Learning-Based Control Algorithms
Learning-Based Servo Control Algorithms
At micron- to nanometer-level motion performance and better, the world's most advanced machines can be compromised by numerous effects and conditions including engineering tolerances, environmental operation conditions and part/process/load dynamics. Engineers employ sophisticated tools to optimize motion system performance in a static environment before machine commissioning. In the field, these machines have a long life expectancy and their operators demand consistent, repeatable performance. Over the course of a machine's life from design and commissioning to field installation and prolonged operation, performance can be variable. A machine that once met motion performance specifications might later in life fail acceptance test procedures.
To maximize and maintain the highest level of motion performance, ACS Motion Control has designed LearningBoost, a new servo control algorithm. This learning-based algorithm is used as a powerful tool during machine calibration, and autonomously online during machine process execution, to ensure continuously maximized performance.
LearningBoost is a state-of-the-art control algorithm that autonomously compensates and optimizes servo performance to address disturbances both inside and outside servo bandwidth. Controls engineers will use LearningBoost to solve a range of issues that produce undesirable motion behavior, including but not limited to friction effects, motor temperature variability, mechanical component degradation, oscillations, resonances, cogging and more. Previously unachievable levels of motion system accuracy and throughput can be reached with LearningBoost.
Featuring a variety of powerful tools, LearningBoost's operation modes and flexible user implementation will enhance motion performance across many different advanced industries and applications with high-precision motion control requirements.
Off-line learning operation mode, designed for use during machine calibration, allows control engineers to manipulate and tailor learned compensations, helping address unique application requirements. Learned compensation corrections are applied until nominal performance levels are reached.
Dynamic online operation mode, intended for in-process execution, is designed for continual learning and disturbance compensation corrections. Following the completion of a recipe, compensation corrections are applied, improving subsequent recipe performance.
Motion recipes that range from fractions of a second to 30 minutes will achieve optimal performance after a few motion recipe executions. Stored on an ACS Motion Controller, users can create and save up to 100 different motion recipes, eliminating the need to re-learn recipes after machine power-up and making them simple to deploy during serial machine production.
To learn more about LearningBoost, contact firstname.lastname@example.org. If you are an original equipment machine manufacturer seeking to evaluate LearningBoost, please apply for our new OEM Motion Control System Evaluation Program.