​

The calculations used for the ongoing monitoring of production processes in industry depend on sensors on the production line. If a sensor is faulty, the calculations will be inaccurate. The extrapolation techniques generally used to fill in holes in the data are not always effective.

Researchers at List developed a more sophisticated prediction algorithm to make production monitoring more reliable. The algorithm uses a statistical regression analysis that takes into account the sensor's history and measurement redundancy. The algorithm then determines the best prediction strategy and creates a model. To confirm the method's effectiveness, actual data and data obtained through different algorithms were compared. "When there are not a lot of holes in the data, all of the methods perform similarly," said a List researcher. "However, when the amount of missing data is more substantial or if the process is very irregular, regression analysis is more effective."

The algorithm was validated on a wide range of data and is currently being scaled up for use in a food manufacturing plant. Its use will then be expanded to other industries. This very generic approach can be used for any continuous manufacturing process as well as for energy monitoring.