Chapter 777: You Reap What You Sow

If these rumors about the source of WS-10 technology were only circulated in China, they would not exceed the concept and influence of a short essay.

But if foreign reports, and serious reports, were also involved, it would not be a bad thing.

In fact, for an engine that may soon enter mass production and service, basic information such as structural design cannot remain unknown for a long time.

After all, you have to issue technical information such as maintenance manuals to grassroots units, and these things will be impossible to be leak-proof after a long time at the grassroots level.

The reason why the outside world's perception of WS-10 is so outrageous is mainly because Chang Haonan's actions are too fast.

From determining the design plan, to subsystem design, to component-level testing, bench testing, and the current installation testing, all were completed in less than three years.

Especially the installation testing, which involves a large range, is even concentrated in the half year from the beginning of 1999 to the present.

Moreover, from design to testing, everything is done in one go, without any rework process, which reduces the number of insiders to a minimum.

In other words, those who have seen what Taihang looks like may not even have time to take their annual leave, let alone retire.

The entire project is almost running in a semi-closed environment.

The chances of leaks are naturally less.

As for the outside world, they are also accustomed to the ten-year or even longer development cycle of aviation engines.

There has been no new news about the WS-10 in just three years, and under normal circumstances, no one would think that it has been torn down and started over. Now that a new engine has come out, unreliable people are naturally completely guessing, and even more reliable analysts can only shoot arrows and then draw targets on the only clue of CFM56.

Of course, from a technical point of view, the overall design plan is not that important.

For example, the overall design plan of CFM56, any ground crew serving Boeing aircraft basically knows it.

Even the engine itself is sold all over the world with the passenger plane, and you can dismantle it for research.

But no one has ever been able to reproduce it exactly.

After all, if it is so simple to make aircraft engines, then aircraft engines don’t need to be the jewel in the crown of industry.

But if it comes to WS-10, we can take advantage of this information gap -

Since foreigners think that our model is developed from the CFM56 core engine, and they have analyzed it very well, then we will follow them.

Neither admit nor deny directly, when asked, just hint three times:

Those who understand will understand, no more explanation, you can think about it carefully.

In this way, at least it can also let the Western camp have a round of Werewolf Killing - the Chinese play CFM56 so well, so there must be a traitor among us who sold the relevant technical information.

In the end, even if the matter is left unresolved, or the actual situation of WS-10 is exposed in a few years, it can also achieve the effect of destroying the internal relationship of the opponent.

Moreover, the French approached the Aviation Power Group for cooperation at this sensitive time node, which is probably related to this.

After all, the CFM Group was jointly established by Snecma and General Electric.

It is just not clear whether the other party wants to investigate the "technology leak" or has other purposes...

Thinking of this, Chang Haonan picked up the phone again and dialed the internal number of Zhang Liangping's office...

...

The task he gave to Zhang Liangping was obviously impossible to complete in one or two days. Therefore, after dealing with this small emergency, Chang Haonan returned to school as planned and began to develop a specific manifold learning algorithm.

Compared with the purely theoretical paper submitted to the Annals of Mathematics before, this is the direction he focuses on.

Before the National Day, Chang Haonan had sorted out two basic algorithm ideas, and Yao Mengna and he each chose one to continue research.

Although the result he conceived on the spot could not be the optimal solution in one step, it was at least representative enough.

The first type is the global idea, which maps the adjacent points on the manifold to the adjacent points in the low-dimensional space when reducing the dimension, and at the same time ensures that the distant points on the manifold are mapped to the distant points in the low-dimensional space.

The second type is the local idea, which only needs to ensure that the close points on the manifold are mapped to the neighboring points in the low-dimensional space.

In comparison, the former is more intuitive (of course, it is only relatively intuitive), but the computational complexity is very high, which is a certain challenge for both hardware level and algorithm design.

The local idea is more abstract, and the correspondence between points far away is unclear, but the amount of calculation is relatively small, which seems to be more suitable for the current computer performance.

This time, it was Yao Mengna who took the initiative to find Chang Haonan a few days later.

However, it was not because the former had constructed an algorithm according to the global idea.

In other words, the algorithm was indeed developed, but it was found to be a dead end.

"Mr. Chang, I used the constructed isometric mapping algorithm to perform data point generation optimization tests on the two-dimensional manifold [t, s, X] in three-dimensional space."

Yao Mengna put a few sheets of paper on Chang Haonan's desk:

"For the complete surface, the efficiency of the algorithm is quite good, and the generated coordinates of the complete S-surface are basically restored."

"But if I dig out a square area with a length and width of π on the two-dimensional manifold, it is equivalent to opening a hole on the surface, which is a very common situation in practical applications. Then the generated coordinates will be distorted, causing the area of ​​the hole to become larger, and it will become an approximately elliptical area..."

"..."

In short, it is not easy to use.

"The existence of holes in the manifold means that the subset of the Euclidean space that is equidistant from the manifold is non-convex, and the deviation generated when calculating the shortest path between sample points on the manifold increases..."

The problem discovered by Yao Mengna is also an area that Chang Haonan has not yet studied.

Fortunately, the overall idea is more intuitive, so he can analyze it on the spot.

"That is to say, if you want to use the isometric mapping algorithm, or to expand it, if you want to use the global algorithm, then the manifold object must satisfy the conditions of being isometric to a subset of Euclidean space and that this subset is convex."

Chang Haonan paused for a moment with the ballpoint pen in his hand and concluded.

After all, this algorithm itself was optimized and corrected by Yao Mengna bit by bit, so she followed Chang Haonan's thinking this time.

"So..."

Yao Mengna looked embarrassed:

"You knew this road was not feasible before?"

"Ahem... that's not the case."

Chang Haonan immediately denied:

"I just thought of it after listening to your explanation..."

"In fact, I have been studying how to improve the local linear embedding algorithm (LLE) during this period."

He turned on his computer and took out a piece of paper from the side and spread it next to the keyboard:

"The biggest problem with LLE is that the local weights it uses cannot fully reflect the local geometric structure of the high-dimensional manifold. Therefore, for singular or nearly singular systems, a positive number γ needs to be added artificially, but the selection of γ has a great impact on the results..."

"..."

Strangely enough, after writing that paper, Chang Haonan found that his way of thinking seemed to be a little different from the past, specifically, it became more coherent and smoother.

After a half-hour introduction, not only did he not get stuck at all, but even Yao Mengna, who was listening next to him, did not feel that there were too many things she didn't understand.

"So."

Chang Haonan put the pen aside and said in a firm tone:

"Obviously, using multiple sets of linearly independent weight vectors to construct local linear structures can improve the final embedding results."

Yao Mengna was silent for a while, then nodded:

"That's true."

The two continued to fall into silence.

"Speaking of which, we first started studying manifold learning, it seems to be to...solve the problem of automated detection of pulsating production lines?"

Yao Mengna broke the silence again.

"Well..."

Chang Haonan nodded:

"But the content we are studying now really can't be directly applied to your topic."

Manifold learning is just a process of data dimensionality reduction, at best it solves one of the many obstacles on the road to automated production.

Hearing this judgment, Yao Mengna sighed:

"Sure enough, my idea may still be a bit too radical..."

That's indeed radical.

If her plan is fully realized, then Factory 112 can directly withdraw the workers and turn it into a dark factory.

However, Chang Haonan then changed the subject:

"However, this improved LLE algorithm can play a role in other fields."

It is a case of an accidental success.

"For example?"

Yao Mengna frowned slightly, looking at the densely packed formulas on the paper in front of her.

"For example... information retrieval, data screening..."

Halfway through, Chang Haonan realized that these did not seem to be specific "applications", but only application-level technologies.

So he thought for a while:

"If we want to move closer to the specific production links... probably... equipment status monitoring and automatic fault diagnosis?"

In fact, if this algorithm is promoted, it should still be able to shine in many fields.

But Chang Haonan could only think of his old business for a while -

For example, an airplane has a fault.

In the past, it could only be slowly checked by ground staff.

If the airplane has enough sensors and can effectively screen and analyze the data fed back by the sensors, then the airplane's avionics system will determine the specific location and situation of the fault.

Even when the fault is just a sign, it can be nipped in the bud.

There are still some difficulties to overcome in operation.

For example, the current sensors are relatively large, and it is difficult to fit enough of them into an aircraft.

But in any case, this is obviously a technology with considerable potential.