Chapter 650: Difficulties in the Dual-Core Technology of the Mask Stage? Break It for Me! Zhang Yanjie: The More I Understand, the More I Awe!
At dawn the next day, Chen Yansen slowly opened his eyes.
Ye Qiuping was sleeping beside him, breathing evenly and slowly. Her figure was faintly visible under her translucent silk nightgown, showing her strong strength.
Chen Yansen lifted the corner of the quilt, put on his slippers and walked out.
A set of Slow Sheep and Green Snake cosplay costumes was casually thrown on the floor, with a moving tail next to it.
Walking into the study, he turned on the computer and checked the core operating indicators of each subsidiary as usual. Only after confirming that there were no problems did he start to deal with emails.
With his memory and comprehension, even if it is an email with thousands of words, he can usually remember all the contents with just a glance, and give an accurate, professional and mature reply in just a few seconds.
Therefore, in less than ten minutes, he finished the work at hand.
With a "buzz" sound!
The phone on the table vibrated.
Chen Yansen picked it up and saw that it was a message from Song Yuncheng. He thought for a moment and replied with an "OK".
At 8:50 in the morning, he arrived at the Optical R&D Center on time, held a brief meeting with the members of the algorithm group, and then devoted himself to the next work.
The ultra-precision motion control technology of the mask stage requires support from motion control algorithms and software with nanometer-level precision.
After thoroughly understanding the core of this technology, Chen Yansen designed a composite control architecture with a three-layer structure of "feedforward + feedback + disturbance compensation".
First, the system hysteresis is compensated based on the acceleration feedforward and friction feedforward of the motion trajectory.
To put it simply means to predict in advance and take the initiative!
Just like an experienced driver who sees a slope ahead and steps on the accelerator to accelerate in advance instead of waiting for the car speed to slow down before reacting.
The meaning of acceleration feedforward is to rely on a pre-planned route, such as how much to accelerate and when to decelerate, to first "send instructions" to the motor so that it exerts force when it should, avoiding movement lag.
Friction feedforward is to calculate in advance the additional force required to overcome this friction, since we know that friction will be generated between the parts when the machine is running, similar to the resistance between the tires and the ground, so as to ensure the accuracy of the machine's movement.
Secondly, a hybrid control method of PID plus LQR is used.
In other words, make corrections in real time and adjust according to the target.
It's like staring at the dashboard while driving. If you find the speed is too fast, release the accelerator; if you find it is too slow, press it a little, and correct the deviation at any time.
PID is the most commonly used control algorithm in industrial control. To put it simply, it is a set of rules that "adjust actions based on errors" to stabilize the operating state of the equipment at the target value.
LQR, or linear quadratic regulator, is a more intelligent automatic control algorithm.
Taking the example of accurately stopping a robotic arm at a specified position, ordinary control methods can only make adjustments by closely monitoring the "deviation between the current position and the target position."
LQR can consider multiple factors at the same time, such as how much the current position has deviated, whether the moving speed is appropriate, the torque output by the motor, and whether the parts will shake due to excessive force.
Once this technology breaks through to the nanometer level, it can not only be used in the manufacture of mask stages, but also in scenarios such as robot motion control and satellite attitude adjustment.
Finally, there is the disturbance compensation technology, which estimates the value of external disturbances through the extended Kalman filter and offsets them in real time.
Just like when you are driving and a gust of wind suddenly blows, the steering wheel will shake and the driver needs to immediately make fine adjustments to keep the direction steady.
The extended Kalman filter is like a combination of a "sensitive sensor + an efficient calculator". It can quickly detect sudden external interference and then immediately calculate the force compensation to offset the impact of these interferences and ensure that the accuracy is not affected.
When these three methods are used in combination, the entire control logic is like a super driver who is good at prediction, proficient in fine-tuning and has excellent anti-interference ability. It can not only make the machine move quickly and accurately, but also cope with various small interferences and stably control the precision at the nanometer level.
In fact, a number of domestic universities and research institutes, such as Harbin Institute of Technology, Tsinghua University, and the Institute of Automation of Huazhong University of Science and Technology, have conducted in-depth research on classic control algorithms such as PID, LQR, and Kalman filtering, and have published a large number of related papers.
Most of the current algorithms have only completed simulations or small-scale experiments and have not yet been verified at the application level on high-precision motion platforms.
There is a serious lack of experience in matching theoretical models with actual mechanical nonlinearities!
And as far as compensation for friction and clearance is concerned, it can only be achieved at the micron level at present.
The main reason that restricts the rapid breakthrough of this technology is that the localization rate of key supporting products such as precision bearings, air flotation components, and high-stability power supplies is too low, and there is a large gap in performance compared with foreign products. This makes it extremely difficult to overcome the precision bottleneck when conducting system integration.
Even though Chen Yansen pushed his talent of [Planck Clock] to its limit every day, the road ahead was still full of difficulties.
Just as problem A was solved, problem B popped up, blocking the road again.
But in the eyes of a group of top industry experts, Chen Yansen's ability is already extraordinary. He has developed a more advanced control algorithm on his own.
If it were them, it would take at least seven or eight years to figure it out.
As a result, in less than half a month, this group of doctors, professors and academicians from automation research institutes across the country changed their way of addressing Chen Yansen from "Mr. Chen" or "Boss Chen" to "Mr. Chen".
Some people who were able to put aside their pride even called him "Teacher Chen".
Among them is Zhang Yanjie from Harbin Institute of Technology. He will be 50 years old in four years, but he always hangs around Chen Yansen, calling him "Teacher Chen".
At this time, Zhang Yanjie came up to Chen Yansen with a laptop and asked, "Teacher Chen, I spent half the night thinking about the disturbance compensation algorithm flowchart you talked about yesterday, but there's one thing I still don't fully understand."
He paused and continued, "How does the extended Kalman filter balance response speed and estimation accuracy when estimating external disturbances? I ran a simulation using the parameters you provided. If the response is fast, the error will exceed 200 nanometers. If I try to reduce the error to below 50 nanometers, the response speed will be slow. If we encounter sudden disturbances, there will be no time to compensate."
Chen Yansen stopped and stared at the screen for dozens of seconds, then said, "Old Zhang, look at this disturbance curve. The external interference is not constant. For example, the pressure fluctuation of the air flotation component is sometimes high-frequency and small-amplitude, and sometimes low-frequency and large-amplitude.
You previously used a fixed-gain Kalman filter, which naturally could not take into account both speed and accuracy. You can add an adaptive gain module to allow the filter to determine the disturbance type by itself.
When there is a high-frequency small disturbance, increase the gain to speed up the response; when there is a low-frequency large disturbance, reduce the gain to reduce the error.
I did a simulation yesterday. Using this logic, the response time can be controlled within 10 milliseconds, and the error can be stabilized below 20 nanometers. You can try this set of parameters today."
Zhang Yanjie frowned deeply, and after a long moment, he slapped his thigh and shouted, "That's right! Why didn't I think of adjusting the gain based on the disturbance type!"
Some technical obstacles are actually like a layer of window paper that can be broken with a poke.
But if no one gives you guidance, you may have to spend months or even years wandering around the door.
Hearing the noise, several R&D engineers who were busy working nearby also gathered over.
Li Tuotao from Tsinghua University's Department of Automation pointed to the module on the diagram and asked, "How do I set the threshold for this adaptive gain? If I misjudge the disturbance type, will it negatively impact accuracy?"
The other people also had curiosity on their faces.
Chen Yansen nodded slightly, pulled up a chair and sat down, then directly used Zhang Yanjie's computer to call up a set of experimental data: "We can use the first 500 milliseconds of disturbance data for feature extraction. For example, those with a frequency exceeding 100 Hz and an amplitude less than 0.5 nanometers are classified as high-frequency small disturbances;
Frequency below 50Hz and amplitude greater than 1 nanometer are classified as low-frequency large disturbances. For those in the fuzzy zone, a weighted algorithm is used to balance the gain. This way, the accuracy of judgment can reach over 98%, with little error.
After hearing this, several professors from the Department of Automation at Tsinghua University had their eyes light up and couldn't help swallowing.
Several PhDs from Harbin Institute of Technology looked at each other, with a flash of shock in their eyes.
The more I interact with Chen Yansen, the more I feel the disparity in the world, the gap between people, and the insignificance of my own abilities.
Zhang Yanjie sighed secretly. He knew in his heart that without Chen Yansen, even if their group got together to study, it would be difficult to make any progress in the research of motion control algorithms.
As he was speaking, a researcher from the bearing group ran in with a test report, his face brimming with excitement. "Mr. Chen, the test results for the air bearing guide are in! The flatness error of this ceramic air bearing is controlled to 30 nanometers, 5 nanometers lower than the imported bearings we used previously. Furthermore, after 24 hours of continuous operation, the temperature rise was only 0.6 degrees Celsius, fully meeting the stability standards."
Chen Yansen took the report, turned to the data page, and raised the corners of his mouth slightly, revealing a faint smile.
As long as it meets expectations!
The air flotation component is the core component of the mask stage movement. In the past, it was impossible to buy or manufacture it.
Now, times have completely changed!
Then, Chen Yansen looked at Zhang Yanjie and said, "Consult the motion control team and retest the entire machine using the new air bearing."
"Okay, I'll go right away." Zhang Yanjie picked up the computer and test report, and left happily.
If the doctoral students from Harbin Institute of Technology saw this, they would be shocked.
Because Zhang Yanjie was humble at this time, completely different from his usual serious appearance at school.
As Xingyuan Technology mastered more and more patented technologies, the Anguo Association's investigation into Chen Yansen became more and more detailed.
Li Qingsong, who was far away in Yanjing, casually flipped through Chen Yansen's investigation report. Finally, he smiled lightly and locked the document in the cabinet.
The other side.
A week passed in a flash. At four o'clock on Friday afternoon, Chen Yansen changed out of his lab coat, walked out of the R&D center, sat in the back seat of the Hongqi L5, and then signaled Lao Huang to drive to the airport.
(End of this chapter)
Continue read on readnovelmtl.com
