During the writing of the book Twice As Far, I knew that this software had been utilized.  However, after reviewing this report, it was determined that the program was of no value in this fire investigation for the reasons shown here.  Therefore, it was merely a ‘red herring’ that would have added unnecessary layers to the story.  Too much reliability and accuracy is applied to software programs in the belief by many that because it is a computer program, it must be reliable and accurate.  However, any computer is run by a program, and that program is written by humans.  In this case, this program requires the input of measurements and details that have been determined through questionable circumstances that themselves are unlikely to be reliable and accurate.

ASSESSMENT OF REPORT

As part of the investigative process, the Transportation Safety Board utilized Smartfire, a Computational Fluid Dynamics software program that could determine fire properties and fire progression in a structure.  They looked to Fire Safety Engineering Group at the University of Greenwich in the United Kingdom as it had been used in that country on several major fires.  However, the program was not compatible with the drafting data of the aircraft that had been created for this investigation.  The TSB had used North America’s standard CAD program while the British standard was different.  Because of the volume of CAD data, it was simpler to rewrite the program to accommodate the TSB.  This was advantageous to the FSEG company as they wanted to break into the North American market with their Smartfire software.  Because they did not wish to absorb the cost of rewriting the program, the TSB agreed to pay for the rewrite.  This would allow the program to be utilized for this investigation plus any other fire investigation in North America. 

Eventually, the software was used to produce a fire scenario, and their report is available on the web at:

 https://www.fire.tc.faa.gov/2004Conference/files/materials/F.Jia,M.Patel&%20E.Galea_Simulating_the_Swissair_Flight_111_in-flight_fire.pdf

The company and their results were also mentioned in the final TSB report for this crash as being highly successful.

A review of their report at the above site provided an insight into some potential problems with the project.  Had this matter ever gone to court, it is quite possible that an examination of the mater would have provided numerous arguments.

Since the FSEG company was eager to break into the North American market, they needed a successful experience with a major investigative group in a prominent and important investigation.  The TSB’s Swissair 111 investigation met those requirements.  If they could then be successful, the TSB would be their route to that North American market.

Several avenues for argument could be raised in court.

1.      With the TSB paying for the rewriting of the software, the company now became indebted to the TSB.  After that amount of expense, a failure would have been the death knell for the software in North America.

2.      A line in their report is as follows: ‘It is worth noting that CFD techniques are not error free;’.  The paragraph continues further with ‘Thus it is important to understand the limitations of such simulations.’

3.      A further line: ‘the key goal of the study undertaken was to validate and/or determine the probability that a scenario is “likely” . . . . The rigours of such a complex simulation are quite demanding . . . .’

4.      Yet another line: ‘. . . software was used to predict the “possible” behaviour of airflow as well as the spread of fire and smoke within SR 111.’

5.      Now this line: ‘This assisted investigators in assessing fire dynamics for cause and origin determination.'

This last line is important because of a further area of the report.  In truth, the cause and origin determination was provided to the company as a start-point for their fire modelling scenarios.

6.      ‘SMARTFIRE thus became the first CFD fire model to be used in an air crash investigation. The official TSB report acknowledged the technology’s successful debut in ‘helping to develop better insight into and understanding of the fire’.

7.      ‘A research version of the SMARTFIRE V3.0 [2,3] software was used to perform the fire simulations in this study.’

This, therefore, was not a standard, well established software version, but an experimental version.

8.      ‘Within SMARTFIRE the user can define a range of scalar variables, which can be used to represent the transport of products such as toxic gases and smoke.’

As I understand this statement, the TSB was able determine what values could be used in the program, and as such, they could adjust those variables to arrive at a predefined conclusion if they so desired.

9.      ‘The fire simulations in this study were performed using a special version of the SMARTFIRE V3.0 software with features developed particularly for this project.'

         'Due to the highly complex nature of the geometry in the aircraft, some of “small” features and components had to be either approximated or removed from the geometry to make the SMARTFIRE computer files manageable. The approach to simplify complex geometry is a common CFD practice.’

Again, this was not an off-the-shelf standard version, but a special version.  There would have been no actual testing to determine the accuracy of the results.  For these results to be accepted in a court of law in this country, the results would have to be verifiable.  Since this was a one and only experience, there was no way to verify its accuracy.

10.    ‘Several simulations were performed to ensure that the airflow pattern observed in the model, with the aid of minor changes to the inlet conditions, was similar to that observed in actual aircraft during airflow flight tests.’

         ‘The TSB provided estimated MD-11 cockpit airflow values (e.g., conditioned air supply rates into and out of the cockpit) that were obtained from the Boeing Company for the flight condition that best represented the occurrences. These airflow values were used to set the initial airflow conditions at the various flow components in the CFD model. The model was then run and the velocity vectors and flow rates were studied at different locations within the virtual aircraft. Some adjustments were made to the flow rates until the model’s velocity vector values closely matched those calculated from the flight test video recordings. The airflow patterns in the computer model were then compared to the smoke release flow patterns recorded on the flight test video recordings. The airflow patterns in the computer model were determined to be consistent with those observed experimentally.

The problem with these two statements is that the airflow flight test provided test results in a non-fire scenario.  Obviously, we were unable to determine actual airflow results consistent with a fire scenario without creating a similar fire in the aircraft.  But how can these non-fire scenario results be considered as accurate for a real fire scenario.  Any experienced fire fighter will agree that as soon as the fire begins, the airflows will change dramatically and unpredictably.

11.    ‘The fire ignition site, as provided by TSB, was prescribed with the aid of a volumetric heat source. The source was adjusted so as to sustain the combustion of the material over a short duration. This was necessary to allow the material to reach a critical temperature to allow flame propagation. In the simulation, the fire spread after initial ignition, was only allowed to spread once the neighbouring cell reached the ignition criterion.  This allowed the heat release to be predicted by the amount of surface that was ignited as a result of the flame spread, if the ignition criteria were met. This approach eliminates use of the more complex route of predicting flame propagation via chemical reaction.’

Several things are a problem with this statement.  First, the TSB provided what they thought was the ignition point.  In other words, it was the location determined by Jim Foot with his wire placement of the short-circuited wires.  How he came to these conclusions and his lack of expertise to do so has been described in the book.  His opinion of the source of the fire would never be presented in a court of law in this country as he had no qualifications to make that determination.  Next, the source point was adjusted to provide the required outcome, and fire spread was allowed only when the burning area reached a certain ignition criterion, or temperature.  Yet, in the FAA burn tests, the temperatures failed to reach anything of significance.  The flame slowly progressed along the Mylar material without providing any significant heat.  In the third test, fire spread was due to sparks from short-circuited wires.  There is no indication that this was included in the scenario, and how can anyone or any software predict that outcome.  Furthermore, some of the materials were likely damp due to condensation on the cold surfaces of water vapour exhaled by the passengers.  To what amount is not known, and it seems this was not included in the computer calculations. 

12.    ‘The main scenario considered for simulation was that the fire started within the confines of a relatively small area above the right rear cockpit ceiling, just forward of the cockpit rear wall near STA 383. The TSB determined that it was most likely that the fire started from a wire arcing event that ignited the nearby MPET-covered insulation blankets. These MPET-covered insulation blankets are easily ignited and were prevalent in the area. Of all the wires and cables that were located in the area of interest, the only arc-damaged wire that could be positioned in that area with relative accuracy were a pair of In-Flight Entertainment Network (IFEN) Power Supply Unit (PSU) cables. Although it is possible that other wires from this localised area that were not recovered might also have arced, the only arcing event that is known to have occurred within that area is the forward arc on TSB Exhibit 1-3791, located just forward of STA 383. The TSB requested that this location be used as the fire initiation site.

          A heat release rate, to sustain the fire at the fire ignition site, was prescribed, as CFD simulations cannot predict arcing type ignition sources. To achieve this, it was necessary to ascertain what fire load was required to sufficiently heat the MPET in a small-localised area, so as to create a small self-propagating fire at the initiation site. This method of igniting the fire was agreed to by the TSB.’

There we have it.  A predetermined source as directed by the TSB based on a scenario that they had decided upon.  The software, utilizing this source point and the TSB’s special parameters, was then used to confirm the source point.  Just how objective is that.  This would never stand the test of court.