§1. Welding nickel superalloys [00:03]
Last time I showed you that there are different nickel alloys — we even have pure nickel that you use for some applications. You get solid-solution strengthening for some strength, but the solid-solution and precipitation-strengthened alloys — the ones used for turbine blades and things like that — are difficult to weld. Sometimes we simply do need to weld them to repair them, so we know how to weld them, and I'm going to talk about that. In particular, these things are perfect candidates for hot cracking, solidification cracking.
I showed you pictures of some of that. All these welding metallurgists in the 1950s and 60s and 70s developed all kinds of tests. I've talked about the Varestraint test — you actually bend the metal while you're welding, it opens up cracks, and then you look where the cracks form in the microstructure. When you add certain things like aluminum, titanium, and niobium to your nickel, you form these very high-melting intermetallics.
This is not a real microstructure, but you have essentially no carbides or very few small carbides over here, and you end up increasing the amount of carbides as you come across. And you also start increasing these precipitates, which are gamma prime precipitates — they get larger as you go on. Take Ni₃Al for example: those have melting points, if they were pure, of 2,000 degrees centigrade. The alloy itself, because it's got nickel in between, has a melting point of 1200 centigrade, but these precipitates get very high strengths. They don't dissolve until about a thousand, eleven hundred centigrade, so you can get very high strength in these alloys at high temperatures, which is why we make them into turbine blades.
§2. Why high-temperature alloys matter: jet engines and Navy recuperators [02:10]
Did I tell you what the value of 50 degrees Fahrenheit is in the operating temperature of a commercial jet engine? If you can increase the operating temperature of that engine by 50 degrees Fahrenheit, you'll save two billion dollars a year in fuel costs for all commercial airlines. That's one of the reasons we spend so much money on these superalloys. If you can get one of those just 50 or 100 degrees higher temperature, your engine efficiency goes way up, okay. It goes up enough because they're burning a lot of fuel.
The Navy has similar problems. I worked on a recuperator for surface ships. The carrier actually has the fuel for all the frigates and destroyers and cruisers. When you're being deployed, because you're concerned about the threat of a nuclear blast up in the atmosphere, you actually deploy for a hundred miles or more. So your small ships are out there 100 miles away from the carrier, and they have to come back to the carrier to refuel. The small ships burn diesel fuel, and they've got to come back. It turns out they can spend thirty percent of the time just going off station, okay.
So you'd like to improve the efficiency of those engines. There was a program between Rolls-Royce and an aerospace company — I think Garrett AiResearch Turbines [?]. They built a heat exchanger about the size of this room that was going to take the hot gases coming out of the engine and preheat the air coming in. This is not when you're going at full speed — but ninety percent of the time you're not at full speed. You're saving fuel, and your efficiency is low. If you can preheat the air coming in to combust it, you can save thirty to forty percent of your fuel. So they wanted a recuperator — just a big heat exchanger between your exhaust gases and your incoming air.
This is primarily for the surface ships, on the gas turbines. So they had this billion-dollar program with Rolls-Royce and an aerospace company. They designed the recuperator in Southern California, built it, and sent it to England to test it. They had a land-based turbine test facility where you could test it with a jet engine — a Rolls or an Allison [?]. This thing was supposed to last for a hundred thousand hours. My best estimate: it lasted for 100 seconds, okay.
What happened is, in the early 90s, peace had broken out and everybody was trying to save money. So they didn't do a full three-dimensional heat transfer model of the turbine — they did a one-dimensional. And this is kind of, you know, why do we have these failures? Well, skipped on the engineering up front. The thing just heated up non-uniformly and twisted — we're talking two six-inch-thick steel rods that hold this thing together, twisted like a pretzel within two minutes. I was on the review board for the failure, and I remember doing a back-of-the-envelope calculation that should have told them you have to heat up slowly. The Navy's requirement was you have to go from dead stop to full startup of the turbine within two minutes.
The land-based turbines, which had been using recuperator technology for 10 or 15 years, had a 30-minute startup. The Navy didn't have that patience because you've got to go fast — heat, power, now. They fixed it since then. This is now 20 years ago, and I've been told you have recuperators on your ships now, but they had to go back and re-engineer the whole thing.
Student: [question about thermal signature]
Well, the recuperator's reducing your thermal signature too. Student: [follow-up] You still have a pretty good signature, right, in service against your enemies? Tom: Yes — but instead of exhausting at 1200 degrees or whatever, it's a lot lower. Any high-technology country that can't find a 90,000-ton piece of steel out the middle of the ocean venting hot gases — well, you don't have to worry about fighting back if they can't find you. It's not exactly going to phase you, okay.
§3. Inconel 718 and the GTD-111 patent story [08:16]
If we look at these precipitation-hardened nickel-based alloys, the strengthening element was traditionally aluminum. But a guy I know — his son was a lawyer in California — Larry Hyzak [?]. His dad in the 1950s, up in West Virginia, invented Inconel 718. All the ones before that were the older turbine alloys, and they were really considered almost impossible to weld. You may have to preheat these things to 1600 degrees Fahrenheit — talk about an oven. This has to be automated. But they do weld these alloys in certain cases because the parts are so valuable you can't afford not to try to repair.
Hyzak found that if you replace some of the aluminum with titanium, it's much easier to weld. So he developed Inconel 718, and Inconel 718 is the workhorse today for nickel-based superalloys — not for the turbine blades, but for the disc or the compressor parts of the engine. Things that have to go to — steel sort of poops out at about 800 to 930 centigrade. You need to go to 1800 Fahrenheit and above, you need to get to some of the nickel-based alloys. Inconel 718 is probably 40, 50, 60 percent by weight of nickel-based superalloy usage. There are full conferences just talking about Inconel 718 alone, because it is the workhorse. There are all kinds of variants — to make it more machinable, to increase the temperature capability — but that's the workhorse nickel-based superalloy.
Turbine blades themselves tend to be B-1900, IN-100 (International Nickel 100), Mar-M-200 (Martin Marietta), Astroloy (a General Electric trade name), René (also a General Electric trade name), Inconel (International Nickel). The companies all put trademarks on it. Usually these alloys are pretty well protected by patents for 20 years, which is good for the companies that develop them — you're not going to sell a million tons of these alloys, so you've got to recoup your costs other ways.
There's a story here. There was a General Electric alloy called GTD-111, and it had this very blocky, squarish gamma prime precipitate. The composition was overlapped by lots of other alloys, so when General Electric developed it in the early 1980s they couldn't get a patent on the composition. They did apply for a patent on the heat treatment, and the Patent Office kept denying it.
In the late 1990s, other people had decided this was a good alloy for land-based turbines, and they had started using it — because it didn't have patent protection — and incorporated it into their engine designs. All of a sudden, 18 years after General Electric had announced this alloy, the Patent Office grants the patent. General Electric had been pursuing it for 18 years, and finally the Patent Office issues it. Everybody out there who's using the alloy has now got to pay royalties to General Electric. When you buy something on the open market, you're essentially buying the patent rights to use that particular turbine blade — but now the price went up.
§4. Hot isostatic pressing and turbine blade rejuvenation [13:08]
When I worked at the overhaul facility, the overhaul time on a TF-30 turbofan engine was 500 hours. They didn't work for very long before they'd come back for complete rebuilding. That was a military engine — I don't know what the overhaul times are on a military engine right now, I'm not that close to the repair side in the last four years. But on a commercial engine it's 30,000 hours, okay. 30,000 hours, if you were running 24 hours a day — and a land-based turbine generating electricity nearly is — is three or four years.
What happens to this structure metallurgically is we get something called coarsening. We like these gamma prime precipitates to be very small, very square, but over time they coarsen, become somewhat rounded, and grow maybe three, four, or five times larger. That's just the thing trying to go to a more equilibrium point in its structure. The question is: could one of these utilities go to a repair company — and the company I was working for as a consultant was Chromalloy — to refurbish this blade? These blades have internal cooling passages and so on, so how can you do that? What type of heat treatment can you give it without turning it into a pretzel?
There is something called hot isostatic pressing, developed in the 50s and used more and more extensively since. You can take a casting — for example, Harley-Davidson aluminum cylinder heads for their engines, which are a critical component. Why are they critical? You're sitting right on top of it; it blows up, okay. You take a pressure vessel — this could be a pretty good-sized pressure vessel — it has end caps that are just screwed in, so it's a big cylinder.
One of these vessels blew up by Grafton, the north end of Worcester. It broke the whole thing in brittle fracture into 70 pieces averaging five tons apiece. This was 17 inches thick at the ends, but the center was 10 or 11 inches thick, where the crack started. It sent a 15-ton piece into the supporter. Some of the neighbors in that end of Worcester were not so pleased, but it didn't land on anybody. The guy was operating it at 2 AM. It's a pretty expensive vessel, so we operate them fairly continuously — put in a batch, take out a batch, don't shut it down because it takes 24 hours to heat it up. He was on the night shift, just had to go change the chart paper, was going to come over to the chair, and the thing blew up and landed a five-ton piece crushing his table.
In a hot isostatic press, you put the castings in on a rack. You introduce argon, and you start heating it up, and the argon gets to pressures of 20,000 psi. In fact, you can't even tell the argon liquid from the argon gas at those pressures. You can't compress them any tighter than that, okay. The atoms at that pressure are so close together thermodynamically you cannot distinguish a liquid from a compressed gas.
When you put the parts in, they don't change shape. It's isostatic — the pressure is the same from all directions. So you put in a part shaped like a cylinder head, or shaped like a turbine blade, and you can squeeze all the pores out at 2,000 degrees Fahrenheit operating temperature. The metal becomes slightly plastic, you hold it there for eight or ten hours, and you essentially forge without deforming the shape of the part. You get rid of all internal defects.
We showed, with a student, that we could rejuvenate and get the original microstructure. Because you're taking it up to 2,000 degrees, dissolving the gamma prime in the nickel matrix, and then if you cool it down right, you get the original microstructure back. So it turns out we could rejuvenate the GTD-111 blades. These blades can be quite large on a commercial turbine — they might be twenty-five thousand dollars apiece. So the repair value, not having to take it out of service — you get another 30,000 hours, maybe five years until the patent runs out, okay.
§5. The Grafton pressure vessel failure: metallurgy [19:17]
Remember, we had the three circles. When you're talking stress corrosion cracking, hydrogen embrittlement — you have a microstructure, you have a stress, and you have a corrosive medium. With hydrogen embrittlement, it's hydrogen. The microstructure here — this forging, this big cylindrical tube — weighed 200 tons. With the end caps maybe 25 tons apiece, it was a 250-ton vessel, about 10 inches thick on the wall, end caps 22 inches thick. They designed it in the mid-1980s on a supercomputer that could fit on an iPhone or a laptop today. They did a big finite element model — but for steady-state operation. Just like the Navy did for the recuperator, okay. They didn't go to a three-dimensional program, and they certainly didn't do transient startups — just like they didn't get the recuperator transient startup heating correct.
I got a call — this is the late nineties — and I go up there two days afterwards. The building has been blown apart like a hurricane, with just the structural steel remaining, walls out. You can see the Harley-Davidson pistons sitting on racks ready to be treated, other components sitting in the building. Travelers, the insurance company, had hired me to help with the analysis. John McNichol, head of Chromalloy's research lab, and I were the first two in the basket that went down into the pit where it had blown apart to look at it.
I went down, and the trouble was the transient thermal stresses. I just did some back-of-the-envelope calculations, basically fracture mechanics. First of all, in normal steady-state operating conditions, the outside should be in compression and the inside in tension — except in the transient condition. Include the residual stresses from welding and the stresses flip. The sidewall was tapered, and this is where the crack started — a stress concentration. They had the threaded caps in here, 17 inches thick at the cap, but only 11 inches thick at the wall. So it's the thinnest part, plus a stress concentration. The failure occurred after 30 minutes, okay, when they were starting up.
I went back and did the calculation. We take complex equations — you just have the Einstein equation: distance something travels is √(α·t), where α is thermal diffusivity. Thermal diffusivity of steel I know, I've been using it for years: a tenth of a centimeter squared per second. I plugged it in, and they were at about 75% of the heat-up — still in the transient region. Then I thought about how the stresses patterned: they had tensile stresses on the outside surface during startup. They had never analyzed the transient condition. They had analyzed the inside surface for tensile stresses in steady-state operation, but they never analyzed the outside surface where you can get tensile stresses on startup. So I came back and said the trouble was the transient thermal stresses. I knew there were some other internal stresses contributing too.
Later on, we found out the steel wasn't quite up to snuff, okay. When you write a specification — and this is in the flight of the LNG vessels — all the bottom had an average of like 20.5 foot-pounds. You take the very best plate, cut it up, and make all the test pieces out of that. But when you make a big forging like this, you can't go cutting pieces out of the big piece. So you use what they call a prolongation: the forging has a little end on it that you're supposed to cut off and make all your test specimens, so you know the properties of the steel are okay.
They had taken their prolongation and measured the fracture toughness as a Charpy — it's not exactly Charpy for fracture toughness, but close enough. You're allowed an extra heat treatment — you can put it back in and reheat-treat it — except they had no prolongation left to cut off for testing. So they just put a chunk of steel in with the furnace, a smaller piece — maybe something this size rather than 200 tons. And believe it or not, the smaller pieces cool at a faster rate than the rest of the 200-ton piece. They passed on that external test, whereas the original one, cooling at the slower rate, didn't pass. So they passed their surrogates and said, okay, it passes. This is just like the Coast Guard case — pick the best place to get the best properties, test it, pass your test, say we've got good material in our product, okay.
Afterwards, we tested. The only hot spot out of the furnace was around here. We tested up in here after the failure and found they had lousy properties — it would have failed. This other part had never been overheated; in 22 inches it was far enough away you never got the heat during a 24-hour cycle. There it changed the structure of the steel; here was the original properties of the forging. If they actually had tested a sample from that location, they would have found it flunks too, they would have had to scrap the 200-ton forging. But they didn't — they put it in service.
So they had high stresses from the transient thermal stresses, they had bad microstructure from improper heat treatment, okay. And it turns out Chromalloy used Nalco — at the time a French company, now headquartered in Naperville, Illinois, having been purchased by another company. They're the largest water treatment people. They go into utility plants — like the nuclear folks, you know more about water chemistry and how to control it than just about anybody in the world. Not everybody controls it so well. They were using some molybdenum oxide inhibitors in the recirculator for corrosion protection, and it turns out it pitted this steel.
They didn't know it, because the whole thing was water-cooled in an outside jacket — no way to look at it. When we went in after it blew up, the surface looked like 101 Dalmatians with all the frozen pits, okay. Corrosion like that produces hydrogen. So it was a form of hydrogen embrittlement, stress corrosion cracking — whatever you want to call it.
The thing had been designed as leak-before-break, which you like — what you have on a nuclear submarine pressure vessel. You design such that the critical flaw size is greater than the thickness of the steel. So if you get a crack and it starts to grow, you'll see a leak before the thing explodes. That's comforting to know that NAVSEA has chosen a good steel, that it will leak. You have a thousand psi or whatever shooting in — you take some duct tape, or you go closer to the surface so the pressure goes down. If you have a wall leak in a pressure vessel, you want to design it as leak-before-break.
I did a rough calculation: if this vessel had the proper fracture toughness like they thought they were getting, even with the transient thermal stresses, it should have leaked before it broke. When we actually got at the piece and found the original crack site, the critical flaw size was about an inch, not 11 inches, okay. Because the hydrogen reduces the fracture toughness, and the bad microstructure reduces the fracture toughness. Each one of those three circles was bigger than it was supposed to be — transient stresses, bad heat treatment, and corrosion — and they ended up overlapping. In theory, if it had been designed properly, you could never have trouble — if you never had to start it up, okay. But you do have to start it up every day.
It's a good story of how all the metallurgy comes together in all three circles, in the blow-up of something. It didn't land on someone's lawn, fortunately.
§6. The Grafton failure: the insurance story [29:31]
Student: [question about liability] You worked for an insurance company — did they find someone liable?
They found all of them liable, okay. But I had been deposed early — I was there on the initial investigation, before the filings. So I ended up not being involved in the lawsuit, because they couldn't expose me as a fact witness and as an expert in the litigation. Professor Pelloux, who was a professor with me at MIT, ended up doing it. It went on for seven or eight years before people finally settled out. Initially they estimated a forty-million-dollar loss; I think I heard numbers of over 60 in the end. The insurance companies just got together.
There's another problem which is sort of interesting, which has nothing to do with this course, but I'll tell you part of it. This is one of the only times I ever had to go down to Travelers and meet with the underwriters — the inner guys who wrote the original insurance. This is not your standard homeowner's policy, okay. Historically, in the commercial world, there's Hartford Steam Boiler — you may have heard their name. They insure pressure vessels around the world out of Hartford, Connecticut. They have a policy that insures against mechanical damage to pressure vessels and heavy machinery, okay. They do not insure against explosions, okay.
But there's something that came along after Hartford Steam Boiler in the 1860s or whatever, called an all-risk policy. This is sort of like your comprehensive on your automobile — you've got collision and you've got comprehensive, and the two mesh together. So if you have both types of insurance, between them, anything bad is going to be covered. Well, they have boiler-and-machinery and they have all-risk, and the two are supposed to mesh together perfectly so whenever you have a loss it's covered by one or the other. Except the brokers who sold this multi-million-dollar policy let Hartford Steam Boiler change their policy so it didn't quite mesh. One thing was missing, and it was the one that this loss fell in.
So we went down to meet with the Travelers underwriter — I was sitting in this meeting with all these other business people, supposed to answer any technical questions, and there really weren't that many technical questions. They pointed out that, technically, according to the policy, there was no insurance coverage for the company, because Hartford had recently changed their policy and the broker who sold it didn't notice the change. They were going to put the brokerage company — a multi-billion-dollar financial firm — on notice for effectively selling a contract that didn't cover them, okay.
I won't name the multi-billion-dollar financial company that was potentially on the hook. They did send a letter to the chairman of the board and president of that company saying, by the way, you may be liable, because you let Hartford rewrite their policy and excluded this one area that has traditionally been covered. But I'll say something good about Travelers: the head underwriter at Travelers said, doesn't matter — we are going to pay for this loss, the company will be made whole, and we'll sort it out with Hartford and the big financial broker later. It wasn't the company's fault.
I was out of it because I was a fact witness — I'd been in the hole, one of the first two in, and they didn't want to put me forward as the metallurgist who had investigated the failure, because they thought I might go a little further afield with my expertise, okay. They didn't want to tell the other side what all the theories were. So I got a couple of days of consulting, but I didn't finish it, and I can't tell you all the story for it. It's a good story from the metallurgy point of view; it's also a good story from the business point of view. Make sure you know what's in your insurance policy — what the exclusions are.
So we have four presentations tomorrow, and we will start aluminum. We didn't quite finish nickel.