Predictive models can be used to predict an industrial process response to an input parameters change. Usually those models are a basis of advanced control and on-line optimization algorithms. We have made more than 150 implementations of software optimizers in different types of plants. Our systems operate in the North America, Europe and Asia. We are using different classes of predictive models, from linear models to sophisticated Fuzzy Artificial Neural Networks. Our customers sees ROI in less than one year, thanks to optimizers that uses accurate, self-adapting models.
Our algorithms are able to detect an unusual process state, that can cause a failure in the nearest future, or is an early symptom of a failure. The inference mechanism is based on on-line analysis of data streams. Our algorithms are able to inform what is the most probable source of a failure and send a recommendation about possible actions that will minimize the risk of a failure. Moreover we usually automatically validate and model measurements signals. In case of sensor failure detection, the algorithm is sending a virtual measurement signal, that will help operators run the plant. We have 5 implementations of such systems in large power plants.
Accurate forecast of electric energy, heat or gas demand is one of key elements that decides about a success of industrial enterprise. We have made 9 implementations of such forecasting systems. Our accurate forecasting models are used by customers in a day to day business.
Over time characteristics of a devices more and more differ from the characteristics provided by the manufacturer. Our algorithms update characteristics based on the set of most recent measurements. It allows for better production planning and plant optimization.
Plant simulators are based on very accurate thermodynamic models of individual devices and their connections. The simulator can be used for operator training, risk analysis and for planning of further plant development. Together with our partner we have implemented 4 simulators of large power plants in USA.
These typical applications of data analysis methods do not exhaust the range of projects that we have completed for the industry. We are always open for new, challeng projects.