Product

Neural Flow is an on-premise advanced software platform designed to optimize the performance of continuously stirred tank reactors (CSTR). By synergistically combining Artificial Intelligence and Process Engineering knowledge, we create an advanced predictive analytics software tool. Designed by a veteran team of chemical engineers and data scientists with extensive experience in the process industry, Neural Flow platform has been proven to increase production yield and efficiency, and minimize the dependency on the human factor.

Neural Flow software platform continuously gathers and analyzes the data from the reactor sensors, control systems and product laboratory results. Based on machine learning methods, the comprehensive analysis inferences, turns the regular process into the predictive process, and ensures the product to be highly repeatable.

Product
Features

Neural Flow offers an advanced platform designed to optimize the performance of continuously stirred tank reactors (CSTR). Neural Flow platform has been proven to increase production yield and efficiency, and minimize the dependency on the human factor. Neural Flow is on-premise, real time, practical and slim platform, that can be easily applied.

Real-Time Predictive Analytics The algorithm dynamically evaluates each running batch and defines the ‘golden’ batch. Throughout the process the current batch is constantly compared with the ‘golden’ batch. The software predicts the current batch product results and alerts in case that predicted results are biased from the ‘golden’ batch.
Predictive Analytics
Sensors behavior comparing high & low yield batches

The software enables users to gain deep insights from the entire manufacturing process Users can produce customized reports of relevant reactor parameter and explore the process behavior in an advanced and efficient way.
Historical Data Analysis

Machine Learning Clustering
  • Machine learning clustering algorithm, analyzes all historical batches and clusters them in groups after calculating thousands of mathematical features.
  • Each new batch is dynamically evaluated his compliance to valid clusters.
  • In case of irregularity, the platform alerts and identifies the root cause of the incompliance.