All good failure analysis investigations rely on the application of the scientific method. Nearly every time I give a presentation to claims personnel, attorneys, or other engineers as it relates to failure investigations, I discuss the scientific method. Even if you are not a technical expert, you need to understand the process of using “data” to support your claimed failure method. It is one thing to suggest a method of failure for an accident, but another to be able to support that suggested failure mechanism with data.
The figure below outlines the simplicity of the scientific method. The steps may appear to be easy, but one needs to be thorough as well. The steps are
- Identify the Problem: What are you investigating?
- Define the Problem: Make sure you are examining the event of interest and not defining it in a way that is not reasonable.
- Collect Data: Data collection can be as simple as taking a few photos and as complex as taking photos, testing, taking measurements, examining an exemplar, reviewing drawings, collecting process data, etc.
- Analyze the Data: Examine the data to understand the timeline associated with the event, determine what the system or component looked like in advance of the failure, etc. The investigator needs to understand how the system actually works before they can truly develop realistic hypotheses for failure modes.
- Hypothesis: Develop an “educated guess” as to what might have happened. You might actually come up with several hypotheses, which is not a problem.
- Test Hypothesis: This is a critical step in the investigation – step back, examine the data, the analysis of that data and determine if it is consistent with the hypotheses you have proposed for the event.
- This may include recreating the event through controlled tests. This may not always be possible due to safety concerns or prohibitively high costs.
- If the data does not fully support the hypotheses one can collect more data, analyze any new and/or existing data, and determine if the new data suggests any sort of modified hypotheses.
- If you are not satisfied the hypotheses is fully supported by the data collected, you need to consider – Do I need more data? Are there other failure modes that are better supported by the existing data? Do I need to revise my hypothesis?
Once you are satisfied that your hypothesis is fully supported by the data and your analysis of that data, you are done. This is the point where you should be able to say “this is what happened.”
Now, if this is purely a forensic investigation, you write a report as requested and your work is done. But, if this is for an industrial client or situation, you may very well have to develop recommendations for corrective action as well. In that situation, you also need to consider
- Are other hazards created by your suggested corrective action?
- Does it eliminate this and other potential failure modes?
- If a manufacturing defect, what steps need to be taken to prevent this problem from happening again in the future.
- If a design defect, you will need to evaluate the overall design
- Completing a failure modes effects analysis (FMEA) as part of design documentation.
Hopefully, this sheds some light on the importance of properly applying the scientific method.