|
Cross Compound
Compressor Design V The compressor was constructed as designed in the first four design sections and ran pretty well. Recall that along the way the upper head was redesigned and later the the bore of the low pressure air cylinder was reduced. The change on the cylinder bore was due to an engineering error -- if I would have run the math on operating the compressor at ~ 100 psi steam pressure rather then ~ 200 psi steam pressure I'd have recognized the problem before making any chips. Symmetry Problem: All that said, I wasn't completely happy with the operation. The main problem was that after running for a while the operation became asymmetric ---- it would have a fast stroke in one direction and a very slow stroke in the other. This become so pronounced that the output rate decreased and the pump didn't speed up even if the air pressure on the output dropped to 20 or 30 psi. The slowdown was caused by the high pressure side of the air compressor becoming unbalanced and bottoming on one side of the stroke and be way off bottom on the other side. The effect of this unbalance was that the pressure on the low pressure side built well beyond the expected maximum of about 15 psi relative to the atmosphere. (The low pressure cylinder volume is about twice the high pressure cylinder volume. If everything is perfectly balanced, at the end of the low pressure stoke the pressure in the inter cylinder channel should be about 15 psi relative to the atmosphere. If the output piston is not symmetric, then the volume in one side of the high pressure cylinder available to receive the air from the low pressure is less than half and the pressure will be greater than the 15 psi.) The unbalance can be caused by any small leak in a check valve or a steam valve. Anyone familiar with DC amplifiers will recognize the analogy of this problem to amplifier drift which was a constant headache with vacuum tube operational amplifiers (yes, I'm that old). The problem was confirmed by cracking the inter cylinder check valve plug on the side where the pressure was too high. The cracked plug had a small leak which bled off the excess pressure and the pump centered itself, the speed increased and the output rate increased. Westinghouse solved this problem by putting a pressure relief on each end of the low pressure cylinder. I assume the relief was set in the 15 to 20 psi range. The effect of the relief going off is to allow the pressure on the two sides of the high pressure cylinder to equalize and the operation to become symmetric. I chose to ignore this relief when I designed the model (another engineering mistake). It actually ran pretty well with the one plug cracked slightly but that is at best a "third order bug killer". (Note that Westinghouse didn't try to balance the many sources of unbalance but rather chose to fix the effect of the unbalance.) I didn't have room for the pressure relief on the upper air cylinder head. I considered hanging it on the back but that wouldn't look so good, even if it was on the back. Next solution was make a small leak in the system to allow air to bleed between the two sides of the low pressure cylinder. This was accomplished by simply drilling a hole in the low pressure air piston. I first tried a 0.040" hole. The pump ran great! I could easily make the pump unbalanced by introducing a large leak by unscrewing one of the inter cylinder check valve plugs. Once the system was running unbalanced I tightened the plug and found it quickly returned to near symmetrical operation.
The hole of course reduces efficiency. Well --- make the hole smaller. I enlarged the first hole and tapped it 4-40. I then took a 4-40 stainless set screw and drilled a 0.0135" hole (#80 drill) in it and installed it in the piston. That worked too and the loss seemed to be minor since the pressure built as it did without the hole to about 100 psi but then the pump continued to run slowly and the rate of pressure build was very slow ----- the leak was causing a problem, The leak has a good characteristic ---- very little loss if the pump is running fast. When the pump becomes unbalanced it runs much slower and the leak can do the job of balancing the pressure and in turn causing the high pressure piston to return to center. The bad characteristic of the leak --- the pump never completely stops. Not elegant! Pressure Switch: The prototype used a pressure switch to turn the compressor on when the pressure dropped and turn it off when the correct pressure was reached --- similar to the shop air compressor. The switch shut the compressor off well below the stall pressure so the compressor tended to run fast even when near the maximum pressure. That is exactly what I need --- a switch that would shut the compressor off when the pressure reaches an acceptable level and while it is still running pretty fast to minimize the loss due to the leak introduced in the low pressure piston. The compressor ran pretty fast up to about 90 psi output air pressure with 100 psi steam supply pressure. The point of intolerable slowdown tracks the input steam pressure. So .... what I needed was a pressure switch that shut off the steam input when the air pressure reached about 90% of the steam pressure. Sounds like a job for a pressure differential switch--- just like used on some old British roadsters to detect when one half the hydraulic brake system failed. That system used a little shuttle valve with each side fed by one side of the hydraulic system In our case we need a shuttle valve where one side in more sensitive than the other. One way to make one side more sensitive is to use different piston sizes and a stepped cylinder. That sounded like too much work. Another option is to put an external piston rod on one side to reduce the effective area of that piston. I did a crude first try using 5/16" pistons and a 3/16" rod on one side. The effective areas of the two pistons were 0.49 sq in and .077 sq in. This system was balanced when the pressure on the large piston (the air side) was 64% of the small piston (the steam side) For example, if the steam side was 100 psi, then it would balance with the air pressure at 64 psi. It takes some pressure unbalance to move the shuttle --- maybe 20 psi so it was anticipated the shuttle would move when the air pressure was ~84 psi (turn off) and at 44 psi (turn on). This valve actually worked quite well. The valve worked so well that I decided to improve upon it. It would be nice if the low pressure turn on was a bit higher. The high pressure turn off could also be a bit higher --- it was running pretty fast when it turned off. So --- less difference between the effective area of the two pistons. It would also be nice to have less difference between the high pressure turn off and the low pressure turn on----- bigger pistons to reduce the unbalance required to move the piston.. The next attempt was 3/8" pistons and a 3/16" rod on the one side. The two effective areas in this version were 0.110 sq in and 0.082 sq in. The system balance is this case is when the large piston has 75% of the pressure of the small piston. For 100 psi steam supply, balance is at 75 psi air. If we assume the large piston will move at 20 psi differential then the high pressure turn off will be at ~ 95 psi and low pressure turn on will be at ~55 psi. This variation also switched close to these calculated pressures--- about 90 psi at the top and about 55 psi. I used several different pressure gauges for these tests --- the gauges differ more than 5% at the top end so one should read too much into slight variations in the measured switch points.
The sketch above shows a cross section of the switch. The design is similar to the compressor main valve and uses the same 3/8" OD (#010) Viton O-Rings for piston rings. A 5/16" OD (#008) Viton O-Ring is used for the rod seal. The plug in the bottom also uses the #010 O-Ring for the seal --- that was the easiest way I could figure to make a compact reliable seal. Drawings for the individual parts are below.
Check Valves: One of the sources of the unbalance described earlier was very small leaks in the check valves. I decided to replace the Clippard based poppet check valves with the more robust piston type check valves from McMaster-Carr (#7768K15). The piston valves are described at Check Valves . This piston check has a larger passage area then the specific poppet valve used so the pump should run a bit faster and more efficiently. The piston type valves can be turned down to essentially the same dimensions as the poppet valves for the input and inter cylinder valves (see drawings for HS531 & HS532 in Design part II). The output check valves can't be machined to match the poppet valve HS530. What I did was to make the recess for the valve in the Output Check Valve Housing (HM538) deeper --- from 0.484" to 0.562". The hole in the bottom of the recess was expanded to 1/4' MPT. The output end of the check valve was turned to 9/32" and an adaptor turned to fit on the end of the check valve. The adaptor had a female end that accepted the end of the check valve. The two parts were held together with Loctite 672. The other end of the adaptor has a 1/4" MTP male thread. (In hindsight I'd recommend that the existing 1/8" NPT threads on the valve output be used and the housing drilled and tapped for 1/8" NPT.) The input end of the output check valve was turned to match the poppet valve per HS530. These design improvements worked ---- See Compressor Construction Part IV.
|
