Measuring Uncertainty and Conservatism in Simplified Blast Models

Engineers performing assessments or designs of structures to blast effects usually begin by computing blast loads for the explosive event under consideration. In almost all cases, this computation involves the use of one of a number of simplified engineering level tools, including PC codes as well as lookup curves. Unfortunately, the tools only provide a deterministic prediction, without giving engineers any sense of the inherent uncertainties in the loading. To further complicate matters, each technical community seems to prefer one particular tool over the others, which often leads to discrepancies when two such communities with different models are required to interact or cooperate on a single project.

This paper compares blast predictions (both reflected and incident loads, both for pressure and impulse, and both positive and negative phases) from a number of popular simplified models, including BlastX, ConWep, SHOCK, to a wide range of test data spanning three decades and comprising a total of nearly 300 individual measurements. All of these were taken at low heights above the ground, some on small cubicles and others on larger buildings. The comparison is restricted to a scaled range of 3ñ100 ft/lb1/3, a regime where variations in the details of the test arrangement should be more or less irrelevant. The results provide quantitative assessment of the inherent uncertainty in any blast prediction tool, even for these relatively simple geometric conditions. Relative comparison between models and test data support the determination of biases and trends within each of the models. Using these results, it is possible to quantify the bias and uncertainty in the models for each of the load metrics, and also to compare uncertainties between the various metrics.