Chapter 931: Respond to Changes with Constancy

Time passed by minute by minute.

Chang Haonan frowned slightly and stared at the computer screen.

In the huge office, there is almost only the faint "click" sound caused by the occasional rotation of the mouse wheel.

Anthony Jameson's paper contains a lot of content, but it is not long. In other words, the content is a little jumpy.

If you are a reader who has not had much exposure to the finite volume method, you will most likely need to read it several times to understand its meaning.

Fortunately, although Chang Haonan is not specialized in this, he actually used the software COMSOL Multiphysics a lot in his previous life, so he has some understanding of the Swedish company's "problem-solving ideas".

"call--"

After about half an hour, Chang Haonan finally put down the mouse in his hand, leaned back on the chair and exhaled a breath.

Apparently he has finished reading it.

However, his frown did not relax.

This made Wei Yongming and Song Jingming who were not far away feel a little worried——

If Mr. Chang finds it difficult to deal with after reading this, it means that the other party really has something.

After a few more minutes, Wei Yongming, who had not dared to disturb him just now, finally couldn't hold it in anymore——

Extending your neck is the same as shrinking your neck. No matter what the conclusion is, don’t be silent...

Therefore, after hesitating for a moment, he tentatively expressed his doubts:

"Mr. Chang, I think Professor Jameson's idea is to apply the spectral volume method on an unstructured triangular grid to solve the hyperbolic conservation law. However, I have conducted several calculation tests before, and the spectral volume method is used in calculations. Controlling the surface integral requires a huge number of Gaussian integration points, but will it increase the calculation time? "

As the head of the technical department, since Wei Yongming came to see Chang Haonan today, he naturally read this paper in advance.

It's just that some of the content in it doesn't seem to match the promotion of the other party's products.

"Thesis...it's just a purely theoretical process."

Chang Haonan replied:

“From theoretical derivation to specific algorithm implementation, the details in between are the key to software design.”

"For example, my paper introducing the conservation level set method was published in a journal, but others did not have the matching grid generation and dimensionality reduction technology for high-order partial differential equations. Even if they understand the content of the paper, , and there is no way to apply it to the software.”

"As for Jameson's method, it should only draw on the idea of ​​​​the spectral volume method. It does not directly apply the spectral volume method to unstructured grids."

"Although the paper does not involve the subsequent application level, I speculate that they will choose to use the node function value of the spectral unit instead of the volume average to construct the discrete format. In this way, the Level Set function value of the control volume boundary can be used The linear combination of high-order Lagrangian polynomial interpolation and second-order upwind interpolation is estimated, and the total degrees of freedom and the number of Gaussian integration points will be reduced accordingly. This method is also in line with what COMSOL is good at..."

"..."

Facing the paper on the computer screen, Chang Haonan quickly deduced the specific technical route of the other party's algorithm through reverse thinking.

Wei Yongming, who was sitting on the sofa, took out a small notebook at some point and was taking notes with his head down.

"Of course, this is just my speculation from a theoretical level and may not be 100% accurate..."

Looking at Wei Yongming who was writing vigorously, Chang Haonan finally did not say anything and left himself a way out.

But this sentence obviously had no impact.

The movements of the former's hands even became faster.

Song Jingming next to him may have received too much content that he didn't understand at all. At this time, he felt that his brain was a little bloated, so he took advantage of Wei Yongming's lack of time to raise his head and quickly changed the topic from technology to management:

"Then Mr. Chang, under the current situation, do we need to change the original planned update route?"

"Such as... uh... using some newer technologies to deal with it, something like that."

In his opinion, since Chang Haonan just showed a solemn expression, it obviously meant that the other party had indeed put enough pressure on him.

Of course, this is not entirely Song Jingming's own idea, but a conclusion he reached after discussing with Wei Yongming next to him before this trip -

In the field of CFD, Torch Group actually has some unused technical reserves.

Mainly focused on high-precision method level.

It's just because of the development cycle that it can't keep up with the original update schedule.

At present, if we want to continue to maintain Torch Group's advantages in numerical calculations, especially in the field of computational fluid dynamics, we must delay updates to introduce these more promising technologies.

However, Chang Haonan did not nod immediately.

The update to version 2.0 of the TORCH Multiphysics software was something he started to prepare more than half a year ago, involving major improvements in three aspects: multiphase flow, topology optimization and CFD.

Nowadays, it can be said that we are only one step away from the launch of the new version. If you rashly disrupt your rhythm because of the other party's actions, you will become passive.

In fact, in the final analysis, it is still caused by backward data transmission methods.

In 2001, the main means of updating large-volume commercial software was still to physically mail storage media such as CDs or hard drives.

This means that users cannot accept too high an update frequency.

It also means that vicious bugs are completely unacceptable.

Because the cycle and cost of fixing bugs are unacceptably high.

Therefore, the software testing cycle these days is generally relatively long to ensure that no big troubles occur.

And if, like more than 20 years later, the software can be upgraded online at any time, then a 2.0.5 Beta version can be pushed first, and then the official 2.1 version can be pushed after the new technology stabilizes.

Even if there are some bugs in the first version, they can be quickly corrected.

Moreover, there is an even more important point.

That paper.

Although Chang Haonan understood it and even guessed the opponent's general technical route.

But there were still some details that concerned him.

That's right, he had been frowning just now. In fact, he was not worried about the threat caused by friendly businessmen to the Torch Group, but was thinking about these issues.

It's just that they were misunderstood by the two people opposite...

…

After struggling alone for a few minutes, Chang Haonan raised his head again and answered decisively:

"Updated plan, unchanged!"

"No... unchanged?"

This decision was obviously beyond the expectations of the two people on the sofa.

"right."

Chang Haonan nodded, then stood up and came to the window, looking at the city in the distance:

“On the one hand, we cannot be led by others. Even if the new version of COMSOL Multiphysics can achieve the results they advertise in the CFD field, we still have advantages in multiphase flow and structural topology optimization, so users cannot So give up on our software.”

"So, the worst-case scenario is that a considerable number of users start to use our two software programs separately for different problems. This result is not completely unacceptable to us."

Having said this, he paused for a moment, then turned around and gestured in the direction of the computer screen:

"On the other hand, this paper only mentions the discretization method of using the lattice-centered finite volume method for the N-S equation, but in actual operations, there are not many high calculation weight issues that are common when numerically solving viscoelastic constitutive equations. discuss."

“For this type of problem, the LCR method that is currently being studied most is to convert the original stress tensor evolution equation into an evolution equation in the logarithmic form of the conformation tensor matrix. This process requires the calculation of the stress tensor and the conformation tensor. Repeated conversion of the matrix logarithmic form of a quantity can easily cause loss of accuracy. If the calculation weight is too large, it may even cause the algorithm to collapse. "

"Of course, it's impossible for a consultant from a commercial company to write down all the technical details when publishing a paper, but it's still too strange not to mention it at all..."

Chang Haonan said and touched his chin.

Although this sentence was obviously not finished yet, Wei Yongming was already a little unable to hold on:

"Mr. Chang, our group probably hasn't researched the finite volume method as a key technology, right?"

"You... know so clearly?"

"Oh, obviously we can't talk about it."

Chang Haonan waved his hand and replied with an indifferent expression:

"When Torch Group was not officially established, I read some relevant information in order to choose a technical route for the new software, but I didn't expect it would be useful at this time."

"Anyway... they said they will push the update later this month. It only takes two weeks. Let's keep the same to cope with the changes. After their new version is released, let's take a look at what is sold in Hululi. What kind of medicine is it..."