Continuous Data Analytics – Impact of Grade Change

The process operating points of a continuous process can shift with changes in the product grade that is manufactured. One example of this is a continuous reactor in which the target for output composition changes with product grades. To change the output composition, it is often necessary to change the operating point of one or more process inputs. In response to these changes in process inputs, other parameters such as cooling water flow, agitator power, and relief flow must also change to maintain batch temperature, constant agitation, and overhead pressure within specification limits. The variations in continuous reactor operation are shown below for changes in the outlet concentration between 35% to 40.

When applying on-line continuous data analytics the mean values should be changed to match those expected for a given grade. As addressed in Advanced Control Foundation – Tools, Techniques, and Applications, the product grade may be selected as the state parameter to account for variations in process measurement mean values. When the state parameter is selected as product grade (a discrete parameter with enumerated values, for example: 1–5), then in model development the parameter mean value for each state is calculated for data samples that coincide with the state parameter being in that state. For example, the mean values that were established for a continuous reactor might appear as shown below.

During on-line operation, the mean values saved in the model for the current state are used in calculating the deviation values used in prediction and fault detection. In this example, when the state parameter value is ADX215, then the mean value for primary flow is 60. If the state value changed to ADX201, then the mean value for primary flow is 70. Any changes in state will not be immediately reflected in the process parameters. To account for the time required for the process to respond to a state change, the deviation from mean is filtered based on the configured transition time.