Machine Learning is the New Jake

  /  Predictive Analytics   /  Machine Learning is the New Jake
machine learning

Machine Learning is the New Jake

Everyone in, and serving, manufacturing businesses has encountered terms such as Industry 4.0, Digital Transformation, and Smart Factory. Small compute footprints, low-cost, high-availability communication systems, high-capacity–low-cost memory, rapidly evolving sensor technology, and new time-series data structures supporting real-time analytics are all conspiring to empower industrial operators to transform their businesses from “tribal knowledge systems” to data-driven operations.

Jake at International Paper

Twenty years ago — in the dawn of the Internet age, at a converting operation then part of International Paper, we relied on Jake. Jake was a medium age, good natured man who had worked the plant for seventeen years — since high school. Jake, unusual in that he was able to stand the middle ground between Union and Management, epitomized “tribal knowledge.” Whenever the winder would spool loose rolls, Jake would know how to put that winder right. When the polydimethylsiloxane wasn’t curing correctly, Jake could nudge the coater into compliance. Jake as invited onto every troubleshooting team, sought by operators who couldn’t get the web to run straight, and by process engineers who needed to understand what was really happening in their processes.  Jake was Tribal Knowledge.

In 1993, we undertook to digitize our manufacturing processes. Weeks we spent writing volumes of requirements — and in many cases, “requirements” were merely wish lists because at that time, processing power, data communications, and sensor technology were not able to deliver what we could ask. I remember clearly writing into the maintenance section “continuous monitoring of accelerometer measurements on drive shaft couplings and automatic maintenance ticket scheduling based on changes in frequency domain harmonic amplitudes.” We could think it, but we couldn’t do it. Luckily, Jake wasn’t retiring for another twenty…

Jake’s 59 this year. Tribal Knowledge is going away. Newer generations’ expertise is different than that of Jake’s generation. We joke about the iPhone age sometimes, but jokes like those are tinged with true meaning. The entire processing power of that converting operation resides neatly in an iPhone X. Better graphics, more memory, better sensors. The wish list of the digital transformation of yore is carried in the hip pocket of junior high school students around the globe.

Replacing Tribal Knowledge with Intelligent Systems

Now is the time that requirements, not wishes, are possible. Industrial operations can implement digital transformation strategies today marching smartly along a maturity curve moving from Tribal Knowledge to Operational Maturity. New generations of operating experts arrive at work with the expectations and facility to use advanced digital technologies like machine learning, AI assistants, virtual reality, and remote–real-time “lights-out–manufacturing.”

IoT platforms bring connectivity to new advanced ruggedized sensors, secure cloud communications, next-generation time series data stores, and high-resolution user-centric multi-pane UIs. And yesterday’s wishes for advance warning of emerging maintenance needs or predictions of quality outcomes are today’s reality.

Intelligent Systems are the new Tribal Knowledge. The new generations of manufacturing operators are armed with understanding of their machines and processes that is bigger than Tribal Knowledge. Their understanding is augmented by the insights now possible as an outcome of digital transformation.

The new challenge is melding physical knowledge with insights from intelligent systems. Modern operators can turn to insights derived from discovered patterns in trailing histories of complex sensor and control signals to tell them not only what is happening at the moment of manufacture, but also as precursor conditions that facilitate proactive maintenance (PaM) and proactive operational performance management (PaOPM).

At Falkonry, we find PaOPM to be one of the most exciting opportunities of Intelligent Systems.

Machine Learning is the New Jake

Consider a continuous process in Mining & Metals or Oil & Gas. Raw materials of varying quality characteristics enter processes of controlled, and uncontrolled (such as ambient temperature and humidity) conditions. When Jake was operating these processes, his Tribal Knowledge ruled. He would listen, smell, feel, and sense the operation — making adjustments as his Tribal Knowledge dictated. Jake’s best apprentice would learn from Jake, but Jake’s shift would always produce the most, waste the least, and hit highness quality — when Jake wasn’t on vacation, or sick, or distracted. Or retiring.

Falkonry’s ready-to-use, automated machine learning discovers and recognizes patterns in real-time, time series data – rigorously, tirelessly, and repeatedly — better even than what Jake was able to achieve only through decades of human observation. The Falkonry LRS system learns patterns of operation in minutes such as those in the continuous processing of natural materials. These patterns are the digital equivalent of what Jake was intuiting through human sensors of sight, sound, smell, and instinct. Falkonry finds and learns the patterns that matter and it’s these patterns that tell modern operators the condition of their processes.

Mineral processing operators are now able to determine in real-time the quality of ore being processed and take proactive operational measures to effect the desired quality outcome. Oil & Gas refinery operators are able to predict divergence from target quality specifications hours in advance of an off-spec event facilitating dynamic proactive operational adjustments to maintain target outcomes.

Intelligent Systems are the new source of Tribal Knowledge. But this time, the knowledge is not just for Jake — it’s for the entire operation. And this source of knowledge never retires. It just keeps getting better. Find out more about how to capitalize on data analytics.

 

Crick WatersThis article was written by Crick Waters. He leads Customer Success for Falkonry and in that role engages both customers and partners. Crick is an expert at connecting Falkonry’s cognitive capabilities to industrial yield improvement objectives due to his experience on both sides.