1. Welding Models and Computational Welding Mechanics; A Welding Science as a Skill of Engineering
Mahyar Asadi

2. Challenges that are regularly faced by welding engineers;
Are not directly addressed by design standards such as ASME, API, etc
Mitigation of Distortion
Using different techniques to counteract the deflection from welding
Weld Sequence Design
Finding the best pattern in multiple-pass welding
Control of Residual Stresses
Proven that affects the service life, fatigue, creep, corrosion, …
Residual stresses and welding distortion are the frequent Challenges that are regularly faced by welding engineers;
And they are not directly addressed by design standards such as ASME, API, etc
The code requires to mitigate and control it but does not say how
There are may techniques that you can mitigate the distortion or control the residual stress
And weld sequence design is one of them when you deal with multiple welds
This is basically finding the best pattern that gives you as low as possible distortion .
These are what we need to do
But the challenge is how to do it 2

3. Weld Sequence Design
Combinatorial optimization; Permutation of beads and sub-passes plus change in direction
(2^n)xn! possibilities n
(2^n)xn! 2 8 3 48 4
384 5
3,840 6
46,080 7
645,120
Mathematics of such a selection lies in the combinatorial optimization that uses models and algorithms to find the best out of many possibilities.
However, the main challenge appears when the number of welds increases
For example (table)
Typical multi-weld design consists of 10s of welds or in other words countless possibilities
Even if you use symmetry and other similarity, still you need to select out of 1000s of possibilities

4. Surrogate Modeling
Explores the combinatorial design space with relatively few analysis
(2^n)xn! Possibilities, n= 4,  384 possible way to weld
It finds the best sequence with only 28 analysis out of 384 possibilities
Surrogate modeling is a mathematical approach that explores your combinatorial design space with relatively few analysis to approximate and find the best sequence.
Lets learn more about the surrogate modeling by using on a simple case,
A panel welding with 4 weld paths
And we want to find the best sequence out of 384 possibilities such that the diagonal deflection is minimal.
We use a CWM model to solve for deflection
Our surrogate model finds the best with only 28 analysis out of 384 possibilities.
Lets go to the surrogate model and tell you how it works.

5. Surrogate Model By convention,
We define positive sign to weld from front to the back and negative from back to the front
For example a+ c- d- b- means a happens first in positive direction, then c in negative followed by d negative, and finally b-
We selected enough cases to have at least 1 occurrence of at each position as it is shown in what we call it R’ matrix

6. Surrogate Model
We selected more cases to also make sure each pair happens at least once e.g b+ & a+, c+ after a+ , etc.
This is controlled by what we call it R” matrix

7. Surrogate Model
Finally, we ended up in 28 cases out of 384 and we measured the deflection of these 28 cases
Having the deflection result of these 28 cases,
Now we can approximate the deflection of other sequences which is not in this list
For example, d+,b+,c-,a+
We searched for similar sequence happening and collect each pieces from different deflection to put the pieces together and approximate the final deflection

8. Surrogate Model
These are the different comparison between surrogate approximation and real deformation for cases out of the 28 selected
And shows that our surrogate mode can properly capture the deflection.

9. Different Panels
These are different cases that undergone the process of weld sequence optimization in order to find the best and worse sequence for minimal diagonal distortion.
And the surrogate model is one technique, and there are other techniques like graph theory, integer programing and so on
However
the point of my presentation is not the technical understanding of surrogate modeling or other ones.

10. Different Panels
What I showed and explained was not a scientific paper or journal article. These were not a PhD or Master thesis.
These were the term projects conducted by students in my course on CWM and how to use CWM to address current challenges in welding.
These pictures was last April when my students were presenting their term projects.

11. “welding models and computational welding mechanics.”
Different Panels
“welding models and computational welding mechanics.”
A 13 week 39 hour lecture graduate level university course
Blended course including theory and simulation lab .
None of the students have ever experienced welding simulation.
At the end of semester, not only they learned about the software, more importantly they learned and implemented how to address an existing welding challenge by the use of software.
Last Winter, I developed and offered a course titled “welding models and computational welding mechanics.” .
This was for the first time in Canada and as far as I know in North America. At present, no university in Canada offers any course on the use of welding software for welding engineering.

12. Existing Welding Science & Capabilities:
Weld can be made with essentially zero distortion & zero residual stress.
Optimal tack welding
Optimal clamping
Optimal weld sequencing
Optimal pre-bending
Optimal side heating
Optimal trail cooling
Optimal live clamping
Optimal adaptive welding …
Current Industrial Practice of Welding Engineering:
Welding distortion always occurs and can be mitigated by:
Intuition-based designs for tack welding and static clamping
(welding science of 1950s and before)
I want to take this opportunity here to express why we need such a course at all.

13. Existing Welding Science & Capabilities:
Weld can be made with essentially zero distortion & zero residual stress.
Optimal tack welding
Optimal clamping
Optimal weld sequencing
Optimal pre-bending
Optimal side heating
Optimal trail cooling
Optimal live clamping
Optimal adaptive welding …
Current Industrial Practice of Welding Engineering:
Welding distortion always occurs and can be mitigated by:
Intuition-based designs for tack welding and static clamping
(welding science of 1950s and before)
This wide gap cannot be filled without computer model that is capable of predicting the welding

14. Simulation models are now mature and reliable
Existing Welding Science & Capabilities:
Weld can be made with essentially zero distortion & zero residual stress.
Optimal tack welding
Optimal clamping
Optimal weld sequencing
Optimal pre-bending
Optimal side heating
Optimal trail cooling
Optimal live clamping
Optimal adaptive welding …
Current Industrial Practice of Welding Engineering:
Welding distortion always occurs and can be mitigated by:
Intuition-based designs for tack welding and static clamping
(welding science of 1950s and before)
This wide gap cannot be filled without computer model that is capable of predicting the welding
Simulation models are now mature and reliable

15. Simulation models are now mature and reliable
Existing Welding Science & Capabilities:
Weld can be made with essentially zero distortion & zero residual stress.
Optimal tack welding
Optimal clamping
Optimal weld sequencing
Optimal pre-bending
Optimal side heating
Optimal trail cooling
Optimal live clamping
Optimal adaptive welding …
Current Industrial Practice of Welding Engineering:
Welding distortion always occurs and can be mitigated by:
Intuition-based designs for tack welding and static clamping
(welding science of 1950s and before)
This wide gap cannot be filled without computer model that is capable of predicting the welding
Simulation models are now mature and reliable
The lack of computational-welding skills for our welding engineers
So what we come shrt now is …

16. Welding moves from being an “art” to being a manufacturing science with the help of computers.
“Smart Factories” where advanced computer modeling, automation and control are key technologies to help welding become more competitive, energy-efficient and innovative as part of manufacturing
“Incorporate welding and joining considerations early in product-design stage [using computer models]”
“virtual factory” where modeling and simulation tools become commonplace in welding operations
This is not my personal idea,
This is what you can see in reports and vision of IIW, AWS, EWF, CWA, and many other organization around the globe.
The exact wording is “Welding moves from being an art to being a manufacturing science with the help of computers.”
American Welding Society (AWS) vision of 2020 clearly says that “Welding will move from being an “art” to being a manufacturing science with the help of computers” and talks about the term “virtual factory” where modeling and simulation tools become commonplace in welding operations.
Canada’s technology roadmap for the Canadian welding and joining industry mentions “Incorporate welding and joining considerations early in product-design stage [using computer models]”
Europeans picture factory of future as “Smart Factories” where advanced automation and control are key technologies to help welding become more competitive, energy-efficient and innovative as part of manufacturing.
Thes are strong evidence that tells we need to incorporate wleding simulation as part of our welding engineeirn skills
Incorporate welding simulation
as part of our welding engineering skills

17. To wrap-up;
Designer-driven optimization in welded structures is now feasible for routine engineering in industry, and computer simulations are enabling tools to help users' apply their creativity, expertise and skill to be more productive and innovative. If they are trained to do so
The lack of computational-welding skills for welding engineers is an opportunity to improve current educational programs for welding and engineering of welding in industry.
I would like to close my talk by saying that

18. To wrap-up;
Designer-driven optimization in welded structures is now feasible for routine engineering in industry, and computer simulations are enabling tools to help users' apply their creativity, expertise and skill to be more productive and innovative. If they are trained to do so
The lack of computational-welding skills for welding engineers is an opportunity to improve current educational programs for welding and engineering of welding in industry.