Static Friction

26/0

3/20

10

André Carvalho BittencourtStatic Friction

ABB CRCVästerås, SE

Static FrictionLoad torques andTemperaturedependency



Motivation What is the most common researched phenomenon in robotics?

0200400600

80010001200140016001800

Control

Observer

Learning

T. planing

Diagnosis

FrictionFlexibilityBacklashRipple

Why?



Motivation - Diagnosis

Robots eventually break down Increase of wear debris Increase of friction

However! Many other things affect friction- Temperature- Load- Position- Velocity/acceleration- Lubricant, etc



Objectives

How does friction changes with Position Load torques

•Manipulated•Perpendicular

Temperature

Scope IRB 6620 Joint 2



Friction Curve Estimation In steady-state one joint at a

time

Average fwd and bwd movements

and use it as an estimate of the direction independent friction



Joint angles 50 static curves Angles range

8.5 to 60º

Effects Not significant

M0 should do it



Perpendicular load torque Tp range

[0.04, 0,10] Small variations possible

Effects Not significant At joint 1, small increase in

Fs

For joint 2, M0 is enough



Manipulated Load Torque 50 static curves Tm range

[-0.7, 0,4]

Effects (linear) Fc increase (linear) Fs increase cte vs

Proposed model, M0 +2 terms



Temperature 50 static curves (min/max) T range

34-80ºC

Effects (linear) Fs increase (linear) vs increase (exp-like) Fv increase

Load torque co-effects Subtract surfaces Subtract identified gravity

terms from surfaces Independent!



Proposed Model

M0 + 4 terms 7 terms (5 vel-weak, 2 vel-str)

Total of 11 parameters

Tm (3): (linear) increase of Fc and Fs

T (4): (linear) increase of Fs and vs (exp) increase of Fv



Validation

Parameters identified with previous data

Error reduced by a factor of 6

M0:mean=0.027,max=0.071M*:mean=0.004,max=0.018



Summary & Conclusions Empirically motivated model to describe T and Tm

So far, 1 joint! Tp might also affect

Tm affects vel-weak only (inc) T affects vel-weak (dec) and vel-str (nonlinear inc) 11 parameters (4nlin: 3 vs,TVo)

Id requires excitation over large range of T and Tm Can T-params be directly related to lubricant props?

No T meas and Tm estimated* Right now, more useful for design/evaluation

Can the model simply be plugged into a dynamic model (LuGre)?

Static Friction

Documents