Analysis of the failure of a mechanical device requires investigation of the circumstances surrounding the failure, inspection of the mechanical device, interpretation of the information gathered during that inspection, and reverse engineering of the incident, with the end goal of reaching a conclusion as to the cause and mechanism of the failure. Often during the process of post-failure forensic analysis, hypotheses are posited to explain the failure. As an investigative engineer, the formulation of hypotheses to explain the failure often occurs after the initial investigation is conducted and the engineer has a general understanding of the conditions present at the time of the failure. If hypotheses are developed before the investigation begins, there may be too many hypotheses to consider. The hypotheses are then tested to determine their strength and false hypotheses are discarded. True hypotheses are assembled into an explanation of the failure. There is often insufficient physical evidence available to establish the exact configuration of a device after a failure. Regardless of the reason why the failed part is not present, the parties involved may dictate a need to determine the cause of the failure. When the device or failed part is not available, other techniques in addition to a physical test of the failed part must be used to test hypotheses and evaluate the cause of the failure. Examples of such techniques are (1) using information on the performance of the machine to determine possible failure mechanisms and progressions and (2) reported accounts of the machine and its environment by operators or surrounding witnesses. The current study investigates the relationship of witness observations to the analysis of example mechanical failures. Various aspects of how witnesses observe and process their observations are discussed as well as the caution needed in considering witness observations due to incomplete observations or potential benefit to the witness. The use of Failure Mode Effects Analysis in determining the cause of a failure is discussed, most notably in an extension of this technique to analyzing operator error. One particular real world example where the allegedly failed part was not available throughout the investigation is discussed in detail. The results of this study show that witness accounts of a failure should be considered by the engineer when determining the cause of a failure and that ignoring witness accounts can lead to unsubstantiated and possibly incorrect conclusions.
