Background

​

Condition monitoring enables product quality control by detecting combinations of equipment health, such as spindle vibration frequency, engine temperature, cutting speed, and ambient parameters, such as temperature and humidity. Combined, these parameters can cause deterioration in the quality of a product output.

​

A historical data set that contains equipment condition records gathered through a time period (say, a year) is combined with the data bout product quality deviations and context data (for example, equipment maintenance history) from either ERP, PIMS, or DCS systems. The combined data set is then fed into advanced machine learning algorithms, which can then detect causal correlations in the incoming data records. Uncovered correlations are reflected in predictive models, which are then used to identify combinations of equipment condition and environmental parameters that can lead to product quality issues.

For example, in pulp processing some of the quality issues include deviations in the concentration of dissolved alkali.

 

The machine learning component of IoT detects hidden patterns in the data and states that a higher concentration of alkali stems from a deviation in two process parameters: reduced processing temperature and increased white liquor flow.