Machine Learning and predictive analysis: a success story!

A machine learning system for predictive maintenance, smart monitoring and energy consumption forecasting, developed by Omnys

About the Company

The Company operates world-wide in the design, manufacturing and installation of full equipment solutions for the retail industry, with 20 production plants and several international offices.


The Problem faced

The Company provides its clients with a monitoring service in order to guarantee maintenance in the short-term, because of very strict regulations concerning the industry they operate in.

The problem is that this service is mostly based on post-incident support: the technical service normally reacts very quickly, but, obviously, it still takes time once the outage or incident is opened by the customer. The company collects a very large amount of raw data on a daily basis from their connected plants and sensors: a really big historical dataset that need to be analyzed in depth in order to detect valuable and interesting relationships among data or to identify potential issues before they happen.


The Challenge raised

OMNYS has been brought in to analyze this complex scenario and to roll out an AWS-based solution to handle the following requirements:

  • to create a big data repository able to store and process the huge amount of collected data;
  • to launch new services based on ML and AI so as to improve the company productivity in terms of: energy consumption forecasting, predictive maintenance, smart monitoring.


Results achieved

OMNYS designed and developed an end-to-end platform that provides a number of new key indicators, overall point-of-sale scoring based on its performances, ML-based insights, and forecasting services with high accuracy which, even in the worst cases, is always higher than 91%.

Such a platform actually analyzes, transforms and processes millions of records a day by storing them as Big Data and continuously improves itself by self-updating its ML and Forecast models and by daily generating new insights and smart notifications.

The User Interface allows the user to quickly identify any issues and to dig down into potential outage problems before they happen.

10+ million records daily processed

ML model accuracy over 91%