Practical Reflash Tuning: Step 4: MAF/Injector Scaling
Watch This Course
$229 USD
-OR-
Or
8 easy payments of only $28.63
Instant access. Easy checkout. No fees.
Step 4: MAF/Injector Scaling
1.00.24
| 00:00 | We're now at a point in the process where we're going to deal with any changes to injectors, in other words uploading the characterisation data for our new injectors if we've gone to a larger set of injectors. |
| 00:11 | Or alternatively, this is where conventionally in our six step process we're dealing with our mass airflow sensor scaling. |
| 00:18 | Now, in our case, injectors are stock, so obviously there is nothing to do there. |
| 00:22 | There isn't a mass airflow sensor on the Dodge Hemi either and this is an area that has been problematic with Dodge tuning up until recently. |
| 00:33 | Dodge with their late model controllers use what they refer to as an artificial neural network, which I'll admit is a mouthful. |
| 00:41 | This is really a little bit similar to GM's virtual volumetric efficiency and it's a way that most manufacturers are going to and it's because they need a way of simply dealing with varying volumetric efficiency based on things like continuously variable cam control. |
| 01:01 | What this means is that we can't just have a nice simple volumetric efficiency table like we could with a more simple engine with fixed cam timing. |
| 01:08 | Of course, as we move the cam timing, the volumetric efficiency of the engine changes. |
| 01:15 | So, what Dodge have moved to is this artificial neural network which is essentially a range of parameters, biases and weightings that it uses to calculate the volumetric efficiency based on the current cam angle. |
| 01:28 | So, relatively complex from the tuner's standpoint, or at least up until now. |
| 01:34 | Let's just have a look in the software and we'll see what this artificial neural network actually looks like. |
| 01:40 | So, if we start on our airflow tab here, we'll see that we have our neural network tab out here on the right hand side and we can click on that. |
| 01:49 | It really doesn't help us a lot because as I mentioned, it is just a range of parameters that are then used to calculate the volumetric efficiency. |
| 01:59 | So, what do we do if our air fuel ratio isn't tracking our target? Because this really drives the ECU's decisions on what the current mass of air entering the engine is. |
| 02:13 | And of course if the artificial neural network is incorrect, then it's going to be telling the controller that the amount of air entering the engine is either greater or less than what's actually happening and of course we're going to end up with an error in our air fuel ratio. |
| 02:28 | And in order to correct that, we need to have a way of changing the parameters for the artificial neural network to get us back on track. |
| 02:36 | Alright, so again, all of this is relatively similar to the GM virtual volumetric efficiency system but there are a few more complexities here. |
| 02:46 | Until recently, tuners have really had no way of visualising the volumetric efficiency tables and making changes directly to them. |
| 02:54 | And a lot of tuners have sort of taken to the process of actually fudging the numbers in the fuel mass versus injector pulse width table. |
| 03:03 | Now, that is a way of getting a solution, but it's definitely not a very tidy way of doing it and essentially this can have knock on effects in other areas of the ECU's operation. |
| 03:14 | Definitely not a technique that I would be recommending. |
| 03:18 | However, now we do have the ability to do this the correct way with HP tuners bringing to the market their neural network trainer. |
| 03:28 | It's a mouthful I admit but we're going to see exactly how this works. |
| 03:31 | So, the process begins by coming into our editor software. |
| 03:36 | If we come to our edit menu for controllers that are available for neural network training, we will see that we have our VE neural network trainer, we can click on that. |
| 03:46 | And at the moment we don't really have a lot to see. |
| 03:50 | We start this by clicking on the create new file for training from tune icon there. |
| 03:56 | And it'll ask us if we wanna do this, of course yes we do. |
| 04:00 | And then we can give the file a name. |
| 04:03 | So, let's call this base tune. |
| 04:06 | Doesn't really matter what we call it but of course this is helpful if we just have a consistent trend in the way we do our naming. |
| 04:16 | Alright, it will then ask us to go to Tuner Tools which is a website so we'll head there now. |
| 04:22 | Alright, so we're on the tunertools.hptuners.com website and what we can see is on the left hand side there are actually a range of options available to us. |
| 04:33 | So, I'm gonna keep this as simple as I can and we're looking at the neural network trainer. |
| 04:37 | There is a cost to use this service, it works out at around about $50 USD at the time I'm recording this to purchase 25 tokens. |
| 04:46 | This is different to the universal credits that we use to actually licence the vehicle. |
| 04:52 | And this will give you access to the neural network trainer for a period of 30 days. |
| 04:57 | As you can see, we've already done some testing with this and we've got 22 days left on our licence, but we'll go through the process anyway. |
| 05:05 | So, for a start what we want to do is select the file that we want to upload. |
| 05:09 | And we can find the file that we just created, basetune.htt and we'll select that and we'll click upload. |
| 05:17 | So, what it's going to do now is go through and process that file, we can see that the file is here and we can see it was saying pending, it's now saying download. |
| 05:26 | So, it's processed that file and it's given us another file that we can now download by clicking download. |
| 05:33 | Now, what we can do is head back to our VCM editor software. |
| 05:36 | Now, what we want to do is click on the open trained file icon and what we want to do is go to our downloads and find the file which is this one right here, basetune, and you see it puts in parenthesis there, results. |
| 05:50 | So, we'll double click on this. |
| 05:52 | Now, we've actually got a volumetric efficiency table that we can visualise and we can see we've got pressure ratio on the vertical axis and we've got our engine speed of course on the horizontal axis. |
| 06:04 | And these numbers are in values that do make sense to us as volumetric efficiency numbers, we can see we're peaking a little bit over 100% VE there and of course we can actually visualise this, it looks exactly like we'd expect. |
| 06:17 | However, there are some complexities here as well. |
| 06:20 | As I mentioned, why Dodge have gone to this artificial neural network is because they need a way of coping with the variable cam control that the Hemi engine runs. |
| 06:30 | So, what we can see here is that we've actually got a variety of tables and these are based on our intake cam and our exhaust cam position. |
| 06:39 | Let's click on our intake cam and we can see we've got five different cam centrelines here. |
| 06:44 | And likewise on our exhaust cam, we've got five cam centrelines. |
| 06:48 | So, we've actually got 25 different tables which is a lot of work. |
| 06:54 | And the idea here is that we're going to modify each of these tables to represent the changes in volumetric efficiency we need to get our fuelling back on track at the specific cam angles where we've got errors. |
| 07:08 | Now, it makes things a little bit easier for us though. |
| 07:11 | While this does show us variable intake and variable exhaust cam tables, with the Dodge Hemi engine, we actually still have a single cam, albeit it is variable cam position and what this means of course is that as the exhaust cam position varies, so does the intake cam. |
| 07:29 | So, we can't have an infinite range of intake cam to exhaust cam positions. |
| 07:33 | If we are on a specific exhaust cam centreline, that means that we also by definition must be on a specific intake cam centreline. |
| 07:43 | And with the stock cam here, the difference or separation between the intake cam centreline and the exhaust cam centreline is always 228.5. |
| 07:53 | So, that means that instead of 25 possible combinations of tables here, we're actually only going to need to make adjustments to five of them. |
| 08:02 | So, a lot easier than it does appear at first sight. |
| 08:06 | Now, we also have here this little tick box which says use table for training. |
| 08:11 | So, there is a little bit of computing power involved in processing the changes that we make and then basically giving us a new set of parameters that we can load into the tune. |
| 08:21 | So, to make our lives as simple as we can, we can untick this little box for the tables that we don't need to adjust. |
| 08:29 | Alright, so once we've done that, essentially we'll now export another file out of here that we've created with our changed VE tables and we'll go through kind of the opposite process, we'll upload that file to the Tuner Tools artificial neural network trainer and that will process that and then give us another file that we download, which will then give us those new biases et cetera for the neural network. |
| 08:53 | So, it is on face value a little bit complicated but as we go through this, you're going to see that it's actually relatively straightforward. |
| 09:02 | Now, let's have a look at our scanner setup, because this is also absolutely integral with the process of creating our new VE tables. |
| 09:13 | So, we'll head over there now. |
| 09:15 | Alright, there's quite a lot going on with this particular scanner layout. |
| 09:19 | In order to be able to properly calibrate the artificial neural network, as well as all the other elements of the Dodge tuning that we're going to be looking at through this worked example, for this reason the layout that I'm using here is available as a download for you so we've done all of that heavy lifting so you don't have to. |
| 09:36 | So, have a quick look at what we've got going on here and in order to really demonstrate this, I'm just going to get our engine up and running at idle. |
| 09:47 | Alright, for a start here we've got our commanded air fuel ratio versus our measured air fuel ratio from our wideband and this is going to be really important for us when it comes to tuning our wide open throttle or power enrichment areas of the mapping, where we're no longer in closed loop mode. |
| 10:05 | Dodge do not use wideband sensors so I've actually added our own and in this case we're using an AEM X-Series wideband and the reason I really like this particular wideband is that it has an OBD2 pass through connector and it actually outputs the wideband data onto the CAN bus, which is then super easy to access via the VCM scanner and also because it's a CAN message, there is no possibility of ground offsets affecting the accuracy of the data. |
| 10:35 | Alright, so that's the first two that are important, we're also scanning all of the parameters that we're going to need for the axes on our graphs. |
| 10:42 | So, for example here, pressure ratio, so we need that because pressure ratio is the axis for our volumetric efficiency tables. |
| 10:51 | Moving to the next set of graphs, these really are set up to give us a good indication of what's going on with our ignition tuning. |
| 10:59 | We've got our air charge which was the load axis for that particular set of tables and we've got our ignition retard, as well as our actual spark advance. |
| 11:07 | So, this allows us to monitor the ignition timing and also any knock activity. |
| 11:13 | Moving down, the next set of graphs here we've got include our actual and expected torque values. |
| 11:22 | So, we talked about that when we were setting up our base file so from this we'll be able to get an indication of whether we are hitting our torque targets that we've entered into our calibration or whether we're coming up short. |
| 11:36 | We've also got in there the cam angles and those are going to be really important for this artificial neural network tuning. |
| 11:44 | We'll skip the next set of graphs, we're looking at our O2 sensor voltages there, so we don't really need to monitor those too closely but the last set of graphs here, our short term and our long term fuel trims for bank one and bank two. |
| 11:56 | Now, interestingly here, given that this is a stock standard vehicle, we can see that we've actually got quite a negative trim in here, our short term fuel trims are sitting at around about minus 8-10%. |
| 12:08 | We can see our long term trims haven't really started coming into play just yet, but they will over time. |
| 12:15 | So, this indicates straight away that we do have some error in our volumetric efficiency, so this is going to need to be solved with our artificial neural network. |
| 12:25 | How are we going to do that? Well let's pull our graphs over here so we can see what we've got going on. |
| 12:32 | And I've got a couple here that we've set up already, again these will be available to you. |
| 12:37 | First of all, we've got our advance so we can actually see what the ignition timing is. |
| 12:42 | Then we've got our retard so if we've got any knock occurring we'll be able to monitor this in our graphs here. |
| 12:49 | Then we've got this set of tables here. |
| 12:52 | So, I've called these ANN exhaust 85, 95.75, 106.5, 117.5 and 128. |
| 13:00 | So, how have I come up with these and how do these work? Let's head back across to our editor now and we'll have a look at our VE table. |
| 13:09 | Alright, let's click on our exhaust cam dropdown here and we can see now we've got exactly the same cam positions as I have graphs set up in our scanner. |
| 13:19 | I've just chosen each of these centre lines here. |
| 13:22 | And we could just as easily have done the intake cam, it doesn't matter which you choose, the result's going to be the same, because remember of course the cams are fixed in terms of the lobe separation between our intake and our exhaust. |
| 13:36 | Now, a little bit more that we need to understand here, if we look at this value here, 85 degrees, we see that the next VE table we have is at a cam angle of 95.75, let's call it 96 degrees. |
| 13:48 | So, we've got about 10 to 11 degrees between these two tables. |
| 13:53 | Same as we go from 96 to 106.5, again we've got about 10 degrees separating them. |
| 14:00 | Now, this is important because what we're going to want to do is set up these tables with a filter so that it's only going to lock data into the graphs when we're basically within plus or minus about five degrees of the current cam map. |
| 14:16 | So, for example here, the 95.75 degree cam map, or VE table I should say for that, 95.75 degree cam angle, we'd want to be logging data into that particular graph when our exhaust cam position is between about 90 degrees and about 100 degrees. |
| 14:34 | So, basically we're taking that plus or minus five degree spread from our centreline. |
| 14:38 | Now, if that doesn't make sense right now, it will all start to make sense once we actually see how that's done. |
| 14:45 | So, now we know what we're trying to do, let's head back and we'll have a look at the graph layout. |
| 14:50 | So, again we can see we've got our five graphs here and we can see actually we're sitting on our exhaust 128 degree map at the moment and we can see that while I've been talking it has been logging data into that. |
| 15:04 | And the reason for that is that if we just move our graphs out of the way a little bit, we can see at the moment that our cam angle here for our exhaust cam is sitting at 125 degrees so basically pretty close to our 128. |
| 15:20 | So, let's see how we've actually set up these graphs and this way while you will have these, it's always important to actually understand how these work. |
| 15:28 | That way if you need, you can always make changes to them. |
| 15:31 | So, let's start by looking at our ANN exhaust 85 degree graph. |
| 15:38 | So, we've got a label, just a name, whatever we want and obviously just something that makes sense. |
| 15:43 | And what we're doing here is we're actually using our long term plus short term fuel trim math. |
| 15:49 | There's two ways of doing this. |
| 15:51 | Basically, what we want to do is log error between our commanded equivalence ratio or air fuel ratio and our measured air fuel ratio. |
| 15:59 | Under closed loop conditions it's reasonably common to use the short term and long term fuel trims to achieve this. |
| 16:06 | Previously we've done a worked example where I've shown how to disable the closed loop control and force the ECU to run permanently in open loop and then what we can do is use another math channel which is our actual calculated error based on our commanded air fuel ratio and our measured wideband air fuel ratio. |
| 16:25 | Either way is going to get the job done. |
| 16:28 | Typically, I do prefer to use wideband and the math channel but I'm going to demonstrate this technique and you can use either one. |
| 16:37 | If you do want to use the wideband solely then you will have to disable the closed loop control. |
| 16:45 | So, let's have a look at how we do that, head back to our editor, close down our neural network trainer and what we want to do is come through to fuel and oxygen sensors and our closed loop enable here, we can see the coolant temp for enable, if we click on this, down the bottom it says engine coolant temperature above which closed loop will be enabled and obviously at the moment that's minus 15°. |
| 17:07 | We can see in the bottom right hand corner it says that the range for this is minus 64 up to 191°. |
| 17:14 | So, if we just set this to 191°, that will essentially prevent us from ever going into closed loop control. |
| 17:24 | So, that's how you could force it into open loop, I'm not going to do that so we'll just set that back to minus 15. |
| 17:30 | What you will find though is that you will need to use your scanner controls and I'll show you how to do that. |
| 17:37 | If we come into our controls, we come to our fuel system, you can reset the fuel chums because even with the closed loop disabled, what you will find is that any existing long term fuel chums that have been learned will still be active. |
| 17:51 | Alright, so now we know we can do that either way, let's get back into our graphs layout. |
| 17:56 | So, if we start with our exhaust cam position of 85°, let's see what we've got. |
| 18:02 | We've already talked about our long term and short term math. |
| 18:05 | If you do want to use the wideband, then you'll be using the parameter in here that we'll look at a little bit later when we're in power enrichment. |
| 18:13 | This is percentage, we do want to add a little bit of precision here so it's useful to have one decimal place there, no real need to get any more granular than that. |
| 18:23 | And now what we've got is our function. |
| 18:27 | So, this is important because this, as it's sitting right now, means that the parameter it's looking at is the exhaust cam position and it will only log into this particular graph if that exhaust cam position is less than 90°. |
| 18:43 | Now, let's just see how exactly we do that. |
| 18:46 | So, I'll clear that away, we'll click new variable and we'll click on parameter there and I'm going to just start by entering exhaust cam and the particular parameter that we're looking for is our exhaust cam centre position. |
| 19:00 | Double click on that and click OK. |
| 19:02 | And then we want to make sure that in this case this is the lowest cam position, so we've only got one parameter here that we need to make sure that we're ahead of, and we want to make sure that the value that we're logging is less than 90°. |
| 19:16 | Remembering we've got about a 10° spread, 5° plus or minus of the centreline so 85 plus 5 obviously is 90, so our job's done there. |
| 19:24 | Let's look at the rest of the parameters here. |
| 19:27 | What I've done is just added some shading here so basically if we see a value of positive 25, we're going to have a red colour, minus 25, which means that we're pulling fuel out because we're too rich, obviously a little bit safer, is going to be green. |
| 19:40 | This is just important because at a glance we get a sense of the scale of any error that we've got and whether it's in the good direction, we're too rich, or maybe the more dangerous position, which is when we're too lean. |
| 19:53 | Then we've got our column and our row axes. |
| 19:56 | So, what we want to do here is replicate the column and row axes and break points from our VE table. |
| 20:04 | So, let's see how that works. |
| 20:06 | So, let's head over to our VE Neural Network, we'll open the same file that we've already had open and what we can see is we've got pressure ratio and engine speed. |
| 20:15 | Pretty self explanatory there. |
| 20:17 | What we can do is basically right click anywhere here and let's start with the horizontal axis, so that's our column axis. |
| 20:24 | We come down to column axis and we click copy labels. |
| 20:27 | Now, we can head back across to our scanner and what we want to do here is choose the parameter for our engine speed, so in this case we enter engine RPM, the parameter that we're using here is this top one. |
| 20:41 | And then if you're starting from scratch, there'll be no values here but let's just delete them and what we want to do here is Ctrl V and that will paste the break points from our VE table in the editor. |
| 20:52 | Meaning that the break points in this table are now going to match the break points in the artificial neural network trainer. |
| 21:00 | Our pressure ratio, exactly the same, I won't go through that process because it's a rinse and repeat of what I just did, however we would go back to our editor, right click anywhere, this time it's row axis and copy labels and that's simply just going to copy all of the break point data from here. |
| 21:16 | Let's get back to our scanner. |
| 21:18 | So, we've already done all of that. |
| 21:21 | Alright, so this is probably the easier end, because as I've already mentioned, we're only looking for values that are less than 90° cam angle. |
| 21:29 | Let's move up though and look at the next one. |
| 21:31 | So, basically everything here's the same but now we've got a centreline of 95.75° and we want to be looking at logging data that's plus or minus 5° from that, just give or take. |
| 21:45 | So, what we can see here is our variable is now a little bit different. |
| 21:48 | So, what we're looking at is our exhaust cam centre position. |
| 21:52 | It has to be less than 101° and the exhaust cam centreline position must also be greater than 90°. |
| 22:01 | From here it's just a rinse and repeat, basically making sure that as we go and fill these out, our break points just continue to match. |
| 22:08 | So, here we've got between 101 and 111, then we've got between 111 and 122, and finally anything that is over 122. |
| 22:17 | Now, I know that if you're just getting into this for the first time, there's a lot to take in here granted. |
| 22:24 | It is actually relatively straightforward if we approach this in a systematic way. |
| 22:30 | And of course you're going to be able to watch me do this. |
| 22:33 | You've already got all of the scanner layout, so all of the hard work there is being done, now it's just a case of putting it to work. |
| 22:41 | So, let's close that down. |
| 22:42 | So, just to show how that's working, we can see again we've filled data in here at the 128° cam position. |
| 22:50 | But none of these other tables have data because the filter is making sure that that's the case. |
| 22:57 | Alright, so now that we've got a reasonable understanding of how the system works and the process we're going to go through, we can actually go ahead and start logging some data and see how we can then use that to train the artificial neural network. |
| 23:11 | Before we can get started though, let's just shut our engine off and we'll disconnect our scanner. |
| 23:17 | And we'll head back to our editor. |
| 23:19 | Now, as you can see, we haven't flashed our base file into the engine control module. |
| 23:24 | And wherever possible, I like to start by getting a dyno run on the board with our existing tune, and in this case because the car's stock of course we absolutely can do that. |
| 23:36 | And we're going to use that then of course as our comparison when we start making changes to the tune to see whether we're making improvements or maybe we're going backwards. |
| 23:44 | We still can do that, but in order to actually start training the artificial neural network, we're going to start by flashing our file that we've just created in the last step into the engine control module. |
| 23:57 | Reason we're doing this is that it's going to give us some nice solid data to work from, particularly disabling aspects such as the MDS. |
| 24:04 | So, that's important so that we're going to get good solid data, we've also got our cat over temp basically zeroed out as well. |
| 24:12 | So, once we've done all of this, we can still absolutely flash our stock file back in and use that when we get into optimising our tune in order to actually go ahead and get that baseline run that we can compare to. |
| 24:26 | For now we will flash this file into the engine control module. |
| 24:30 | So, let's get our ECU powered up, we'll go into the run position and what we're going to do here is just use the little right vehicle icon, so we'll click on that. |
| 24:40 | Now, if you haven't licensed the controller at this stage, you will have to do that. |
| 24:45 | From memory there's about two credits involved in the engine control module and about four in the transmission control module. |
| 24:52 | In this case we don't need to make any changes to the transmission control module so we'll select do not write, that's just going to speed that process up a bit. |
| 25:01 | And we've got our car powered on, it's in the run position so we'll click write and this will go through the process. |
| 25:09 | Fortunately, much much faster than the process of reading so we'll just let this complete now. |
| 25:15 | Alright, our write's complete there, we can close down the vehicle writer. |
| 25:18 | Now, we can get our engine up and running, head back to our scanner and we can actually start gathering some data and see what we need to do in order to gather the data required under both steady state closed loop conditions, as well as under wide open throttle power enrichment conditions. |
| 25:34 | Alright, we've got our engine up and running on the dyno and I've allowed it to come up to normal operating temperature. |
| 25:39 | That's really important regardless whether we're logging this sort of data on the dyno or out on the road that we're logging under normal conditions. |
| 25:47 | So, what I mean by this is we don't want the engine cold and coming up to operating temperature, and a real problem as well that we often find on the road is that when we're pulling over and allowing the car to idle for a long period of time while perhaps we're looking at our data or making changes, the engine bay will heat soak and that will affect the accuracy of our data. |
| 26:07 | So, we really want to be operating the engine for this particular purpose under as close to normal operating conditions as we can get. |
| 26:14 | And if we look in our scanner software here, just while I've been talking, we can see we've got our coolant temperature sitting at 91°C and we've got our intake air temperature at 15°C. |
| 26:26 | Reasonably cool day here in Queenstown so this feels about right. |
| 26:29 | And it's also worth just having a bit of a look at the sort of data we're getting here at idle as well so if we just circle our short term and long term fuel trims, definitely not what I would expect to see from a essentially stock standard car. |
| 26:43 | We're seeing our long term trims here sitting at around about minus 10 and minus 7% and of course as we'd expect at this point, our short term fuel trims are sort of bouncing backwards and forwards around zero which is exactly what they should be doing. |
| 26:55 | So, essentially what this means is that at this point at idle, we're somewhere in the range of about maybe 8.5% too rich and our long term trims are pulling our fueling down. |
| 27:05 | At the same time if we jump over to our graphs here, we can see we've got all of the graphs we've already talked about and we can see we're about to actually logging data which at the moment is exclusively on our exhaust cam 128° position. |
| 27:21 | And we can see we're logging minus 8.8%. |
| 27:25 | so, what we're going to do here is gather some data under steady state conditions. |
| 27:29 | And here we're wanting to make sure that the controller is staying in closed loop mode. |
| 27:35 | And we can actually monitor this here as well, we can see we've got our fuel system status closed loop normal. |
| 27:42 | Now, we could also actually add this in as a filter, but particularly on the dyno, it's not very difficult for us to make sure that we're staying in closed loop mode. |
| 27:50 | So, my process here is going to be simply to run the car in steady state conditions, fifth gear on the dyno and I'm just going to work my way through from low RPM, low load, I'll be slowly applying more and more throttle until we get up to around about maybe 0.87, 0.91 pressure ratio and then we'll be increasing the RPM and doing the next column. |
| 28:12 | So, in this way we're going to be able to fill out as much of these graphs as possible and we'll be doing this up to around about 4500 RPM. |
| 28:20 | So, let's get started now. |
| 28:22 | Ok, we've got ourselves up and running now and we're already in fifth gear. |
| 28:26 | What I've actually done there is just stopped our data logger while I was getting the engine running and going through the gears and this is just going to stop us gathering any poor quality data during the gear shifts. |
| 28:36 | So, we've got ourselves down to about 1200 RPM, it's probably about as low as we're actually going to be able to go so there's going to be a little bit of extrapolation required out into the other areas of the table that we're not going to be able to get to. |
| 28:49 | So, what I've done with the cursor on the left hand side in our channels list is I've just highlighted our pressure ratio and what we want to do is basically get that down as low as we can before our RPM starts to drop away, which is probably around about that 0.30, 0.28 range and what I want to do is just slowly and smoothly increase the throttle, I want to bring this up to around about that 0.87, maybe 0.91 and once we've done this row, we'll just continue increasing our RPM as I explained. |
| 29:17 | This is a reasonably slow process, we're not seeing any data being gathered over here on our graphs at the moment, but of course that is just due to our current cam angle and we can see that our exhaust cam angle at the moment is sitting at 88 so if we come down here, yeah we can see that we are filling in the correct table. |
| 29:35 | So, let's continue this process now, we'll speed this up a little bit because it is quite time consuming. |
| 30:09 | Alright, we've gathered our data there so let's just stop our scanner and we'll talk about our results. |
| 30:31 | So, firstly I should have mentioned the driving techniques while we're gathering this data are really important regardless again whether you're on the dyno or out on the road. |
| 30:38 | We want to be really smooth with our throttle input so we're not bringing in acceleration enrichment, we also want to be quite mindful of the amount of time we spend in each cell, and this is quite important because we want to make sure we've got enough hits in each of these cells to actually have reliable data. |
| 30:56 | And a good example of this is just here at idle, we can see that we're on the 128 degree exhaust map and what we can see is we've got very little data here, we've only accessed a few cells. |
| 31:09 | So, we've got two cells with minus 10 in them and then we've got this outlier here at 963 RPM.24 pressure ratio of positive 5.7. |
| 31:19 | And generally I would not expect to see such a large jump, when we've got two severely negative cells beside this one and then all of a sudden we go positive. |
| 31:28 | So, what we can do is have a look and see how many actual counts we had in those cells. |
| 31:33 | So, we've got some controls here in the scanner. |
| 31:36 | At the moment we're on the A box which is the average of the values that we've had in that particular cell. |
| 31:42 | If we come over to click on this little button that says C, this is the number of counts we've had. |
| 31:48 | So, what we can see is that we've got 484 at idle, we've got 100 at.19 and 963 RPM and only 63 in that cell that was a little bit of an outlier. |
| 32:00 | We can also look at the last value and that's a bit more telling. |
| 32:03 | So, what we can see is the last value that we actually had in here was a negative 5, which is a bit more fitting with what I'd expect to see. |
| 32:10 | The point of this is we do need to apply a little bit of common sense to the numbers that we're logging and not just blindly copy and paste these values and expect results. |
| 32:20 | In this case I would absolutely be ignoring this positive 5.7. |
| 32:26 | Anyway, let's have a look at the rest of our data. |
| 32:28 | So, we'll click on our 85 degree map and we see we've got actually quite a lot of data here. |
| 32:34 | Interestingly, reasonably consistently around about the negative 4-5%. |
| 32:41 | And there are again some variations here as we'd absolutely expect. |
| 32:45 | We've got data down here at sort of minus two to minus zero, so pretty close. |
| 32:51 | And again if we look at our cell counts we can sort of see the cells that we can get a bit more confidence on the accuracy of it. |
| 32:58 | So, the one that I mentioned there, minus zero point something, 38 counts, probably not really as much data as I'd like. |
| 33:05 | So, the more data we have, the more we can trust the results. |
| 33:09 | Anyway, stepping up through these and again we've got this outlier here at 95.75°, this positive 2.6 surrounded by negatives, probably not going to trust that so much. |
| 33:21 | 106, 117.25 and then 128. |
| 33:27 | Ok so, at this point clearly we've only logged data up to about 0.91 and we have remained in closed loop the entire time we've been doing that. |
| 33:35 | Key, as we've already discussed, because we are relying on that short term and long term fuel trim to actually fill in this table. |
| 33:42 | So, if we went into open loop where our long term and short term fuel trims dropped to zero, everything's going to be looking really good, but it's absolutely anyone's guess at this point as to what we're actually getting. |
| 33:53 | So, we have to do this in two ways. |
| 33:55 | One is a steady state capture there, now I did mention we'll go out to 4500 RPM. |
| 34:00 | The reality is, just for our worked example here, our engine coolant temperature was getting up around about 103, 104°C, so I did decide to cut this a little bit short but the process is exactly what you've just watched and you would just extend this out. |
| 34:14 | Generally when we are doing this, particularly at higher RPM, obviously the engine is under sustained load for a longer period of time and will tend to build up temperature so one of our jobs as a tuner is to monitor this and often we're going to have to do this in a number of goes and basically come back to idle, let the engine cool down and then go again. |
| 34:33 | So, it can be a laborious job in some instances, particularly if your cooling fan is not quite up to the task. |
| 34:40 | Regardless though, the process that I've just shown you is essentially exactly the same. |
| 34:44 | Alright, so now we've got this data, what are we going to do with it? Well what we're going to do is use this to help us modify our virtual VE or artificial neural network VE table. |
| 34:56 | So, let's start with our first table here which is actually our last table. |
| 35:01 | And this as we've already seen, predominantly is coming in at this stage of idle. |
| 35:04 | And we can see we've got these negative 10% trims. |
| 35:08 | Ok so, normally, and you'll see me do this, I'd use the copy and paste special function but again, just being mindful of this outlier here at positive 5.7, what I'm going to do is just actually make some manual changes to this particular table. |
| 35:24 | So, let's head over to our editor and we'll see how we can do this and this starts again by opening up our VE neural network and we'll also load in the base file that we created previously. |
| 35:37 | Alright, so what we want to do now is come to the correct cam angle and in order to do this, let's have a look at the scanner again. |
| 35:45 | So, we already know our exhaust cam angle, 128 degrees. |
| 35:48 | We also need to know which of the cam angles for the intake corresponds to this. |
| 35:54 | I've already mentioned that the load separation between intake and exhaust has to be fixed because it is a single cam engine. |
| 36:01 | And we can get a sense of what this is using our chart versus time. |
| 36:05 | Now, remember that we've got in here our intake cam and our exhaust cam angles. |
| 36:09 | And all we need to do is basically click anywhere, and these are our commanded angles, not our actual measured angles, and what we'll do is we'll bring up our calculator. |
| 36:18 | And if we simply add 103.4 plus 125.1, 228.5, this is always going to be the angle of separation between our intake and our exhaust cam. |
| 36:34 | And this is important because this allows us to select in the artificial neural network, which of the two tables we're going to be training or making changes to. |
| 36:44 | So, now that we know that our load separation angle's always going to be 228.5, if we simply put that value in and we subtract, in this case 128 for our exhaust cam, and click equals, that's going to show us that we'll be at 105, 100.5 degrees for our intake cam. |
| 37:03 | So, let's go back to our editor and make sure that is the case. |
| 37:06 | So, we are going to be going on 128 and we are just checking 100 for our intake cam. |
| 37:15 | So, there we go, this is the correct combination and we want to be using this table for training. |
| 37:20 | Alright, so now what are we going to do here? Well we only had a very small amount of data down in this region here and it was showing a negative trim of around about minus 10%, remember we're getting rid of that outlier there. |
| 37:34 | So, what I can do here is make an across the board change. |
| 37:37 | But we haven't gone up above .90 at the moment in our wide open throttle operating area, we haven't gone into open loop. |
| 37:45 | So, we can make a bit of an assessment here and for the moment what I'm going to do is simply choose the cells down to around about .50 , .48 in this case, and up to about 1284 and I'm going to make a change of 0.9. |
| 38:01 | So, this is the same as removing 10%. |
| 38:04 | So, we'll press the multiply symbol and that will take 10% out. |
| 38:08 | Now, what I'm going to do as well, I'm not going to change anything above .91, because again we are going to be doing this in wide open throttle in power enrichment, we may or may not end up accessing this combination of tables, but we'll see. |
| 38:21 | So, what I'm going to do is just basically extrapolate the changes out and I'll make them a little bit less aggressive now, so we took out 10%, now what I'm going to do is multiply this by 0.95, which of course is going to take out 5% and I'm going to do the same in the untouched area here. |
| 38:41 | Between 1600 RPM and 6100 and 0.14 and .48 , again 5%. |
| 38:49 | So, we could extrapolate this out into this wide open throttle area, but we'll come back to that because once we've made these changes, we're going to come back and look at the wide open throttle area and see what changes if any are needed. |
| 39:01 | So, that's our first change here. |
| 39:03 | Let's head back in and look at our next set of tables which is for, let's actually just move this back across so we can see the whole thing. |
| 39:12 | So, this is for our 117 degree cam angle, exhaust cam angle. |
| 39:16 | And here we've got a little bit more of an anomaly. |
| 39:19 | We obviously haven't accessed very much of this and this can be problematic, because of course we're trying to make an assessment here of what individual changes are required, and while we haven't managed to maybe capture a cell here or maybe some of the cells here, that's not to mean that we wouldn't possibly get to those on the road or if we were a little bit more thorough on the dyno. |
| 39:41 | So, what we want to do is sort of look at the type of change we're seeing here and make an assessment as to what's required. |
| 39:47 | Ok so, we've got between minus 3 and minus 5% here. |
| 39:50 | We do have this little outlier here at 0.8. |
| 39:53 | Let's again just see how realistic that is. |
| 39:56 | Ok, and we see that we've only got 18 counts here so I'm not really going to put much weight on that. |
| 40:00 | We'll come back to our average. |
| 40:02 | Remembering our aim here is not perfection. |
| 40:05 | It'd be nice but the reality is you're never going to get your combination short term and long term fuel trims to sit at zero. |
| 40:11 | And that is why the car has closed loop anyway. |
| 40:14 | I'm generally trying for plus or minus 5% and if I'm in that range, which we're already pretty well there, I'm going to be reasonably happy. |
| 40:21 | Generally, I'm a bit fussier than that, I'd like to be maybe plus or minus 2% to 3%. |
| 40:26 | So, there's some work that we can do here. |
| 40:28 | But how are we going to do that? Let's head back to our editor and what we're going to do is go to our next exhaust cam angle and remembering again we want to make sure that this is also matched with our intake cam angle. |
| 40:42 | So, 0.8.5 minus 117.25 gives us our intake cam angle of 111.25, so let's go ahead and select that. |
| 40:52 | And what we can see is we're essentially working backwards with those two sets of tables, which of course makes sense. |
| 40:57 | What I'm going to do here is highlight everything up to 0.91 pressure ratio which is what we had and I'm not going to be quite as aggressive as the 5% trim that it showed. |
| 41:08 | I'm just going to sort of split the difference here and just multiply the entire table by 0.97. |
| 41:12 | Obviously, that is removing 3% so we should be pretty close there and certainly within my plus or minus 5% range. |
| 41:20 | Ok,our next table, 106.5 here. |
| 41:23 | Now, we're a little bit all over the place here. |
| 41:25 | Not too bad, again we've sort of got a massive outlier here at minus 7. |
| 41:30 | We've got a positive trim here of 0.8, again we can use our count and see how much weight we should put on these actually. |
| 41:37 | So, we're at 382 counts at plus 0.8 so we are probably pretty close here. |
| 41:41 | This one here, only four counts and that was the 7.6, so not really going to pay too much attention to that. |
| 41:48 | Essentially, down in this lower load area, we're generally negative by maybe between 1% and 3%, I'm actually pretty happy with that. |
| 41:56 | But let's just pull 1% out. |
| 41:59 | Then in the higher load areas, we're sort of between negative 5% and negative 3%. |
| 42:05 | So, what I'm going to do is sort of split the difference here so let's say up to around about 0.53 pressure ratio, I'm going to pull 2% out. |
| 42:13 | So, let's go ahead and do that for a start. |
| 42:15 | So, we're at 106.5 so we need to of course change our exhaust cam position to 106.5, that's going to correspond with 122 on our intake cam. |
| 42:24 | And what we wanted to do was go down to let's say 0.53 will be close enough. |
| 42:29 | Again we don't need to be absolutely pinpoint accurate. |
| 42:32 | When our errors are already reasonably close. |
| 42:35 | So, we're going to pull 2% here, 0.98, press the multiply symbol. |
| 42:39 | Let's go back to our scanner. |
| 42:41 | And our error's got a little bit greater once we get out sort of above 0.58. |
| 42:47 | So, let's just average that and let's call it minus 3%. |
| 42:51 | Again it's not going to clean up all of the error, that's going to be very close. |
| 42:54 | So, we'll highlight 0.58 down to 0.91. |
| 42:58 | And this time of course we're multiplying by 0.97. |
| 43:02 | Job done, ok, that's our next set of tables. |
| 43:05 | Back to our scanner again and we'll come to our 95.75 set of tables. |
| 43:10 | And looking at the error here, again probably pretty close in a lot of areas with some outliers. |
| 43:18 | Again we can click on our count and see how close these were. |
| 43:21 | So, we've got one here at 28 and one here at four. |
| 43:24 | This count of 28 also corresponds to 7.9. |
| 43:27 | So, I'm not gonna trust these numbers. |
| 43:30 | Basically, looking at all of this, I'd be pretty happy multiplying the entire table out to probably an area of around about let's say 0.62 by let's say, we'll take 3% out so let's go ahead and do that. |
| 43:47 | So, first of all 95.75. |
| 43:50 | And of course we're stepping up our intake cam to 132.75. |
| 43:55 | And we'll come out to 0.62. |
| 43:58 | 3% multiply, job done. |
| 44:02 | And then looking above 0.62 we do get a little bit more negative, we've got this again, this outlier here, but if we look at our counts, we've only got eight, I am gonna completely ignore that. |
| 44:12 | We're sort of more like four to 6%, but again we'll split the difference here because we do also have negative three, negative 0.8. |
| 44:20 | So, let's say we take out 4% out of that area. |
| 44:24 | So, we'll highlight out to 0.91. |
| 44:28 | And of course to remove 4%, that's the same as multiplying by 0.96. |
| 44:33 | Job done there, let's head back to our scanner and our last set of tables here. |
| 44:38 | Now, we've got actually quite a lot of data so the controller is actually relying on this cam angle quite heavily through this mid range area. |
| 44:46 | And our data looks pretty consistent. |
| 44:49 | We're sort of looking at values in the, maybe we do have a 0.2 negative, but again let's check our count, not something I'd be too worried about. |
| 44:59 | So, we're sort of more in the three to 6% range. |
| 45:03 | And again we can split the difference here, I'm not gonna get too granular with this but I will show you how we can use the pace special function as well. |
| 45:11 | So, let's do exactly that for a start. |
| 45:14 | So, what we wanna do here is we can highlight this table, right click and click copy. |
| 45:20 | Now, we'll come across to our editor and we want to first of all go to the correct table, which is of course 85 degrees exhaust cam and 143.5 intake cam. |
| 45:30 | So, what we can do here is right click on the table and go to our pace special function and then multiply by percent. |
| 45:39 | Now, what that's gonna do is apply that percentage change that we just got from our scanner directly into here. |
| 45:44 | And this can be a really powerful way of making quick changes to our VE table. |
| 45:49 | Problem of course is it's going to take into account any outliers. |
| 45:54 | This is where a little bit of common sense from the tuner is required. |
| 45:57 | And of course it misses all of these areas that we haven't been able to actually access and get data in while we were doing our scanning there. |
| 46:05 | So, powerful technique but we need to understand its implications and its limitations. |
| 46:10 | So, we're just gonna undo that though and we'll come back to our scanner, just to refresh my memory on this. |
| 46:16 | And yeah I'm gonna take out 4% from the entire area up to .91 and that should get us in pretty good shape. |
| 46:24 | So, let's do exactly that. |
| 46:26 | We'll highlight up to.91. |
| 46:29 | And we're multiplying by .96, our job's done there. |
| 46:32 | Ok so, that's created some changes to our VE table so we can then train and create some new values for our neural network. |
| 46:41 | However, of course as we've explained, we haven't looked at our wide open throttle area yet and that's going to be pretty important. |
| 46:48 | So, let's go ahead and gather some data in order to do that. |
| 46:51 | Let's head back to our scanner. |
| 46:53 | Alright, so we're going to need to create a new set of graphs, and I'm going to show you how to do this but again these will be available in the download so you don't have to, again really important I think to just understand what I'm doing so you can do it yourself. |
| 47:06 | So, let's right click here and go to our graphs layout and all of these graphs here are using our long term and short term fuel trim math. |
| 47:14 | Of course, when we're in power enrichment we don't have our long term and short term fuel trims active because this only runs narrowband lambda sensors so we are going to now switch the parameter we're going to log and this is going to be another math channel which is our equivalence ratio error. |
| 47:30 | Simply the difference between our measured air fuel ratio from our wide band and the standard air fuel ratio that the controller is asking for. |
| 47:37 | So, how do we do this? Pretty easy, let's start with our, we'll start actually at the start with our 85° map and what we're going to do is use the clone graph icon here which does exactly what you'd expect, it clones the graph and creates a new one. |
| 47:51 | So, what I'm going to do now is just click on our parameter and we'll change that and the rest is pretty straightforward. |
| 47:57 | We'll scroll down to the bottom to our math channels and we want to come down to our lambda and AFR, and we're going to click on EQ ratio error. |
| 48:06 | Now, when we do this it does, it inconveniently relabelled the graph, so what we're going to do is just relabel that and we're going to call it ANNEXH85 and we'll just add PE to the end of it, so we know this is relevant for power enrichment and there our job's done for our first graph. |
| 48:26 | This of course is just a rinse and repeat for the remaining four graphs so we'll jump ahead. |
| 48:30 | Alright, so we've got our graphs here. |
| 48:32 | Now, there's one obvious problem that you might have already picked up in that so far these graphs are going to log data irrespective of whether we are in closed loop mode or power enrichment. |
| 48:44 | So, to see that in effect, let's just start our scanner here, obviously we're at idle at the moment and if we come down to our 128° map, here we can see we are logging data into our graph and of course because of our closed loop trims it's showing that everything is pretty much perfect. |
| 49:03 | Couple of ways of dealing with this, we can add another filter, so that it will only log data into this set of graphs if we are in fact in power enrichment. |
| 49:13 | Variety of ways of doing this, we could log it relative to our closed loop status or we could include a target lambda or commanded air fuel ratio, make sure that it's any time that that is not our stoich air fuel ratio. |
| 49:27 | Or in this case I'm just going to control the way we log data and understand the data that I'm actually going to be using for the purpose of training our artificial neural network. |
| 49:40 | So, with that out of the way, what we're going to do now is perform a ramp run on the dyno. |
| 49:44 | Again, really important that we make sure that our engine operating conditions, coolant temperature and intake air temperature are normal before we do so. |
| 49:52 | So, let's get our engine running, get ourselves ready to create a run on the dyno and log some data. |
| 49:57 | Alright, so we've got rid of most of our heat soak here, I'm ready to perform our run in fifth gear. |
| 50:02 | Not going to worry about the actual dyno screen at the moment, we're not in the least bit interested or worried about our results, we're just looking at our air fuel ratio and how well this tracks. |
| 50:10 | So, let's just get our graphs out of sight for a moment, we'll scroll those across to the right so we'll be able to actually watch in real time what's actually happening with our scanner log data and we'll go through to full throttle and get our run underway. |
| 50:41 | Ok so, we've gathered our data here and what we really want to do is be comparing these two graphs here, which is our commanded air fuel ratio in red and our wideband air fuel ratio in yellow. |
| 50:54 | Now, we can see that we are a little bit richer than our target, just about everywhere, which admittedly is a safe place to be. |
| 51:01 | However, this does give us some room to go ahead and make some changes. |
| 51:06 | So, this probably tracks pretty consistently with what we actually saw in the part throttle areas, essentially everywhere we were a little bit too rich. |
| 51:14 | No surprises there, there's no big steps in our VE table. |
| 51:17 | Let's just scroll our graphs back into view and what we'll do is have a look and see what we actually had here with the areas that we were capturing. |
| 51:27 | So, if we click on our 128 for a start, so again we've got data down here, which I'm absolutely going to ignore, we really are only interested in .91 and above or .96 and above even. |
| 51:39 | At 128 degrees cam angle we're basically .8, sorry 8-10% too rich. |
| 51:47 | We'll scroll back through these, again probably in the region of sort of 3 or 2% through to 6 and 8, 9%, keep coming through here. |
| 51:59 | 3% in the higher RPM ranges here. |
| 52:03 | And we'll keep coming down, really we haven't captured any data at 95.75. |
| 52:08 | Likewise really nothing here at 85 degrees that's worthwhile. |
| 52:14 | And we've got a little bit more at 128 degrees, but realistically this is all in the idle area. |
| 52:19 | So, what are we going to do here? Let's start with our 128 degree cam angle. |
| 52:26 | So, we'll head back to our scanner, we can see we're around about 8-10%. |
| 52:29 | So, we're going to split the difference here, but first of all we're going to go back to 128 degrees and we're going to come to 100 degrees. |
| 52:37 | So, we've gone up to .91 already and what we'll do here is highlight .96 and above, actually we haven't done .91; .91 and above and what we're going to do is take out 8%. |
| 52:51 | So, just undershooting a little bit. |
| 52:53 | Generally, this is an iterative process which I haven't mentioned before, you're unlikely to get a perfect result at your first set of changes. |
| 52:59 | So, generally I will tend to undershoot the magnitude of the change that the data is showing us. |
| 53:06 | So, let's just multiply by and we've got our first table complete there. |
| 53:10 | Looking at it, pretty smooth there. |
| 53:13 | We do have a bit of a step between that, around that .91 which I just made that larger change to and if you really want to get fussy here, what you can do is actually extrapolate that, interpolate that a little bit. |
| 53:26 | So, let's just interpolate that between vertical bounds there and if we look at that again, it's going to make it a little bit smoother. |
| 53:35 | So, you could use the smooth selection. |
| 53:37 | Anyway, so that's our first table done. |
| 53:39 | What we'll do now is step to our 117.25 and of course we match that by changing our intake cam as well. |
| 53:47 | And let's have a look at that particular table. |
| 53:50 | Alright, so we've already made changes down in this region and again we'd have been transferring back into closed loop, I'm ignoring that data. |
| 53:57 | What we're looking for is sort of the magnitude of change that we're seeing here. |
| 54:01 | Quite high down here below about 2600 RPM and then in the higher RPM ranges, it sort of gets a little bit better. |
| 54:11 | So, what we're going to do is sort of match that. |
| 54:13 | So, let's open our scanner and we'll start in our lower RPM range, sorry our editor I should say, from let's say 3200, I'm going to take 4% out, maybe 5%. |
| 54:27 | So, 3200, let's go to that point here. |
| 54:32 | 3200 and below and yeah we'll multiply that by .96 for 4%. |
| 54:40 | Let's head back to our scanner now. |
| 54:43 | And above 3200, I'm going to take 2% out. |
| 54:49 | Now, it seems like it's a diminishing error so we'll go 2%, that should be again at least close enough. |
| 54:55 | Again, remembering we're not aiming for perfection here, we're just aiming for plus or minus 5% which we're already within, just being a little bit fussy and showing you the technique. |
| 55:06 | Alright, second set of tables is complete, let's drop our exhaust cam to 106. |
| 55:10 | We'll change our intake cam to 122, come back to our scanner and have a quick look at our data. |
| 55:17 | Not really much going on here, we've only got a couple of cells being hit, but it sort of follows the trend that we saw from our previous so I'm just going to take 2% out across the board there in the areas that we haven't actually changed. |
| 55:29 | .98 is our 2% change, we'll change that, make that change. |
| 55:33 | 95.75 and 132.75, back to our scanner, look at our next set of data. |
| 55:41 | Obviously, we're not accessing this table at all in the wide open throttle or power enrichment area so there's really no necessary changes to be made here. |
| 55:50 | If you want you could change the high pressure ratio areas by what we've already changed the rest of the map, but we're not accessing it, I'm not going to be too concerned about that so we'll have a look at our last set of graphs here and we'll head back to our scanner and again we're really not accessing this set of tables at all. |
| 56:14 | So, again for simplicity I'm simply going to leave that alone. |
| 56:17 | Ok so, now what we want to do is train the neural network. |
| 56:21 | So, we do that by simply clicking this button, export for training, and that will create a file for us and we need to name this so we'll call this BaseTuneANN1, something of that nature, just so we can follow our iterations and we will save that. |
| 56:41 | Ok, now we need to go to our tuner tools and upload that file so let's head over there now. |
| 56:49 | We'll find that particular file which was saved to my downloads and we'll upload that. |
| 56:55 | Now, the process here of actually processing the data and creating our new file does tend to take a little bit of time and I've actually just made that worse, because I haven't unticked the box to not use the series of cam position tables that we're not actually accessing, so this is now going to take longer than it should. |
| 57:15 | So, while this is processing, let's just come back here. |
| 57:18 | We already know the combination of tables that we are using. |
| 57:21 | So, let's say if we stick to 143.5 intake cam and 85 degree exhaust cam. |
| 57:26 | So, we know that this is a combination of tables that is being used, remembering again, the cam lobe separation is always going to be that 228.5. |
| 57:35 | So, what we could do is leave our intake cam set there and then move through the remaining tables and just untick that little box. |
| 57:47 | This obviously is a time consuming process but it is going to save you in the end. |
| 57:52 | So, that now covers the one combination of these tables that we are using. |
| 57:57 | We can then step to our next one and for this one, this is the only combination of tables we're using, so we'd untick the box for the remainder. |
| 58:05 | So, that will speed up the process. |
| 58:07 | Anyway let's head back and see if the changes have been made and it has, we've got this file to download so we'll download that file. |
| 58:15 | Head back to our editor and now we're going to find that particular file. |
| 58:20 | We will save our current progress as well. |
| 58:26 | Ok so, now we've got our base tune ANN1 and it always puts result in parenthesis there. |
| 58:33 | So, this is the file, we're going to double click on that and then it will ask if we want to update the tune. |
| 58:38 | So, we're going to click yes and that will then update the calibration changes. |
| 58:42 | So, in order to see what that's actually done, let's go to our airflow and our neural network and we can see the values that have been changed there, the first layer, second layer and outer layer weights and biases have been altered. |
| 58:55 | We can click on these but again they're not particularly useful for us, they are not going to make a huge amount of sense as to the changes that have been made. |
| 59:03 | But what we're going to do now is we will save this file and I'm going to actually give it a new name and we'll call this ANN1, just so again we can track those changes. |
| 59:15 | What we'll do now is we'll flash this into the ECU, we'll start up and we'll have a look around and see how much that's fixed the errors that we were seeing. |
| 59:24 | Ok so, we've got our engine back up and running after our reflash and we've got rid of our heat so straight away you can see that our results are looking really promising. |
| 59:32 | Obviously, we're only at idle at this stage but you'll recall that our short term and long term fuel trims were negative around sort of 7-10%, so now we're sitting right bang on zero, obviously still moving around a little bit, but I'm really happy with the results that we've seen there just for one iteration of these changes. |
| 59:50 | Of course, the next step is to check what those trims look like under our steady state conditions just like you've seen, as well as under a full throttle ramp run power enrichment condition. |
| 01:00:01 | This is an iterative process, you're not going to get this right in one round of changes. |
| 01:00:06 | So, generally I would expect to have to do this maybe two to four times, maybe a little bit more, but the process that you're going to use is just a rinse and repeat of what we've already seen. |
| 01:00:18 | So, let's head now to the next step of our process. |
