We’re about to follow a journey where it, in total, took nearly a year to find the solution. The journey began not long after I had completed a hydraulics course where I met Roy Sturges from RCT Power Services. He had heard of PicoScope and was looking to get more involved with scope diagnosis. I figured this sounded like a good opportunity to show what the scope could do while also broadening my hydraulics knowledge and experience. It would also give me a good idea on how to adapt the scope for use with hydraulic faults.
I soon had a phone call about visiting a Manitou Telehandler. The customer reported that the winch accessory was slower than that of other machines. Several other people had already taken a look at this machine but my plan was to get a more detailed understanding of the fault with the PicoScope and hopefully be able to pinpoint the problem to the customer. When I arrived on the site it was, quite possibly, one of the coldest days I’d experienced for some time. We had some snow and the wind chill factor when we were out in the open made it painful to be outside. As we were warming the machine up the ice melted and then refroze! I even had to wear a hat!
As with any diagnostic job, we started the job with the customer interview. By making sure we know how to repeat the fault we can ensure an accurate confirmation and validation after the repair. In this case, it is very simple, when operating the winch, it was noticeably slower in operation than that of the sister machine and it became increasingly difficult to operate when lifting heavier loads.
To determine and record the rotational speed of the hydraulic motor, we used the optical pick up to capture the measurement in PicoScope. We were very fortunate to have another machine of the same specification with the same winch on site, which meant we could compare against a known good machine. We captured the measurement with the engine RPM at 1400 RPM (or as close to as possible), and with both machines at operating temperature. When commanding the winch to go ‘up’, the faulty machine speed was 18 RPM and the known good was 29 RPM.
With the fault confirmed and the evidence to support it, it was decided to get some further information from the known good machine while we still had it at our disposal. As the customer complaint concerned the speed, we had to be thinking about a possible flow issue – flow makes it go. The winch is powered by a hydraulic motor, and to control the speed of the hydraulic motor, you have to control the flow. Proving what it isn’t is often equally important as proving what it is, so to eliminate the hydraulic motor we decided to introduce a flow meter into the winch circuit. Flow meters can provide valuable diagnostic information as to where the oil is going, and with the 300lpm flow meter Pico supplies, we could also mimic a load by restricting the oil flow with the loading valve. With four channels available we could also include the engine speed, hydraulic oil pressure at the flow meter and the temperature of the oil at the flow meter. We could then use this information to compare the known good machine with the faulty machine. The downside to using flow meters is that you are opening up the system introducing a high risk of contamination. This can be minimised with due care and consideration for the parts and hoses used, by ensuring that the equipment is clean both before and after any measurements are taken. This is particularly important if you connect to a machine which has already been affected by contamination. In order to protect yourself, and your equipment, it is always worthwhile taking a sample of the oil for particle counting. We tested the known good machine and the faulty machine and the results for both were within specification. One thing to note down was the readings from the faulty machine were 19/14/09, being high on four microns but this was still within specification for the pump. This made us confident that we were not about to contaminate the equipment.
The machine had quick couplers to attach the winch, which, providing you have the correct adaptors, made the connection fairly easy. Straight away, though, the quick couplers were an area of concern as any restriction here could slow the oil flow. Observing the data would help us better understand if this could be the case.
From the known good machine, we saw the following:
We increased the loading valve until we had a pressure of 221 bar with a flow rate of 21.35 lpm. This was taken with an oil temperature measurement of 39°C at 1400 rpm. This was important information as we wanted to make sure we carried out the measurement on the faulty machine under the same conditions. To help us graph the RPM more smoothly, we used a math channel on the crankshaft sensor. For more information on graphing crank speed, see the following forum post.
When carrying out the same connections on the faulty machine, we could not load the accessory line to the same pressure as the known good, as this machine was only able to produce 183.8 bar. The temperature of the oil and the engine RPM was similar enough to make a comparison, but the key thing to note is that the flow rate had dropped dramatically to just 9.435 lpm. This was a speed-related issue which we now knew was linked to flow, proving the complaint. To get a fairer comparison we looked back through the known good machine’s capture to find the same pressure as the faulty machine to compare the flow rates.
When we were looking at the same conditions, it gave us a much better view of the problem. When we went back through the buffers to when we had 183 bar on the good machine, there is a flow rate of 36.67 lpm. This was significantly more flow at this pressure compared to the faulty machine, which gives us some direction. As the winch motor for the flow meter had been replaced, we could move our attention away from the motor. What was next?
To control the machine, the work is done through a number of valve blocks. These distribute the flow of oil to the various parts of the machine depending on how the operator commands this from the controls in the cab. The directional control valves (DCV), on this machine, are controlled electronically over CAN and adjust the position of the spool inside to send the oil in the direction that is required. To help with smooth operation, the shape of the spool allows a greater level of control by restricting the flow of oil as it moves from neutral to fully open. We knew that there was communication as all the services worked, plus there were no warning lights on the dash. But, what if the control of the spool wasn’t moving all the way to fully open? Fortunately, many DCVs have the option for a lever control, which will allow you to manually control the valve. This means that you can quickly determine if it’s a control issue or if the fault still is with the flow of oil into the valve. With the flow meter still connected, we manually operated the valve and still had the same results. I’m not one to miss an opportunity to get some captures, so we used the breakout lead set (preventing any faults being recorded) to allow us to look at the signals during valve operation.
Breaking out a 10-wire plug is nothing for the faint-hearted, but it did allow us to see some interesting signals with CAN present. This led to some more investigation into the valve block and the individual modules, but understanding how this worked was a job for another day as the light was fading. We looked at the CAN signal, in the first instance, to make sure that the communication to the winch DCV was good and not obscured by any interference that could have an effect on the control of the winch. When we reviewed the connections to these DCVs, we could see that they were all linked together. We needed to make sure that the CAN signal to the winch slice was OK.
Looking at the raw CAN signal, you might say it is not a clean pattern. However, using the A-B math channel to see the physical layer, it all looks OK. No invalid packets of data either, which is good to see. When we were operating the winch there was no interference present in the physical layer.
Something I did notice here was the ID of the packets. Typically, with many of the Off-highway machines, the network used follows the J1939 protocol. With the decoded data in this capture, it looks more like J2534, which we see in automotive. Looking further into this, it is actually another protocol known as CANopen, more widely used for industrial control systems. These valves also have an LED indicator for a quick visual check of the valve communication and all were illuminated in green indicating everything was as it should be.
Happy that the communication to the PVG controllers was OK, it was time to call it a day. The light was fading fast now and the temperature was continuing to drop.
Conclusions from day one:
It might not look like much had been achieved on one day, but when working with hydraulics, rushing simply isn’t an option. With the risks of contamination being high at all stages, care and consideration in each step of the diagnostic process is an absolute must. Along with this, we were working with a faulty machine. This seems like a safe one to be working on but you still need to be extremely cautious with hydraulics.
The evidence found the first day showed us we were looking at a flow issue with this machine, so we created an action plan. Firstly, to check if the pump was physically capable of producing enough flow to meet the demands of the winch. Secondly, to check if there were somewhere else the oil could be going if the pump was good. Time to find the schematic.
The image above has been redrawn from the original schematic for simplified viewing. We can see that there is a 51cc fixed gear pump supplying oil to this valve block. The oil passes through the rotating coupler on the machine. The point the oil enters the valve block is via the inlet slice, which also houses a steering priority valve. It is then passed towards the outlet slice, where we find a dual-stage relief valve along with the pilot gallery used to control the spools in the DCVs. As we follow the oil from the outlet slice, we come to a manifold where a number of return pipes (not shown) from various other components join and go back to the tank. Establishing if the pump was up to the challenge was an important step, but not one taken lightly.
The test for pumps is often termed a PQ test, where P = pressure and Q = flow. Where possible it is good to have the known PQ test for a new pump, which you can then use for comparison to check its health during operation. If PQ tests are carried out regularly, it could be possible to catch a pump before it fails completely. However, in reality, PQ testing is extremely tricky to do and as with any test in hydraulics diagnostics, breaking open the circuit should only occur if it is an absolute must due to contamination risks. This was something we needed to know, though, as we could not really continue testing without knowing if the pump was good or not.
Looking back to the schematic, we can see that the dual-stage relief valve is our protection should the pressure increase beyond that of the working pressure of the machine. Adding the two springs together we have a relief pressure of 285 bar. What is also interesting about the way this machine works is that there is a load sense circuit, not usually found with a gear pump. Load Sense, LS, is used to remove some of the effort required to maintain a particular pressure and flow rate if there is no service being requested. This means that the standby pressure can be much lower than that of the working pressure and so reducing the load on the prime mover, in our case the diesel engine. This makes machines a lot more fuel-efficient.
Here is where we run into our first problem. When carrying out a PQ test, you want to be able to load the circuit but also have the protection of the relief valve should something go wrong. To achieve the maximum working pressure of 285 bar, an LS pressure has to be sent to the main stage of the relief valve (the 10 bar spring), to keep the valve shut so that the pressure can build. In normal operation, this LS signal is sent when a service is commanded. However, doing this will send oil through the valve block, introducing more variables. Ultimately, we wanted to remove the hydraulic pump from the circuit without removing it, if that makes sense.
We decided on a plan that I will admit was not good practise. I don’t want to sound too cliché, but please, do not try this at home. I was with a very experienced hydraulic technician who guided us through the whole process. We placed the flow meter between the outlet of the rotary coupling and the inlet slice of the valve block, with the intention of loading the pump using the flow meter. We would observe the pressures with the PicoScope and set limits to make sure we never went over the system’s working pressure set by the relief valve.
Again, this is definitely not a safe method for testing this, but with limited options and the PicoScope to keep an eye on the pressures, we felt that we kept the risks as low as possible. We brought the oil temperature up to 50°C and with the RPM just under 1500 used two pressure transducers to ensure we could monitor the oil pressure more precisely. One transducer was connected to the pressure sensor port at the pump and the other was connected to the port on the flow meter.
Having to stretch the pipework out from the machine meant we had some pressure drop due to distance and passing through the rotatory coupling. We did, however, have a consistent pressure drop of around 5.5 bar between the pressure measured at the pump and the pressure measured at the flow meter. 5.5 bar was not a huge loss, and the fact we were the other side of the rotary coupling meant we could also move this down the list of possible suspects. The most important fact to note from this capture is that the pump is still able to produce 70.75 lpm of oil at a pressure of 250 bar. With the PicoScope, we could also add in the desired flow rate, which was based on the size of the hydraulic pump and the pump speed. This allowed us to look at the efficiency of the pump. For more information on this, see this forum post for the technique and the formula’s used: https://www.picoauto.com/support/topic21311-10.html#p98250. This came in at 89%, which for a gear pump is just under the expected 90% ,but when we look back at the results from the winch circuit, we could only produce 190 bar with 9 lpm. The pump looks to be able to produce more than what we have already seen so why do we have such a dramatic drop in oil flow when the pressure starts to increase at the winch?
Happy that the pump is capable of producing more than we had recorded so far, we moved on to the next point on our action list. Having asked for some assistance from the trainers at the NFPC, we were recommended to check the LS pressure at the outlet slice. Being able to measure this will give us the pressure signal that is used to close the main stage relief valve. It could be that the LS pressure is only reaching a limited value and so allowing the main stage to open sooner and reducing the flow to the winch.
We also still had the sister machine at our disposal and managed to obtain some known goods. The first test we wanted to complete was a simple deadhead on an accessory, in order to see if maximum pressure can be obtained as from our PQ test. In theory, we are demanding flow of oil but restricting it so the pressure should make maximum pressure before overcoming the LS stage of the relief valve, and therefore, opening the main stage so the oil will pass back to the tank. This should be somewhere near the 285 bar that is set by the dual-stage relief valve, but we have to allow for the cracking pressure which will be slightly under this.
As you can see the readings are even lower when we get to the winch slice, recording just 163.4 bar of pressure at the pump. Some of this could be down to pressure drops between slices but it could also be absolutely normal, so we needed to do the same test on the sister machine. Another thing to mention is the difference between the pump outlet pressure and the LS pressure. By applying a simple math channel taking the pump pressure away from the LS pressure, we can see the difference clearly. This is always approximately 30 bar. This doesn’t make an awful lot of sense as we have a 10 bar spring on the LS relief. However, before getting too carried away with working this out we decided to check the sister machine carrying out the same test with the same connection.
The same connections were made to the sister machine, connecting a WPS600C pressure transducer to the pump test port and then connecting to the LS port on the outlet slice. Bringing the oil up to temperature and the engine speed up to 1500 RPM we deadheaded the Tele part of the system and saw the pressure at the pump up at 283.2 bar, over 100 bar more than the faulty machine!
As you can see, there is a significant difference between the donor machine and the faulty one.
We can appreciate that there will be some pressure drop between the tele section and the winch due to the length of pipework and the number of hose size changes that take place along the way, but this was a significant difference between machines. We also had evidence that the LS line on the outlet slice should be around 25-30 bar difference to the pump outlet pressure, which was the same for both, but the faulty machine was just not creating enough pressure to begin with. We knew that pressure is resistance to flow and the pump is capable of producing 250bar, so where was the oil going? If we just focus on the results from deadheading the teleservice, we can see that the faulty machine reached a maximum of 197 bar and no more. We would assume the oil flows over the relief valve and back to the tank. Could the issue lie within the relief valve opening too early and allowing oil back to the tank too soon? Could the oil be going somewhere else? So many questions, still no direction and both machines needed to be put back together before we left so they could be used. What should our next measurement be?
To be continued…
May 28 2020
Is there a way to measure flow in the return or possibly use a piezo on the return and compare return flow or return vibrations to see if you are getting return flow before reaching relief valve pressure, at least you will have a known good relief valve to swap into the faulty machine
May 28 2020
Is there a way to measure flow in the return or possibly use a piezo on the return and compare return flow or return vibrations to see if you are getting return flow before reaching relief valve pressure, at least you will have a known good relief valve to swap into the faulty machine 😊