The images in Fig。 6 indicate that there is no distinct steady state condition reached in the simulations。 This can be seen inFig。 7 where the mass flow is plotted over the simulation time。 In both cases the batch of feed particles was too small to achieve steady state operation。 In the 20 Hz case there is a slight tendency of relatively steady operation between 0。5 and 1。5 s。

The same phenomena of non-steady state condition can be seen in the power draw data, see Fig。 8。 Each peak in the data corresponds to a compression event where the mantle has rolled overthe material bed in the simulated section of the crusher。 In the 20 Hz case there is a longer period of equal power draw level than in the 10 Hz case。 The predicted power draw levels are lower than anticipated and the reason for this discrepancy with the experimental data is not yet fully understood。 To some degree the difference can be explained by the idling power losses of the machine not present in the simulation, however there is still an order of magnitude difference。

One explanation of the low power draw estimation may be due to the calculation method in the DEM software。 The power is calculated as the product of the torque and angular velocity added to the product of the linear velocity and force。

It is currently unclear if this calculation is done on each particle to geometry interaction and then summed together or if the total torque is otherwise calculated on selected geometry。 In order to investigate this further, all force vectors for all particle to mantle interactions should be exported to facilitate external calculation of the actual torque on the mantle around the vertical axis。

In order to further evaluate to what degree the crusher has been simulated at steady state the net power draw data from Fig。 8 has been recalculated and presented as the cumulative energy in Fig。 9。

For each compression event it can be seen that each peak in Fig。 8 corresponds to a step increase in energy。 As can be seen the 10 Hz case is never in a condition where each step is similar for several steps in a row。 A higher tendency of linearity can be seen for the 20 Hz case。

From the simulations it is possible to extract numerical datasets that enable comparison with experimental data。 It is also possible to qualitatively evaluate the behaviour of the particle flow and how well it conforms to the assumptions made regarding, for instance, the dynamics of the compression event。 In Fig。 10 an attempt is made to show an example of such qualitative insight drawn from simulations。 In the images the trajectory stream is presented for a randomly selected  cluster of particles。

The clusters were selected from the surviving discharge in order to track it backwards to the feed。 The colour represents the vertical velocity component and cold colours correspond to the particles falling downwards in the crusher。 When the stream colour turns green or red the particle is hence subjected to a compression event interaction。 In the 20 Hz case it can be seen that an additional parintended cluster。 As a consequence this additional particle trajec- ticle was unintentionally selected which did not belong to the the mantle begins to rotate with the main shaft。 This phenomenon is normally denoted as head-spin。 When operating the crusher at unusually high eccentric speeds the head-spin effect becomes an increasingly troublesome issue。 Attempts have been made to limit the relative mantle rotation during idling power tests however it is unknown to what degree this limiting action affects the power draw。 Due to these issues the net power is not yet calculated or presented here。 It is a fair assumption to make that the idling power is relatively significant, especially in the 20 Hz case。 The bearing assembly pre-loading and lubrication viscosity is not opti- mized for these high speeds; hence, this should be looked into for

future test campaigns。