Siemens SINUMERIK 840D sl Function Manual page 417

● Number of learning passes
● Section-by-section learning active yes/no (TRUE/FALSE)
● Range boundaries for "Section-by-section learning" (minimum acceleration, maximum
● Time taken for one test motion (to and fro)
Requirements
In the learning phase, the neural QEC requires a speed feedforward control without jerk
limitation (BRISK):
(MD32620 $MA_FFW_MODE=1; FFWON (feedforward control mode))
The feedforward control must therefore be correctly parameterized and optimized. When the
learning process is started a check is made to see whether the speed feedforward control is
activated. If not, the learning process is canceled and an error message is generated.
Extended Functions
Function Manual, 01/2008, 6FC5397-1BP10-3BA0
Default value = 15; range > 0
The effect of this parameter depends on whether "Detailed learning active" is set or not.
a) Detailed learning not active (= FALSE):
The number of test motions (back and forth) is defined for each acceleration stage. The
higher the number, the more accurate learning is, but the longer learning takes.
With directional compensation ($AA_QEC_DIRECTION = TRUE), the parameterized
number of test movements for every direction is generated.
b) Detailed learning not active (= TRUE):
The number of complete runs, from maximum to minimum acceleration and vice versa, is
activated with fine incrementation, i.e. with a setting of 1, all acceleration stages are
executed once starting at the maximum value. For every acceleration stage, two test
movements are generated if there is no directional compensation
($AA_QEC_DIRECTION = FALSE), otherwise four test movements are performed per
acceleration stage.
A reduction of the "Number of learning passes" can be made if data blocks for the
machine type already exist (series machines) and these are to be used as a basis for
further optimization.
"Section-by-section learning" in certain acceleration ranges is especially interesting for
"Detailed learning" e.g. in technologically important ranges of the machine. By defining
the ranges appropriately it is possible to reduce the learning duration.
Default value = FALSE
acceleration); only relevant for "Section-by-section learning active".
Default value = 0; format: mm/s
Default value = 0.5; format: s (seconds) (corresponds to a frequency of 2 Hz)
5.7 Neural quadrant error compensation
2
Compensations (K3)
417