Data Acquistion

Based on the latest networking technologies, our system can acquire 4-20 mamps, Volts. It has been running for one year continuously 24/7 without interruption.

When a DCS is in place, we can acquire directly through the industrial network / bus with an OPG client-server or we can  re-acquire the signal on a IO from the DCS (a solution priviledged by some client for security).

Our system is particularly well suited for small units and pilots. It is unexpensive, use traditional PID control with a decentralised mini computer.

Our system is installed on a DIN rail in Biowanze in an ethanol production plant.


With analog and digitital I/Os, we design start-up/shut-down sequences. The sequence is written in python and easily understood by any engineer with some programmation knowledge. So it can be modified easily to include in a later stage observations from the operators.

Sequences written by our company have been outstanding and decreased the number of costly mistakes. For the time being, sequences relies on logical intelligence and is not yet learning.

The control valve in Prayon plant

Artificial intelligence control

We demonstrated in an industrial environment that our control is more stable and reacts better to process deviation.

The learning is is excellent on a unit where we control the quantity of dry matter and the level in a falling film. 

A too high concentration will burn the product and foul the exchanger. This is the graph of all data collected by the ABCSU (Artificial Brain Control Supervisor) and the 2 valves controlled by the level and dry matter (in blue)

Thanks to the learning process, we can change easily the number of variables that we want to consider in the control. It is just a matter of one cell change in an excel file. It is very easy to test and optimise the results.

Data are collected and saved on external computer. Statistics show clearly the improvement of the control but also it the speed of correction on external deviation is also clearly visible.

The following two graphs shows the results of a learning artificial neural network controlling a level (here on the left) and the dry matter of the product (just down).

Big process deviation are visible sometime and the control is taking back the controlled variables much faster than the usual PID "modified" control.

The time scale shows the control during a period of 6 hours.