Delta Robot Parameters for Predictive Maintenance

Published: 31 August 2021| Version 1 | DOI: 10.17632/fddp3dvvzr.1
Contributors:
Kiavash Fathi,
,

Description

____________________ Data points' description: ____________________ This dataset contains 11 parameter readings acquired from a delta robot and is stored as shown below: p0 := Motor inertia p1 := Gravity p2 := Mass p3 := Spring offset p4 := Coulomb friction 0 p5 := Coulomb friction 1 p6 := Coulomb friction 2 p7 := Spring constant p8 := Viscose friction 0 p9 := Viscose friction 1 p10 := Viscose friction 2 The aforementioned delta robot is a part of a smart factory and performs pick and place of barrels and springs in a customized pen production line. The parameters are attained from the feed forward controller and are movement dependent. The learning rate for different parameters is as follows (lambda k is the learning rate for the kth parameter): lambda0 = 0.0005 lambda1 = 0.08 lambda2 = 0.09 lambda3 = 0.07 lambda4 = 0.1 lambda5 = 0.1 lambda6 = 0.1 lambda7 = 0.2 lambda8 = 0.01 lambda9 = 0.01 lambda10 = 0.01 ____________________ Datasets: ____________________ BarrelMovement.csv contains parameter readings from barrel pick and place movement. Similarly, SpringMovement_Converged.csv and SpringMovement_withTransientData.csv contain parameter readings for spring pick and place movement. However, SpringMovement_withTransientData.csv includes data points which have not converged to the final steady state values (the parameters are still being updated given the calculated error in the control loop). Plot p10 in both datasets to see the difference. SpringMovement_Converged.csv on the other hand is similar to BarrelMovement.csv and contains the converged parameter values for a specific movement. ____________________ Hints: ____________________ It is best to normalize the data so that the patterns in the parameter readings can be acquired for different models. The authors have used scikit learn library for pre-processing the data.

Files

Institutions

Zurcher Hochschule fur Angewandte Wissenschaften

Categories

Control Theory, Robotics

Licence