Sprinting an ADR Sport 2

[andylaurence]

Yes, there's lots on eBay but what's the resolution like? Can I do a whole car accurately? If so, it's worth £35 to give it a punt, although the software is comparatively quite expensive if I want to export the data. I could do a scan of the car without bodywork, then a further scan with the bodywork to get a good approximation of the whole car with the internal flows modelled slightly more accurately. There's a lot of basic work that needs to go in before I worry about CFD though.

[RicME85]

http://www.youtube.com/watch?v=ofZWv6AczM0

[stez90]

I don't think you need such precision right now. I was thinking about a simplified model, with no detailed internals (maybe only major ducts and "box" shaped placeholders for engine and biggest components) and filled rotating wheels. At the moment you need qualitative datas and trends (i.e. this.. is better than...). For reliable quantitative datas you would need superior knowledge about solver, meshing and very big computational power.

A good starting point would be something like this, with a realistic exterior shape "not an extruded plane with sharp edges", filled wheels and not much more.
http://andrecars.com/wp-content/uploads ... -cfd-1.png
http://andrecars.com/wp-content/uploads ... -cfd-2.png

[andylaurence]

Agreed. There's always Amazon's cloud processing for doing more in-depth analysis.

What's been done with the Inverter is exactly what I'd like to get onto with mine. I'd also like that lovely shade of blue on the bottom and pockets of deep orange on the top! Hopefully, I'll have a good size diffuser by the weekend...

[andylaurence]

I'd competed at the Abingdon Carnival before but in the AutoSolo several years ago. This was my first time at the sprint. The event consists of two sprint courses and the results take the fastest run from each course, which totals around 90-100 seconds. The entries are split into two groups with one group running the Abingdon course and the other running the Bentley course with a switchover after lunch.

I was first on the Bentley course and the convoy run was not a good start. The course was simple; a straight, a 90-right, a hairpin left, 90 right, a straight and a right-left-right chicane before the finish. It was all about power and mechanical grip. I found out on the convoy run that steering lock was important too. I didn't have enough for the last left-hander in the middle of the chicane. As a result, I ran out onto the grass on the exit. Some creative driving would be required.

In first practice, I took it steady to set a reasonable time, but at the chicane tried stamping on the throttle to get some oversteer around the corner. At that point, I realised I was in 2nd gear and the car just accelerated and understeered two wheels off. In second practice, it was much the same, but I gave it beans in 1st gear rather than 2nd. The grip at Abingdon is good and 999cc isn't enough to generate oversteer at will. I went four wheels off in an attempt to emulate a Flymo. I also learned that I was braking far too early for the first corner. More to come in the timed runs!

In the first timed run, I braked too early at the first corner, then out-braked myself at the third corner, diving through the cones and rejoining a couple of cones later. Frustratingly, I managed to get around the chicane with a bit of oversteer helping to turn the car. Despite the off, I was fastest, but that run wouldn't count with four wheels off. On the final timed run, I made sure I didn't make any mistakes and put in a good run to lead the class.

Let's look at the data for the Bentley course. Off the line, I'm up into 4th gear at 102mph before the first corner. 0-100mph takes 10.05 seconds - not bad for a 999cc car! The G-meter shows braking is variable and peaks at about 1.1G, reducing slightly as I align the car for the left kink just before the turn. I made a pretty good job of that on cold tyres and the nice upwards slew shows I trail braked into the corner. The smooth curve shows I got on the throttle right at the apex, but the bunching of plots in the top-left shows some fluctuation on the exit. I expect this is a little clumsily corrected oversteer on the exit, most likely as a result of giving too much throttle. You can see that we're up over 0.6G acceleration here, which suggests full throttle before the corner exit.



Moving onto turn two, I reached 74mph before jumping on the brakes. 1.1G of braking compared to 1.3G in cornering suggests I could have braked later. Also, the concave curve from braking at the bottom to cornering on the right suggests I didn't fully utilise the traction circle and could have carried more speed in on the brakes. The number of points on the right is because I balanced the car on the throttle around the long left-handed hairpin. There was clearly a bit of oversteer on the exit as there's another hiccup on the curve in the top right.



67mph achieved before braking for the right-hander and again there's not enough pressing of the brake. There was a new master cylinder fitted for this event and the different pedal ratios means the balance isn't quite right and the feel is very different. Again, I've braked too early as I've rolled off the brakes quite quickly on turn-in, resulting in the concave graph. True to form, I've also stamped on the throttle on the exit and got a mild case of oversteer on the exit! It's worth noting that as I'm still on the edge of the traction circle, this is unlikely to have cost much time.

Heading down the straight to the final chicane, I reach 94mph before jumping on the brakes. Again, I'm braking at 1.1G, which probably means my brake balance is pretty far out and I'm feeling a lockup and not braking any harder. There's a jink left under braking before trail-braking into the right-hander. The lateral G peaks at 1.5G in the right-hander, at which point I'm off the brakes before getting back on the brakes lightly for the left-hander in the middle of the chicane. I kept on the brakes past the apex on the left before stamping on the throttle and this is where the G meter gets all messy. The trail braking past the apex was to keep the nose tucked in and keep make it round, but this meant losing speed on the exit of the corner and not making the most of the traction circle. Needs must! The final right-hander was also scrappy. Definitely lots of time to make up next year.



Moving onto the Abingdon course after lunch, this was a much faster course. From the line, accelerate through the first left, upshifting along the way and into 6th down the bumpy perimeter road. Turn in flat out to the right-hander, then pick a braking point for the multi-apex left-right-left chicane, then accelerate down the straight and scrub off a little speed into the ultra-fast left-hander before the finish. It's a tricky course as the corners are blind with lots of false apexes on the approach to the chicane and bumps the whole way round. I'll definitely walk the course on the Friday next year as it'll take a lot of learning.

In first practice, I was tentative into the chicane and the last corner. I even lifted on the flat right-hander on the approach to the chicane. In second practice, I was again too slow into the chicane, and still wondering if I should keep flat over the right into it. The apex of the final corner is earlier than it looks! Let's move onto the data from the last timed run.

Away from the line, the left-hander slows acceleration down and 0-100mph takes 10.9 seconds as a result. You can see the gearchanges through the corners with the dips as the lateral G increases and decreases. You can also see the upshift into 6th happened after turning into the right-hander, hitting 119mph in the process. It was a scrappy turn-in, not helped by the bumps and peak braking was only 0.9G, so much more to come there.



There's also some indecision about the direction I'm taking under braking, again the bumps causing me problems. Turning into the left-hander, I'm not using the brakes as hard as I could and the cluster of points around that area shows that I took a couple of bites at the corner. Clearly, I need to walk the course! I carried on braking lightly into the right-hander in the chicane, pulling just over 1.2G in the corner. It looks like there was again a bit of oversteer as I got on the throttle towards the final left of the chicane and what looks like a bit more as I turned left.



I took two more gears up to 4th down the straight, hitting 113mph before braking at 1G and dropping into 3rd. I took two bites at the corner, suggesting turn-in oversteer. The G plot moves about all over the place, so it was pretty scrappy, but part of that will be down to the bumpiness. Clearly, the car doesn't ride the bumps as well as it should, so there's something to delve deeper into there. Interestingly, there was a peak of 1.4G during a moment of no longitudinal acceleration. A short burst of acceleration took me over the finish line at 94mph.



Here's the video to go with it.

[andylaurence]

After the unfamiliar lines of Abingdon, it was back to Castle Combe a fortnight later for the familiar lines of my home circuit at the Dick Mayo Sprint. This time, the start line had been moved to the pit lane rather than the race start line as has been the case in previous years. This meant the winner of each class would get a class record to go with it. The day started wet with the rain falling on the way to the circuit, but it stopped almost immediately. By the time the drivers' briefing was over, the sun was out and by the time I was on track, it was definitely slicks weather.

So slicks it was, but it wasn't half slippy! Despite a tentative disposition given it was first practice, still damp and I was on slicks, the lack of grip caught me out at Quarry. The data shows I barely broke 1G through Quarry. The tyres warmed up through the lap though, and I was back to 1.4G through Bobbies. Second practice was dry and I pushed on to lop a few seconds off. I was comfortably fastest at this point, so I was confident for the afternoon. The first run was quicker still and I thought that was a good run. Clive Wooster and Luke Trotman had both gone quicker than me. I needed a big improvement to beat Clive and I didn't think I could, so I was quite pleased when my last run was seconds quicker and enough to win the class by four tenths.

So, how did it happen? Off the line, I made a good enough start and made a point of going flat over Avon Rise. You can see the lateral G increases whilst the car is still accelerating and 1.2G laterally was generated over Avon Rise. It's actually quite a corner! Coming off the rise, I moved straight to the brakes without completely straightening the car.



Turning into Quarry, I took a couple of bites at the corner as I tried to slow the car and missed the apex. Peaks of 1.5G show the car had good grip and show how much better the surface is than Abingdon where the cornering forces were lower, despite the much warmer weather. The exit was clearly not too smooth, but I made good use of the traction available. In the panic to slow the car for the corner, I'd not made enough downshifts and was still in 4th gear, where 2nd would be more appropriate.



Braking for The Esses was too light and I released the brakes partially before turning in. This was because I needed to keep the speed up as I'd realised my mistake. There's more speed that could be carried into and through the right-hander, whilst the left-hander was pretty much what the car could do. I was off the throttle during the transition from right to left, but the data suggests I could carry more speed and trail brake to between the two apices. I was clearly heavy on the throttle coming out of The Esses as there's a two-stage horizontal trend line on the G-circle.



Through Old Paddock, it's clear to see that I was flat out, despite the strong wind coming from the right side of the car on turn-in and the opposite side on the exit. This was causing oversteer on the way in and massive understeer on the exit. I unexpectedly came through the corner on opposite lock with two wheels on the grass. The marshal on the post said I looked like I was about to crash. I didn't disagree, but I was still flat!



Tower is the corner I never get right and, as usual, I've braked too early, rolled off the brakes and taken a couple of bites into the corner. What's striking is how early I got on the throttle, most likely because I realised I'd braked too early before I'd reached the turn-in point. The maximum lateral G was seen with the foot on the throttle, so there's lots of time to be found here.



Into Bobbies, the braking is as strong as anywhere on the circuit, most likely because of the uphill approach to the corner. Good use of the tyres made here, but I was clearly on the throttle before turning into the second apex, although there appears to be a lift on the transition. The finish is on the exit of Westway, which is flat, hence the horizontal trend line to the left of the graph, peaking at about 0.9G. Lots of time to be lopped off next year!



Here's the video to go with it.

[Blanchimont]

Nice graphs!

I'm really wondering about the lateral G-forces being higher than the longitudinal ones.

Has this always been the case with this car since you measure the accelerations?
What was the highest acceleration under braking you recorded so far?

[andylaurence]

The graphs come straight out of Xoomcentre. Only the words are mine. As for the longitudinal G, there's a couple of reasons for that. First and foremost, I've not yet got the brake balance sorted with the new master cylinder fitted just before Abingdon. Secondly, I've not been pushing hard enough on the brakes and have not found the limit. It's a confidence thing - Abingdon was new to me and Combe was wet/dry, so pushing hard wasn't really on the cards given the absence of run-off. Bear in mind that I only get 4 laps in a day; two to learn the track and/or test things and two laps that count. Test braking to work out the brake balance can only be done in practice and preferably not in the wet!

[andylaurence]

It was all or nothing at Clay Pigeon. To keep my championship hopes alive, I needed a class win and a new record, both by a second. With only a Mini and a Seat Marbella in class, the win was a formality, provided I finished. The record was also pretty likely, given that I'd broken it earlier in the year in near freezing conditions with a tentative run after the sticky throttle I'd had that day.

The weather was scorching and with the 1.75 lap format and no straights, keeping the tyres cool was going to be a problem! Straight out of the box, I put in my fastest ever run with a 73.37. That was several seconds inside the record. All I had to do was the same in the afternoon. In second practice, I dropped down to a 72.42 and hopes were high that I could do a 71 in the afternoon.

When the afternoon came around, I was perplexed by a 74.03. Still, that was maximum points and there was a chance to do it properly in the second run. In true fashion, I went a tenth quicker than second practice in the first lap and pushed it through the chicane more than I had done all day. A 71 was looking probable. The extra speed through the chicane caught me out at the next corner though and I spun to the inside under braking. Still, that leaves just one ASWMC record to get; Llandow.

I'd been well under the record at Llandow during testing in 2012, but never had a dry run in competition to prove it. The day looked dry. In first practice, I made a solid start. In second practice, I was 0.99 clear of Clive Wooster in his Radical SR4 and dipped under the record. If I could maintain the gap and lop a few tenths off in the afternoon, that would be maximum points and enough to keep the championship alive.

Over the line, I heard a clunk. On investigation, it appeared the chain had gone. I whipped out my spare and got to work. Once the chain was out, I realised the real situation. The sprocket was missing from the chain tensioner and was clearly near the track somewhere around the finish. Chance of finding it? Nil! Day over and with it any chance of the championship. Congratulations to Ian Parr on his championship win.



When I got home and took it all apart, it turned out the bolt that holds the idler sprocket axle in place had sheared as a result of a stress fracture at the point of the split in the sleeve that acts as the axle. I swiftly ordered a new sprocket, but not knowing the size of the original, there was a certain amount of guesswork. My guess was not as educated as I had hoped and the sprocket needed an extra tooth or two. I was not going to make Curborough in the ADR, but thanks to a very nice man, I was able to play in a Dax Rush with a 'Busa engine and a turbo on road tyres. Awesome fun!

[flynfrog]

ouch looks like that bolt has been cracked for a while.

[andylaurence]

Yep. I don't know how long as I've never crack tested it. The part will be engineered out over the winter, otherwise I'd be lifing the part.

[KeithYoung]

What material was that made of? If you're limited by size consider something more resilient to fatigue, or keep a close eye on that part and cycle it out over time with a replacement.

By the way, thanks for the bump. I saw all the data above and I will be looking at them shortly!

[andylaurence]

I'm not sure of the material spec - it's the original item I had with the car that I've never crack tested. It's been running in the car for 3 years of my ownership, so if it's retained, I'll life the part at less than that.

You're welcome to have a thumb through the raw data if that'll be interesting/useful to you.

[Smokes]

Look like brittle fracture failure to me. Which means the material failed before it reached it elastic limit. This may be induced by cold temperature or improper heat treatment. If you work in oil and gas Hydrogen will do this to metals if the material is over hardend.

[andylaurence]

I've been doing CFD simulations on my own computer for some time using Sketchup and the Khamsin plugin to do the meshing and processing using OpenFOAM. It's really matured this year with the latest Khamsin plugin, which is much improved. Despite this, it's still hard to setup and the processing can take a very long time, especially if you want to model a whole car. The tool to view the results, Paraview, is also pretty tricky to learn. This is where Hibou Scientific's latest product, Aerodynamic on Demand, comes in.

The web application takes your model in standard STL format, which can be exported from any CAD program, including the free and easy to use Sketchup. There are currently four options to choose from; Geometry Check, Discovery, Detailed and Premium. The first option is just to make sure your model is good and pointing in the right direction. It's worth doing this to be sure you have it right, especially as it's free and results don't take long to come back. The other options give you increasing accuracy and a few extra features such as graphs of downforce/drag over the length of the car. Needless to say, I had to give this a go and delved in with a Premium analysis for $15AUD (just under a tenner), currently discounted from $40AUD.

I fired up Sketchup and drew a quick model of the ADR by tracing a side-on photo and extruding. It's rough, but all I need for now. After I added in the Export STL plugin from the Sketchup website, I was able to generate the STL file. I tapped in my email address, the speed I wished to simulate at, the yaw angle (0 degrees), selected the STL file, pressed the button for Premium analysis and I was able to immediately checkout with Paypal. Simples!

It's worth noting that these simulations take a long time with big models and it takes 24-36 hours for the results to come back. When they arrived the next day, I was presented with some summary data and a large number of 3D views that could be navigated in the web browser (Chrome works, but my version of Internet Explorer does not as it uses WebGL). The left mouse button rotates the car, the right button zooms in/out and the middle button moves around, which is where my first problem arose - my laptop only has two mouse buttons!

So what did I receive back? Why don't you take a look? On the Summary tab, it tells me what options I selected when the model was uploaded and how big the model is on the left. On the right, it shows the complexity of the model, downforce and drag values and the all important centre of pressure. At this point, I gravitated to the downforce value and was surprised to see 5kg of lift with a CoP near the front wheels. Then there's the L/D ratio of 0.19:1. Not good news! We'll see why later.

The next tab has three 3D views; Geometry, Streamlines and Pressure Map. The first is the same view you get in the free geometry check. The second is a pretty view for showing people that might show something interesting if you're lucky. The third is the pressure map and shows the car colourised by pressure with grey being ambient, blue being low pressure and red being high pressure. What we're aiming for here is a blue bottom and a red top! This is the first point of analysis, so what did I learn?



The first thing that's striking is the dark red area on the top of the splitter. That shows there's good downforce on the splitter as it's a nice shade of blue on the bottom. It also shows there's a nice amount of drag there too as the vertical area behind the splitter is also red. Going back to the underside of the splitter, it's blue at the leading edge because the air is separating from the splitter. It might be good for downforce at that exact point, but it's not much help for the rest of the car that has to utilise that air!

Also immediately obvious is the blue areas on the side of the car ahead of the wheels. Again, that's separation of the flow caused by lazy modelling as the edges of the car are sharp. Whilst the same is true of the side of the wheels, you'll note an interesting pattern which shows the generation of vortices from the high pressure area ahead of the tyre to the low pressure area at the side. This is often coined as tyre squirt.

Moving to the bottom of the car, there's red areas in the wheel arches caused by the high pressure area ahead of the tyre spilling into the wheel arch. That's why cars have vents on the top of the wheel arches! Looking at the floor as a whole, there's a blue tinge to in, which means there is some low pressure there, albeit not much.

Turning the car over, there's clear low pressure over much of the body, notably the convex front wheel arches, rear wheel arches, rear deck and the rollbar. The rollbar is mostly due to separation and the flow at that point is slightly upwards, hence why the underside is grey. I don't believe the real car has this much lift from the rollbar as it's not got sharp edges, so separation isn't going to be an issue. The wheel arches and rear deck most likely do have this low pressure, but without re-designing the body, I can't do much about it. What's important here is that most of the top side of the body is a darker blue than the underside.

Moving rearwards, the rear wing shows strong high pressure on the top and low pressure on the underside. This explains how the centre of pressure is so far forwards as the rear wing is producing solid downforce to counteract the lift from the bulk of the body. We'll see that graphically later on.

The next tab is the 3D streamlines and has 13 views with streamlines from a vertical line ahead of the car. These streamlines show the path of the air around the car. This view shows clearly how the air rolls over the sides of the car, generating a vortex that runs down the side of the car into the rear wing end plate. This view shows some air going under the car and through the diffuser, whilst the air over the top hits the rollbar. The rollbar is again causing a vortex and some of the air goes both under and over the rear wing.



The next tab is a series of 13 vertical cross sections showing the pressure or velocity of the air all around the car. From this, you can see the wake of the car and also clearly see the separation at the top of the rollbar. The following tab is the same, but with 6 horizontal cross sections.



On the last tab is the distribution of drag and downforce over the length of the car. Taking the drag first, the nose causes the largest single point of drag. The rollbar is also causing a large amount of drag, most likely more than on the real car because of the separation on the top side. The final two peaks are the rear of the car, which is a flat back in the model and coloured blue in the pressure map, and the rear wing.



The downforce is also interesting (negative numbers are downforce, whilst positive are lift) with most of the downforce coming from the front splitter, moving to lift where the high pressure under the wheel arches combine with the low pressure over the top. Low pressure on the floor and ambient in the cockpit means net downforce, whilst the huge lift from the rollbar is the cause of the biggest single point of lift. Clearly, this is skewing the overall figures with over 100N of lift from the rollbar alone! Steady lift from the rear clam is countered by 146N of downforce from the rear wing, which appears to have a lift/drag ratio of about 4.4:1. The lift from the rear clam has been shown true when the rear body clip broke at Llandow and the whole rear clam lifted at around 100mph!



So, what have I learned from this? It's likely that the car doesn't produce much downforce. Taking off the 100N of lift from the rollbar, that's only about 5kg of downforce at 60mph. Even allowing for the simplistic nature of the model and the lack of wheel rotation, there's not much downforce there. How can I improve that? Simplistically, that's a case of reducing pressure on the underside and increasing pressure on the top. We must also be mindful of the balance, which is currently very much rearwards.

We know that there's high pressure in the wheel arch, so venting that will help reduce lift, bringing the CoP forwards. With lower pressure in the wheel arch, we can fit a diffuser to the trailing edge of the splitter to reduce pressure under the splitter and increase the area of the splitter (force is equal to pressure x area). To try and reduce pressure on the whole of the underside, we can enlarge the diffuser to try and draw air under the car at higher speed, thus reducing pressure. We can also try to relieve the pressure ahead of the front wheels by reducing the pressure of the air going past the wheel arch. This could be achieved by vortex generation from end plates on the splitter. So the next step is to model that...