Cake compression and data truncation

There are a few occasions when the Premetz model is still unable to achieve a good fit whether using the basic model, or the 2-stage analysis. There are generally found to be 2 distinct causes for this behaviour.

The most common is that the flow rate through the filter slows virtually to a stop (effectively the filter is plugged), or there is insufficient metal used in the test. In these cases, the weight vs time plot shows a distinct plateau as shown.

In this case the algorithm searches for a good fit by disposing of later data points: in this example it only uses the first 80 seconds worth of data to find a solution (the data is said to be truncated).

A second less common phenomenon which results in data truncation occurs when the cake of inclusions which builds up on top of the filter begins to compress towards the end of the test, thus changing the permeability of the cake with time. The propensity for the inclusion cake to compress depends on the inclusion loading in the metal, and the type of inclusions present.

The onset of cake compression is detected by the Premetz analysis, and the model prediction is then only applied to that portion of the test before compression. The set of results adjacent to the chart indicates both the total number of data points, and the number used for the Premetz analysis.