that you are not mixing apples and oranges when analyzing the data?

```Domain owners know what signals they are collecting for a given asset or process. When one is monitoring failure modes of an iron ore furnace, one only need use the signals that are associated with the process. Extra signals are simply ignored by the Falkonry AI. Falkonry does engage with customer to assist grouping, including, and discarding signals.```

Does Falconry only work with automatically generated data? What do you do about manually generated, collected, and reported data? What if the historical data are in paper form only?

```Falkonry is a computational AI and as such needs value in electronic, digital form. Valu7es recorded on paper need to be converted into electronic format. (Do people still use slide rules and ledgers ? :-).

The data do not have to be generated or coll=extend automatically; however, the data do have to be time-stamped. It is not possible to discern patterns across multiple signals and mark this patterns to event occurrences without time as a base.

Inspection and event log data can be included in pattern detection. Of course, manual data entry becomes a response time problem when the data are not collected and entered at the time of their relevance.```

Does Falconry only work in real time? What do you do with historical data?

```Falkonry can work with historical data. Most commonly customers start with historical data in cvs file format. These data are dropped into Falkonry, Falkonry perorms unsupervised learning to present "first approximation" of patterns over time, then the domain expert uses historical information to label periods of time with text names that are meaningful in the domain (e.g., low-agitation, high fermentation, VOC release, etc.). Once a model is complete for given asset, the "pipeline is opened" and real-time data is flowed into Falkonry. Falkonry applies the generated model and returns the condition of the asset (name of the pattern occurring) as it is observed.```


The Oil & Gas industry is now subject to extra federal regulations regarding "climate change." In point of fact, Marathon Oil is being required to provide data back to 2006. And a lot of these data are not easily accessible. Can Falconry help with this problem?


```Sounds like a data forensics problem, which isn't where we provide value. Now if Marathon needed to comb through historical data to find periods of time during which a specific pattern of events occurred (across one or more variables), then Falkonry is the technology for the job.```


Ping me on Crick@Falkonry.com if you see a fit.
Regards,
Crick

RSS

Branding + Energy + Strategy = Brandergy

Blog Posts

MicroEssays

Posted by Vincent Wright on November 27, 2016 at 11:00am 120 Comments

You

Posted by Vincent Wright on June 4, 2015 at 1:30am 16 Comments

Financial Troubles

Posted by Vincent Wright on April 1, 2010 at 4:30am 14 Comments

YESLinking!

Posted by Vincent Wright on September 16, 2015 at 12:30am 13 Comments

© 2017   Created by Vincent Wright.   Powered by

Badges  |  Report an Issue  |  Terms of Service