Spontaneous Ignition ofHydrogen Leaks:

A Review of Postulated Mechanisms

S.J. Hawksworth and G.R. Astbury

Health & Safety Laboratory

Buxton, U.K.

Introduction

• Hydrogen has reputation for spontaneous ignition

• Major Hazard Incident Database Service (MHIDAS) searched

• 81 incidents reported – 4 delayed ignition– 86% no ignition source identified

• Compare with non-hydrogen releases– 65% no ignition source identified

• Zero non-ignitions not significant – – no reports?

Introduction

• Frequency of occurrences of ignition sources:

Ignition Hydrogen Non-hydrogensource incidents incidents

Number % Number %Arson 0 0 37 2.6Collision 2 2.5 121 8.4Flame 3 3.7 113 7.9Hot Surface 2 2.5 56 3.9Electric 2 2.5 114 7.9Friction Spark 2 2.5 33 2.3Not identified 70 86.3 942 65.5Non-ignition 0 0 21 1.5Total 81 100 1437 100

Specific Incidents

• 1922 – Work by Nusselt – Germany

• 1926 – Fenning & Cotton – U.K.

• 1930 – Fenning & Cotton – U.K.

• 1991 – Bond – U.K.

• 1991 – Bond – U.K.

• 1964 – Reider, Otway & Knight – U.S.A.

• 2004 – Work at HSL Buxton – not yet reported

Nusselt - 1922

• Spontaneous ignitions occurred – releases at 21 bar

• Test work releasing hydrogen through different nozzles – no ignitions

• Cylinders contained rust although apparently dry

• Potential for electrostatic charging• No ignitions using many fine powders• Only fine iron oxide and manganese oxide

caused ignitions• Rust then thought to catalyse oxidation

Nusselt Experiments

• Hydrogen and oxygen mixtures stored at

• 11 bar – pressure fell but no explosions– 24 hrs at 100°C

– 9 hrs at 200°C

– 1 hrs at 380°C

• Subsequent trials in dark revealed corona discharge – fine rust present

• Tapping equipment caused ignitions – disturbed rust?

• Corona discharge probable cause

Fenning & Cotton - 1930

• First ignition 1926

• Only reported after second occurred in 1930

• Fine dust present

• Thought to be electrostatic ignition

• Charging of dust due to high velocity

• Many experiments – no ignitions at all

• Review suggested electrostatic ignition

• Inconclusive – probably electrostatic ignition

Fenning & Cotton - 1926

• First ignition 1926 – reviewed

• Fine spray of mercury at atmospheric pressure

• Thought to be electrostatic ignition

• Now known mechanism of bursting bubbles and sprays igniting hydrocarbon/oxygen mixture

• Hydrocarbon/oxygen ignition energy similar to hydrogen/air

• Sufficient charge to ignite sensitive atmospheres

Bond - 1991

• First release– 110 bar release from flange

– Ignition reported to occur on second strike of hammer wrench by fitter

– Not apparent whether impact spark or diffusion ignition

• Second release– “Snifting” gas cylinder (230 bar)

– Attributed to diffusion ignition

Reider, Otway & Knight - 1964

• Release at 230 bar through nozzle

• After 10 seconds, valve closed• 3 seconds after starting to closing valve, ignition

occurred• System cleaned prior to test to eliminate static

generation from loose dust• After event: velocity far higher than previous

runs• Bar across nozzle detached at one end –

possible ignition source

Health & Safety Laboratory

• Releases from storage at 150 bar

• Various nozzles from 1 mm to 12 mm

• No ignitions occurred

• Attempts to induce ignition by entraining dust in the jet (externally) did not produce ignition.

Health & Safety Laboratory

Postulated Mechanisms

• Reverse Joule-Thomson Effect

• Electrostatic ignition– Spark discharges from isolated conductors

– Brush discharges

– Corona discharges

• Diffusion ignition

• Sudden adiabatic compression

• Hot surface ignition

Reverse Joule-Thomson Effect

• Joule-Thomson inversion temperature 193 K

• Above inversion temperature, temperature rises on expansion (opposite to air at ambient)

• Known data partly experimental, part calculation

• Isenthalpic lines very non-linear at very high pressures

• At 2500 bar, coefficient is 0.53 K MPa-1

• Maximum temperature rise typically only 132 K

• Unlikely to cause ignition – AIT is 560°C

Electrostatic Discharge Types

• Three possible types of discharge:

– Spark discharges from isolated conductors• Discrete plasma channel

– Brush discharges• Typically from plastics and insulators

– Corona discharges• Continuous discharge with no plasma channel

Spark Discharge

• Energy calculated from:

• E = ½ C V2

• Typical hydrocarbon (propane) E is 0.29 mJ

• For 100 pF person, voltage required is ~ 2kV

• Breakdown of air is 30 kV cm-1, so gap is 0.8 mm

• Quenching gap typically 2 – 3 mm

• Gap of 2 mm needed for ignition gives required voltage as 6 kV – not easy to achieve

Spark Discharge

• For hydrogen E is 0.017 mJ

• Breakdown strength for hydrogen 17.5 kV cm-1

• Assuming linearity, breakdown is 26.25 kV• Quenching gap is 0.69 mm• For 100 pF person, voltage required is 1810

volts• Easy to reach 2 kV on person• Cannot feel such small energy discharges• Higher risk of ignition of hydrogen than petrol

vapour

Brush Discharges

• Typically discharge from insulating plastic – cannot measure energy as capacitance cannot be measured

• Gibson and Harper determined “incendivity” using flat polyethylene sheets

• Brush discharge equivalent to about 4 mJ

• New work by Ackroyd shows “incendivity” greater than Gibson's work

• Higher incendivities with fluorinated polymers and thin layers on metal substrate

Corona Discharges

• Silent – usually continuous

• Tip radius determines corona or spark discharge

• Small tip radius gives corona rather than spark• Incendive to hydrogen – air mixtures• Atmospheric electrical activity:

– high field strength – starts corona from sharp edges

• Hydrogen vents known to ignite during frosty weather, rain, sleet and falling snow

• Assume hydrogen vents will always ignite

Diffusion Ignition

• Theory postulated by Wolański and Wójcicki

• High pressure ignition in shock tube

• Confirmed theory with confined shock tube

• No experimental work for open ignition

• Initial conditions high temperature for experiments

• No indication that atmospheric releases would be ignited by diffusion ignition

Sudden Adiabatic Compression

• Temperature rise when gas compressed adiabatically

• For compression volume ratio 10:1– theory pressure rise ratio 25.7– theory temperature rise 428 K

Adiabatic Compression

• Work by Cain indicates compression ignition occurs at about 1050K for H

2/O

2/He mixtures

• Relatively constant ignition temperature irrespective of pressure rise ratio starting at 300K

• Ratio of 80 needed in theory for adiabatic temperature rise from 300K to 1050K

• Much lower ratio needed by Cain ≈ 35 to 70

• Suggests another mechanism present

Hot Surface Ignition

• At high temperature– Oxidation generates heat

– Heat lost to surroundings

– If less lost than generated, chain reaction occurs

• Under turbulent conditions ignition occurs at lower temperatures

• Also ignition occurs at lower temperatures under shock conditions

Turbulence

• Neer suggests ignition speed rather than temperature – ignition under shock conditions needs lower

temperature than classical stationary conditions

• Bulewicz showed position and mode of heating affected ignition temperature

• Heated surface down – longer delay

• Impulsively heated plate – higher temperature

Discussion - 1

• No one mechanism explains all ignitions

• Potential for electrostatic ignition to occur– Demonstrated by some incidents

• Confined heated surfaces act as ignition sources– Unconfined hot surfaces – not well understood

• Joule-Thomson Effect needs high initial temperature

• Diffusion ignition only demonstrated in shock-tube apparatus

Discussion - 2

• Shock-tube theory and experiments for diffusion ignition appear non-specific to hydrogen– But, no other gases appear to exhibit spontaneous

ignition on release from high pressure

• Adiabatic compression requires confinement

• Difficult to separate diffusion ignition from adiabatic compression – both unlikely with discharge direct to atmosphere

Discussion - 3

• Electrostatic charging of pure gases negligible

• Particulates present can charge

• Corona known to be able to ignite hydrogen

• Possible erosion of metal of pipes – particles then able to charge

• Expansion increases temperature – lowers ignition energy

• Potential for corona to ignite more sensitive atmosphere

Conclusions - 1

• Hydrogen does not necessarily ignite when released at high pressure

• Compression ignition, Joule – Thomson expansion and diffusion ignition unlikely mechanisms for releases at ambient temperature

• Possible electrostatic charging is part of mechanism of ignition of high pressure releases

Conclusions - 2

• Mechanisms postulated in literature do not account for all ignitions and non-ignitions

• Possibility that ignitions of hydrogen are a combination of two or more postulated mechanisms

• Further work is required to establish conditions under which hydrogen release ignites – particularly electrostatic phenomena

• Stuart.Hawksworth@hls.gov.ukTel: +44 (0)1298218139Fax: +44 (0)1298218160

• Graham.Astbury@hsl.gov.uk Tel: +44(0)1298218145 Fax: +44(0)1298218160

Further Information