Probabilistic Factors in Emc of Complex or Ill-defined Systems: Experience from Rf Hazards Analysis

As with most other EMC standards, standards for the avoidance of radio frequency ignition hazards, such as BS6656 [1], have an apparently-deterministic set of go/no-go thresholds to assess whether a hazard is present in a given circumstance. These standards are in fact based on intuitive consensus judgements by members of a committee which discriminate between concatenated sets of circumstances which are significantly probable, and those which are negligibly probable. The ‘significantly probable’ scenarios are then effectively assigned a probability of unity and the ‘negligibly probable’ scenarios are assigned a probability of zero. This ‘binary’ evaluation of probabilities is arguably rather unsatisfactory, being based heavily on guesswork. It incorporates the obvious risk that an unusual set of circumstances that has been assigned a probability of zero may yet still occur, leading to potentially catastrophic consequences. More commonly, it is likely that the assignment of unit probability to a very wide range of scenarios will lead to over-engineering of protective measures which might compromise other aspects of the system, or simply be an extravagance in relation to other safety issues.