During initial testing, it was found that the spark generated by the piezoelectric igniter was inadequate to ignite the propane/air mixture. No form of ignition could be achieved using this method so it was decided to upgrade the system to use a motorcycle ignition coil and an old motorcycle battery as a power source.
A coil was purchased and wired to the battery via a push-to-make switch. The circuit was then tested by connecting the coil lead to the spark plug. A spark was observed but it appeared weak and unreliable. When the circuit was connected to the engine, this spark also proved unable to ignite the fuel/air mixture in the engine.
The spark plug which was being used up to this point was an NGK BM6A plug. This plug has a standard thread reach of 9.5mm. It was decided to replace this plug with one with a longer reach thread. This would place the electrode further into the engine and increase the chances of ignition. An NGK BR9EH plug was purchased as a replacement. The replacement plug had a 19mm thread reach and also had a higher heat rating which would allow the engine to withstand higher engine temperatures and therefore last longer.
A 5kV power supply was connected to the spark plug as a temporary solution to the ignition problem. This system produced a continuous spark across the plug gap. A continuous spark is undesirable in a pulsejet engine as it can disrupt the pulsating combustion and cause the engine to stop. It was decided to use this method anyway to see if the engine would at least ignite with the current spark plug position. Ignition with the continuous spark was achieved but the jet did not pulsate at all. The result of this test is discussed in more detail in section 7.3. It was decided that the intermittent spark
46 which could be produced using an induction coil was much more desirable for pulsejet ignition.
The motorcycle coil in the old circuit was replaced with an old-type distributor coil from a car and a short length of silicone HT lead was purchased to provide the connection to the plug. However, on testing, the spark was again weak and very unreliable. The circuit was checked over with a multimeter and the impedance of the coil and spark plug were found to be 4 kΩ each. The total resistance of 8 kΩ between the coil and plug electrode was much too high and the HT lead and plug were replaced.
The spark plug was replaced with a non-resistor type B9ES NGK plug and the silicone HT lead was replaced with a length of standard copper-cored spark plug wire.
The performance of the new system was found to be very satisfactory with a strong reliable spark being produced across the plug gap each time the switch was pushed.
Although this system was adequate for ignition, it proved awkward to have to push a button each time a spark was needed. This system meant that more people were needed to run a test; one person to provide spark and a second to vary the fuel flow until ignition was achieved. An automatic system would solve this problem by allowing the operator to simply switch on the ignition circuit, vary the fuel flow until ignition was achieved and then switch off the ignition circuit. An ideal automatic system would be switched on using a toggle switch and then discharge the coil at preset regular intervals to send a steady stream of sparks across the plug gap until the system was switched off again.
After contacting the Electronic Engineering department in the university to help with automating the circuit, two possible solutions were determined:
1. Use a 555 timer circuit with a large transistor which would act as a switch to cut the current to the induction coil at regular intervals which would be controlled by the 555 circuit. This had the advantage of being completely portable with all power to the circuit being provided by the motorcycle battery. The downside
47 was that the timing of the spark was dependant on the 555 circuit and could not be easily changed.
2. Use almost the same circuit as above but instead of using a 555 timer to control the transistor, a signal generator would provide a square wave signal to do the same thing. This system had the advantage that the timing of the spark was easily adjustable by varying the frequency of the output square wave on the signal generator. The disadvantage was that the signal generator needed an A/C power source and so the portability was reduced.
The second solution was chosen over the first as it was simpler to set up and the ease of adjustment was attractive. The Electrical Engineering department also had such a circuit already made up for demonstration purposes which was made available to this project and could easily be integrated into the existing circuit.
The final circuit provided a reliable and adjustable ignition source for the pulsejet during testing. It also allowed tests to be conducted more easily and with minimal personnel. The final ignition setup is shown in figure 7.1. A circuit diagram can also be seen in appendix B.
Figure 7.1 Final Ignition Circuit Signal Generator
Battery
Ignition Coil On/Off Switch &
Transistor Circuit
48 7.2. Fuel Mixing
During the initial tests when the engine was igniting but was acting almost like a simple propane burner (section 7.3), a video clip taken looking up the tailpipe showed that the burning in the combustion chamber appeared over-rich and uneven. This can be seen in a still image from the video clip in figure 7.2 below.
It was thought that the fuel may have been introduced too far into the combustion chamber for adequate mixing of fuel/air to take place before combustion. To attempt to solve this, a new fuel injector nozzle was machined as detailed in section 6.6.2. The new fuel injector would be threaded in place between the original injector nozzle and the propane hose and would move the point of injection forward into the intake diffuser.
This would give the fuel a much longer time to mix with the air as it passed through the intake orifices and over the valve tips.
The new injector nozzle proved to be very effective. The engine was never tested with the old injector nozzle after the valve frequency tuning had allowed the engine to operate correctly as the performance of the engine with the new fuel nozzle was significantly improved. There was no evidence of inadequate mixing with the new nozzle.
Figure 7.2 Uneven Burning in the Combustion Chamber (left) & Burning With New Nozzle Fitted (right)
49 7.3. Valve Frequency Ratio Tuning
The engine test using the 5kV power supply to provide spark resulted in ignition of the fuel/air mixture at a certain fuel pressure setting. The engine would not resonate and the sound of burning was very low. If gas flow was decreased, the burning would stop and if gas flow was increased, yellow flames would appear from the tailpipe. This led to the conclusion that the continuous spark had set up a standing flame front inside the engine which would only sustain at a certain fuel/air setting. The engine was acting as a simple propane burner. It was this conclusion that led to the desire to create the intermittent spark ignition system detailed in section 7.1.
However, the new improved ignition system did not improve the quality of burning in the engine. Even with the sparking frequency turned down to under 0.5Hz, the engine would still ignite the fuel/air mixture in the same manner as before. With the sparking frequency that low, it ruled out that the problem was a standing flame front being set up in the engine.
Further visual comparison of the movement of the petal valves before and after ignition concluded that the engine was taking in air by itself and was therefore attempting to resonate.
Since the tailpipe had been left oversized, the excess length was trimmed back to the designed length of 1.1m and the test was run again. The change in length did not affect the quality of burning in the engine at all.
When comparing the project engine to the previous engine which had been built in the university, it was noticed that the basic jet body dimensions were almost identical. The previous engine had achieved resonant combustion, albeit with an external supply of air.
The only major difference in design was the valve plate and petal valves. This detail, coupled with the failed test following the length reduction, prompted an investigation into the vibration of the petal valves.
50 In Part II of “The Propulsive Duct”, C.E. Tharratt explains how a mechanical reed valve made up of two identical metal reeds sandwiched together provides “added stiffness whilst retaining, as closely as possible, the response characteristics of a single metal reed.” (Tharratt, 1965)
In order to narrow down the problem, the engine was tested using a double set of 0.006”
petal valves in place of one. This would increase static stiffness of the valves but keep the natural frequency of vibration as close as possible to that of a single petal valve.
Using the double valve setup, the engine ran almost exactly the same way as it had in previous tests with a single 0.006” valve. The burning characteristics were very similar but the engine required a much higher air supply to be started and sustain burning.
These results suggested that successful resonant combustion was reliant on the natural frequency of vibration of the petal valves.
The theory necessary to calculate the natural frequency of the petal valves is detailed in section 5.2. An attempt was made to theoretically plot the response of a petal valve to the forcing frequency of the jet with varying frequency ratios. However, the analysis was regarded inconclusive due to the following reasons:
The motion of the valve cannot be modelled as a simple spring/mass system without damping. This is due to the effect of the valve plate damping out one-half of the valves motion. This means that the momentum of the valve does not carry through from one cycle to the next and therefore renders conventional modelling inaccurate.
Although the valve motion cannot be regarded as being damped, it cannot be modelled as a damped system either. Viscous damping and coulomb damping both restrict the motion of a spring system regardless of whether the amplitude is positive or negative. In a reed valve system, the valve is not restricted at all when the amplitude is positive but the valve plate does not allow the amplitude to become negative at any time. (Figure 7.3) Essentially the valve is returned to initial conditions [ (0) = 0 ; (0) = 0] before the beginning of each negative pressure cycle.
51
Figure 7.3 Valve Motion Sign Convention (left) & Simplified Valve Motion Plot (right)
These issues prevented an accurate theoretical solution for the response of the valve to be obtained without considerable further work. It was decided to carry out various tests, varying the natural frequency of the valves each time and observe the results.
The next step in testing was to use a petal valve with a higher natural frequency of vibration than the original. Using the theory in section 5.2, the original 0.006” valve was calculated to have a natural frequency of 66Hz. If the thickness of the reed was increased to 0.010”, the natural frequency would rise to 110Hz.
Due to the unavailability of additional sheet spring steel in Ireland, it was decided to carry out testing using valves cut from shim steel. The shim steel valve would have the same vibrational characteristics as a spring steel valve of equal thickness but would be more prone to deformation. The shim steel valves would help determine whether or not the engine would resonate with different frequency ratios.
A test was carried out using a 0.010” thick shim steel valve. The engine achieved resonance immediately but would only sustain for 15-20 seconds. Additionally, the engine would not sustain combustion without an external supply of air. Several
52 subsequent attempts were made to start the engine. Engine started each time but failed to sustain for more than 10 seconds.
When the engine was disassembled following the test and the valves were examined, it was found that one petal had been bent so much that it no longer seated against the valve plate. (Figure 7.4) The sections of valves which covered the intake orifices were also visually deformed from the pressure of combustion.
Figure 7.4 0.010" Deformed Shim Steel Valve Following Engine Run
The audio was extracted from a video clip of the test and analysed using Audacity sound editor to determine the operating frequency of the engine. An operating frequency of ~150Hz was measured from the audio file. This is very close to the frequency of 142Hz which was estimated in section 5.1.1.
Although the engine started with a valve frequency of 110Hz, the performance was not satisfactory. It was decided to try and run the engine with a valve frequency of roughly double that of the previous test. The simplest method of doing this was to insert a steel washer of a certain size behind the petal valve. This would shorten the effective length of the valve and thereby increase the natural frequency of vibration. It was calculated that a 47mm diameter washer would reduce the length of the valves by 9mm and increase natural frequency to ~250Hz.
53 The engine did not fire at all with this valve in place. It was determined that the static stiffness of the valve was too high to allow the air from the external supply to open the valves and create an air flow through the engine. It was calculated that using one of the 0.006” petal valves with the same diameter washer would result in a natural frequency of 150Hz but that the stiffness would be much lower. The valve was changed immediately and another test was run. The engine ran very erratically using this setup and was not able to sustain at a constant setting. It was determined that the frequency of the valves was too close to the forcing frequency of the engine for normal operation to be achieved.
To solve this problem, it was decided to make a second washer which would increase the natural frequency of a 0.006” petal valve to 250Hz. It was calculated that a 53mm diameter washer was needed for this. Using this setup, the engine fired and sustained for over 1:30 minutes. The external air supply was shut off about five seconds after starting with no noticeable difference in running. When the engine cooled and the valves removed, visual inspection showed that the tips of many petals were broken and one petal had cracked along its line of flexure with the washer. (Figure 7.5)
Figure 7.5 Impact and Fatigue Damage on a 0.006" Spring Steel Petal Valve
54 Due to time constraints with the project, it was decided to discontinue further valve frequency testing and use the remaining two valves to attempt to get pressure plots and inlet velocity data from the jet.