I started with a ship full of max-level veterans, and I'm just a complete noob.

The Soviet experts were using a ruler to hold down the rolled-up blueprints. After seeing the records of Furnace No. 3 brought by Jiang Ming, they first flipped to the step response above 1200 degrees Celsius, and then circled the point where the heating and cooling curves separated with a pencil.

"You treat it like a resistance bar that changes with temperature, so you always try to catch up with it by adjusting the parameters, but its grain boundaries are also changing, and there is hysteresis in it."

Jiang Ming laid out the power curve for the third temperature zone on the table, pointing to the section where the integral correction always lagged behind the temperature drop.

"The segmented PID has already reduced the fluctuation to ±0.6 degrees. Adding more derivative will amplify the temperature measurement noise. I need to compensate for the gain of the rod outside the controller."

The expert picked up a piece of chalk and wrote the sustaining power on the blackboard, then added the PID correction value next to it. The two were added together and both pointed to the actuator.

"If the PID controller is to handle all the heating power, it will have to deal with static nonlinearity, grain boundary hysteresis and furnace thermal inertia at the same time. Any change in any of these factors will affect the parameters."

"First, the reference power required to maintain the furnace temperature is given according to the temperature. The PID only corrects the small difference between the actual value and the reference value, so that the object being controlled by the controller will be close to linear."

Looking at the reference power item, Jiang Ming immediately thought of the compensation method for the demagnetization curve of the Yuanlei No. 1 permanent magnet. At that time, the correction value was also pre-stored according to the temperature range, and then the corresponding output was selected by real-time signal.

"Table lookup feedforward?"

"Yes, two tables, one for rising temperature and one for falling temperature. Don't mix them together."

Experts pinned a measured curve of a silicon carbide rod on a blackboard. The resistance in the mid-temperature zone decreased with temperature, while the direction turned positive in the high-temperature zone. The two paths formed loops of varying widths in the transition zone.

"Both the grain boundary barrier and the internal conductivity mechanism of the material require time to adjust. When the temperature is raised to the same level and when it is lowered to the same level, the equivalent resistance will fall at different positions."

"First, mark your heating element from room temperature to 1,500 degrees Celsius. Record the voltage, current, and stable power at each temperature point, and then create tables according to the direction of temperature rise and fall."

Jiang Ming drew the signal path on his notebook. One path of the thermocouple temperature signal was sent to the lookup table, and the other path was sent to the PID controller. The former provided the reference power, and the latter output the deviation correction.

He continued to inquire about the selection method for the transition zone, and the expert suggested adding a hold logic based on the direction of temperature change. When the temperature rises, the system runs along the heating table, and only switches to the cooling table after the temperature begins to drop continuously, thus avoiding noise-triggered switching back and forth.

"Are there enough points on the table?"

"First, see how many points your furnace can stably maintain. Industrial furnaces use twenty to thirty points. Then do piecewise interpolation, which is usually more reliable than a fancy but wrong formula."

The experts unearthed old blueprints of the Ural silicon carbide furnace and pointed out to Jiang Ming a function generator composed of a resistor network, where each node corresponds to a temperature range and the output voltage represents the required reference power.

"If you lack a servo multiplier, use a voltage divider network to approximate the curve. It doesn't matter if the line is thicker; the PID will compensate for small errors."

Jiang Ming then drew out the low-frequency oscillations in the positive temperature coefficient region and asked how much of the differential element should be retained.

Experts used chalk to erase the overly strong differential coefficients, leaving only a small portion for prediction.

"After the feedforward takes on a large change, the differential only needs to suppress the overshoot caused by the furnace body's inertia, so the noise in the signal does not need to be amplified."

As the time limit for communication with the spirit was nearing its end, Jiang Ming had already compiled a complete plan in his notebook, with calibration, dual tables, segmented interpolation, direction determination, and a dual loop of feedforward plus PID arranged in sequence.

The blueprints of the furnace control room gradually faded away. When he returned to his desk in the dormitory, the kerosene lamp wick still held its original flame, and time outside remained frozen at the moment he entered the psychic realm.

Jiang Ming then pulled out a blank appendix from the twelve-page technical proposal and rewrote the table lookup scheme into a technical route that 502 could publicly explain.

The demagnetization curve compensation of Yuanlei-1 permanent magnet provides experience in table lookup, the nonlinear control chapter in Qian Xuesen's lectures provides the feedforward principle, and the measured data of silicon carbide rods are responsible for establishing specific functions.

The three sources are interconnected, and anyone who follows the records can find a ready-made entry point for the experiment.

Before dawn, the proposal had been completed, titled "Compensation Scheme for Feedforward Lookup Table of Silicon Carbide Heating Body of Furnace No. 3," while the original psychic records were locked in a safe.

When Da Liu read the table showing the temperature rise and fall separately, he drew two broken lines in the transition zone with a pencil.

"This is equivalent to first telling the controller how much power to provide, and then letting the PID control compensate for the difference. But how do you connect the circuitry to 29 points?"

"First calibrate, then determine the resistor network. If the actual curve has obvious inflection points, the number of nodes can be reduced."

The furnace room then entered a daily calibration phase. Old Sun was responsible for controlling the temperature rise rate. Every 50 degrees Celsius increase, the temperature was kept constant. Big Liu adjusted the input to keep the furnace temperature at the current point for ten minutes.

Fang Xudong simultaneously recorded the voltage, current, thermoelectric potential, and stable power at both ends of the silicon carbide rod, and then calculated the equivalent resistance at each temperature point by dividing the voltage by the current.

There are 29 calibration points from room temperature to 1,450 degrees Celsius. After exceeding 800 degrees Celsius, each point needs to wait for the heat flow in the furnace to redistribute before the entire roll of paper is slowly laid across the edge of the recording table.

Below 1100 degrees Celsius, the equivalent resistance continues to decrease with increasing temperature, and the negative temperature coefficient characteristic is clearly shown on the graph paper.

Between 1,100 and 1,200 degrees Celsius, the slope of the curve gradually flattens out, and the response of several adjacent calibration points to power changes varies considerably.

Above 1200 degrees Celsius, the equivalent resistance begins to rise, and the higher the temperature, the more significant the input correction required to maintain the same temperature range.

After the temperature calibration was completed, Lao Sun gradually reduced the temperature according to the same nodes, while Fang Xudong used a different record sheet to avoid mixing the two sets of data.

When the two curves are drawn together, there is a lateral difference of about 30 degrees in the transition zone, and the heating path and the cooling path form a clear hysteresis loop.

Da Liu held the coordinate paper down with a ruler for a long time and finally understood why the integral correction in the third temperature zone always followed the wrong direction.

"Even if it's labeled 1,150 degrees, the resistance of the rod depends on which side it comes from. No wonder one set of parameters can't cover everything."

Jiang Ming divided the 29 points into two sets of data according to the temperature direction, and then used linear interpolation to connect adjacent points to form a broken line that the hardware lookup table needs to approximate.

Da Liu added a voltage divider sampling network to the front end of the control circuit to obtain the voltage and current signals at both ends of the silicon carbide rod, and then used a resistor network to form a simple function generator.

Each inflection point corresponds to a temperature range, and the output voltage represents the reference power required for that range. The heating gauge and cooling gauge are selected by a directional relay.

The thermocouple signal is simultaneously fed into the original PID loop. The lookup table output and the PID correction are superimposed by the summing network to jointly drive the actuator.

After the circuit was set up, Da Liu first used a calibration source to input the simulated temperature point by point, and Fang Xudong checked the output of each node and found that there was a large broken line error near 1100 degrees.

Jiang Ming removed two additional nodes from the transition zone to make this function approximate the hysteresis curve, while the remaining temperature zones maintained their original spacing.

When the fourth no-load heating began, everyone's attention was focused on the output of the lookup table and the temperature paper tape. The reference power of 800 to 1200 degrees Celsius changed segment by segment with the temperature range, and the PID correction amount was significantly reduced.

After entering the highest temperature zone, the resistance of the silicon carbide rod turns positive, and the reference current is lowered in advance by referring to the table network. The upward surge caused by the thermal inertia of the furnace body leaves only a narrow section.

After the temperature stabilized at around 1420 degrees, the paper tape fluctuations gradually decreased to ±0.2 degrees and did not expand again for several consecutive cycles.

Fang Xudong took out a red pencil, circled the entire stable curve, and wrote down next to it: "Look up the table for feedforward input, PID only corrects for deviation."

The ±0.2 degree is still higher than the ±0.08 degree in the germanium era, but this number has already crossed into the range that silicon regional purification can attempt.

When Huang Kun arrived outside the furnace room, he first looked at the four rolls of paper, one old and one new, and then walked to the record table and wrote a line of annotations.

The temperature is 0.2 degrees Celsius under no-load conditions; it needs to be re-verified when carrying materials.

Jiang Ming listed the potential increase in heat capacity, radiation, and molten zone movement disturbance after adding material in the annotation below.

"The framework can be reused, but the reference power needs to be recalibrated based on the data of the material being transported."

Huang Kun nodded, turned to look at the double spring clamps at both ends of the silicon carbide rod. The nickel-chromium alloy wire had undergone multiple high-temperature cycles, and the tablet pressing position remained symmetrical.

"How long can this set of springs last?"

Old Sun handed over the measuring instruments from the recent disassembly and reassembly, pressing his palm on the cumulative power-on time column.

"If it holds for 500 hours, then we need to check the elasticity decay after that, and check it separately in the furnace time record."

Huang Kun put away the record sheet, leaving only one sentence.

"Let's start running."

The furnace room door was then opened, and the primary silicon material brought by Lin Lanying was placed on the weighing platform. The first test of the No. 3 furnace after its modification officially entered the furnace loading process.