Review of Specific Calculation Methods for the Rijnmond Safety Study

Summary

Cremer and Warner (C&W) were commissioned to conduct a study on the risk of six selected industrial installations in the Rijnmond area. Battelle-Institute e.V. was requested to prepare a counter expertise on this study which was expected in particular to consist of a critical review of C&W’s models for release, dispersion, and explosion. In addition, calculations on selected cases were to be made to show possible differences in the consequence analysis which might arise from different model approaches.

Our engagement in this risk study boils down to two major observations: first, in general we are in full agreement with C&W’s choice and application of most models to describe the consequences of unwanted events. Points of major disagreement are discussed below. Second, from some of these points of disagreements we conclude that the overall uncertainties of the results have been underestimated by C&W.

Points of disagreement:

• For unknown physical or statistical reasons C&Wexcluded two events which might make a significant contribution to the overall risk: ‘simultaneous failure of tank and bund of ACN or LNG storage tank’ and ‘tank rupture due to polymerisation of ACN’.

• Two-phase flow release should be calculated according to the Moody theory which would in many cases result in lower consequences.

• The uncertainties which result from the initial temperature of an ammonia cloud and from the initial dilution of material with air following a rupture of a pressurised tank have been underestimated or not analysed.

• The calculations of dense gas dispersion are reliable for dilutions down to lower flammability limits of explosive gases but have not been validated experimentally for dilutions down to interesting limits of toxic concentrations. In the risk study the uncertainties for this dilution range affect only the worst case chlorine scenario, which is the tank rupture.

• C&Wshould have considered and included in detail the effects of the local topography which critically influence the spreading of liquid on the ground and the spreading of heavy gas clouds. Both effects reduce in most cases the consequences.

• C&Wshould have estimated the chances of people to escape unharmed a toxic vapour cloud when they smell it (warning time).

• C&Wshould have modelled the explosion of a pancake-shaped heavy vapour cloud more realistically. The effects on the results are not significant for the investigated cases if C&W’s consequence criterion is applied: all people indoors and within the 0.3 bar overpressure contour are killed. If casualties from broken window glass were considered, C&W’s explosion model would greatly overestimate the consequences.