Top Tweets for #DefectMonitoring

over 2 years ago

To the brilliant folks her tweaking atoms, something to consider - How AI and Atoms collaborate... New #AI vs old AI & why solutions like #vibration analysis or #defectmonitoring will struggle to transform operations or usher in #I40... tl/dr; - watch the video - #manufacturing data architecture was built for old AI, is not suited for new AI - as a result, the 4th IR has been delayed - while many impressive incremental advances to I30 are deployed - explaining why we're stuck in the Industry 3.9 Rut #I39R Thanks to @AndrewWLo for inspiration, h/t to @sarbjeetjohal for digging it up. Longer clip: https://t.co/YPoNuSVyd5 🚫 the Mythical Magical Manual #GenerativeAI example - AI scans a manual and we ask it questions about blinking lights - makes the point well: - a manual is an expert system, “an exhaustive enumeration of complex scenarios” related to optimal design, installation, operation - having sparse real-world data related to actual performance - and no real “storage” mechanism for deployed operational mods - as explained in the clip; complex math, little real data ✅ New AI will transform ops BUT requires a shift in how the existing IO and resultant data is handled. 👋IMPORTANT: Little new to invent, just reconfiguring, which is the big opportunity for a solution vendor... Why the architectural shift? 1) the existing data is siloed / tightly controlled to ensure reliability and security- perfectly logical 2) nodes only expel pre-defined summaries to systems of record - current state, start time, count, deviation alarms, etc 3) and too much human interaction is left un-documented (see also #EAM) 4) resulting in new AI leveraging summary systems with less and/or incomplete data How we got here: - storage used to be expensive, manufacturing was cost sensitive, an entire class of software (the historian) was developed to address this - manufacturing data is huge in aggregate, small at the node - leadership thinks every data point = protected IP - nodes were built to pub/sub status, not insights - who can afford to expert systems on every pump outside of our showcase? We’re getting close accidentally, not intentionally: - we run ‘tinyML' on pumps - cheap connectivity gives the illusion of data accessibility - cloud ingress and egress are expensive - visibility to aggregate node data is elusive unless you buy the OEM subscription - which IDs pump wobble, but why do I care? Today? On a Snowy Friday? At the end of the month/quarter/year? I40 = new AI as a distributed congress of simple(r) #SPI (little brains) controlling nodes while expelling summary + insight to a #CI (big brains) trained by aggregate data AND enterprise expectation AND cyber-aware systems. The solution requires a different form of leadership & partnership. It will all start with the small-mids because despite over a decade of decline in industrial performance, large enterprise shops all have it figured out, I’m told. New AI will radically alter the path, and create rapid, sustainable growth, once the foundation is laid for it. ..and it will be revolutionary!

543

Klera is now part of Gathr @Klera_io

over 4 years ago

#CodeConflict #DefectMonitoring #DevOps #ALM #CodeCommits #DataSync #Klera

Top Tweets for #DefectMonitoring

Last Seen Hashtags on Sotwe

Trends for you

Most Popular Users