Pattern Recognition

Pattern Recognition Definition

WARNING

UNEXPECTED FREQUENCY VALUE

Ensure that the MPU sensor provides the minimum input voltage for pattern recognition.

Failure to follow these instructions can result in death, serious injury, or equipment damage.

At a stabilized speed, a gear can generate a dedicated pulse delay for each tooth. During repetitive cycles, the machinery generates irregular pulses. The module functionality can correct the calculation mode of the frequency value for irregular pulse.

To be corrected:

The same pattern of irregular pulses must be reproduced at each cycle. That means, one cycle of the pattern equals one revolution of the gear.
The pattern of irregular pulses must be learned with a learning sequence.

NOTE: The velocity of the machinery must be stable during the learning sequence.

For best accuracy:

The faster pulse should not exceed 40 kHz
The ratio of faster to slower pulse should not exceed 20

The figure below show a repetitive cycle:

Sampling period

Pattern of the repetitive cycle

Integration period

When the pattern recognition is activated, the frequency value derives from a corrected integration period instead of the measured integration period.

For each sampling period, the integration period is corrected according to the received pulses within the measured period.

As for conditions:

A fixed number of teeth have been configured.
The correction ratio of each pulse has been internally memorized during the learning sequence.

Pattern Recognition Configuration

The parameter Pattern Resolution defines the quality of the pattern recognition correction:

when the parameter is set to 1:1000 (Default):
- Accuracy of the frequency value is impacted by 0.1% during the run time.
- The minimum speed must be 60 rotations per minute during the learning sequence.
when the parameter is set to 1:10000:
- Accuracy of the frequency value is impacted by 0.01% during the run time.
- The minimum speed must be 600 rotations per minute and the maximum recommended frequency is 4 kHz during the learning sequence.

The parameter Pattern Number defines the number of pulses within the pattern. According to the number of pulses declared, the pattern recognition is configured into one of the following modes:

Continuous mode when the number of pulses within the pattern is set to 2...256.

In this mode the pattern learning cycles while the control bit FORCE_PATTERN_LEARN is active.
Triggered mode when the number of pulses within the pattern is set to 257...512.

In this mode the pattern learning is executed once on rizing edge of the control bit FORCE_PATTERN_LEARN.

How to Control the Pattern Recognition

The following figure shows how to control the pattern recognition with the application:

Sequence	Description
1	When the pattern recognition is not required: ENABLE_PATTERN_RECOGNITION is false or The configuration parameter Pattern Number is set to 1. The frequency value is calculated without corrected integration period (The pattern recognition function is disabled). The VALIDITY_FREQUENCY bit is true (that is frequency, acceleration, and jerk values are guaranteed).
2	The pattern recognition is required but not possible: ENABLE_PATTERN_RECOGNITION is true and The configuration parameter Pattern Number is set to a value > 1 and The pattern learning is not completed. The frequency value is still calculated without corrected integration period. The VALIDITY_FREQUENCY bit is false (uncertain) and the application has to trigger the pattern learning process.
3	The pattern recognition is required and possible: ENABLE_PATTERN_RECOGNITION is true and The configuration parameter Pattern Number is set to a value > 1 and Pattern learning is done The frequency value is calculated with corrected integration period. The VALIDITY_FREQUENCY bit is true (that is frequency, acceleratrion, and jerk values are guaranteed).
4	The step 2 is not mandatory, the application can enable the pattern recognition after the pattern learning done (PATTERN_LEARNT is true).

Pattern Learning System

Pattern learning must be executed once after the machinery have started and before applying the pattern recognition. The velocity of the machinery must be stable during the learning sequence.

The variable COUNTER_STATUS (%IW.r.m.c.0) reports the state of the pattern learning system to the application. The application implements the recommended sequence to control the pattern learning before enabling the correction.

How to Control the Pattern Learning

Application can implement the following sequence to control the pattern learning:

Sequence	Description
1	The machinery has started and reached a stable velocity conducive to learn the pattern.
2	Application triggers the pattern learning. After the pattern learning is started, the application can reset the control bit FORCE_PATTERN_LEARN.
3	Module is executing the pattern learning once for the first time.This is the learning sequence.
4	Pattern learning is completed according to the number of pulses defined in the configuration. The ratio of each pulse within the pattern is now held until the next power cycle. The frequency value can be corrected with the pattern recognition and reports the velocity of the machinery into the regulation area.
5	Application can trigger again the pattern learning. Depending on the mode set in the configuration screen, the pattern recognition can be interrupted: In continuous mode, the pattern recognition is possible while the pattern learning cycles. In triggered mode, the pattern recognition stops during each new learning sequence.