102
2 Experiments in Pebble Flows
first frame by comparing the Voronoï stars. It is necessary to check whether particle
P b,2 is the matching partner of P a,1 after P a,1 has an avalanche. Firstly, the two stars
must share several same extremities whose centers are P a,1 and P b,2 , respectively.
Secondly, it is apparent that particle P b,2 should be located below P a,1 in the vertical
direction, and the distance between the two particles must be less than a few particle
diameters. Moreover, the voids between particles are needed to provide enough room
for particle P a,1 to move toward P b,2 below. Then, a real avalanche motion will be
detected. This forms the improved RM matching method and the flowchart is shown
in Fig. 2.36.
As a numerical test, the improved RM matching algorithm is compared to the
DEM simulation results for verification. Since the test velocity, particle positions,
and numbers are known in advance, similarly, the reliability Φ r and the total measurement yield Φ y are used to evaluate the performance of this tracking method.
Let N c be the number of links correctly identified. The reliability y is defined as
Φ r =
N c
N m
, where N m is the measured link numbers. The total measurement yield
is defined as Φ y =
N c
N a
, where N a is the actual linking number in the simulation.
One hundred and sixty frames of the DEM simulation data are used for comparison (Fig. 2.37). In general, the improved method shows higher reliabilities, which
means the unpaired or noise pebbles can be reliably discriminated against and dismissed. Meanwhile, the mean improved Φ y equals 96.3% which is more significant
Fig. 2.36 The flowchart of the improved RM matching
2 Experiments in Pebble Flows
first frame by comparing the Voronoï stars. It is necessary to check whether particle
P b,2 is the matching partner of P a,1 after P a,1 has an avalanche. Firstly, the two stars
must share several same extremities whose centers are P a,1 and P b,2 , respectively.
Secondly, it is apparent that particle P b,2 should be located below P a,1 in the vertical
direction, and the distance between the two particles must be less than a few particle
diameters. Moreover, the voids between particles are needed to provide enough room
for particle P a,1 to move toward P b,2 below. Then, a real avalanche motion will be
detected. This forms the improved RM matching method and the flowchart is shown
in Fig. 2.36.
As a numerical test, the improved RM matching algorithm is compared to the
DEM simulation results for verification. Since the test velocity, particle positions,
and numbers are known in advance, similarly, the reliability Φ r and the total measurement yield Φ y are used to evaluate the performance of this tracking method.
Let N c be the number of links correctly identified. The reliability y is defined as
Φ r =
N c
N m
, where N m is the measured link numbers. The total measurement yield
is defined as Φ y =
N c
N a
, where N a is the actual linking number in the simulation.
One hundred and sixty frames of the DEM simulation data are used for comparison (Fig. 2.37). In general, the improved method shows higher reliabilities, which
means the unpaired or noise pebbles can be reliably discriminated against and dismissed. Meanwhile, the mean improved Φ y equals 96.3% which is more significant
Fig. 2.36 The flowchart of the improved RM matching