Measurement of parameters in the process industry is key for optimizing process control and increasing OEE. To achieve this, sensors are one of the most common methods. The installation and integration of a sensor on the processes line however can be costly or might even lead to downtime. The responsibility of automating such process falls on the shoulders of automation and controls engineers.
Digitalization is now becoming a key area for many companies in the process industries and automation engineers are at the forefront of this field to enable other colleagues to do more with less. Automation engineers must take care of topics like measurement (e.g. measurement of pressure), control (e.g. setting up logic based on pressure value), and actuators (e.g. activating the safety pressure relief valve if pressure is too much).
The latest study from Accenture (source) shows that the chemical plants are still thinking solely of measuring parameters using physical sensors level. Even though new technologies and methods are available.
We know that concentration is a function of the conductivity of the liquid. So Instead of directly measuring the concentration of a liquid with physical methods, and by using this correlation of concentration and conductivity, the concentration of the liquid can be measured “virtually”.
Conductivity needs to be measured physically and this measured conductivity gives the virtual sensor the input needed to simulate concentration.
To deploy the Virtual Sensor, first, the target KPI and the parameters influencing it are defined considering the physical sensor data and relevant manual measurements. The virtual sensor is deployed in such use cases to simulate the state of the parameters for the areas where there is a shortage of information and where a physical sensor cannot be applied (blind spots).
Continuously changing parameters need fast decision-making to cope up with the changes to avoid negative cascading effects, otherwise, the obtained data cannot be converted into the next action items. Taking the best decisions in such environments can be very stressful and lead to fatigue. This leads to a question “If a machine is measuring the parameters, would it not be a good idea for the same machine to also tell what to do?” So-called recommender models can help in such situations by modeling the adjustment of process parameters based on historical data, just as experienced employees do. Thus, knowledge from experience is de facto "democratized" and expanded in a data-based manner.