Simulation Fidelity - and Uncertainty
CM and others,
Perhaps I might be allowed a short observation here.
CM's post about the complexity of the simulations now being used by the US and the UK for F-35 is illuminating. There are literally hundreds of parameters being fed into these programmes, and literally hundreds of different dynamic and tactical situations being investigated. Plus hundreds of different starting positions.
As well as these, there are literally thousands of assumptions being fed into these things. Many of them centre on the attributes of the 'foe' or 'adversary' weapon systems, plus their tactics. Some are intelligence fed, some are pure estimates.
Some of these simulations run without pilot input, many have 'pilot in the loop'. Some have other personnel 'operating' various friendly or 'foe' synthetic systems.
All these inputs have varying degrees of sensitivity. To exaggerate to make my point, I think it's a safe bet that your own missile's warhead effectiveness is a bit more significant than the thickness of your enemy's cockpit framing. But a lot of these inputs have a high degree of sensitivity.
So what? The 'so what' has to be uncertainty. Your aircraft might not achieve some of the parameters you initially assumed. (Such as F-35 sustained turn rate). Or your assumptions about enemy equipment or their tactics might be wrong. So how reliable are these simulations? Running them thousands of times, and feeding the results into what I understand are called 'Monte Carlo' algorithms, is claimed to reduce the uncertainties, but I always worried about the simple certainty of errors in the simulations. Plainly put, given the size of these things, the chance of incorrect code being entered has to be very high.
I had personal experience of a Farnborough simulation that was using wholly incorrect input data, tactics and assumptions. Plus some bad code. These had gone undetected for three years, and the results from it were influencing MoD decisions. The only reason they were found was that someone who had experienced similar issues in the US advised us to ask the same questions of the UK team doing the sim work.
Yes, these simulations can be used to estimate changes in combat effectiveness, and to derive estimates of loss ratios. But that's all they are. Estimates. Really numeric, complicated, high end guesses. But for aircraft like F-35, where we can't really send four of them up against four real Shenyang superjets, and see who lives, simulations are all the people driving these programmes have.
I'm not 'pro' or 'anti' sim. They are an essential tool, and very valuable if used well. In many areas, they are all the decision makers have. But I do worry that some people think that the more complex a simulation is, the better it is. I'd be more worried about errors.
As far as F-35 goes, (and I'm not a pilot, so this is just my own opinion), the direction of travel that the F-35 has followed is the right one. Outright aerodynamic performance is no longer the dominant attribute required for effective combat - it's the ability of the platform to sense, gather information, use it, and deliver an effective weapon, while denying as much of that information as possible to the foe. The Sea Harrier FA2 was a very, very effective fighter in the 90s despite its very limited airframe capability, due to a great radar, good radar/missile integration, and a great weapon. Same went for the Tornado F3, once it had JTIDS and AMRAAM.
The F-35 has, from day one, gone for a balance of platform and systems capabilities that is not the same as that pursued for legacy aircraft. Is that right or wrong? I don't know, and I'm certainly not qualified to judge. Perhaps this thread could calm down a bit if a few more of us admitted that.
Best regards as ever to those who have to make the calls for real in the air and on the ground
Engines