## 16 Applying the Scientific Method to Forensic Engineering

In a laboratory setting, it is usual to design experiments where the variable being studied is not obscured or complicated by other effects acting simul taneously. The variable is singled out to be free from other influences. Various experiments are then conducted to determine what occurs when the variable is changed. Numerous tests of the effects of changing the variable provide a statistical basis for concluding how the variable works, and predicting what will occur under other circumstances.

In this way, theoretically, any accident, failure, crime, or catastrophic event could be experimentally duplicated or reconstructed. The variables would simply be changed and combined until the "right" combination is found that faithfully reconstructs the event. When the actual event is experimentally duplicated, it might be said that the reconstruction of the failure event has been solved.

There are problems with this approach, however. Foremost is the fact that many accidents and failures are singular events. From considerations of cost and safety, the event can not be repeated over and over in different ways just so an engineer can play with the variables and make measurements.

It can be argued, however, that if there is a large body of observational evidence and facts about a particular accident or failure, this is a suitable substitute for direct experimental data. The premise is that only the correct reconstruction hypothesis will account for all of the observations, and also be consistent with accepted scientific laws and knowledge.

An analogous example is the determination of an algebraic equation from a plot of points on a Cartesian plane. The more data points there are on the graph, the better the curve fit will be. Inductively, a large number of data points with excellent correspondence to a certain curve or equation would be proof that the fitted curve was equal to the original function that generated the data.

Thus, the scientific method, as it applied to the reconstruction of accidents and failures, is as follows:

• a general working hypothesis is proposed based on "first cut" verified information.

• as more information is gathered, the original working hypothesis is modified to encompass the growing body of observations.

• after a certain time, the working hypothesis could be tested by using it to predict the presence of evidence that may not have been obvious or was overlooked during the initial information gathering effort.

A hypothesis is considered a complete reconstruction when the following are satisfied:

• the hypothesis accounts for all the verified observations.

• when possible, the hypothesis accurately predicts the existence of additional evidence not previously known.

• the hypothesis is consistent with accepted scientific principles, knowledge, and methodologies.

The scientific method as noted above is not without some shortcomings. The reality of some types of failures and accidents is that the event itself may destroy evidence about itself. The fatigue fracture that may have caused a drive shaft to fail, for example, may be rubbed away by friction with another part after the failure has occurred. Because of this, the fatigue fracture itself may not be directly observable. Or, perhaps the defective part responsible for the failure is lost or obscured from discovery in the accident debris. Clean up or emergency repair activities may also inadvertently destroy or obscure important evidence. In short, there can be observational gaps.

Using the previous graph analogy, this is like having areas of the graph with no data points, or few data points. Of course, if the data points are too few, perhaps several curves might be fitted to the available data. For example, two points not only determine a simple line, but also an infinite number of polynomial curves and transcendental functions.

Thus, it is possible that the available observations can be explained by several hypotheses. Gaps or paucity in the observational data may not allow a unique solution. For this reason, two qualified and otherwise forthright experts can sometimes proffer two conflicting reconstruction hypotheses, both equally consistent with the available data.

0 0