AI/ML
Modcon’s Industrial AI solutions complement our analytic hardware and leverage the data it generates together with other data sources operated by our customers. One aspect of our Industrial AI capabilities relates to maintaining the spectrometric models for our NIR analysers. A proprietary patent-pending Machine Learning application has been developed to automatically identify the most suitable samples for model building while excluding the outliers.
Rich process data, including the measurements provided by our NIR analysers and other online analytic solutions, allows the building of the Digital Twin of the process, precisely replicating the actual process dynamics. These high-fidelity process models allow answering the “what if” questions that lead to finding the optimal process setpoints for various process states. The Energy Conservation solution leverages this modelling ability through an online, DRL-based, optimising multivariate process controller.
CDU Optimisation Suite uses our MOD 4100 Crude Oil analyser to allow the AI-based CDU controller to adapt to changing feedstock. The controller relies on a high-precision rectification column model, complementing the first principles model with data-driven correction based on historical process data. Working in the paradigm of optimising control, the CDU controller, implemented using the Deep Reinforcement Learning technology, iteratively nudges the process from its present state to the one aligned with the optimisation criteria.
The in-depth understanding of the process obtained through its precise data-driven model is a solid foundation for detecting divergencies between the model prediction and actual measurements of various process parameters. The trends in these divergencies over time, coupled with process parameters anomalies detection, indicate the process health and serve as process health markers.
Modcon leverages its decades-long expertise in smart sensors to create a range of medical solutions based on the contactless evaluation of vital signs. Coupled with advanced algorithmic and Machine Learning models, we allow the end-user multiple self-service and triage capabilities and niche applications developed in cooperation with clinicians from the leading hospitals.
