It should be noted that the projection of the bonded clusters issmaller than if the clusters are represented by the fractionparticles。 This means that the size of each cluster should becom pensated for by a size class specific compensation factorcorresponding to approximately one radii on each side of the cluster。 This size class compensation factor has not been applied so far in this work。 Hence it should be noted that the estimated size distributions are slightly finer than the actual clusters surviving after crushing。

 Since the product size distribution is estimated from the surviving clusters all liberated fraction particles are disregarded。 This means that the size of the small fraction particles limits the capability of predicting the fine region of the product size distribution。  The fraction particle size distribution in itself could be added inorder to calculate the total discharge distribution。 However, this distribution is fixed and is not influenced by the interaction with the crusher as fraction particles are non-breakable。 As a consequence of this size limitation only the corresponding coarse region of the experimental size distribution is compared with the DEM

results。

 The power draw is calculated from the total torque on the mantle from all particle interactions multiplied by the angular velocity。 This calculation is performed as a default output in EDEM and additional validation of the accuracy should be performed in future work。

3。 Experiments

 The particle size distributions for the two tested eccentric speed levels are presented in Fig。 4。 As expected the high speed level of 20 Hz results in a finer size distribution than for 10 Hz。 In cone crushing operations it is normally seen that the P80 corresponds roughly to the close side setting。 In this case this rule of thumb corresponds well to the 10 Hz case where the P80 was calculated to 2。09 mm which is close to the CSS of 2。2 mm, especially considering the variance from calibrating and measuring the CSS。 In the 20 Hz case a considerable amount of fine material is generated and relatively few particles larger than the CSS survive the crushing chamber。 Those particles that were retained on larger sievedecks were normally relatively flaky。

 The measured gross power draw for the two speed levels is presented in Fig。 5。 The power draw level was relatively stable during the tests with a slightly higher variability in the 20 Hz case。 The feeding arrangement to the crusher is based on a vibrating feeder with a manual potentiometer for controlling the rate。 This means that variance in the incoming feed mass flow rate may cause some of the variance in the power draw data。

 If the crusher would be fully choke fed this variance would possibly be dampened。 However, during the tests the crusher was operated at near choke feeding conditions。 When attempting full choke feeding the motor and frequency drive tended to overload。 This issue needs to be solved for future experiments。

Attempts have been made to estimate the idling power of the crusher and it is believed that it ranges from around 0。5–2 Kw depending on the speed。 It has proven to be practically very difficult to carefully measure the idling power due to the issue of mantle head spin。 When operating the crusher with no feed material mantle more frequently squeezing particles upwards more than in the 10 Hz case。 The resulting breakage illustrated in Fig。 6,may seem finer in the 10 Hz case, however the particles in the20 Hz case will receive more compressions because of the higher eccentric speed。 The number of compressions usually increases the amount of fines in the final product。

It can be observed from animations of the simulations that theparticles move in a very erratic way due to the high frequency。Since there is no heat loss accounted for in the breakage events all stored elastic energy is released into kinetic energy for the fragments。 This is an effect that needs to be investigated further and possibly a heat loss model needs to be incorporated into the bonded particle model。