User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Boeing TX: A Precursor to Boeing NMA?

Thu May 16, 2019 11:18 pm

Very interesting article from AvWeek ( https://aviationweek.com/defense/inside ... la-win-t-x (email reg required) ) gives us some of the first detailed insights about the TX program.

As I was reading it, I didn't have a hard time seeing how the same approaches are probably being used for NMA right now.

One interesting quote:

The team decided to adopt an extreme approach to model-based engineering.

It’s a discipline of product design that calls for constructing an elaborate, three-dimensional digital model of the entire aircraft. The model allows engineers to analyze aerodynamic flows and loads, create a manufacturing plan and distribute the design seamlessly throughout the supply chain. Although the concept of model-based engineering is not unique to Boeing in the aerospace industry, Niewald believes that Boeing took it further than the competition.

“Yes, 3D design has been out there. But to be able to put it all together through the process, all the way through production, this gave us the opportunity to prove it out and show that it does have time-savings,” he said.

Then the article goes on to say:

    :arrow: They used agile software development using eight week 'scrum' cycles which they claim reduced software development time by 50%.

    :arrow: They don't consider the first two models to be prototypes, they consider they are very close to being the same as the first EMD aircraft.

    :arrow: "The mold line doesn’t change in the airplane. That’s why we call it ‘EMD-ready.’”

    :arrow: The precision of the modelling allowed the initial assembly process to take 80% fewer hours than a traditional process would take

    :arrow: They developed all the mission and control computers in-house to eliminate the overhead of dealing with external vendors

    :arrow: This is the first acknowledged use of Boeing's in-house and newly branded AvionX technology

Pretty exciting stuff!

Will this be the tech that leads Boeing out from under the shadow of MAX and gets used on the NMA and/or a new NSA?

I think it very well could be.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
Weatherwatcher1
Posts: 266
Joined: Sun Mar 03, 2019 5:14 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 12:33 am

Very interesting. I didn’t realize that the T-X program could translate to an all new commercial airplane. Perhaps Agile is how they will meet the standard 5 year from launch to entry into service window preserving 2025 delivery dates. The 777 was able to go from launch to delivery in 5 years. Agile is a common software development methodology, so it is pretty cool to see it applied to aviation
 
PEK777
Posts: 418
Joined: Thu Jun 28, 2012 12:56 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:01 am

Interesting take, quite the REVELATION
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:19 am

Weatherwatcher1 wrote:
Very interesting. I didn’t realize that the T-X program could translate to an all new commercial airplane. Perhaps Agile is how they will meet the standard 5 year from launch to entry into service window preserving 2025 delivery dates. The 777 was able to go from launch to delivery in 5 years. Agile is a common software development methodology, so it is pretty cool to see it applied to aviation

I also found the Agile stuff interesting, in particular the eight week scrum. I attended in-depth training and was told that scrums should be one week long, and if not, at most two weeks long, to get the agility that came from Agile. Yet in the kind of work we were doing (embedded systems, lots of legacy "C" code that was impenetrable to modern IDE tools) most people in my organization felt we could never do one or two weeks. It's interesting to see Boeing end up with eight week scrums, since two months isn't a lot of "agility", IMHO. We ended up with four week scrums that often were just parts of bigger eight or sixteen week scrums. I was pushing for two week scrums but that required more scrum masters and other support people to engage in the process, so it never happened.

Overall, I wasn't a huge fan of Agile as deployed in my organization. It largely ended up being a methodology to generate metrics for middle managers to pour over and reach absurd conclusions from. Most of the interim scrums were not usable in the way the methodology projected they should have been. Maybe Boeing had a different experience than I did. Maybe they were able to spend more on training, tools and manpower than my outfit did. Still, I can't wrap my head around eight week scrums, that is just not much Agility.

In the big picture view, I think in one of Leeham's articles we read some caution around Boeing's next move due to the pall cast by the MAX, but I imagine they have to keep the engineering talent busy somehow, and presumably till someone comes up with a different plan, they're working on NMA.

I found the "extreme approach to model-based engineering" sentence to be enticing.

At some level it reads like the infamous "drug like rush" associated with the 787 program.

On the other hand, the 787 was launched a dozen years ago (2007) and technology sure has come a long way since then, especially when it comes to warehouse scale computing, the kind one would associate with the ability to "analyze aerodynamic flows and loads, create a manufacturing plan and distribute the design seamlessly throughout the supply chain" that is quoted in the article.

Boeing has hinted about great leaps in their development in past descriptions of NMA, and it seems TX is a proof point.

On the other hand, I think much of the same tech is available to all of their competitors.

It should be interesting to see the next clean sheet designs from all the industry players.

Tle last big one was A350, and insiders said it largely reused systems designed for the A380.

It'll be way cool when Airbus starts talking more about their next clean sheet.

Hopefully we see fruits from their research on laminar flow.

PEK777 wrote:
Interesting take, quite the REVELATION

Thanks, PEK777, I have enjoyed many of your posts over the last several years.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
TMccrury
Posts: 93
Joined: Sun Oct 09, 2005 10:24 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:26 am

Help me in my ignorance. What is a "SCRUM"? Thanks folks for educating an old man.
 
BoeingGuy
Posts: 6314
Joined: Fri Dec 10, 2010 6:01 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:33 am

The scrums are 15 minutes long. It’s the Sprints that are 2-4 weeks long.
 
BoeingGuy
Posts: 6314
Joined: Fri Dec 10, 2010 6:01 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:39 am

TMccrury wrote:
Help me in my ignorance. What is a "SCRUM"? Thanks folks for educating an old man.


It’s part of the Agile process. A Scrum is a daily tag-up of about 15 minutes to review progress. The Sprint is the time period of about 2-4 weeks where the work is done. A Demo is a longer status review after each Sprint ends.
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 1:51 am

BoeingGuy wrote:
The scrums are 15 minutes long. It’s the Sprints that are 2-4 weeks long.

Thank you for the correction.

It seems AvWeek described the 8 week scrums, and I followed their mistaken lead.

TMccrury wrote:
Help me in my ignorance. What is a "SCRUM"? Thanks folks for educating an old man.

The main idea behind Agile is that instead of one big deliverable at the end, the project generates lots of incremental "deliverables" as it progresses.

The Sprints are the entities that generate the deliverables.

In many cases of more agile environments like web sites, the incremental deliverables actually do get pushed out to users and do go live.

In other cases of more regulated environments, they end up being internal deliverables.

The main idea is each "deliverable" delivers incremental functionality to the end users, be it the public at large, or internal consumers.

The theology says that these incremental deliverables allow for lots of mid-course corrections as the usefulness of the incremental deliverables becomes clear.

Hopefully that's what other organizations see.

In my case it seemed none of that incremental feedback ever got delivered, so we ended up developing what we would develop in a traditional process, with greatly increased overhead due to the frequent meetings and the overbearing tools that used garbage inputs to generate all kinds of silly metrics that made countless middle managers angry.

As always, it sucks to be at the pointy end of the spear.

BoeingGuy wrote:
It’s part of the Agile process. A Scrum is a daily tag-up of about 15 minutes to review progress. The Sprint is the time period of about 2-4 weeks where the work is done. A Demo is a longer status review after each Sprint ends.

Care to share any thoughts about an "extreme approach to model-based engineering"?
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
MrBretz
Posts: 400
Joined: Tue Jun 14, 2016 9:13 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 2:10 am

Years ago, we tried to replace “classical” software design and development with rapid prototyping. It didn’t work too well in anything complicated. The words you are using sounds like new buzz words for rapid prototyping deliverables. I hope it goes better this time around. I moved on from large organizations to much smaller development organizations.
 
Ozair
Posts: 4277
Joined: Mon Jan 31, 2005 8:38 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 2:54 am

Revelation wrote:
I also found the Agile stuff interesting, in particular the eight week scrum. I attended in-depth training and was told that scrums should be one week long, and if not, at most two weeks long, to get the agility that came from Agile. Yet in the kind of work we were doing (embedded systems, lots of legacy "C" code that was impenetrable to modern IDE tools) most people in my organization felt we could never do one or two weeks. It's interesting to see Boeing end up with eight week scrums, since two months isn't a lot of "agility", IMHO. We ended up with four week scrums that often were just parts of bigger eight or sixteen week scrums. I was pushing for two week scrums but that required more scrum masters and other support people to engage in the process, so it never happened.

Overall, I wasn't a huge fan of Agile as deployed in my organization. It largely ended up being a methodology to generate metrics for middle managers to pour over and reach absurd conclusions from. Most of the interim scrums were not usable in the way the methodology projected they should have been. Maybe Boeing had a different experience than I did. Maybe they were able to spend more on training, tools and manpower than my outfit did. Still, I can't wrap my head around eight week scrums, that is just not much Agility.

In my time with Agile I found it really depended on the type of project and environment you were working on to how successful it was. Having worked in small commercial orgs where the rapid prototyping and feedback drove the design in different directions it worked well. On projects where the size and scale were simply too big Agile became a burden as it was impossible to break the tasks down into small enough chucks, a legacy of the software and the staff available (perhaps also not enough understanding of the agile process). I managed a sustainment team and found 4 week sprints worked really well in allowing for both continuing development and housekeeping but for the small rapid stuff two weeks worked best.

Revelation wrote:
Then the article goes on to say:

They used agile software development using eight week 'scrum' cycles which they claim reduced software development time by 50%.

They don't consider the first two models to be prototypes, they consider they are very close to being the same as the first EMD aircraft.

"The mold line doesn’t change in the airplane. That’s why we call it ‘EMD-ready.’”

The precision of the modelling allowed the initial assembly process to take 80% fewer hours than a traditional process would take

They developed all the mission and control computers in-house to eliminate the overhead of dealing with external vendors

How much of this is possible because the requirements, function and design of the T-X, and civilian airliners, is very static compared to more complex military aircraft?

For example the T-X used an off the shelf 30 year old engine and re-used parts and components from existing programs to get that rapid designing completed. When some/many of those components are unique to the aircraft, the example of a 5th gen fighter because of the LO considerations, how much saving would this process provide?

I see the benefit for NMA and future civilian platforms, not as sold on the military side when the process gets complicated and includes stakeholders that change their minds frequently.
 
JayinKitsap
Posts: 1489
Joined: Sat Nov 26, 2005 9:55 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 3:42 am

I posted this over in the Military thread, but it seems appropriate here:
_______________________________________________________

An article in Airforce Magazine on the T-X, some notable highlights:
- same cockpit display as on the F-15EX's being produced for Qutar.
- Saab building the aft fuselage while Boeing is building the cockpit forward, the wings, and empennage.
- Build rate 48 per year, with other sales could reach 60 per year.
- Airplane and trainer to have the same software, updated at the same time. Apps can be added simply.
- Plane pulls more than 8g.
- Most access panels open tool free.

http://www.airforcemag.com/Features/Pag ... craft.aspx
 
JayinKitsap
Posts: 1489
Joined: Sat Nov 26, 2005 9:55 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 3:52 am

The T-X design will not appear in the NMA, but the design process will. They went from conception to flying in 2-3 years, they wrote off about $750M on both the MQ-25 and the T-X. The Air Force liked the performance better than 2 already in production planes. It really says a lot about the approach. Imagine working a design for several years with 1,000 doing it.

Could the 8 week scrum be more a stage review where all of the inputs since the last are reviewed and ensures the model is cleaned up, excess code etc purged and a high powered full calculation done to see where the design is. Are we overweight by 2%, did the change to the fairings make the anticipated improvement or should we look at a different faring. etc.

One reason the NMA seems paused is the focus on the T-X is debugging the design process for the NMA.
 
ELBOB
Posts: 295
Joined: Sun Jun 21, 2015 6:56 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 6:04 am

It is bemusing to see Boeing boasting about Agile when the software world is now in the post-Agile era; certainly, some aspects of it are useful for developing a smartphone app but as mentioned above it doesn't 'fit' in big-scale, complex-dependency projects. It just leads to more meetings and more useless metrics and less work.

There's probably some axiom we could invent along the lines of "By the time Boeing adopts a methodology, all the consultants have retired to Florida"
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 11:58 am

JayinKitsap wrote:
The T-X design will not appear in the NMA, but the design process will. They went from conception to flying in 2-3 years, they wrote off about $750M on both the MQ-25 and the T-X. The Air Force liked the performance better than 2 already in production planes. It really says a lot about the approach. Imagine working a design for several years with 1,000 doing it.

It will be interesting to see if the process used by two relatively small projects scales up to something the size of NMA with so many firm requirements to satisfy in a timely fashion.

Could the 8 week scrum be more a stage review where all of the inputs since the last are reviewed and ensures the model is cleaned up, excess code etc purged and a high powered full calculation done to see where the design is. Are we overweight by 2%, did the change to the fairings make the anticipated improvement or should we look at a different faring. etc.

The real question I have is if the need to converge every eight weeks or so inhibits certain processes or optimizations that take more than eight weeks to achieve.

The old line is that a baby takes nine months and there's no way to get one baby every month for nine months. In my experience with Agile, the middle managers kept asking why each deliverable wasn't a baby, when looking at a bunch of cells in a uterus. They couldn't be arsed to engage in the process and understand why each deliverable wasn't a baby, they just wanted their bleeping baby right now.

ELBOB wrote:
It is bemusing to see Boeing boasting about Agile when the software world is now in the post-Agile era; certainly, some aspects of it are useful for developing a smartphone app but as mentioned above it doesn't 'fit' in big-scale, complex-dependency projects. It just leads to more meetings and more useless metrics and less work.

There's probably some axiom we could invent along the lines of "By the time Boeing adopts a methodology, all the consultants have retired to Florida"

I agree with your point of view, but it'd be interesting to see exactly what parts of their traditional methodology needed to be improved, and which parts Agile and/or other modern approaches could help.

For instance, most people feel continuous integration / continuous test is a big win, and while that's not necessarily an Agile thing, most people got the funding to do CI/CT as part of a shift to Agile.

From what I've read, the key is the extreme approach to model-based engineering that this article describes.

I think we can see how this could unlock CI/CT to a higher degree than non-digital and old-school digital approaches.

As I said above, the last clean sheet BCA did was a dozen years ago, and warehouse scale computing was in its infancy in 2007.

It will be interesting to learn more about TX and how it influences future BCA programs.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
musman9853
Posts: 835
Joined: Mon May 14, 2018 12:30 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 2:19 pm

didn't we already know this with the "project diamond" thing?
Welcome to the City Beautiful.
 
User avatar
JetBuddy
Posts: 2225
Joined: Wed Dec 25, 2013 1:04 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 2:47 pm

It's the model-based engineering that allows the use of the Agile way of doing things. If there was a need to physically manufacture parts, test them in complex rigs and wind tunnels, and rethink the process with every scrum, it wouldn't work.

Now they're designing the entire aircraft in the software. Every last bit of it. Changes and developments are done continously with a review every 8 weeks. An engineer can see the whole "picture" in the software while working on smaller things like nuts and bolts, and then run the simulation to see how it affects everything else.

I'm wondering if they've tried to implement machine learning in the process. That would be a game changer. Airbus is doing that on a smaller scale already.
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 3:13 pm

musman9853 wrote:
didn't we already know this with the "project diamond" thing?

Thanks for bringing that up.

I had forgotten about Project Black Diamond, manly because the articles that told you what it was required subscriptions that I don't have ( ref: https://aviationweek.com/awin-only/boei ... -cost-cuts and https://theaircurrent.com/aircraft-deve ... lo-to-nma/ and https://leehamnews.com/2018/10/01/how-b ... connected/ ).

I think this is the most detail we've gotten about the Project Black Diamond technology a non-subscription article.

In retrospect, the Leeham article had some interesting teasers in the non-subscription content such as:

As a former SAAB and Gripen fighter program employee, I asked if Boeing’s reason for the tie-up was Gripen’s systems and software (which is a big part of today’s aircraft programs). Chadwick answered, “Boeing’s primary interest in cooperating with SAAB for the T-X is their know-how to develop aircraft programs more efficiently and to produce them at lower cost.” I was surprised by the answer and didn’t get it at first.

Then articles appeared on how SAAB developed the Gripen New Generation (NG) fighter demonstrator faster and at a lower cost than budgeted. The full-scale development of the Gripen D, a 60% new version of the Gripen I worked on, was following the same pattern.

The key to the improvements was in changed work methods, more than in any technological advances. SAAB had tried out and perfected more efficient work methods and proofed them in several projects.

Boeing has also done trials with changed work methods, similar to SAAB’s, but lacked the program proof they worked. The T-X cooperation with SAAB became the proof project.


And:

Boeing has declared the successful T-X development as a raw model for all future Boeing development projects. The $9.2bn offered price for producing 351 T-X trainers and their ground training simulators, $10 billion below what the Air Force estimated it would cost, shows Boeing and SAAB are convinced about the Black Diamond advantages.

In summary, the T-X project is Boeing’s pilot for how to develop and produce future aircraft. The first BCA project using Black Diamond to the full is NMA. Boeing aims for a six years development period followed by a production phase at record low costs.

Given what we've been told about NMA (in active development, model being iterated to provide more data to help close the business case), Boeing must already have a pretty detailed digital model of NMA in house.

This is what made me skeptical about all the various a.net discussions about NMA's form factor. Boeing has always said it will be an ovoid small widebody, and everyone who has been briefed by Boeing has confirmed that. It seems obvious to me that they had already done detailed modelling to convince themselves about the form factor. If you see the picture in the AirCurrent article it's pretty clear they sort that out early in the process as they use the customer needs to drive the mission and business models.

JetBuddy wrote:
It's the model-based engineering that allows the use of the Agile way of doing things. If there was a need to physically manufacture parts, test them in complex rigs and wind tunnels, and rethink the process with every scrum, it wouldn't work.

Now they're designing the entire aircraft in the software. Every last bit of it. Changes and developments are done continously with a review every 8 weeks. An engineer can see the whole "picture" in the software while working on smaller things like nuts and bolts, and then run the simulation to see how it affects everything else.

That makes perfect sense to me.

It makes me wonder if they (like many others) are using Agile terminology for something that really is just state-of-the-art digital modelling and CI/CT infrastructure combined with (relatively) short periods between milestones/gates.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
Weatherwatcher1
Posts: 266
Joined: Sun Mar 03, 2019 5:14 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 3:22 pm

Thank you for opening this thread. It is one of the most informative thread about how engineering work is being done (as opposed to the other NMA threads).

It sounds like scrum teams are breaking work off into smaller pieces which are the sprints limited to a few weeks. I can imagine that this would work for various engineering drawing releases. Breaking down the frames, stringers, chords etc into specific sprints with deliverables could be a way to progress through the engineering work faster with fewer errors. I also imagine them going System by System as they create the assemblies and modules. Fundamentally airplanes consist of thousands and millions of parts just like software is thousands and millions of lines of code.

What I don’t understand is what model based engineering is. Can someone explain more?
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 4:53 pm

Weatherwatcher1 wrote:
Thank you for opening this thread. It is one of the most informative thread about how engineering work is being done (as opposed to the other NMA threads).

It sounds like scrum teams are breaking work off into smaller pieces which are the sprints limited to a few weeks. I can imagine that this would work for various engineering drawing releases. Breaking down the frames, stringers, chords etc into specific sprints with deliverables could be a way to progress through the engineering work faster with fewer errors. I also imagine them going System by System as they create the assemblies and modules. Fundamentally airplanes consist of thousands and millions of parts just like software is thousands and millions of lines of code.

What I don’t understand is what model based engineering is. Can someone explain more?

Wiki ( https://en.wikipedia.org/wiki/Model-bas ... ngineering ) seems to give a good definition (and lots of references too):

Model-based systems engineering (MBSE) is a systems engineering methodology that focuses on creating and exploiting domain models as the primary means of information exchange between engineers, rather than on document-based information exchange.

More recently, the focus has also started to cover aspects related to the model execution in computer simulation experiment, to further overcome the gap between the system model specification and the respective simulation software. As a consequence, the term modeling and simulation-based systems engineering (M&SBSE) has also been used along with MBSE.[1]


https://www.3dcadworld.com/why-you-need ... gineering/ expands on this some more:

Until recently, most engineering and manufacturing activities relied on hardcopy and/or digital documents (including 2D drawings) to convey engineering data and to drive manufacturing processes. With the advent of new manufacturing data format standards and more powerful engineering software, it is now possible to perform all engineering functions using data models. The model-based engineering (MBE) approach uses these models rather than documents as the data source for all engineering activities throughout the product life cycle. The core MBE tenet is that models are used to drive all aspects of the product lifecycle and that data is created once and reused by all downstream data consumers.

A model is a representation or idealization of the structure, behavior, operation, or other characteristics of a real-world system. A model is used to convey design information, simulate real world behavior, or specify a process. Engineers use models to convey product definition or otherwise define a product’s form, fit and function. In MBE, models can be applicable to a wide range of domains (systems, software, electronics, mechanics, human behavior, logistics, and manufacturing).

It's pretty clear that one of our suppliers is using MBE. All their info from the bit/signal level through register level to subsystem level is so detailed it's clear the data was created once and reused by all downstream consumers (nice web based search engine along with old school PDF documents).

It's pretty clear my current employer has nothing like this. Walking around their building one would find most tools (especially software) were modern 15 or so years ago and haven't changed much if any since. Makes me glad that I'm not that far away from choosing to retire.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
JayinKitsap
Posts: 1489
Joined: Sat Nov 26, 2005 9:55 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 7:38 pm

Weatherwatcher1 wrote:
Thank you for opening this thread. It is one of the most informative thread about how engineering work is being done (as opposed to the other NMA threads).

What I don’t understand is what model based engineering is. Can someone explain more?


The digital model is set to real dimensions (not a scaled drawing) with every part being correctly dimensioned with the real material properties included. Each part is a block (in autoCAD parlance), very small parts might have representative performance in the model (things like bolt threading is bearish to model with little value added, so it is designed for certain capacities and thread strength reviewed later). On an airplane above the bolt thread level the parts are true digital including stress properties (forgings are harder and less ductile on the surface than inside).

This model has the surfaces defined so the aero performance can be evaluated, concurrent with this, the model is structurally evaluated with its deflected shape fed back to the aero performance, the loads to each actuator are provided. It would be a step up but not too hard to check the actuator design in this model too. Output would confirm the electrical & hydraulic draws. Things like cables, ducts, and tubing are in the model so vibration on these can be assessed etc. Basically it is a real plane existing within the computer, not in reality. Things like thermal expansion, the bounce on landing, the cabin pressurization.

Anyone that has done linear algebra understands the idea of solving for n variables with m equations, it is simple to solve a 2 variable 2 equation case (a 2x2 matrix) but by hand inverting (solving the equations) the matrix is quite tough in a 4x4 matrix). The nodes in a Finite Element Model (FEM) have 6 variables and need 6 equations to solve. My RISA 3D program I have for building design can handle 100K nodes, 32k members, and 100K plate elements, basically a matrix that is a million variables in a million lines. To solve it goes thru and 'tugs' each node 1 unit, the 2nd run compares to the first, some models it takes 40-50 iterations to get the answer to close withing 1%. For a full airplane with a super hot parallel computing mainframe I would guess it is an overnight run (possibly a day or two) to check all load cases. All of these results would be stored and be available as possibly 500 stations are concurrently working the model. One station might be working changes to the rudder hinge, other working the frame around a window, another a door frame, etc. The computer would check the changes to an accuracy of 2 or 3% immediately based on the last main run results, the 40# part gets a 1 or 2# diet, the decision to use a new fastener for a heavy loaded part so most fasteners can be lighter are made etc. The faring skin is thinned by a few mils etc. At the next main run these preliminary results are confirmed, the model cleared and the aero / stress is checked fully for all load cases.

The first cycle will be far off, overloads of 60+%, excess weight everywhere, but by the 5th or 6th cycle the weight improvement might be just a few percent. Getting the manufacturing part analysis done I still would think is done outside of the model, but the model would be changed to reflect the part being made by 3D printing, cut metals, composites etc. The full analysis would be doing the CFRP layup design based on the inputted parameters.

The really cool thing with this is the model can address tolerances too, so the produced parts come out fitting perfectly the first time. That fitting perfectly makes the assembly hours drop dramatically. Anyone that has assembled a gas grill realize that the few parts that it was a fight to fit took most of the assembly time.
 
cpd
Posts: 6017
Joined: Sat Jun 28, 2008 4:46 am

Re: Boeing TX: A Precursor to Boeing NMA?

Fri May 17, 2019 8:09 pm

BoeingGuy wrote:
TMccrury wrote:
Help me in my ignorance. What is a "SCRUM"? Thanks folks for educating an old man.


It’s part of the Agile process. A Scrum is a daily tag-up of about 15 minutes to review progress. The Sprint is the time period of about 2-4 weeks where the work is done. A Demo is a longer status review after each Sprint ends.


The retros will be enormous for something like this...

I work in agile at the moment (I’m a scrum master) - it’s great for what we do. But using it for an aircraft, that’s interesting.
 
User avatar
kitplane01
Posts: 1350
Joined: Thu Jun 16, 2016 5:58 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 7:23 am

JayinKitsap wrote:
Weatherwatcher1 wrote:
Thank you for opening this thread. It is one of the most informative thread about how engineering work is being done (as opposed to the other NMA threads).

What I don’t understand is what model based engineering is. Can someone explain more?


The digital model is set to real dimensions (not a scaled drawing) with every part being correctly dimensioned with the real material properties included. Each part is a block (in autoCAD parlance), very small parts might have representative performance in the model (things like bolt threading is bearish to model with little value added, so it is designed for certain capacities and thread strength reviewed later). On an airplane above the bolt thread level the parts are true digital including stress properties (forgings are harder and less ductile on the surface than inside).

This model has the surfaces defined so the aero performance can be evaluated, concurrent with this, the model is structurally evaluated with its deflected shape fed back to the aero performance, the loads to each actuator are provided. It would be a step up but not too hard to check the actuator design in this model too. Output would confirm the electrical & hydraulic draws. Things like cables, ducts, and tubing are in the model so vibration on these can be assessed etc. Basically it is a real plane existing within the computer, not in reality. Things like thermal expansion, the bounce on landing, the cabin pressurization.

Anyone that has done linear algebra understands the idea of solving for n variables with m equations, it is simple to solve a 2 variable 2 equation case (a 2x2 matrix) but by hand inverting (solving the equations) the matrix is quite tough in a 4x4 matrix). The nodes in a Finite Element Model (FEM) have 6 variables and need 6 equations to solve. My RISA 3D program I have for building design can handle 100K nodes, 32k members, and 100K plate elements, basically a matrix that is a million variables in a million lines. To solve it goes thru and 'tugs' each node 1 unit, the 2nd run compares to the first, some models it takes 40-50 iterations to get the answer to close withing 1%. For a full airplane with a super hot parallel computing mainframe I would guess it is an overnight run (possibly a day or two) to check all load cases. All of these results would be stored and be available as possibly 500 stations are concurrently working the model. One station might be working changes to the rudder hinge, other working the frame around a window, another a door frame, etc. The computer would check the changes to an accuracy of 2 or 3% immediately based on the last main run results, the 40# part gets a 1 or 2# diet, the decision to use a new fastener for a heavy loaded part so most fasteners can be lighter are made etc. The faring skin is thinned by a few mils etc. At the next main run these preliminary results are confirmed, the model cleared and the aero / stress is checked fully for all load cases.

The first cycle will be far off, overloads of 60+%, excess weight everywhere, but by the 5th or 6th cycle the weight improvement might be just a few percent. Getting the manufacturing part analysis done I still would think is done outside of the model, but the model would be changed to reflect the part being made by 3D printing, cut metals, composites etc. The full analysis would be doing the CFRP layup design based on the inputted parameters.

The really cool thing with this is the model can address tolerances too, so the produced parts come out fitting perfectly the first time. That fitting perfectly makes the assembly hours drop dramatically. Anyone that has assembled a gas grill realize that the few parts that it was a fight to fit took most of the assembly time.


Question: Are you describing real technology that is actually used, or what "could be"? Can something like CATIA do everything you wrote in so seamless a fashion?
 
User avatar
bikerthai
Posts: 2965
Joined: Wed Apr 28, 2010 1:45 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 2:32 pm

kitplane01 wrote:
Question: Are you describing real technology that is actually used, or what "could be"? Can something like CATIA do everything you wrote in so seamless a fashion?


It have been used to varying degrees on various programs in the company. CATIA by itself can not do it all. There is a suite of software that is used to bring it all together.

While the Algile part was something new to me, the manufacturing aspect of the Diamond Project have been implemented on various sub components of current commercial aircraft production. The TX and Tanker drone will probably be the first to fully employ the process.

The process is IP property and is guarded, so it would take any competitor time and money to come up something on their own without running into pattent issues.

bt
Intelligent seeks knowledge. Enlightened seeks wisdom.
 
JayinKitsap
Posts: 1489
Joined: Sat Nov 26, 2005 9:55 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 4:51 pm

kitplane01 wrote:
JayinKitsap wrote:
Weatherwatcher1 wrote:
Thank you for opening this thread. It is one of the most informative thread about how engineering work is being done (as opposed to the other NMA threads).

What I don’t understand is what model based engineering is. Can someone explain more?


The digital model is set to real dimensions (not a scaled drawing) with every part being correctly dimensioned with the real material properties included. Each part is a block (in autoCAD parlance), very small parts might have representative performance in the model (things like bolt threading is bearish to model with little value added, so it is designed for certain capacities and thread strength reviewed later). On an airplane above the bolt thread level the parts are true digital including stress properties (forgings are harder and less ductile on the surface than inside).

This model has the surfaces defined so the aero performance can be evaluated, concurrent with this, the model is structurally evaluated with its deflected shape fed back to the aero performance, the loads to each actuator are provided. It would be a step up but not too hard to check the actuator design in this model too. Output would confirm the electrical & hydraulic draws. Things like cables, ducts, and tubing are in the model so vibration on these can be assessed etc. Basically it is a real plane existing within the computer, not in reality. Things like thermal expansion, the bounce on landing, the cabin pressurization.

Anyone that has done linear algebra understands the idea of solving for n variables with m equations, it is simple to solve a 2 variable 2 equation case (a 2x2 matrix) but by hand inverting (solving the equations) the matrix is quite tough in a 4x4 matrix). The nodes in a Finite Element Model (FEM) have 6 variables and need 6 equations to solve. My RISA 3D program I have for building design can handle 100K nodes, 32k members, and 100K plate elements, basically a matrix that is a million variables in a million lines. To solve it goes thru and 'tugs' each node 1 unit, the 2nd run compares to the first, some models it takes 40-50 iterations to get the answer to close withing 1%. For a full airplane with a super hot parallel computing mainframe I would guess it is an overnight run (possibly a day or two) to check all load cases. All of these results would be stored and be available as possibly 500 stations are concurrently working the model. One station might be working changes to the rudder hinge, other working the frame around a window, another a door frame, etc. The computer would check the changes to an accuracy of 2 or 3% immediately based on the last main run results, the 40# part gets a 1 or 2# diet, the decision to use a new fastener for a heavy loaded part so most fasteners can be lighter are made etc. The faring skin is thinned by a few mils etc. At the next main run these preliminary results are confirmed, the model cleared and the aero / stress is checked fully for all load cases.

The first cycle will be far off, overloads of 60+%, excess weight everywhere, but by the 5th or 6th cycle the weight improvement might be just a few percent. Getting the manufacturing part analysis done I still would think is done outside of the model, but the model would be changed to reflect the part being made by 3D printing, cut metals, composites etc. The full analysis would be doing the CFRP layup design based on the inputted parameters.

The really cool thing with this is the model can address tolerances too, so the produced parts come out fitting perfectly the first time. That fitting perfectly makes the assembly hours drop dramatically. Anyone that has assembled a gas grill realize that the few parts that it was a fight to fit took most of the assembly time.


Question: Are you describing real technology that is actually used, or what "could be"? Can something like CATIA do everything you wrote in so seamless a fashion?


I am in building design, my RISA 3D costs for a single station $1,600 /year. I do 3 and 4 story building designs, I have to enter myself the wind and floor loads because I have not bought all the added modules. But it does the seismic dynamic analysis fast (minutes what used to take hours), resizes the members as needed, then exports to either AutoCAD or Revit the model, in Revit it has the shapes included, an added module details the steel including the connections ready for shop fabrication. It is working only on the elastic state in its modelling, I am sure B's includes full plastic state on the deflected shape.

A quote from the OP's link follows, that pretty well confirms what I noted:
The team decided to adopt an extreme approach to model-based engineering. It’s a discipline of product design that calls for constructing an elaborate, three-dimensional digital model of the entire aircraft. The model allows engineers to analyze aerodynamic flows and loads, create a manufacturing plan and distribute the design seamlessly throughout the supply chain. Although the concept of model-based engineering is not unique to Boeing in the aerospace industry, Niewald believes that Boeing took it further than the competition.


I see the results of this in the 777X, it is probably a half digital model, the original hull is in the model but not in full definition as its parts are staying as is. But the wing box and wings are in the model. Anyone notice how it went thru assembly slow but with precision, no glitches popping up where there is a 3 month delay, like on the 787.

On the T-X the first 2 planes were assembled in rather quick time as all the parts fit. An 8g supersonic fighter was flying with the design performance and good cost data to bid on the design in 3 years. Yes the engine is tried and true but design for subsonic and supersonic flight as well as the transition is tough fluid dynamics. As BikerThai noted this the Boeing's crown jewels and is being secured like state secrets should be. This design suite I am sure cost several hundred million to pull together.
 
FluidFlow
Posts: 306
Joined: Wed Apr 10, 2019 6:39 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 5:10 pm

They will probably use a 3D finite Volume model for calculating the boundary conditions (wing/air) and then use the finite element method to calculate the load in the wing. As this will need a lot of computational power it could easely take multiple days up to weeks to get a solution.

We needed multiple weeks when we did planetary formation models. But this was 10 years ago so it is probably faster now.
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 5:57 pm

JayinKitsap wrote:

The digital model is set to real dimensions (not a scaled drawing) with every part being correctly dimensioned with the real material properties included. Each part is a block (in autoCAD parlance), very small parts might have representative performance in the model (things like bolt threading is bearish to model with little value added, so it is designed for certain capacities and thread strength reviewed later). On an airplane above the bolt thread level the parts are true digital including stress properties (forgings are harder and less ductile on the surface than inside).

This model has the surfaces defined so the aero performance can be evaluated, concurrent with this, the model is structurally evaluated with its deflected shape fed back to the aero performance, the loads to each actuator are provided. It would be a step up but not too hard to check the actuator design in this model too. Output would confirm the electrical & hydraulic draws. Things like cables, ducts, and tubing are in the model so vibration on these can be assessed etc. Basically it is a real plane existing within the computer, not in reality. Things like thermal expansion, the bounce on landing, the cabin pressurization.


Thanks for providing an understandable framework of model based engineering..... "real plane existing within the computer", indeed !

JayinKitsap wrote:
For a full airplane with a super hot parallel computing mainframe I would guess it is an overnight run (possibly a day or two) to check all load cases. All of these results would be stored and be available as possibly 500 stations are concurrently working the model. One station might be working changes to the rudder hinge, other working the frame around a window, another a door frame, etc. The computer would check the changes to an accuracy of 2 or 3% immediately based on the last main run results, the 40# part gets a 1 or 2# diet, the decision to use a new fastener for a heavy loaded part so most fasteners can be lighter are made etc. The faring skin is thinned by a few mils etc. At the next main run these preliminary results are confirmed, the model cleared and the aero / stress is checked fully for all load cases.


I work with some respectable computing iron, but this makes my head hurt.
 
User avatar
kitplane01
Posts: 1350
Joined: Thu Jun 16, 2016 5:58 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 6:22 pm

bikerthai wrote:
kitplane01 wrote:
Question: Are you describing real technology that is actually used, or what "could be"? Can something like CATIA do everything you wrote in so seamless a fashion?


It have been used to varying degrees on various programs in the company. CATIA by itself can not do it all. There is a suite of software that is used to bring it all together.

While the Algile part was something new to me, the manufacturing aspect of the Diamond Project have been implemented on various sub components of current commercial aircraft production. The TX and Tanker drone will probably be the first to fully employ the process.

The process is IP property and is guarded, so it would take any competitor time and money to come up something on their own without running into pattent issues.

bt


Can you tell more about this? The CATIA model gets reformatted to the other modules? How hard is this in practice? Can I just click a few menus and get flutter/noise/electrical analysis?
 
User avatar
zululima
Posts: 460
Joined: Sat Jul 14, 2007 10:21 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sat May 18, 2019 7:23 pm

I would be interested in having the magnitude of change from the original 777 to this explained. In the early 90s the 777 was hailed as the first all-digital design airliner, where a 3D model was used and could be manipulated to check clearances, the ability to fit MX tools and hands, etc. all before being built. Nearing three decades later, the improvements must be astounding. I wonder how much those old CAD designs were a pain to work with compared to today; although at the time I'm sure everyone was just thrilled to bin their slide-rule.

Another potential implication for massive digital workflow would be on the certification process. It seems that this approach would not just allow reduced time and cost to manufacture, but could also trim the ridiculous year-plus test flying campaign. How involved is the FAA in the digital design process? Would they allow a higher percentage of "computer-proving" vs. real-world proving? How much time do Boeing expect to save by not having as many post-production problems to solve such as the APU airflow issue on the 787, the gear-door thing with the 747-8, or the dreaded "flutter" from the 747-100?
I didn't get a 'Harumph' outta that guy!
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 12:31 am

zululima wrote:
I would be interested in having the magnitude of change from the original 777 to this explained. In the early 90s the 777 was hailed as the first all-digital design airliner, where a 3D model was used and could be manipulated to check clearances, the ability to fit MX tools and hands, etc. all before being built. Nearing three decades later, the improvements must be astounding. I wonder how much those old CAD designs were a pain to work with compared to today; although at the time I'm sure everyone was just thrilled to bin their slide-rule.

It seems most of us don't know the current setup, and those who know won't say.

I've found a few references to what tech the 777 program used:

The Boeing 777 is the first jetliner to be 100 percent digitally designed using three-dimensional solids technology. Throughout the design process, the airplane was "preassembled" on the computer, eliminating the need for a costly, full-scale mock-up. The 777 Division is using CATIA (computer-aided, three-dimensional interactive application) and ELFINI (Finite Element Analysis System), both developed by Dassault Systemes of France and licensed in the United States through IBM. Designers also use EPIC (Electronic Preassembly Integration on CATIA) and other digital preassembly applications developed by Boeing.

Ref: https://www.mura.org/websites/me39c.me. ... catia.html

And:

Instead of producing drawings on reams of paper, designers store their work inside the memory of eight IBM mainframes. In the Puget Sound area, approximately 1,700 individual computer workstations are linked to the largest mainframe installation of its kind in the world, consisting of four IBM mainframe computers. This mainframe cluster also is linked to mainframes and workstation installations in Japan; Wichita, Kan.; Philadelphia; and other locations.

Ibid.

And:

At the height of the design process, Boeing had more than 2,2000 workstations networked into eight IBM 3090-600 mainframe computers.

Ref: https://books.google.com/books?id=OEniu ... 90&f=false

So if we go with the high end estimate, we see eight IBM 3090-600 mainframes serving 2200 workstations.

Each one looked like:

Image

Each of the six extensions of the "main frame" holds what we would now refer to as one "core" that ran at 17.2 nanosecond clock rate, or in modern terms, 60 MHz (megahertz, not gigahertz).

http://www.chilton-computing.org.uk/ccd ... s/p009.htm suggested it could access a total of 256 megabytes (not gigabytes) of central (directly addressed) storage and 1024 megabytes (i.e. 1 gigabyte) of expanded storage (basically bulk storage that needed special instructions to access so was only somewhat useful).

So in very round terms, we have a six core server running at 60 megahertz with 256 megabytes of memory serving 2200/8 = 275 workstations.

In round terms each one of these mainframes cost $12 million, before you bought any of the other stuff you needed to go with it, such as disk drives, workstations, network connections, etc.

Each of the eight mainframes could have easily cost twice the $12M figure by the time you bought all the stuff you needed to run it, and they consumed massive amounts of electricity.

Each designer working with the system would (if they were lucky) would be using this kind of terminal (IBM 6090):

Image

If they weren't lucky, they were given an IBM 5080, which is so old I'm having a hard time finding a picture of it.

I do have a connection to all this stuff. In the late 80s / early 90s I was an IBM employee and worked for the part of that giant corporation (over 425,000 employees!) that sold this kind of stuff.

I remember seeing internal news releases when Boeing decided they needed to add two more mainframes to their cluster.

We loved hearing that news, it was in essence job security for all of us.

Please do keep in mind Boeing was buying this stuff around 30 years ago.

Soooo much has changed how we do things.

And yes, a lot has changed since 2007 when Boeing launched the 787.

2007 was the year of the first iPhone (yes, generation 1) and the Intel Core Duo.

Interestingly enough, single core computer speed really hasn't increased since 2007, see below.

What has increased is the number of cores, memory bandwidth, network bandwidth, and cheap computes from GPUs.

Instead of mainframes, we now talk in terms of warehouse scale computing.

Image
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
User avatar
Super80Fan
Posts: 1622
Joined: Thu Jul 06, 2017 4:14 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 12:35 am

Glad to see some innovation happening back at Boeing as opposed to slapping larger engines on a 1960's design and calling it a day.
RIP McDonnell Douglas
RIP US Airways
 
JayinKitsap
Posts: 1489
Joined: Sat Nov 26, 2005 9:55 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 6:15 am

There is also quantum computing entering the market, they are approaching the best supercomputer today with far higher growth rates in speed. In 1998 at 2, 2006 at 12 and 2018 at 72 qubits, an expected 2019 machine may hit 128. The quantum computer is quite powerful but has a higher error rate. Modelling fluid flow looks quite promising in it, but not so sure about the structural calculations. Possibly design the model on the supercomputer, but insert that model into the quantum for the more detailed analysis.

https://www.zerohedge.com/news/2019-05- ... ing-growth

https://www.technologyreview.com/s/6128 ... computing/

https://www.technologyreview.com/s/6135 ... supremacy/

Zulima - I am old school in a lot of ways, we should physically test the hell out of the planes, just to confirm the design and manufacturer process. If the design process is correct, certification testing should be a breeze. But we all have seen breezes that turn into tempests.


Revelation wrote:
zululima wrote:
I would be interested in having the magnitude of change from the original 777 to this explained. In the early 90s the 777 was hailed as the first all-digital design airliner, where a 3D model was used and could be manipulated to check clearances, the ability to fit MX tools and hands, etc. all before being built. Nearing three decades later, the improvements must be astounding. I wonder how much those old CAD designs were a pain to work with compared to today; although at the time I'm sure everyone was just thrilled to bin their slide-rule.

It seems most of us don't know the current setup, and those who know won't say.

I've found a few references to what tech the 777 program used:

The Boeing 777 is the first jetliner to be 100 percent digitally designed using three-dimensional solids technology. Throughout the design process, the airplane was "preassembled" on the computer, eliminating the need for a costly, full-scale mock-up. The 777 Division is using CATIA (computer-aided, three-dimensional interactive application) and ELFINI (Finite Element Analysis System), both developed by Dassault Systemes of France and licensed in the United States through IBM. Designers also use EPIC (Electronic Preassembly Integration on CATIA) and other digital preassembly applications developed by Boeing.

Ref: https://www.mura.org/websites/me39c.me. ... catia.html

And:

Instead of producing drawings on reams of paper, designers store their work inside the memory of eight IBM mainframes. In the Puget Sound area, approximately 1,700 individual computer workstations are linked to the largest mainframe installation of its kind in the world, consisting of four IBM mainframe computers. This mainframe cluster also is linked to mainframes and workstation installations in Japan; Wichita, Kan.; Philadelphia; and other locations.

Ibid.

And:

At the height of the design process, Boeing had more than 2,2000 workstations networked into eight IBM 3090-600 mainframe computers.

Ref: https://books.google.com/books?id=OEniu ... 90&f=false

So if we go with the high end estimate, we see eight IBM 3090-600 mainframes serving 2200 workstations.

Each one looked like:

Image

Each of the six extensions of the "main frame" holds what we would now refer to as one "core" that ran at 17.2 nanosecond clock rate, or in modern terms, 60 MHz (megahertz, not gigahertz).

http://www.chilton-computing.org.uk/ccd ... s/p009.htm suggested it could access a total of 256 megabytes (not gigabytes) of central (directly addressed) storage and 1024 megabytes (i.e. 1 gigabyte) of expanded storage (basically bulk storage that needed special instructions to access so was only somewhat useful).

So in very round terms, we have a six core server running at 60 megahertz with 256 megabytes of memory serving 2200/8 = 275 workstations.

In round terms each one of these mainframes cost $12 million, before you bought any of the other stuff you needed to go with it, such as disk drives, workstations, network connections, etc.

Each of the eight mainframes could have easily cost twice the $12M figure by the time you bought all the stuff you needed to run it, and they consumed massive amounts of electricity.

Each designer working with the system would (if they were lucky) would be using this kind of terminal (IBM 6090):

Image

If they weren't lucky, they were given an IBM 5080, which is so old I'm having a hard time finding a picture of it.

I do have a connection to all this stuff. In the late 80s / early 90s I was an IBM employee and worked for the part of that giant corporation (over 425,000 employees!) that sold this kind of stuff.

I remember seeing internal news releases when Boeing decided they needed to add two more mainframes to their cluster.

We loved hearing that news, it was in essence job security for all of us.

Please do keep in mind Boeing was buying this stuff around 30 years ago.

Soooo much has changed how we do things.

And yes, a lot has changed since 2007 when Boeing launched the 787.

2007 was the year of the first iPhone (yes, generation 1) and the Intel Core Duo.

Interestingly enough, single core computer speed really hasn't increased since 2007, see below.

What has increased is the number of cores, memory bandwidth, network bandwidth, and cheap computes from GPUs.

Instead of mainframes, we now talk in terms of warehouse scale computing.

Image
 
YIMBY
Posts: 630
Joined: Tue Sep 20, 2016 4:32 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 9:06 am

TMccrury wrote:
Help me in my ignorance. What is a "SCRUM"? Thanks folks for educating an old man.


Scrum is a project managing protocol commonly used in software development, instead of PRINCE2/PMP etc that are more commonly used in hardware development. Their main purpose is to give middle management a sense of control. They work quite well in controlled environments where surprises and deviations are to be minimized at the cost of inhibiting innovations. Agile methods like Scrum assume that the customers do not know what they want in the beginning but the requirements and priorities may change during the course of the project. (Truly the programmers do not know how they will be building the code until they learn what it is all about.) Non-Agile protocols assume that the specifications and timelines are fixed at the start and the goal is to fulfil those as accurately as possible.
 
smokeybandit
Posts: 1077
Joined: Mon Mar 17, 2014 3:24 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 11:26 am

The times I've been on projects that claimed to use agile, they just used agile terms to describe traditional waterfall concepts.

If they had an 8 week scrum, then I'm guessing the same is true here, too.


"Agile methods like Scrum assume that the customers do not know what they want in the beginning but the requirements and priorities may change during the course of the project. "

In my 20 years as a software engineer, I've never been on a project where clients truly knew what they wanted and didn't have changing requirements/priorities throughout.
Last edited by smokeybandit on Sun May 19, 2019 11:28 am, edited 1 time in total.
 
RJMAZ
Posts: 1639
Joined: Sat Jul 09, 2016 2:54 am

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 11:27 am

I've been telling everyone on here about this for the last year. Here is a post from 8 months ago.

viewtopic.php?f=10&t=1027931&p=20695851

"The new design software allows you to design in parallel. 100 parts can be open on 100 different computers at once with 100 engineers working on it simultaneously. The complex design gets updated in real time.

Previously computer design you could open up a single part, build it, shape it and save it. Effectively designing in series. You would have to break the big design up into sections to get different teams working in parallel on it, but each big section would still get built in series. But then you get the A380 where the two sections didnt fit together. High risk.

Also this new design software with the design coming together in real time can be digitally tested for load. Fatigue testing in the digital world. Previously you could only optimising the individual parts for weight but now you could optimise the full combined structure. Real world result, lower weight, longer life, less repair work and design fixes to restore fatigue life.

Fluid dynamic guys can also be digital wind tunnel testing the design. You want 9G and high alpha? You got it. No real flight testing, no prototyping, you'll turn like an F-35 with the very first production ready aircraft off the line.

Then you have stealth. Planform alignment the design with a click of a button. So all doors and panel join lines you can simply have a snap to function at a certain angle. Test the radar cross section in the digital world."


And another post by me:
viewtopic.php?f=3&t=1402809&p=20693889


"In the last 5 years there has been expoential advancement in digitaldesign of structures. You can now have networked teams working parallel on hundreds of computers. Fluid dynamic guys can be updating the model while the structure guys are working on it. Flight testing can be done digitally which brings the schedule forward. The Boeing trainer aircraft went from contract signed to flying production ready prototype in 12 months.

Cost of development will fall for something of equal complexity. It would not surprise me if the A350 cost less than half to develop as the 787 as they were 4 years apart."
 
User avatar
zeke
Posts: 13995
Joined: Thu Dec 14, 2006 1:42 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 12:23 pm

RJMAZ wrote:
Previously computer design you could open up a single part, build it, shape it and save it. Effectively designing in series. You would have to break the big design up into sections to get different teams working in parallel on it, but each big section would still get built in series. But then you get the A380 where the two sections didnt fit together. High risk.


You don’t know what you are talking about. CATIA has been in use for decades and thousands of engineers and technicians concurrently access the digital models all the time.

It has CFD, FEM/FEA etc built in. It can drive robots and machines for manufacturing directly. It can do quality control for parts built.

The A380 didn’t have problems with sections not fitting. The had a wiring routing issue due to one site using V4 and another using V5.

CATIA was used by Airbus well before the A380, remember both airbus and Boeing using CATIA on IBM RS6000 workstations in the 1980s.

The only competition CATIA had was Pro/ENGINEER and Unigraphics (whatever they are called these days) but never matched the functionality or capability or CATIA.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
Noshow
Posts: 1021
Joined: Wed Jun 15, 2016 3:20 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 12:32 pm

They later had some material's temperature tensions issue as well with early wing ribs made of both CFRP and metal.
Today's software can anticipate this.
 
frmrCapCadet
Posts: 3152
Joined: Thu May 29, 2008 8:24 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 2:04 pm

I've been telling everyone on here about this for the last year. Here is a post from 8 months ago.

viewtopic.php?f=10&t=1027931&p=20695851


You need to give a post number
Buffet: the airline business...has eaten up capital...like..no other (business)
 
User avatar
Revelation
Topic Author
Posts: 21475
Joined: Wed Feb 09, 2005 9:37 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:10 pm

zeke wrote:
RJMAZ wrote:
Previously computer design you could open up a single part, build it, shape it and save it. Effectively designing in series. You would have to break the big design up into sections to get different teams working in parallel on it, but each big section would still get built in series. But then you get the A380 where the two sections didnt fit together. High risk.

You don’t know what you are talking about. CATIA has been in use for decades and thousands of engineers and technicians concurrently access the digital models all the time.

It has CFD, FEM/FEA etc built in. It can drive robots and machines for manufacturing directly. It can do quality control for parts built.

The A380 didn’t have problems with sections not fitting. The had a wiring routing issue due to one site using V4 and another using V5.

CATIA was used by Airbus well before the A380, remember both airbus and Boeing using CATIA on IBM RS6000 workstations in the 1980s.

The only competition CATIA had was Pro/ENGINEER and Unigraphics (whatever they are called these days) but never matched the functionality or capability or CATIA.

All you say is true, yet we have a prototype for a supersonic trainer being flown in a remarkably short period of time that will be sold for significantly cheaper than the competition can offer, so clearly we've moved on beyond the use of CATIA in the 90s.

As we read in #23:

bikerthai wrote:
kitplane01 wrote:
Question: Are you describing real technology that is actually used, or what "could be"? Can something like CATIA do everything you wrote in so seamless a fashion?


It have been used to varying degrees on various programs in the company. CATIA by itself can not do it all. There is a suite of software that is used to bring it all together.

While the Algile part was something new to me, the manufacturing aspect of the Diamond Project have been implemented on various sub components of current commercial aircraft production. The TX and Tanker drone will probably be the first to fully employ the process.

The process is IP property and is guarded, so it would take any competitor time and money to come up something on their own without running into pattent issues.

So an insider is telling us this about more than CATIA, and my earlier post shows it was more than just CATIA even back in the 777 days.

From the main article:

The team decided to adopt an extreme approach to model-based engineering. It’s a discipline of product design that calls for constructing an elaborate, three-dimensional digital model of the entire aircraft. The model allows engineers to analyze aerodynamic flows and loads, create a manufacturing plan and distribute the design seamlessly throughout the supply chain. Although the concept of model-based engineering is not unique to Boeing in the aerospace industry, Niewald believes that Boeing took it further than the competition.

So we can presume many of the original tools are updated to new versions or replaced by new ones that allow for the scope of the work to cover design, manufacturing, and supply chain.

I can also see where what RJMAZ writes is true, that earlier versions of the tools did not allow for as many engineers to be updating the same parts of the model at the same time whilst also keeping the overall model intact.

I can also see how, relative to the days of the 787 in 2007, the huge increase in compute power and network speed provides to power to do things incrementally and continuously instead of the old days of locking things down while doing an update cycle.

And yes, we read that Boeing has some proprietary IP in this space and the article feels they have advantages over their competition (and in this context that is LM rather than Airbus, but clearly the same tech can be apply to airliners too), but that doesn't mean other industry players don't have similar stuff, or can't get similar stuff fairly quickly, presuming they're willing to spend on a similar scale.

This article gives some actual data about the impact these approaches are having.

It'd be nice if someone would share info about what the competitors are doing instead of just nay-saying.
Wake up to find out that you are the eyes of the world
The heart has its beaches, its homeland and thoughts of its own
Wake now, discover that you are the song that the morning brings
The heart has its seasons, its evenings and songs of its own
 
User avatar
zeke
Posts: 13995
Joined: Thu Dec 14, 2006 1:42 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:26 pm

I can guarantee you that thousands more engineers and technicians work collaboratively on civil aero projects like the 787 than on the TX.

Civil projects as far more complex than many military projects due to the size and complexity of the aircraft.

With all due respect, the T-X is very simple project compared to a 787, and even the NMA.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:27 pm

zeke wrote:
RJMAZ wrote:
Previously computer design you could open up a single part, build it, shape it and save it. Effectively designing in series. You would have to break the big design up into sections to get different teams working in parallel on it, but each big section would still get built in series. But then you get the A380 where the two sections didnt fit together. High risk.



The A380 didn’t have problems with sections not fitting. The had a wiring routing issue due to one site using V4 and another using V5.



and you have just proven his point. having separate sites, running separate versions of a software application fundamentally means that CATIA, (as utilized at that time) was not performing parallel design against a single model in real time; which is the crux of the advancement being touted in this TX claim. RJMAZ description of the process at that time is accurate.

and the A380 most certainly had sections that did not fit..... from a design and manufacturing perspective, wiring that did not mate up correctly in real life (due to the lack of centralized software modeling) are "sections" ; the semantics of what constitute an airframe section are irrelevant. No wiring, no airplane.
 
User avatar
sassiciai
Posts: 1091
Joined: Mon Jan 07, 2013 8:26 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:32 pm

I am retired after 40+ years in the software industry. Much of the above goes over my head. but there are elements from my own experience that worry about all of this, at least a wee bit!

I was "initiated" into software production on an RAF air/defense radar system in the early 1970s, and that was pure rugby, without the scrums, and without the reviews! Not a good professional start - even my university had better QC, and even in those days, a bit of QA! Cost plus 8% contract conditions!

For the majority of my software production career, waterfall methodology was king, along with tight specifications and costed Change Control. It lead to frequent arguments about things being in or out of scope, but generally got the job done! More relaxed contracts on a Time&Materials basis gave the customer total flexibility, but often led to the project failing to deliver

Trying the short-cuts of iterative development (aka scrumming) proved in my days to be a shortcut that lead into the swamp, that then led off into an entirely separate historical category that belongs on a different forum!

The waterfall process, tightly linked such that changes going down the left side are mirrored by changes in the right side for future testing, has stood the test of time. Maybe the bit at the bottom of this V-model can be turned into a rugby game, but as long as the V-model rules are not broken! The V-model at its higher levels includes the customer, and the real end client!
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:37 pm

zeke wrote:
I can guarantee you that thousands more engineers and technicians work collaboratively on civil aero projects like the 787 than on the TX.

Civil projects as far more complex than many military projects due to the size and complexity of the aircraft.

With all due respect, the T-X is very simple project compared to a 787, and even the NMA.


You underestimate Boeing (and Airbus) understanding and ability to scale their own design/manufacture/supply process from small to large.
This is primarily driven by the rapid evolution in compute power and hyper scale analytics.

What I am read out of this is that design systems today are capable of near realtime modeling / simultaneous component design at a scale far beyond just a few years ago; which makes perfect sense, given the massive increases we see in other computationally heavy industries.
 
User avatar
zeke
Posts: 13995
Joined: Thu Dec 14, 2006 1:42 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:43 pm

CATIA does have the functionality for that, and it is how Airbus and Boeing work with the one digital model, both in house, and with suppliers.

CATIA V4 was the early 1990s version, if I remember correctly the first version ported off mainframes. V5 is late 1990s. It did have the capability to work off one model, they made a budget decision not to upgrade from V4 as you needed to buy the IBM workstations as well. To get models from V4 to V5 they needed to conversion. If the grey matter is still functioning, the problem was to do with the mainframe version being big-endian while the new version was little-endian.

It bears little resemblance to today’s R28 which runs on windows PCs, 16 upgrades from V5.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
User avatar
sassiciai
Posts: 1091
Joined: Mon Jan 07, 2013 8:26 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:50 pm

It appears that there is a certain implicit trust (to 100%) in the ultra-rapid shiny graphic all singing design software as witnessed here.

MCAS V1 probably came out of that environment

It might be great, quick, exact in what it is instructed to do. But that system was also built by humans, and should never be considered as being infallible. I might even venture that a product from such an automated design process requires much more testing than any design developed "manually in the 21st centuary" by engineers with a bit of cynical blood in their veins. The testing must prove the product, and also the build process
 
User avatar
smithbs
Posts: 317
Joined: Wed May 17, 2017 6:09 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:51 pm

One of the reasons to use Agile is that you don't entirely know what your customer wants - or better yet, your customer doesn't know what they want until they see it. So you break off the project into time-sliced chunks and present more incremental changes. If the customer wants a change of direction, the team is better able to adapt because the plan didn't go very far in the first place.

This is great for commercial software. For machinery, it's a bit awkward because of the physical nature of the product - changing steel takes time. For aircraft, I'd have to wonder since a lot needs to be defined up front. A host of incremental changes in the middle of the program is sure to happen, but the framework for a highly complex machine is different than coding someone's web page. Changes have to be closely analyzed, vetted and coordinated because their effects ripple like waves.

Sprint periods of two weeks are fine for a lot of types of software. For mechanical engineer teams, it's hard and I'm not surprised Boeing chose 8 weeks. It has very much to do with the turn-around time of what you are designing, combined with how far out your team is able to forecast their work. I know software teams that can hardly think beyond two weeks, and mechanical engineer teams that can forecast out to the end. They are very different fields.

3D CAD is great, but there is still a computer horsepower problem. When you model a hugely complex machine with high parts count, the average desktop computer can't handle it. You need screaming fast network speeds and every engineer's computer needs to be bleeding edge, and even then you get huge delays in trying to get the model and more delays as you computer wheezes as it tries to render it.
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 3:52 pm

Revelation wrote:
zululima wrote:
I would be interested in having the magnitude of change from the original 777 to this explained. In the early 90s the 777 was hailed as the first all-digital design airliner, where a 3D model was used and could be manipulated to check clearances, the ability to fit MX tools and hands, etc. all before being built. Nearing three decades later, the improvements must be astounding. I wonder how much those old CAD designs were a pain to work with compared to today; although at the time I'm sure everyone was just thrilled to bin their slide-rule.

It seems most of us don't know the current setup, and those who know won't say.

I've found a few references to what tech the 777 program used:

At the height of the design process, Boeing had more than 2,2000 workstations networked into eight IBM 3090-600 mainframe computers.


So if we go with the high end estimate, we see eight IBM 3090-600 mainframes serving 2200 workstations.

So in very round terms, we have a six core server running at 60 megahertz with 256 megabytes of memory serving 2200/8 = 275 workstations.


Seriously, that's almost funny. I'd be asking myself what I could NOT do with that that system as opposed to what I could do, today.

To me, the real question is what Airbus had in place to design the A350, being the most recent example of similar scale design to presumed NMA.
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 4:06 pm

sassiciai wrote:
It appears that there is a certain implicit trust (to 100%) in the ultra-rapid shiny graphic all singing design software as witnessed here.

MCAS V1 probably came out of that environment

It might be great, quick, exact in what it is instructed to do. But that system was also built by humans, and should never be considered as being infallible. I might even venture that a product from such an automated design process requires much more testing than any design developed "manually in the 21st centuary" by engineers with a bit of cynical blood in their veins. The testing must prove the product, and also the build process


I was wondering how long it would take for someone to invoke this.
all designs are (of course) ultimately validated in the physical world, but computer modeled calculations can be run repeatedly, more quickly and cheaply, than can iterations of physical testing. That does not mean the physical testing won't take place- it will, and will do so to validate the virtual model. obviously, you can model in a virtual wind tunnel much more quickly and cheaply than in real life, and the beauty is you can make nearly realtime tweaks to optimize (not possible in the physical) ; this doesn't mean there won't be real tunnel testing... there will. It should be clear that this is precisely how new gens like CS300/A350/B787 are radically more efficient than predecessors.

re MCAS - that isn't relevant here. 737Max issues are institutional failures surrounding failure mode analysis, documentation/training, mgmt decision making. Computer design and modeling is hardly a factor, aside from fact that modeling would have identified the need for MCAS in those edge cases.
 
User avatar
zeke
Posts: 13995
Joined: Thu Dec 14, 2006 1:42 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 4:11 pm

FlyHappy wrote:
To me, the real question is what Airbus had in place to design the A350, being the most recent example of similar scale design to presumed NMA.



787 and A350 are both CATIA, it moved off propriety IBM hardware to be a windows application a long time ago.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
Weatherwatcher1
Posts: 266
Joined: Sun Mar 03, 2019 5:14 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 4:16 pm

smithbs wrote:
One of the reasons to use Agile is that you don't entirely know what your customer wants - or better yet, your customer doesn't know what they want until they see it. So you break off the project into time-sliced chunks and present more incremental changes. If the customer wants a change of direction, the team is better able to adapt because the plan didn't go very far in the first place.

This is great for commercial software. For machinery, it's a bit awkward because of the physical nature of the product - changing steel takes time. For aircraft, I'd have to wonder since a lot needs to be defined up front. A host of incremental changes in the middle of the program is sure to happen, but the framework for a highly complex machine is different than coding someone's web page. Changes have to be closely analyzed, vetted and coordinated because their effects ripple like waves.

Sprint periods of two weeks are fine for a lot of types of software. For mechanical engineer teams, it's hard and I'm not surprised Boeing chose 8 weeks. It has very much to do with the turn-around time of what you are designing, combined with how far out your team is able to forecast their work. I know software teams that can hardly think beyond two weeks, and mechanical engineer teams that can forecast out to the end. They are very different fields.

3D CAD is great, but there is still a computer horsepower problem. When you model a hugely complex machine with high parts count, the average desktop computer can't handle it. You need screaming fast network speeds and every engineer's computer needs to be bleeding edge, and even then you get huge delays in trying to get the model and more delays as you computer wheezes as it tries to render it.


Thanks for sharing. It sounds like the idea of scrum and sprints is to better deal with changing and evolving requirements, Dealing with changing requirements in a waterfall is a challenge for both software and airplane design. As airplane loads are being generated, new requirements come in and the change process can be cumbersome. I think the idea of 8 week developmental sprints makes a lot of sense. If rework comes, it is limited to a few weeks of work.
 
FlyHappy
Posts: 1048
Joined: Sat May 13, 2017 1:06 pm

Re: Boeing TX: A Precursor to Boeing NMA?

Sun May 19, 2019 4:24 pm

zeke wrote:
CATIA does have the functionality for that, and it is how Airbus and Boeing work with the one digital model, both in house, and with suppliers.

CATIA V4 was the early 1990s version, if I remember correctly the first version ported off mainframes. V5 is late 1990s. It did have the capability to work off one model, they made a budget decision not to upgrade from V4 as you needed to buy the IBM workstations as well. To get models from V4 to V5 they needed to conversion. If the grey matter is still functioning, the problem was to do with the mainframe version being big-endian while the new version was little-endian.

It bears little resemblance to today’s R28 which runs on windows PCs, 16 upgrades from V5.


There's nothing wrong with admitting you're wrong once in a while, Zeke.
Airbus simply did not make wise technical choices at that time, and paid a heavy price for (an amazing simple) mistake - the delay of A380 EIS, lost sales and in retrospect, likely contribution to the premature termination of A380 program.

I imagine there were a few early retirements due to this decision.
Obviously, Airbus would not make the same mistake again, A350 cycle was smooth, but not notably more rapid than generations past.
Without being a high placed insider, its impossible to know if new CFRP tech added time, without which the EIS could have been shorter (implying more effective "parallelization") , or not. The fact is, we don't know what Airbus capabilities are in this area, then or now.

We do know that the TX program has been uncommonly successful for the reasons being discussed, and it seems reasonable that this success can be scaled to a new clean sheet civil program. Not a guarantee - just not unreasonable.

Popular Searches On Airliners.net

Top Photos of Last:   24 Hours  •  48 Hours  •  7 Days  •  30 Days  •  180 Days  •  365 Days  •  All Time

Military Aircraft Every type from fighters to helicopters from air forces around the globe

Classic Airliners Props and jets from the good old days

Flight Decks Views from inside the cockpit

Aircraft Cabins Passenger cabin shots showing seat arrangements as well as cargo aircraft interior

Cargo Aircraft Pictures of great freighter aircraft

Government Aircraft Aircraft flying government officials

Helicopters Our large helicopter section. Both military and civil versions

Blimps / Airships Everything from the Goodyear blimp to the Zeppelin

Night Photos Beautiful shots taken while the sun is below the horizon

Accidents Accident, incident and crash related photos

Air to Air Photos taken by airborne photographers of airborne aircraft

Special Paint Schemes Aircraft painted in beautiful and original liveries

Airport Overviews Airport overviews from the air or ground

Tails and Winglets Tail and Winglet closeups with beautiful airline logos