In the process industry, control and optimisation of proces equipment is fundamental to ensure energy preservation, increase profitability and avoid hazardous suboptimal functioning. Advanced control strategies require high-quality measures of output quality, in order to steer plants’ operating conditions towards optimality.

Some of these measures are however impossible or impractical to measure with physical sensors. These tend to be therefore measured sporadically through laboratory analyses, which creates delays that affect the performance of the plant control system, reducing the quality of the outputs as well as plant safety.

However, the vast amount of sensor data collected in a typical process plant offers an excellent opportunity for a transformative, data-driven solution.

In a collaborative effort with Alpha Process Control Srl. (experts in advanced process control technology) we developed VPI Soft Sensors - a groundbreaking software suite for the design, calibration, and real-time integration of ML based predictors for unmeasurable industrial process data.

A novel, hybrid soft sensor architecture: Our approach centers around a hybrid soft sensor architecture that combines first-principles thermodynamic process models with a statistical and optimisation engine that incorporates historical process and laboratory data. An advanced scenario-specific machine learning model is trained on both datasets, yielding a high quality predictor that goes beyond the state of the art.

Design, calibrate and integrate: The software suite provides a comprehensive set of tools to design and calibrate software based soft sensors and integrate them for real time prediction in a plant’s APC architecture including with any existing APC software.

In environments where every percentage point counts, VPI Soft Sensors delivers game changing results without requiring any changes to existing physical plant infrastructure. A representative deployment can expect €1M in benefit within the first year, between increased revenue and lower costs.

Now installed at some of the largest petrochemical plants in the world (Shell Singapore and GS Caltex), end-users of VPI Soft Sensors benefit from unparalled predictive accuracy, improved yield, reduced waste, lower energy costs and improved operational efficiency.

The project was awarded the prestigious and hyper-competitive Innovate UK SMART Grant- a glowing endorsement of its wider impacts.