12 - Fatigue of Axles

Question 1

History of axle failures

Answer 1

Versailles accident of 8th May 1842
Broken axle caused derailment of engine; fire in wooden carriages; doors locked and around 70 killed
First railway accident to cause major loss of life

Question 2

Modern cases of axle failure

Answer 2

Broken axle caused derailment near Andheri station, Mumbai Suburban Railway network, India, September 2015
Derailment of acid train 9T90 , Queensland, Australia, September 2017 - in-service impact damage initiated a crack, anomalies in inspection procedures

Question 3

Wohler’s experiment

Answer 3

German railway engineer conducted tests, published between 1858 and 1871
Combination of rotation and bending used
Identified ‘endurance limit’ of steels; importance of gradual profile transitions (i.e. sharp changes in surface profile can be stress raisers) and that micro-slip (‘fretting’) between press fitted wheels and axles leads to fatigue failure of axle

Question 4

Define endurance limit

Answer 4

Level of stress for which the material can survive for 10^7 loading cycles

Question 5

Experimental output of Wohler’s experiment

Answer 5

Summarised by an S-N curve (stress amplitude against cycles to failure - log scale)

Question 6

How can modern axles be designed?

Answer 6

Inboard bearings (i.e. bogie lies between wheels); hollow axles; 30% reduced unsprung mass
More conventional bogie frame; for heavy locomotives; solid axles
Reduced unsprung mass and reduced overall weight contribute to lower track damage and lower fuel consumption

Question 7

What is the most common trigger for crack growth in axles?

Answer 7

Press-fit of wheels and gearbox onto axle

Question 8

How are axles designed to avoid fatigue?

Answer 8

Consider stress analysis, input loads and material properties
Degree of certainty with which crack growth can be predicted depends on input with highest degree of uncertainty

Question 9

How is stress analysis conducted?

Answer 9

Traditionally with beam theory
Now with FEA for hollow axles and those with ‘inboard’ bearings
Forces and moments due to vehicle weight can be identified in the plane of its cross-section

Question 10

Stress analysis - moments

Answer 10

Bending moment due to vehicle weight is uniform between two wheel seats (i.e. bearing locations)

Question 11

Stress analysis - uneven loading

Answer 11

If the vehicle is cornering, a greater load is taken at the outer wheel
Produces a linear variation of bending moment between the wheel seats, highest at the outer (more heavily loaded) wheel

Question 12

Real world complications to stress analysis

Answer 12

Driving/braking produces additional bending
Predicted failure location is at the wheel seat (wheel location), wheel will support axle above the assumed strength

Question 13

Input loads - loading spectra

Answer 13

Loading can be measured or predicted
Axle experiences a fully reversed sinusoidal stress cycle each revolution
Highest stresses at the surface

Question 14

What does the severity of loading depend on?

Answer 14

Track geometry and quality (curved line gives greater axle stress than straight ones)
Vehicle type (determines suspension type and configuration)
Static axle load (vehicle by itself ‘tare load’)
Passenger load spectrum (loading of passengers, how much time vehicle is full, partially loaded or empty)G

Question 15

Generating material properties data - full size

Answer 15

Full scale tests
Large scale equipment
Very long durations
Expensive
Result has a lot of credibility as it’s on the actual component

Question 16

Generating material properties data - small scale

Answer 16

Cheaper, quicker, but not real axle
Careful selection of specimen location taken from real axle
Rotating bending - grips hold and spin specimen, slight angle between grips produces bending
Multiple tests run until failure with different degrees of bending
Plotting gives S-N curve
Interrupted tests, inspected to measure crack growth at intervals throughout can give growth data
Growth rate increases noon-linearly with life

Question 17

Axle life - Palmgren-Miner’s rule

Answer 17

For each of j loading levels, find the life from the SN curve for the axle
Damage fraction at any stress level is linearly proportional to number of cycles that would produce failure at that stress level
Major drawback is that it doesn’t recognise the order of application of various stress levels
Damage assumed to accumulate at same rate at a given stress level without considering past history
In reality, sum does not alway total to 1 - range of 0.7-2.2 is normal

Question 18

Fracture mechanics - stress range

Answer 18

Several stages of analysis needed: formula to convert stress range into stress intensity factor range; crack growth law to convert this range into crack growth rate and a way to combine predictions for full range of applied loads to predict axle life

Question 19

Geometry factor for a semi-elliptical crack

Answer 19

0.6

Question 20

Combining lots of different loads

Answer 20

Fracture mechanic analysis can be repeated for spread of experienced loads
Probability distribution of time to failure can be generated
Technique is ‘Probabilistic Fracture Mechanics’
Is a risk assessment technique, some failures may still happen

Question 21

Corrosion and damage

Answer 21

Surface damage and corrosion can greatly reduce fatigue life
Surface damage is caused by ballast or other debris hitting the axle
SN curve moves down (i.e. life reduced at any given stress range)

Question 22

NDT - preventing problems

Answer 22

Fatigue failures in axles are very rare - around 1-2 per year
Ultrasound and magnetic particle inspection techniques used (both NDT)
Careful planning needed around ‘human factors’ of inspections, as so few cracks are ever found that it’s difficult for inspection staff to stay alert and watching for defects

Question 23

Surface inspections

Answer 23

Visual inspection
Ultrasonic
Magnetic particle inspection

Question 24

Sub-surface or not accessible inspections

Answer 24

Ultrasonic

Question 25

Magnetic particle inspection

Answer 25

Magnetise axle
Any defects on/near surface create leakage field
Apply iron particles
Particles are attracted to cluster at flux leakage fields, forming a visible indication, often viewed under UV light
Carried out at overhaul with wheels and brake discs etc. removed (inspects normally inaccessible areas)
Operator dependent (equipment set up/interpretation/alertness)
Expectation of detecting cracks less than 1mm deep on good surfaces

Question 26

Ultrasonic inspection

Answer 26

Scans from axle end or axle body (hollow axles)
Occur relatively frequently
Ultrasonic wave is applied to axle and reflections measured
Results compared with standard reflection trace measured in structural sound axle and assessment made of any deviations
Generally poorer detection capability than surface methods
Manual operation, highly skilled

Question 27

Inspection intervals

Answer 27

Inspection frequency depends on time for assumed crack to grow to failure
Can be timed so next inspection is an opportunity to detect a crack should it be missed during the first inspection in which it becomes visible
Conducting inspection introduces some risk through disassembly and reassembly
Cost of replacement axle can be similar to cost of conducting magnetic particle examination