Software helps to optimise the next generation of aircraft
Making trade-offs is fundamental to being an engineer. Taking different attributes and weighing them against each other to reach a solution is fundamental to the profession. However, there are limits to how much can be quantatively compared before more senior engineers need to make a judgment call that is based as much on experience as calculation.
This approach, however, has obvious limitations. First, it can rely on experience and instinct as much as it does anything else. It also ties up that knowledge, ability and judgement in a single individual, which has limitations in terms of access, availability and security. After all, what happens if that person retires, changes jobs or is hit by the proverbial bus?
It was the ability to take many different parameters and accurately evaluate them that was the driver for software produced by engineering consultancy Fraser-Nash. Established in the 1970s, the company has a historic pedigree in mechanical engineering that can harness the power of computer simulation.
"A couple of years ago we were doing work for the renewables sector looking at wind farm optimisation layout," says Glyn Norris, civil aerospace business manager at Frazer-Nash. "We developed a software tool that enabled trade-off studies to be preformed. This took in to account the aerodynamic and structural performance, but also cost and reliability. What the tools did was allow some of the knowledge typically locked up in senior engineer's judgement to be captured.
"Humans are pretty good at trading off two or three inputs in our head. But, once you get more than five, it becomes difficult to visualise. This tool will help visualise complex problems that might have 20 or 30 different parameters."
Fraser Nash's technical manager of the modelling and simulation group who did the work on wind farm optimisation had previously worked at Airbus and knew the challenges it had in solving very complex, multidimensional problems.
The tool essentially takes vast reams of data and then presents them to a user in a much simpler x-y-z graph. This allows them to be much more usefully interrogated and offer the chance to make engineering decisions and judgements on designs, trade-offs or optimisations.
It is a way of taking key parameters and seeing how those interact with each other. This is normally up to a chief engineer or decision maker to weigh up and pick the optimum point within the graph, which in the past was done largely by experience.
David Marles, aerospace engineer, modelling and simulation, Airbus says: "Mass is always the main driver. It is the load experience on the aircraft that drives weight, and that basically drives performance. Ultimately we are trying to increase the performance of the aircraft and give a minimum mass. Landing gear may weigh 5 tonnes and if you can make it 4.9 tonnes, that is a fantastic result.
"The tool allows us to spot and realise which factors matter and which don't, much earlier in the design process. We call this 'parameter sensitivity'. It allows us to play around with a number of key parameters and say this is key, or the interaction between another two parameters is really important."
This is the benefit that Airbus has successfully applied to part of the design of the landing gear of the A350. It has been able to use existing processes such as the load-generation process of the landing gear, and automate them so that the tool will automatically be updated.
The software is a multidimensional design space, which basically translates to having 20 or 30 key parameters. This is just for the landing gear and can be the wheelbase or wheel track, fore-aft frequency or the main bearing. The tool allows these to be 'wrapped' together so that the response surface of a graph is a multidimensional response surface and that is what is plotted.
"It is plotting big, polynomial equations in the order of hundreds of terms," says Norris. "The main parameters shown on the graph are fore-aft frequency and main landing gear bearing friction. But, behind that there are essentially another 20 parameters that the user can vary in real time. The tool will then recompute the response surface based on changing parameters that you are not actually plotting. So if you are changing wheel radius that could have an effect on your plot on bearing friction and fore-aft frequency. However, this is not obvious and you wouldn't know this unless the tool gathers that information together and allows you to visualise it."
The function of the polynomial equation is immensely complex and it can take extremely complicated factors and plot a meaningful response surface that links them all. Each time the user varies something it uses this equation to recalculate the response surface. In different parts of that surface, something might vary, but others might not.
"Every time you change something it is all those 20 or 30 parameters interacting with one another," says Norris. "They might be coupled within the design space. They might be coupled within one part of the design space, but not the other. So that is why every time you play with something, it uses a single equation – albeit an enormous one – that then plots a new response surface and that is what you are visualising and what the user sees in real time."
This ability means that Airbus can explore many more possibilities, more openly and easily than ever before. A chief engineer might say, 'what if we do this?', or 'can you move that?', and the output will update. The chief engineer has essentially been doing all of this in his head.
Now, however, the tool allows a relatively junior person to use it and find what the real key parameters are and what the drivers should be. He or she can ask the questions in real time, play around with parameters and quickly become confident they understand what is sensitive and what is not.
The other aspect of the tool is that it can perform a Monte Carlo analysis (a class of computational algorithms that rely on repeated random sampling to obtain numerical results). Airbus has to set the load that they think an aircraft will experience and be confident that by changing some parameters later on, it is not going to exceed that load.
Designing an aircraft is obviously a massive activity involving thousands of stress engineers and people designing components, all of whom need certain parameters to be set early in the process.
If the load is too high it means the aircraft will come out too heavy, if its too low and they need to move a landing gear back slightly, it will mean a redesign of the entire section. The Monte Carlo analysis gives confidence that the load level set will ultimately be designed to.
The tool can take various parameters from other programs. For example, it could read in key outputs from a FEA analysis such as mechanical property. However, what comes out of an FEA analysis in a single design run would be a load that gets read in the tool, allowing Airbus to see how stress varies when all these parameters are varied.
The tool can be used anywhere that has multiple parameters that interact with one another. And it is not limited to engineering. You could use it for cost or financial modelling, anything used to visualise a multidimensional problem.
It is hoped the tool can be deployed more widely on Airbus projects to see links that perhaps has, so far, not been obvious. "By changing the landing gear design, you change the loads that the fuselage will experience," says Norris. "So by changing the damping on the landing gear when it lands you are going to change the force that the landing gear puts into the fuselage, thus changing the moments and peak moments the fuselage will experience."
This material is protected by MA Business copyright
see Terms and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the