CASE STUDY
Xaba xCognition Robotics Control System
Learn how xCognition enables a standard ABB IRB4600 to perform high accuracy drilling operations.
Having an accurate and repeatable robotics system is paramount in order to improve upon current robotics applications and to enable industrial robotics to penetrate into new applications currently dominated by custom and expensive CNC machine tools.
The aim of this case study is to demonstrate how Xaba’s xCognition can augment accuracy, repeatability, consistency and consequently job execution quality of any standard commercial industrial robot.
In the first part of this case study, we integrated a compacted industrial controller (NI CRIO), containing our proprietary xCognition machine learning algorithm into the standard control cabinet of a commercial ABB IRB4600.
In the second part of this case study, we trained the xCognition controller by collecting a “quasi-stochastic” dataset of points using our proprietary algorithm and toolsets for industrial robots data collection and machine learning model training.
In the third part of this case study we have run the ISO 9283 tests and compared position accuracy results for an ABB IRB4600 with and without our xCognition controller.
Last, we performed a drilling test using an ABB IRB4600 and compared hole positions accuracy results for an ABB IRB4600 with and without our xCognition controller.
ISO 9283 test
The ISO 9283 test has been performed comparing the accuracies of the standard ABB rigid model and Xaba’s Machine Learning (ML) model. The volume covered by the TCP is represented as a red cube in the following picture, where the measured points are also reported (on the diagonal plane of the cube).
Results
The FARO Vantage laser tracker has been used to acquire and validate the data reported in this case study.
Without xCognition
Relative error between 0.697 and 1.078 mm
With xCognition
Relative error between 0.098 and 0.187 mm
Drilling test
As a representation of performance in a realistic manufacturing scenario, drilling tests have also been performed by drilling an actual metal workpiece. During these tests, a Laser SMR target was continuously tracked in order to capture the trajectory followed by the drill spindle.
In the following pictures, measurements are reported for two holes drilled at 20mm distance from each other. For each hole, a line has been fitted to the measured points, and then the distance between the fitted lines has been computed.
These tests are performed with the ML model, and the hole positioning error is less than 0.02 mm as reported below.
Mean distance between drilling trajectories
19.981 mm
Measured via Laser Tracker
Mean distance between holes
19.95 mm
Measured via Digital Caliper
Contact us
Looking to implement new intelligent robotics, create an automated factory or develop a disruptive new process or material? Let's talk.