More and more exhaust Aftertreatment systems are found on all types of off-highway machinery and this brings about more complications. We wondered if there was a test we could carry out to quickly check the operation and health of the main actuating components. The system fitted to the Komatsu is the Bosch Denoxtronic 2.2. It can also be found on some light and medium commercial vehicles. You can see a very basic layout of these actuator components in the illustration below.
Fundamentally, there are always a set number of components for an AdBlue Aftertreatment system. These consist of:
Dosing control module
DEF metering unit
SCR - Selective Catalytic Reduction unit including Ammonia filter
Heater elements for pipework, tank and supply module
AdBlue quality sensor
AdBlue level sensor
The actuating part of the circuit is easier to explain with a hydraulic diagram:
1. Dosing supply module
2. AdBlue reservoir
4. One way check for the valve and throttle. Throttle is used to generate pressure by restricting the flow
5. Pressure sensor
6. 4/2 flow control valve
7. Motor-driven single direction, fixed displacement pump
8. 2/2 solenoid valve (DEF Injector)
When the ignition is switched on, the pump will prime the system once the fluid is up to temperature. The pump will draw fluid from the tank where the pressure will begin to rise to around 9 Bar (130 psi). This pressure is created by the throttle in the return line to the reservoir as pressure is the resistance to flow. A one-way valve prevents the fluid from entering the supply lines from the reservoir on the return side.
During operation the injector will be commanded to open by the Aftertreatment ECU, allowing the solution to enter the exhaust system to begin the chemical reaction. When AdBlue is introduced to a high-temperature environment the urea breaks down to form ammonia and Isocyanic acid. This combines with the water vapour in the process of hydrolysis and creates CO2 and NH3 (ammonia). In an environment containing a catalyst and high levels of oxygen, found in lean-burn engines following combustion, the ammonia will combine with NOx present in the exhaust gas to form nitrogen, carbon dioxide and water.
When the engine is switched off, the Aftertreatment ECU will command the 4/2 flow control valve (6.), shifting its position so that the pump is now pulling the fluid from the supply line and returning it to the reservoir. Due to the one-way valve and the closed position of the injector, a vacuum will start to build in the pipework. To prevent the collapse of the pipework, the injector is pulsed at a high frequency to allow air into the pipework and control the amount of vacuum present. The pressure sensor will detect vacuum changes in the pipework, which can be useful when you are looking for blockages as shown in the case study. Below is a capture taken from a known good system using an active test function from a scan tool.
It was really interesting to see that while the injector current was flowing we could see a drop in the AdBlue pressure. This indicates that the AdBlue was indeed being delivered into the exhaust system. As this is a known good, we can use this drop in pressure to help highlight issues with blockages in the future. If there were problems with NOx levels despite this pressure drop, we would have to turn our attention to the spray pattern or the actual SCR catalyst.
I hope you agree that this is a valid test for this particular system, especially as it's becoming increasingly common. My understanding is that the wiring is the same for all units and as such below is a pin out of the plug which will hopefully help with the connection -
There is one aspect of this system that is still a slight mystery to me and despite trawling the internet, I've yet to find out what it means. It involves the PWM signal from the pump motor. When the pump is running it is a traditional PWM signal which we can graph with maths should we wish too. The unknown aspect is when the pump isn't running and the strange waveform that comes with it.
Above is the known good capture where we have the oddity at the beginning and then where the signal changes to the conventional PWM signal which is highlighted with a math channel. This pattern looks very similar to that which Steve found on the BMW fan as seen here - viewtopic.php?p=101043#p101043. I believe what we have here is something very similar. We know from the circuit diagram that temperature is also present on this signal but it is not known how this measurement is interpreted. If there is some sort of data then this certainly stops when the pump is running and we have our PWM signal. What's more is we can see the voltage levels are different between the PWM signal and the potential data which could be seen as something similar to LIN!
Here is where I ask for a little help as it would be great to find out if we can interpret this signal and convert it to something meaningful.
I hope this helps.
Great post, in regards to the strange waveform capture..
This is only a guess based on experience, the system once primed needs to keep pressure in the lines between pump and injector, the Denox pump utilizes a diaphragm pump, with this type pump (from my understanding) there is no mechanical device internal to the pump to hold pressure in the system, the pulsing of the pump in conjunction with pulsing the injector is used to "control" (couldn't think if a better word) the pump output i.e. supplying DEF or during reversion.
Thanks again for the post.
Thank you. I'm hoping that I can get it to a point where it's a guided test. Watch this space!
I've not had one of these pumps apart (yet) but on all the drawings I've seen, there is a small check valve in the return to tank which I would imagine is there to regulate the pressure. You're right that these are a small diaphragm pump which is used to generate the flow but as you say as soon as the pump stops there is nothing to 'hold' pressure in the system.
I'm still convinced that the odd part of the pump control is temperature data over PWM in order to prevent the pump from starting should it be potentially frozen. I've still yet to find anything about it though!
Thanks again for the support Neil. Hope all is well with you.
I came across this tutorial document from Audi entitled "Audi 3.0-liter V6 TDI With Clean Diesel System - 941803", where there is a description of their exhaust gas after-treatment system and operation.
What got my attention was the description of how the reducing agent pump V437 works regarding the PWM signal.
I know it is not the same exhaust gas after-treatment system fitted in this Audi as in your Komatsu (Bosch Denoxtronic 2.2) but basically they work the same way.
Here is a quote from this document.
"By reversing the reducing agent pump's direction of rotation (via PWM signal), the system switches
to evacuation mode whenever the engine is shut off.
A differential pressure sensor is integrated into the pump for pressure control purposes.
It recognizes the pump's direction of rotation by sensing positive or negative pressure. "
And here I guess - could this be what you saw in the PWM signal !!!.
Thanks as always for your input and great find on the Audi system. Looking at the image with the diagram it would appear to work in a similar way using a bi-directional pump to change the flow direction rather than a valve. Still uses a check valve and orifice to restrict the flow. As we know, pressure is just the restriction to flow. Nice to see the statement about the PWM controlling the pump as well which is definitely what we see when the pump is active. However, when the pump isn't active there is something strange going on.
Below is an image of the priming and then purging of the Denox 2.2 system and I've highlighted the points of interest.
I've zoomed in on an area where the pump isn't active but we have a type of PWM signal. This change is also very noticeable in the duty cycle math channel I've added. When the pump is active we can see how the PWM % changes depending on what the pump is doing. During the priming stage the pump is commanded to run high to generate the pressure quickly and then as pressure builds it backs on the speed to maintain the pressure. Then as the purging cycle starts the pump is commanded back to high to draw back the fluid as quickly as possible.
I've added the file below as well for anyone that is interested in viewing. Thanks for everyone's input so far, always appreciated.
- (12.91 MiB) Downloaded 118 times
The signal prior to the PWM is indeed temperature data from the pump unit to the emissions controller. I believe it may be a communication language that is proprietary to Bosch but I'm not sure on that.
When a Bosch SCR emissions system is first turned on, it enters a standby phase where it initiates communication on the CAN Bus back to the vehicle - it's first message generally includes DEF Level Sensor data to get the dash gauge up and running.
Once the standby phase is complete, the system enters into a "No Pressure Control Phase". When this happens, the Denoxotronic Pump Module sends temperature data from one of it's internal temperature sensors (it has two). The temperature sensor that is sending data is measuring the temperature of the DEF inside the pump unit. If the temperature is too low, the emissions controller will initiate a defrost mode - the exact temperature of this varies in calibration from OE to OE but 25°F is a pretty common number here.
The pump's internal heater is wired in-series with a PTC temperature sensor. The PTC sensor is measuring the temperature of the heater. Therefore, as the pump's temperature increases, the current through the heater decreases. It's sort of a self-regulating system, although the emissions controller ultimately has control over whether it's on or off.
Once the pump is defrosted and enters it's standard pumping mode, the emissions controller will generally turn the heater on and off - based on ambient temperature (this data comes either (A) over a CAN Bus to the emissions controller or (B) from an ambient temp sensor inside the emissions controller). Because there is a PTC sensor wired in-series with the heater, we really don't have to worry about overheating the pump.
First of all, we see the same two messages repeating over and over. If you zoom in on that section, it will be clear what I'm referring to.
The first positive pulse is always the same length (around 100 ms). I suspect that's more or less an initialization signal to the emissions controller - saying "hey, look at me, I'm about to send you temperature data".
Then we see the low pulse varies in length and the next high pulse varies in length. This section is likely the data - what temperature the DEF actually is inside the pump).
Then we see the same length low pulse of 105 ms, this is likely an end-of-message or reset pulse.
I suspect that this isn't truly "communication" like a LIN or CAN bus. The emissions controller is ONLY looking for temperature data here. Bosch has absolutely no reason to include all the overhead from CAN, LIN, K-Line, etc. that comes at the start of a message. All Bosch needs is a specific length pulse (so it can initialize the receiver) and then the data.
I suspect if you took one of these modules and put different temperature liquid in them then took these measurements, we could very rapidly reverse engineer that signal. I suspect it might even be as simple as a straight conversion rate between length of pulse (ms) and DEF temperature.
Thank you for this information as it does help to back up the theory. As you say though any interpretation of the data will need a little more reverse engineering.
I have used the 1-Wire serial decoder which although doesn't give the data does show the pulse timing. You can see the first 3 pulses after the pump has stopped are all around 100ms in length followed by a much smaller pulse. This reminds me of how K-Line functions where communication is established by this initial messaging.
I like the idea of stripping down the pump to find the temperature sensor and then altering the temperature in a controlled manner. As the signal is just analogue I think it would be there all the time if the module has power so it could be run on the bench just to see the output. This could/should be relatively straight forward although nothing is easy in this world!
Thank you for your input as I'm sure it's not just helped me but many others.
Yes, the first few pulses are unique (when compared to the rest). I concur, it's likely bus initialization.
Do you have any idea what the temperature might have been when you captured this recording? A rough estimate?
I often go for the 1-wire or the UART decoder depending on the data. You just never know what might come up!
The capture I posted isn't one that I took but one I was sent for reviewing against my known bad from the case study. The file was captured in November here in the UK which means outdoor temperatures were sitting around 10C (50F).
Not sure if this helps to attribute the pulse width to an actually temperature without further data. I still think getting hold of a unit and power it up to submit different temperatures to the sensor and start to plot from there.