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.

Neural Flow is an Industry-4.0 startup, established within MGT Liquid & Process Systems Group.

Opportunity
The modern manufacturing process produces an enormous amount of raw data. Today the data is visualized mostly in a two-dimensional graph and each batch is compared to the process specifications and set points.

Neural Flow offers a next-generation solution that revolutionizes data analysis and provides predictive analytics to the industry. The software analysis, based on machine learning methods, uniquely combines the profound process engineering with the raw data exploration.

As a result, the manufacturers reach a significant increase in the production yield and efficiency, and the dependency on the human factor minimizes.

Features
We develop our platform by incorporating the following features:

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.

Machine Learning Pattern Recognition
The ‘Isolated Forest’ algorithm reveals hidden patterns in the data, that lead to beneficial insights concerning the reactor yield and batch quality.

Historical Data Analysis
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 behaviour in an advanced and efficient way.