Technology

Beyond Six Sigma: Process Improvement with Machine Learning

Alp Kucukelbir
Ph.D.
Chief Scientist

Modern factories are massively complex systems that depend, by and large, on humans. They can only operate smoothly under the attention of highly specialized experts, like plant managers, quality engineers, and process improvement specialists. These experts have years of invaluable domain experience and on-the-job wisdom. They are the heroes of modern manufacturing, regularly making high-stake decisions to maintain and improve their plant’s operations.

When heroes are faced with a challenge, they choose the best weapon for the task from their arsenal. The usual weapon of choice for factories is “Six Sigma”—a set of statistical techniques and procedures from the 1980s used to systematically improve industrial processes. Six Sigma was popular in the 1990s. By the end of that decade, it was being used by two thirds of the Fortune 500. But in today's digitized, data-rich factories, this weapon is starting to show its age. The arsenal is due for an update.

Six Sigma is slow and expensive.

Consider a two-phase chemicals process. The plant manager and her operators know how to control each reaction individually, but when they go to improve their process, they can only focus on a few operating regimes, centered around typical set points. The process may have over 1000 sensor measurements and parameters to tinker with, but with her small team and Six Sigma’s multiple month iteration process, the plant manager needs to carefully handpick only a few to explore. To leverage the rest of this data and extract more insights, she needs a more powerful approach, and along with it, a more powerful tool.

It is hard for the Six Sigma approach to sift through thousands of parameters and expose the subtle relationships between all of them. Six Sigma improvement is too slow and expensive.

Fero brings a new weapon to the arsenal, one that lets industrial process managers take advantage of all their data. Fero’s machine learning software empowers human experts to discover surprising relationships in their process, identify the root cause of issues, and make data-driven decisions on a continual basis.

With Fero at her fingertips, the plant manager may find that the temperature at an unexpected part of the process exhibits a significant impact on yield, but only under a specific flow setting. This explains why the yield has been below-average for the past few weeks! The amount of data required to identify this specific relationship is so overwhelming that she would never have had the time or money to test them using Six Sigma. Fero enables continuous improvement experts to make the best out of their data.

Configuring the process to operate at this new temperature setting is simply the first step. As the plant manager enters this new operating regime, Fero’s software updates automatically and continuously. Fero may identify that the pressure actuator is no longer at its optimal setting. As she explores this uncharted operating regime, she continuously relies on Fero’s up-to-date machine learning models to provide a steady stream of improvement insights. Finally, “continuous improvement” can truly be continuous.

Your company deserves to have the best weapons in its arsenal. With Fero, you can leverage the power of machine learning to improve your industrial processes better and more cheaply than ever before. Ask for a demo today!

Stay in touch

Signup to receive news and updates.