Special rolling mill controller without PLC (Programmable Logic Controller) architecture and integrated control system

Abstract

The invention relates to the technical field of rolling mill automatic control, in particular to a special rolling mill controller without a PLC framework and an integrated control system, and aims to solve the problems that in rolling of a rolling mill, the sampling rate difference of pressure and thickness sensors and material plastic deformation accumulation cause the time sequence dislocation of a rolling force peak value and a thickness valley value, the collaborative utilization of data by a thickness control algorithm is influenced, and the cost is reduced. In order to solve the problems that rolling force, thickness data and roller pulse signals are collected through a special controller, and a rolling length observation benchmark is converted; the dynamic phase compensation unit calculates a plastic deformation cumulant, after triggering compensation, a basic phase difference is established based on a sampling rate difference, a compensation coefficient is dynamically increased, a compensation data segment is generated through linear interpolation, and a time sequence is forcibly aligned; and the integrated control unit inputs the alignment data into a thickness control algorithm, the rolling force serves as feedforward, the thickness data serves as feedback, a precise roller gap adjusting instruction is generated, and the material thickness rolling precision is improved.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for rolling mills, and more specifically, to a dedicated controller and integrated control system for rolling mills without a PLC architecture. Background Technology

[0002] Automatic control of rolling mills is an important technology, specifically applied to the material thickness control stage of the rolling process. Its core is to improve the synergy of rolling data through timing calibration, meeting the mill's high-precision material thickness control requirements. During rolling, pressure sensors and thickness sensors have a fixed sampling rate difference, and material plastic deformation accumulates with increasing rolling length. These factors cause a misalignment in the physical response timing between the peak rolling force data and the trough material thickness data, affecting the thickness control algorithm's ability to utilize both types of data effectively. This makes it difficult to accurately generate roll gap adjustment commands and ensure the rolling accuracy of the material thickness. To solve this technical problem, we provide a PLC-free dedicated controller and integrated control system for rolling mills. Summary of the Invention

[0003] The purpose of this invention is to provide a PLC-free rolling mill dedicated controller and integrated control system to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, a PLC-free rolling mill dedicated controller is provided, including: The dedicated controller unit for the rolling mill collects rolling force data and material thickness data through pressure sensors and thickness sensors, and receives orthogonal pulse signals from the encoder at the end of the roll shaft through the roll rotary encoder. The pulse signals are converted into rolling length in real time by measuring the number of roll rotations, providing a rolling length observation benchmark for dynamic compensation. The dynamic phase compensation unit uses the rolling length observation benchmark to calculate the cumulative amount of plastic deformation in real time. The increase in rolling length is positively correlated with the cumulative amount of plastic deformation. When the cumulative amount of plastic deformation reaches a preset threshold, a phase delay trigger signal is generated. Upon receiving the phase delay trigger signal, the dynamic compensation engine is activated. The dynamic compensation engine establishes a basic phase difference compensation amount based on the fixed sampling rate difference between the pressure sensor and the thickness sensor, applies an initial timestamp offset to the rolling force data, and increases the dynamic compensation coefficient according to the increase in the cumulative amount of plastic deformation using a preset step function. An incremental time delay is injected into the material thickness data. Finally, a linear interpolation algorithm is used to dynamically generate compensation data segments between adjacent sampling points of the thickness sensor. By reconstructing the data stream in real time, the physical response timing of the peak moment of the rolling force data and the valley moment of the material thickness data are forcibly aligned. The integrated control output unit inputs the rolling force data and material thickness data after forced alignment of the physical response timing into the thickness control algorithm, and generates roll gap adjustment commands to the actuator.

[0005] The second objective of this invention is to provide an integrated control system for implementing a PLC-free rolling mill dedicated controller, characterized in that: the PLC-free rolling mill dedicated controller is mounted on the control board of the integrated control system, wherein the control board is provided with multiple connection interfaces corresponding to the PLC-free rolling mill dedicated controller, and a name identifier corresponding to each connection interface.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention collects multi-source data through a dedicated rolling mill controller unit, generates a precise rolling length observation benchmark through quadruple frequency counting and length conversion, providing a reliable basis for dynamic compensation. The dynamic phase compensation unit establishes a basic phase difference compensation amount based on sampling rate differences, and dynamically improves the compensation coefficient by combining the increase in the cumulative amount of plastic deformation. Through linear interpolation, continuous compensation data segments are generated to achieve time-series forced alignment of the peak rolling force and the valley thickness, effectively offsetting the time-series misalignment caused by sampling differences and deformation accumulation. The integrated control output unit inputs the aligned data into the thickness control algorithm, using rolling force as a feedforward term to predict the trend and thickness data as a feedback term to correct deviations, generating precise roll gap adjustment commands. It does not rely on a PLC architecture. Through closed-loop compensation and precise control, it improves the rolling accuracy of material thickness, ensures the stability and reliability of rolling mill operation, and perfectly meets the core requirement of high-precision thickness control of rolling mills. Attached Figure Description

[0007] Figure 1 This is an overall block diagram of the present invention.

[0008] The meanings of the labels in the diagram are as follows: 1. Rolling mill dedicated controller unit; 2. Dynamic phase compensation unit; 3. Integrated control output unit. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] This invention provides a PLC-free rolling mill dedicated controller; please refer to [link / reference]. Figure 1 As shown, it includes: The dedicated controller unit 1 for the rolling mill collects rolling force data and material thickness data through pressure sensors and thickness sensors, and receives orthogonal pulse signals from the encoder at the end of the roll shaft through the roll rotary encoder. The pulse signals are converted into rolling length in real time by measuring the number of roll rotations, providing a rolling length observation benchmark for dynamic compensation. The dynamic phase compensation unit 2 calculates the cumulative amount of plastic deformation in real time using the rolling length observation benchmark. The increase in rolling length is positively correlated with the cumulative amount of plastic deformation. When the cumulative amount of plastic deformation reaches a preset threshold, a phase delay trigger signal is generated. Upon receiving the phase delay trigger signal, the dynamic compensation engine is activated. The dynamic compensation engine establishes a basic phase difference compensation amount based on the fixed sampling rate difference between the pressure sensor and the thickness sensor. An initial timestamp offset is applied to the rolling force data, and the dynamic compensation coefficient is increased according to the increase in the cumulative amount of plastic deformation using a preset step function. An incremental time delay is injected into the material thickness data. Finally, a linear interpolation algorithm is used to dynamically generate compensation data segments between adjacent sampling points of the thickness sensor. By reconstructing the data stream in real time, the physical response timing of the peak moment of the rolling force data and the valley moment of the material thickness data are forcibly aligned. The integrated control output unit 3 inputs the rolling force data and material thickness data after the physical response timing is forcibly aligned into the thickness control algorithm, and generates a roll gap adjustment command to the actuator.

[0011] The generation of the rolling length observation benchmark includes: The orthogonal pulse signal from the encoder at the end of the roll shaft is received by the roll rotary encoder. A dual-channel counter is used to count the rising and falling edges of the orthogonal pulse signal by four times the frequency to obtain the cumulative number of pulses. The cumulative number of pulses is multiplied by the preset material travel length corresponding to each rotation of the roll, and the rolling length is output to the dynamic phase compensation unit 2.

[0012] The calculation of cumulative plastic deformation includes: In the dynamic phase compensation unit 2, a mapping relationship between rolling length and cumulative plastic deformation is established. The mapping relationship is realized by the embedded processor calling the material deformation coefficient lookup table. The material deformation coefficient is dynamically adjusted according to the instantaneous value of the rolling force, so that the increase in cumulative plastic deformation corresponding to each unit increase in rolling length increases stepwise with the increase of rolling force.

[0013] The generation and response of the phase delay trigger signal include: When the cumulative amount of plastic deformation exceeds the preset threshold, the processor comparator module outputs a high-level trigger signal to the enable port of the dynamic compensation engine. After receiving the high-level trigger signal, the dynamic compensation engine activates the three-stage pipeline processing architecture. The first-stage pipeline reads the basic phase difference compensation amount, the second-stage pipeline calculates the step function increment, and the third-stage pipeline initializes the interpolation buffer.

[0014] The establishment of the basic phase difference compensation includes: The dynamic compensation engine pre-stores the integer multiple relationship between the sampling period of the pressure sensor and the sampling period of the thickness sensor. After dividing the sampling clock signal of the thickness sensor, it compares the phase of the sampling clock signal with that of the sampling clock of the pressure sensor to generate an initial timestamp offset and writes it into the index value of the circular buffer.

[0015] The improvement of the dynamic compensation coefficient includes: The incremental value of the compensation coefficient is output by querying the step function mapping table based on the cumulative increase of plastic deformation. The initial timestamp offset is then superimposed with the incremental value of the compensation coefficient to generate a composite offset, which is then reloaded into the ring buffer index register. This ensures that the overall delay of the rolling force data stream on the time axis is equal to the duration of the composite offset.

[0016] Incremental delay injection includes: Material thickness data is written to the front end of the circular buffer according to the sampling timestamp, and data is read from the back end of the circular buffer according to the composite offset index value. When the composite offset index value crosses the adjacent sampling point, a transition data segment generated by the interpolation algorithm is inserted into the read data stream to realize the continuous translation of the material thickness data time axis.

[0017] The generation of the compensation data segment includes: A linear interpolation algorithm is used to process adjacent sampling point data of the thickness sensor. The value of the previous sampling point is used as a reference, and the value is linearly and gradually changed to the value of the next sampling point according to the time ratio. Interpolation data points with equal time intervals are generated within the original sampling interval to fill the data gaps caused by the index value crossing the sampling point.

[0018] Physical response timing forced alignment includes: The local maxima and local minima of the differential curve of rolling force data and material thickness data are monitored in real time. When the time deviation between the local maxima and local minima exceeds the preset tolerance, the incremental value of the compensation coefficient in the composite offset is adaptively adjusted until the difference between the peak rolling force timestamp and the valley thickness timestamp is stable within the tolerance range.

[0019] The generation of roll gap adjustment commands includes: The timing-aligned rolling force data and material thickness data are input into the data fusion layer of the thickness control algorithm. The rolling force data is used as a feedforward term to predict the thickness change trend, and the thickness data is used as a feedback term to correct the prediction deviation. The roll gap adjustment amount is then output to the servo motor drive module.

[0020] Further explanation is needed: to provide the dynamic phase compensation unit with an accurate rolling length observation benchmark, the mill-specific controller unit must first acquire rolling length data through sensor signal acquisition and processing. The core is to convert roll rotation information into material travel length. The specific implementation process is as follows: The orthogonal pulse signal from the roll shaft-end encoder is received by the roll rotary encoder. The roll rotary encoder is a sensor that converts the mechanical rotational displacement of the roll into an electrical signal. Its core function is to capture the rotational state of the roll in real time and output pulse signals related to the number of rotations and the rotational speed. The roll shaft-end encoder is an encoding component directly installed on the roll shaft end, rotating synchronously with the roll to ensure that the output signal accurately reflects the actual rotation of the roll. The orthogonal pulse signal refers to two pulse signals with a 90-degree phase difference (usually marked as phase A and phase B) output by the encoder. It can not only calculate the number of rotations by the number of pulses, but also determine the direction of roll rotation by the phase relationship between the two signals, providing a basis for subsequent accurate counting. The specific receiving process is as follows: The roll rotary encoder is connected to the pulse signal interface of the rolling mill's dedicated controller unit via a dedicated signal cable. The interface uses a differential signal transmission mode to resist electromagnetic interference in the industrial environment. When the roll rotates, the shaft-end encoder rotates synchronously and outputs two orthogonal pulse signals, A and B. The signals are transmitted to the controller's signal conditioning module via cable. This module filters and shapes the original pulse signal, removes noise interference, and normalizes the signal into a standard square wave pulse to ensure that the subsequent counting module can stably identify it. After signal conditioning, a dual-channel counter is used to perform quadruple frequency counting on the rising and falling edges of the quadrature pulse signal to obtain the cumulative pulse count. The dual-channel counter is a dedicated counting module integrated within the controller, containing two independent and synchronously operating counting channels, corresponding to phase A and phase B of the quadrature pulse signal respectively. It can process two signals simultaneously and achieve coordinated counting. The rising edge of the pulse refers to the instant when the pulse signal jumps from a low level to a high level, and the falling edge is the instant when it jumps from a high level to a low level. Capturing these two moments improves the counting resolution. Quadruple frequency counting is a counting method that utilizes the phase characteristics of quadrature signals. By simultaneously detecting the rising and falling edges of the A and B phase pulses, a counting pulse is generated for every 1 / 4 revolution. Compared to the counting method that only detects a single edge, the resolution is improved by four times, allowing for more accurate quantification of the minute rotation of the roll. The cumulative pulse count refers to the total number of all valid pulses recorded by the counter from the start of counting to the current moment, which is directly proportional to the roll rotation angle. The specific counting process is as follows: The two channels of the dual-channel counter are respectively connected to the conditioned A-phase and B-phase pulse signals. The internal logic circuit of the counter simultaneously monitors the rising and falling edges of each signal. Whenever an edge transition of either phase is detected, the counter automatically increments by 1. At the same time, the rotation direction of the roll is determined by the phase relationship between the two signals. If the A-phase pulse leads the B-phase pulse, it is a forward rotation, and the counter increments in the forward direction. If the B-phase pulse leads the B-phase pulse, it is a reverse rotation, and the counter increments in the reverse direction (to avoid counting distortion when the material retracts). During the counting process, the counter stores the accumulated pulse count in real time to the controller's buffer unit to ensure that the data can be retrieved at any time. Finally, the cumulative number of pulses is multiplied by the preset material travel length corresponding to each rotation of the roll, and the rolling length is output to the dynamic phase compensation unit 2. The preset material travel length corresponding to each rotation of the roll is the effective circumference of the roll, which refers to the circumference of the part of the roll that contacts the material. This value is calculated from the design diameter of the roll (considering the elastic deformation of the roll during actual operation, it needs to be preset and stored in the parameter configuration module of the controller after factory calibration, and can be fine-tuned according to the actual working conditions). Its core is to establish a fixed mapping relationship between the number of roll rotations and the material travel distance. The rolling length refers to the cumulative travel length of the material after being rolled by the roll from the start of rolling to the current moment. It is the core reference data for the dynamic phase compensation unit to calculate the cumulative amount of plastic deformation. The dynamic phase compensation unit 2 is the functional unit in the system responsible for processing data timing alignment and deformation compensation. It needs to rely on the rolling length data to establish the relationship between the amount of deformation and the rolling distance. The specific conversion and output process is as follows: The controller retrieves the preset material travel length per roll from the parameter configuration module and reads the current cumulative pulse count from the buffer unit; the two values ​​are multiplied to obtain the current rolling length (for example, if the cumulative pulse count is 1000 and the material travel length per roll is 0.5 meters, then the rolling length is 500 meters); after the calculation is completed, the controller transmits the rolling length data to the input interface of the dynamic phase compensation unit 2 in real time through the internal data bus, along with a data validity indicator, to ensure that the compensation unit receives accurate and usable observation reference data, laying the foundation for the subsequent calculation of the cumulative amount of plastic deformation.

[0021] After the rolling mill dedicated controller unit outputs a precise rolling length observation benchmark, one of the core tasks of the dynamic phase compensation unit 2 is to establish the correlation between this benchmark and the cumulative amount of plastic deformation, providing a core basis for the generation of subsequent phase delay trigger signals. This correlation needs to conform to the mechanical properties of the material rolling to ensure that the deformation calculation can respond in real time to changes in the rolling conditions. The specific implementation process is as follows: In dynamic phase compensation unit 2, a mapping relationship between rolling length and cumulative plastic deformation is established. Rolling length is the cumulative material travel length obtained by pulse counting and length conversion, directly reflecting the material's rolling process. Cumulative plastic deformation is the sum of irreversible deformations occurring under rolling pressure. This value continuously increases with the rolling process and is a key indicator for judging the rolling state of the material. The mapping relationship refers to establishing a fixed correlation between the change in rolling length and the increase in cumulative plastic deformation through a quantitative model, ensuring that the other variable can be accurately derived when one variable is known. The specific construction process is as follows: Based on the mechanical experimental data of the target rolled material, a basic correlation model between rolling length and cumulative plastic deformation is fitted. The core logic of the model is that the longer the rolling length, the greater the cumulative plastic deformation. Simultaneously, a coefficient adjustment interface is reserved to allow dynamic correction of the correlation strength based on real-time rolling force, avoiding the limitation of a fixed model being unable to adapt to complex working conditions. This mapping relationship is achieved by the embedded processor calling the material deformation coefficient lookup table. The embedded processor is a dedicated computing core built into the dynamic phase compensation unit 2, featuring low latency and high reliability, meeting the real-time data processing needs of industrial scenarios. The material deformation coefficient lookup table is a two-dimensional data table pre-stored in the processor's storage module. The horizontal axis represents the rolling material type, and the vertical axis represents the rolling force range (a continuous range divided by the magnitude of the rolling force). Each cell in the table stores the material deformation coefficient for the corresponding material type and force range. This coefficient is based on empirical values ​​calibrated from numerous rolling experiments and directly determines the correlation between rolling length and deformation. The specific calling process is as follows: The embedded processor first reads the type of the currently rolled material from the system parameter configuration module, and simultaneously receives rolling force data transmitted from the pressure sensor in real time. Using the material type and rolling force data as indexes, it precisely locates the corresponding cell in the material deformation coefficient lookup table, retrieves the material deformation coefficient stored in that cell, and substitutes it into the previously constructed basic correlation model. This allows the model to output the cumulative amount of plastic deformation matching the current working condition. The material deformation coefficient is dynamically adjusted based on the instantaneous value of the rolling force. The material deformation coefficient is the core parameter for quantifying the degree of plastic deformation of the material per unit rolling length; the larger the coefficient, the greater the deformation for the same rolling length. The instantaneous value of the rolling force is the rolling force data collected in real time by the pressure sensor at the same sampling frequency as the thickness sensor, accurately capturing the dynamic fluctuations of the rolling force. Dynamic adjustment refers to automatically switching the corresponding material deformation coefficient as the instantaneous value of the rolling force changes, ensuring that the deformation calculation can respond to changes in force in real time. The specific adjustment process is as follows: The embedded processor compares the instantaneous value of the rolling force received in real time with the rolling force range in the lookup table. If the instantaneous value falls into a new rolling force range, it immediately switches to the material deformation coefficient corresponding to that range. To avoid frequent coefficient switching caused by small fluctuations in rolling force, a hysteresis threshold is set. The coefficient is only triggered when the instantaneous value of the rolling force crosses the hysteresis threshold, ensuring the stability and reliability of the adjustment process. Through the above dynamic adjustment mechanism, the cumulative increase in plastic deformation per unit rolling length increases stepwise with the increase in rolling force. The increase in rolling length per unit refers to each fixed reference value (calibrated to 1 mm, taking into account both calculation accuracy and efficiency) for each increase in rolling length. The increase in cumulative plastic deformation is the increment corresponding to the increase in cumulative plastic deformation when the rolling length increases by that unit. The stepwise increase means that as the rolling force enters a higher range, the increase jumps by a preset ratio, rather than changing continuously, which conforms to the deformation law of materials under different stress intensities (the greater the rolling force, the easier it is for the material to undergo plastic deformation, and the more significant the increase in deformation per unit length). The specific implementation process is as follows: In the material deformation coefficient lookup table, several continuous intervals are divided according to the rolling force from low to high (such as three intervals: low, medium, and high). The material deformation coefficient corresponding to each interval is configured in a step-by-step manner (such as the coefficient k1 for the low force interval, k2 for the medium force interval, and k3 for the high force interval, with k3>k2>k1). When the instantaneous value of the rolling force moves from the low force interval to the medium force interval, the material deformation coefficient switches from k1 to k2, and the increase in the cumulative amount of plastic deformation per unit rolling length jumps from the increment corresponding to k1 to the increment corresponding to k2. When the rolling force further enters the high force interval, the coefficient switches to k3, and the increase jumps again. In the end, a step-by-step improvement effect is achieved, where the greater the rolling force, the greater the increase in deformation per unit length, ensuring that the calculation of the cumulative amount of plastic deformation can truly reflect the deformation characteristics of the material under different rolling forces.

[0022] After the dynamic phase compensation unit 2 completes the real-time calculation of the cumulative amount of plastic deformation, when the cumulative amount exceeds the preset threshold, the system will initiate the phase delay triggering and dynamic compensation engine activation process to lay the foundation for subsequent data timing alignment. The specific implementation process is as follows: When the accumulated plastic deformation exceeds a preset threshold, the processor comparator module outputs a high-level trigger signal to the enable port of the dynamic compensation engine. The preset threshold is a critical value calibrated based on the plastic deformation limit of the target rolled material, rolling process requirements, and equipment safety operation parameters. This value is pre-stored in the parameter storage module of the dynamic phase compensation unit after extensive experimental verification and can be flexibly adjusted according to material type or rolling specifications. The processor comparator module is a dedicated logic unit built into the dynamic phase compensation unit. Its core function is to receive the calculation result of the accumulated plastic deformation in real time and compare it with the preset threshold at high speed. The comparison response delay is controlled at the microsecond level to ensure the real-time performance of the trigger signal. The high-level trigger signal is a signal state that represents a valid signal state in digital circuits (usually 3.3V or 5V voltage), which is different from the invalid state of low level and can clearly activate the subsequent execution unit. The dynamic compensation engine is the core functional module responsible for phase difference calculation, time delay injection, and data reconstruction, and has high-speed data processing capabilities. The enable port is the activation control interface of the dynamic compensation engine. It will only switch from standby state to working state when a high-level trigger signal is received to avoid meaningless resource consumption. The specific process is as follows: The processor comparator module reads the current value of the accumulated plastic deformation in real time and continuously compares it with a preset threshold. When the accumulated amount is detected to be greater than the threshold, a high-level trigger signal is immediately output. This signal is transmitted to the enable port of the dynamic compensation engine through internal hardware circuitry, completing the transmission of the trigger command. After receiving the high-level trigger signal, the dynamic compensation engine quickly activates the three-stage pipeline processing architecture. The three-stage pipeline processing architecture is a parallel processing mechanism that decomposes the core task of dynamic compensation into three independent and continuous processing stages. Each stage is handled by a dedicated computing unit. The processing result of the previous stage is transmitted to the next stage in real time. The three stages work synchronously and in parallel, greatly improving the overall processing efficiency and avoiding the latency caused by single-threaded processing. The specific activation process is as follows: After the enable port receives the high-level signal, the control logic inside the dynamic compensation engine immediately releases the standby state of each pipeline unit, initializes the working parameters of each unit, establishes a data transmission channel between units, and simultaneously starts the clock synchronization signal to ensure that the three-stage pipeline works collaboratively in a unified timing sequence. The time from activation to entering a stable working state does not exceed 10 microseconds, meeting the time requirements of real-time compensation. The core task of the first-stage pipeline is to read the basic phase difference compensation value. This value is an initial phase compensation value established in advance based on the fixed sampling rate difference between the pressure sensor and the thickness sensor. Its core function is to compensate for the inherent timing deviation caused by the different sampling periods of the two types of sensors. This value is stored in the dedicated register of the dynamic compensation engine and does not require real-time calculation; it only needs to be read and called directly. The specific process is as follows: After activation, the first-level pipeline directly accesses a dedicated register based on a preset address index to read the stored basic phase difference compensation data. Simultaneously, it verifies the data's validity. Upon successful verification, the compensation amount is transmitted in real-time to the second-level pipeline, providing the foundational data for subsequent composite offset calculations. The entire reading and verification process is completed within one clock cycle, ensuring processing efficiency. The second-level pipeline simultaneously calculates the step function increment. This step function increment is a dynamically generated compensation coefficient increment value based on the increase in the cumulative amount of plastic deformation. Its core principle is to allow the compensation amount to increase stepwise with the increase in deformation, conforming to the correlation between deformation and phase deviation during material rolling. The step function mapping table is a data table pre-stored in the dynamic compensation engine. The horizontal axis of the table represents the increase range of the cumulative amount of plastic deformation, and the vertical axis represents the corresponding increment value. Each increase range corresponds to a fixed increment, presenting a stepwise distribution. The specific process is as follows: The second-stage pipeline receives the basic phase difference compensation amount from the first-stage pipeline while simultaneously acquiring the real-time increase data of the current cumulative plastic deformation (i.e., the difference between the current cumulative amount and the previous cumulative amount). This increase data is matched with the interval in the step function mapping table to accurately locate the corresponding increment value, which is the step function increment. The basic phase difference compensation amount and the step function increment are then temporarily stored to prepare for the generation of the subsequent composite offset. The entire matching and calculation process is parallel and synchronous with the reading process of the first-stage pipeline. The third-stage pipeline synchronously initializes the interpolation buffer. The interpolation buffer is a high-speed cache area in the dynamic compensation engine specifically used to store interpolation data and transition data segments. Its capacity is calibrated according to the sensor sampling rate and interpolation density to ensure it can accommodate all the transition data required for a single compensation, avoiding data overflow or loss. Initialization refers to clearing old data, setting the data storage format, and calibrating the start position of the read / write pointers before using the buffer, ensuring that the buffer is in a clean and usable state. The specific process is as follows: After the third-stage pipeline is activated, a clear command is first sent to remove residual historical data from the interpolation buffer. Then, the data storage width of the buffer is configured (to adapt to the accuracy requirements of thickness data), the initial position of the read / write pointer is set (starting from the buffer's starting address), and a data transmission interface is established between the buffer and subsequent interpolation algorithm units. After initialization, the buffer enters a standby state, waiting to receive the transition data segment generated by the interpolation algorithm. The entire initialization process is completed in parallel with the processing of the first two stages of the pipeline, realizing the synchronous advancement of the three types of tasks and making full preparations for subsequent incremental delay injection and compensation data segment generation.

[0023] After the dynamic compensation engine activates the three-stage pipeline, the basic phase difference compensation amount read by the first-stage pipeline needs to be constructed by comparing the pre-stored sensor sampling period relationship with the clock phase to ensure that the compensation amount can accurately offset the inherent timing deviation of the two types of sensors, laying the foundation for subsequent composite offset calculation. The specific implementation process is as follows: The dynamic compensation engine pre-stores the integer multiple relationship between the sampling periods of the pressure sensor and the thickness sensor. The dynamic compensation engine is the core computing module of the dynamic phase compensation unit, possessing data storage and real-time computing capabilities, and is specifically responsible for parameter processing and signal adjustment related to timing compensation. The pressure sensor sampling period is the time interval between two acquisitions of rolling force data by the pressure sensor, and the thickness sensor sampling period is the time interval between two acquisitions of material thickness data by the thickness sensor. Both are fixed values ​​calibrated at the factory. The shorter the sampling period, the more frequent the data updates, and the higher the timing accuracy. The integer multiple relationship means that the longer period is an integer multiple of the shorter period (e.g., the thickness sensor sampling period is 2 milliseconds, and the pressure sensor is 1 millisecond, i.e., a 2-fold relationship). This relationship ensures accurate synchronization after subsequent clock signal frequency division. The specific pre-store process is as follows: The sampling period parameters of the pressure sensor and thickness sensor are entered into the non-volatile storage module of the dynamic compensation engine through the configuration interface. The engine's built-in verification unit automatically calculates the ratio of the two periods to verify whether it meets the integer multiple requirement. If it does not meet the requirement, an alarm is triggered, and the matching sensor needs to be reselected or the sampling parameters need to be adjusted. After the verification is successful, the engine stores the integer multiple relationship together with the two original sampling periods to form a fixed parameter configuration. In subsequent rolling processes, there is no need to repeat the configuration; it can be called only. Based on a pre-stored integer multiple relationship, the thickness sensor sampling clock signal is frequency-divided and compared with the pressure sensor sampling clock signal in phase to generate an initial timestamp offset, which is then written into the circular buffer index value. The sampling clock signal is a pulse signal that controls the timing of data acquisition by the sensor. Each time the sensor receives a clock pulse, it completes one data acquisition. The frequency of the clock signal is inversely proportional to the sampling period. Frequency division is the process of reducing the frequency of the thickness sensor's sampling clock signal to match the sampling clock frequency of the pressure sensor using a digital frequency divider. For example, if the thickness sensor clock frequency is 500Hz (2 milliseconds period) and the pressure sensor clock frequency is 1000Hz (1 millisecond period), then a 2-fold frequency division is used to reduce the thickness sensor clock frequency to 500Hz, ensuring synchronization between the two clock frequencies. Phase comparison provides the prerequisite; phase comparison refers to comparing the phase difference of two synchronous frequency clock signals in real time using a phase comparator. The phase difference is the time difference between the rising edges of the two clock pulses, which directly reflects the inherent timing deviation of the data acquired by the two types of sensors. The initial timestamp offset is a time compensation value calculated based on the phase difference, used to compensate for the timing deviation caused by the phase asynchrony between the two types of sensors, ensuring that the starting reference of the data on the time axis is consistent. The circular buffer is a high-speed storage area in the dynamic compensation engine used to cyclically store the data stream. It adopts a first-in-first-out circular read / write mechanism to avoid data overflow and ensure high-speed data transmission. The index value is the identifier of the data storage location in the circular buffer. By adjusting the index value, the time offset of the data stream can be achieved. The specific process is as follows: The dynamic compensation engine calls a pre-stored integer multiple relationship to control the digital frequency divider to divide the thickness sensor sampling clock signal, generating a synchronous clock signal with the same frequency as the pressure sensor sampling clock signal. Then, the synchronized thickness sensor clock signal and the pressure sensor sampling clock signal are input to a phase comparator to measure the phase difference between the two in real time. This phase difference is converted into a corresponding time difference value, which is the initial timestamp offset. The engine writes the initial timestamp offset into the index register of the circular buffer through the internal bus, updating the starting position of the index value. This ensures that the rolling force data and thickness data stored in the circular buffer can achieve initial timing alignment based on this offset. For example, if the initial timestamp offset is 0.5 milliseconds, the index value of the rolling force data in the circular buffer will be shifted to the corresponding data position to ensure that it is consistent with the acquisition timing reference of the thickness data.

[0024] In the three-stage pipeline processing of the dynamic compensation engine, the second-stage pipeline needs to optimize the compensation parameters based on the real-time changes in plastic deformation, so that the timing compensation can accurately respond to the dynamic deformation during the rolling process. The specific implementation process is as follows: First, the step function mapping table is consulted based on the cumulative increase in plastic deformation. The cumulative increase in plastic deformation is the difference between the cumulative plastic deformation at the current sampling time and the cumulative plastic deformation at the previous sampling time, directly reflecting the growth rate of material plastic deformation per unit time. A larger increase indicates faster deformation and a greater need for dynamic compensation. The step function mapping table is a standardized data table pre-stored in the dynamic compensation engine's storage module. The horizontal axis divides continuous intervals according to the magnitude of the cumulative increase in plastic deformation, and the vertical axis represents the fixed compensation coefficient increment value corresponding to each interval. The core feature of the table is that the increment is fixed within each interval, and there are step-like jumps across intervals, closely matching the non-linear relationship between material deformation and compensation requirements. The specific query process is as follows: the embedded processor calculates the current cumulative increase in plastic deformation in real time, compares this value with each interval on the horizontal axis of the step function mapping table, finds the target interval to which the increase belongs, and accurately locates the vertical axis value corresponding to the target interval, ensuring that the query result matches the current deformation growth rate. After the query is completed, the incremental value of the compensation coefficient is output. The incremental value of the compensation coefficient is a dynamic adjustment used to enhance the compensation strength. Its value is positively correlated with the increase of the cumulative amount of plastic deformation. The larger the increase, the larger the incremental value, which can inject a dynamic adjustment component into the basic offset. The output process must ensure the real-time performance and validity of the data. Specifically, after the processor extracts the incremental value of the compensation coefficient corresponding to the target interval from the step function mapping table, it first verifies whether the value is within the preset reasonable range through the data verification module to eliminate outlier interference. After the verification is passed, the incremental value is transmitted to the arithmetic operation unit through the internal high-speed data bus to prepare for the subsequent offset superposition. Then, the initial timestamp offset is superimposed with the incremental value of the compensation coefficient. The initial timestamp offset is the basic compensation value obtained by comparing the clock phase of the pressure sensor and the thickness sensor. It is used to offset the timing deviation caused by the inherent sampling rate difference between the two types of sensors and is the basis of compensation. Superposition refers to the addition operation of the two parameters in the arithmetic operation unit to combine the basic deviation and the additional deviation brought by dynamic deformation to form a more comprehensive compensation amount. The specific process is as follows: The processor retrieves the stored initial timestamp offset from the circular buffer index register, ensuring that its time unit is consistent with the compensation coefficient increment (both are in the microsecond range). Then, it performs addition in the arithmetic unit to precisely sum the two values, avoiding compensation deviations caused by unit confusion or insufficient calculation precision. After superposition, a composite offset is generated and reloaded into the circular buffer index register. The composite offset is the final offset obtained after the superposition operation, containing both the basic components to offset fixed sampling rate differences and the incremental components to handle dynamic deformations; it is a core parameter for adjusting data flow timing. The circular buffer index register is a hardware component in the dynamic compensation engine that stores the circular buffer read / write address index. The index value directly determines the starting position of data reading, and the timestamp of the data can be indirectly adjusted by modifying the index value. Reloading means overwriting the index value corresponding to the original initial timestamp offset in the register with the newly generated composite offset, ensuring that subsequent data readings are performed based on the new compensation parameters. The specific process is as follows: After generating the composite offset through addition, the processor sends a write enable signal to the circular buffer index register, converting the composite offset into the corresponding index address value and writing it to the register, overwriting the original index value. Simultaneously, the register's status flag is updated to inform the system that the index has been updated, ensuring that subsequent data read / write processes can respond promptly to the new compensation parameters. Ultimately, this results in the rolling force data stream being delayed on the time axis by a duration equal to the composite offset. The rolling force data stream is a sequence of rolling force data continuously acquired and transmitted by pressure sensors. Each data point has a corresponding acquisition timestamp, forming a continuous time axis distribution. The overall time axis delay refers to postponing the timestamps of all data points in the entire rolling force data stream by a fixed duration, achieving a temporal shift of the data stream, rather than a local adjustment of a single data point. The duration equal to the composite offset is the specific delay length; the value of the composite offset directly corresponds to the delay duration, with consistent units. The specific process is as follows: The circular buffer performs data reading operations according to the reloaded index value. The rolling force data that was originally read from index address X is now read from the address corresponding to index address X + composite offset. This is equivalent to delaying the reading time of all rolling force data by the duration corresponding to the composite offset. Reflected on the timeline, the timestamps of the entire rolling force data stream are synchronously shifted backward, and the delay duration is completely consistent with the value of the composite offset. This accurately offsets the additional timing deviation caused by dynamic deformation and provides a guarantee for the forced alignment with the thickness data in the future.

[0025] After the dynamic compensation engine generates the composite offset and reloads it to the circular buffer index register, the next step is to inject incremental delay into the material thickness data through the read and write operations of the circular buffer. The core is to shift the time axis of the thickness data with the composite offset while ensuring data continuity. The specific implementation process is as follows: Material thickness data is written to the front end of the circular buffer according to the sampling timestamp. The material thickness data is a quantitative data reflecting the current thickness of the rolled material, collected in real time by the thickness sensor. Each data point is accompanied by a unique sampling timestamp to ensure time traceability. The sampling timestamp is a precise time record of when the sensor collected the thickness data, generated synchronously by the system clock, and directly related to the data acquisition time sequence. The circular buffer is a high-speed storage area in the dynamic compensation engine that uses a circular read-write mechanism. Its storage space is divided into fixed sizes. After data is written to the end, it will automatically overwrite the old data at the beginning to avoid storage overflow and ensure high-speed response of data read and write. The front end of the circular buffer refers to the writing end in the buffer used to receive new data. It forms an independent channel with the reading end at the back end and supports parallel read and write operations. The specific writing process is as follows: Each time the thickness sensor acquires material thickness data, it binds the data to the corresponding sampling timestamp and transmits it to the front-end interface of the circular buffer via an internal high-speed data bus. The buffer's write controller stores the data sequentially into consecutive storage units of the buffer according to the sampling timestamps, while simultaneously updating the write pointer position to ensure that subsequent data is written sequentially without data corruption or loss. The entire writing process is synchronized with data acquisition, with latency controlled at the microsecond level. Simultaneously, data is read from the back end of the circular buffer according to the composite offset index value. The back end of the circular buffer is the read terminal used to output stored data, executing in parallel with the front-end write operation without interference. The composite offset index value is the final index address obtained by superimposing the initial timestamp offset and the compensation coefficient increment value, and its value directly corresponds to the required delay time length. It has been reloaded into the index register of the circular buffer. The specific reading process is as follows: The read controller of the circular buffer retrieves the composite offset index value from the index register in real time and converts the index value into a physical storage address in the buffer. Starting from this address, the stored material thickness data is read sequentially from the back end of the circular buffer. The reading order is consistent with the writing order to ensure the timing integrity of the data. During the reading process, the read pointer and the write pointer work together. Through the circular mechanism of the buffer, even if the write pointer exceeds the read pointer, the corresponding historical data can still be read accurately, realizing continuous data reading and providing a stable data stream for subsequent delay injection. When the composite offset index value crosses adjacent sampling points, a transition data segment generated by the interpolation algorithm is inserted into the read data stream to achieve continuous translation of the material thickness data time axis. The composite offset index value crossing adjacent sampling points means that the physical storage address corresponding to the composite offset does not fall on a complete sampling point storage unit, but is between the storage addresses of two adjacent sampling points. Direct reading in this case would cause data timing breaks. The transition data segment is a continuous data sequence between two adjacent sampling point data generated by the interpolation algorithm. Its value changes linearly with time, filling the gaps between sampling points. The interpolation algorithm used here is a linear interpolation algorithm, which is simple to calculate and has strong real-time performance, suitable for dynamic compensation in industrial scenarios. Continuous translation of the material thickness data time axis means that by offsetting the data reading position and inserting the transition data segment, the overall timing of the thickness data stream is delayed by the duration corresponding to the composite offset, without data breaks or jumps, maintaining continuous smoothness. The specific process is as follows: The controller reads the physical address corresponding to the composite offset index value in real time. When the address is detected to be between the storage addresses of two adjacent sampling points, the interpolation algorithm is immediately triggered. The interpolation algorithm retrieves the thickness data and corresponding sampling timestamps of these two adjacent sampling points. Using the data of the previous sampling point as the starting value and the data of the next sampling point as the ending value, it generates equally spaced transition data points according to the time ratio. These data points form a transition data segment. The transition data segment is inserted between the data of two adjacent sampling points to replace the originally broken reading data. In this way, the material thickness data stream is shifted as a whole on the time axis. The shift duration is completely consistent with the composite offset, and the data is continuous and distortion-free. The incremental delay is successfully injected, laying the foundation for subsequent timing alignment with the rolling force data.

[0026] When inserting transition data segments into the material thickness data stream to achieve continuous time axis shifting, the core is to use a linear interpolation algorithm to process adjacent sampling point data from the thickness sensor. This algorithm has the advantages of low computational cost and strong real-time performance, and can accurately fit the changing trends of adjacent data, ensuring seamless connection between the transition data and the original data. The specific implementation process is as follows: First, a linear interpolation algorithm is used to process the adjacent sampling point data of the thickness sensor. The linear interpolation algorithm is based on establishing a linear relationship between two known data points (adjacent sampling points), and then deriving the data value at any position between the two points. Its core assumption is that the change in material thickness between adjacent sampling points follows a uniform linear trend, which can maintain data smoothness while ensuring computational efficiency. Adjacent sampling point data of the thickness sensor refers to two consecutive valid data points collected by the thickness sensor at a fixed sampling period. Each data point contains a specific thickness value and a corresponding sampling timestamp, serving as the basic data source for interpolation calculations. These two points must ensure temporal continuity and the absence of outliers (such as jump data caused by sensor malfunctions, which have been filtered out in advance). The specific processing procedure is as follows: The interpolation unit of the dynamic compensation engine accurately retrieves the data from two adjacent original sampling points spanned by the composite offset index value from the circular buffer. First, the data verification module confirms that the thickness values ​​of these two points are within a reasonable range, and that the sampling timestamps are continuous and non-repeating, avoiding invalid data from affecting the interpolation results. After verification, this data is used as the reference data for interpolation calculation. Then, using the previous sampling point value as the reference, the data is linearly and gradually transitioned to the next sampling point value according to the time ratio. The previous sampling point value refers to the thickness value of the data point with the earlier time sequence among the two adjacent sampling points, serving as the starting reference for the linear transition and ensuring the continuity of data change. The time ratio refers to the proportion of the difference between the timestamp corresponding to a certain interpolation position and the timestamp of the previous sampling point to the time interval between two adjacent sampling points. This proportion directly determines the magnitude of the interpolated data value, and the proportion range is between 0 and 1. Linear transition means that the interpolated data value increases or decreases uniformly from one sampling point to the next, without abrupt changes, perfectly matching the stable change law of material thickness over a short period. The specific process is as follows: The interpolation unit first calculates the time interval between two adjacent original sampling points (i.e., the timestamp of the later sampling point minus the timestamp of the earlier sampling point), and then determines the target time range corresponding to the transition data segment (i.e., the time period without original data spanned by the composite offset index value). For each potential interpolation position within the target time range, the time difference between it and the previous sampling point is calculated one by one. This time difference is divided by the time interval between adjacent sampling points to obtain the corresponding time ratio. Based on this time ratio, with the thickness value of the previous sampling point as the starting point and the thickness value of the later sampling point as the ending point, the thickness data value corresponding to the interpolation position is obtained through linear calculation. For example, if the value of the previous sampling point is 5.2mm and the value of the later sampling point is 5.8mm, and the time ratio of a certain interpolation position is 0.3, then the interpolation data value at that position is 5.2mm + (5.8mm - 5.2mm) × 0.3, thus achieving a smooth transition of values ​​from front to back. Finally, interpolated data points with equal time intervals are generated within the original sampling interval to fill the data gaps caused by the index value crossing the sampling point. The original sampling interval is a fixed sampling period preset by the thickness sensor (e.g., 2 milliseconds), which is the time length between two adjacent original sampling points and is the fixed time boundary for the generation of interpolated data points. Interpolated data points with equal time intervals refer to continuous data points generated within the original sampling interval at uniform small time intervals (calibrated to 1 / 10 of the original sampling interval, such as 0.2 milliseconds). High-density interpolated points can ensure the continuity and smoothness of the data. Data gaps refer to the absence of thickness data corresponding to that time period in the original data stream when the composite offset index value does not fall on the storage address of a certain original sampling point, but is between two adjacent sampling points. If these gaps are not filled, the thickness data stream will jump, affecting the subsequent timing alignment accuracy. The specific process is as follows: The interpolation unit calculates the number of interpolation data points to be generated within the target time range based on the original sampling interval and the preset micro-interval (e.g., if the original sampling interval is 2 milliseconds and the micro-interval is 0.2 milliseconds, then 10 interpolation data points are generated); a unique timestamp is assigned to each interpolation data point at equal time intervals to ensure that the timestamps are continuous and non-overlapping; the thickness value corresponding to each timestamp is obtained one by one through the linear gradient calculation method described above, and these interpolation data points are combined in the order of timestamps to form a complete transition data segment; finally, the transition data segment is inserted between two adjacent original sampling data points to accurately fill the data gaps, so that the entire material thickness data stream presents a continuous and smooth change trend on the time axis without breaks or jumps, providing high-quality complete data stream support for the forced alignment of the physical response timing of the subsequent rolling force data and thickness data.

[0027] After generating compensation data segments using a linear interpolation algorithm and achieving continuous translation of the material thickness data time axis, the dynamic phase compensation unit needs to further complete the forced alignment of the physical response timing. The core is to monitor the key timing points of the rolling force and thickness data, and dynamically fine-tune the compensation parameters to ensure that the physical response laws of the two are accurately matched. The specific implementation process is as follows: First, the local maxima and minima of the differential curve of rolling force data and the differential curve of material thickness data are monitored in real time. The differential curve of rolling force data is obtained by differentiating the original rolling force data stream. Differentiation highlights the rate of change of rolling force, making the peak position of force easier to identify. The local maxima are the critical points in the differential curve of rolling force data where the value changes from positive to negative, corresponding to the peak moment of the original rolling force data. At this time, the rolling force is the largest, and the extrusion effect on the material is the strongest. The differential curve of material thickness data is the curve obtained by differentiating the original thickness data stream, used to highlight the trend of thickness change. The local minima are the critical points in the differential curve of thickness data where the value changes from negative to positive, corresponding to the valley moment of the original thickness data. At this moment, the material is compressed to its thinnest due to the maximum rolling force, which conforms to the physical law that the thickness is thinnest when the force is maximum during rolling. Therefore, the time alignment of these two points is the core of achieving overall data time matching. The specific monitoring process is as follows: The dynamic compensation engine has a built-in differential calculation module that performs real-time differential processing on the time-adjusted rolling force data and material thickness data to generate corresponding differential curves. At the same time, it uses a sliding window method (the window size is calibrated to 5 sampling points) to traverse the two differential curves. When it is detected that the values ​​of the adjacent points before and after a certain point in the rolling force differential curve are all less than that point (i.e., a local maximum point), and the values ​​of the adjacent points before and after a certain point in the thickness differential curve are all greater than that point (i.e., a local minimum point), the sampling timestamps corresponding to these two points are recorded, completing the real-time capture of key time points. The entire monitoring process is synchronized with data processing, with a delay of no more than one sampling cycle. When the time deviation between local maxima and local minima exceeds the preset tolerance, the incremental value of the compensation coefficient in the composite offset is adaptively adjusted. The time deviation refers to the difference between the captured peak rolling force timestamp and the valley thickness timestamp, directly reflecting the degree of timing misalignment between the two key physical response moments. The preset tolerance is the maximum allowable time deviation set based on the rolling process accuracy requirements (calibrated to 50 microseconds). Exceeding this tolerance will affect the accuracy of thickness control, requiring adjustment. Adaptive adjustment means that the adjustment range and direction of the incremental compensation coefficient value are automatically determined according to the magnitude and direction of the time deviation, without manual intervention. The incremental compensation coefficient value in the composite offset is a dynamic adjustment component superimposed when generating the composite offset. Its value directly affects the data stream delay. Therefore, by adjusting this incremental value, the timing deviation can be accurately corrected. The specific adjustment process is as follows: First, calculate the difference between the current time deviation and the preset tolerance. If the time deviation is positive (the peak rolling force lags behind the valley thickness), increase the compensation coefficient increment by the proportion of the deviation exceeding the tolerance to further extend the delay time of the rolling force data stream. If the time deviation is negative (the peak rolling force leads the valley thickness), decrease the compensation coefficient increment by the corresponding proportion to shorten the delay time of the rolling force data stream. The adjustment step size is dynamically set according to the rule that the larger the deviation, the larger the step size (e.g., the step size is 2 microseconds when the deviation exceeds the tolerance by 10 microseconds, and 4 microseconds when it exceeds 20 microseconds). At the same time, set the upper limit of the maximum adjustment range (not exceeding 50% of the initial compensation coefficient increment value) to avoid over-adjustment that could cause timing oscillations. The above monitoring and adjustment process is repeated until the difference between the peak rolling force timestamp and the valley thickness timestamp stabilizes within the tolerance range. The peak rolling force timestamp is the precise time record corresponding to the local maximum point, and the valley thickness timestamp is the precise time record corresponding to the local minimum point. The difference between the two is the final time alignment accuracy indicator. "Stabilizing within the tolerance range" means that the difference between the two timestamps is within the preset tolerance for 10 consecutive sampling periods, rather than meeting the standard only once. This ensures the stability of time alignment and avoids misjudgments caused by accidental factors. The specific verification process is as follows: After each adjustment of the compensation coefficient increment, the dynamic compensation engine regenerates the composite offset, updates the circular buffer index value, and then adjusts the delay duration of the rolling force data stream. Subsequently, it re-monitors the key timing points of the two differential curves and calculates the time deviation. If the time deviation for 10 consecutive sampling cycles does not exceed the preset tolerance, the timing alignment is considered satisfactory, and adjustment stops. If it still exceeds the tolerance, the compensation coefficient increment continues to be fine-tuned according to adaptive rules until the stability condition is met. Ultimately, through this closed-loop adjustment mechanism, the physical response timing of the peak rolling force data and the valley material thickness data is forcibly aligned, ensuring precise matching of the two types of data in the time dimension. This provides high-quality input data with consistent timing and corresponding physical meaning for the thickness control algorithm of the subsequent integrated control output unit.

[0028] After the dynamic phase compensation unit completes the physical response timing alignment of the rolling force data and the material thickness data, the integrated control output unit needs to generate precise roll gap adjustment commands based on these two sets of time-consistent data. The core is to achieve closed-loop control of the material thickness through the feedforward prediction and feedback correction mechanism of the thickness control algorithm. The specific implementation process is as follows: First, the time-aligned rolling force data and material thickness data are input into the data fusion layer of the thickness control algorithm. The thickness control algorithm is the core computational logic of the integrated control output unit, specifically used to calculate the roll gap adjustment based on real-time data during the rolling process, ensuring that the final material thickness meets the process requirements. The data fusion layer is the front-end processing module of the thickness control algorithm. Its core function is to standardize the format of multi-source input data, filter outliers, and extract features, eliminating data redundancy and interference, and providing a unified, high-quality data source for subsequent calculations. The specific process is as follows: After time alignment, the rolling force data and material thickness data are transmitted to the integrated control output unit via an internal high-speed data bus. The data fusion layer first adapts the two types of data to a unified floating-point data format that the algorithm can recognize. Then, it removes instantaneous noise from the data through sliding window filtering (window size is 3 sampling points) while retaining the core trend of change. Next, it extracts key feature parameters, such as the real-time value and rate of change of the rolling force, the current value of the material thickness, and the deviation from the target thickness. These processed feature data are then integrated into an input vector that the algorithm can directly call, completing the data preprocessing and fusion. Based on the fused feature data, the thickness change trend is predicted using the rolling force data as a feedforward term. The feedforward term refers to the control term that predicts the change trend based on the input signal and intervenes before the system output changes. Its core advantage is that it can avoid lag and respond to impending thickness changes in advance. The thickness change trend refers to the prediction, based on the current rolling force state, of whether the subsequent material thickness will increase or decrease, as well as the rate and magnitude of change, providing a basis for adjusting the roll gap in advance. The specific process is as follows: The thickness control algorithm incorporates a feedforward prediction model trained on extensive rolling experimental data. This model has established a correlation between rolling force and thickness changes (e.g., as rolling force increases, the material is subjected to stronger compression, leading to a subsequent decrease in thickness). The model inputs the fused real-time rolling force value and its rate of change. By querying a pre-defined force-thickness correlation mapping table and combining real-time calculations, the model outputs the predicted material thickness and its trend for the next 1-2 sampling periods. For example, if the current rolling force suddenly increases by 10%, the model predicts that the subsequent material thickness will decrease by 0.05 mm, providing a forward-looking basis for generating subsequent adjustment commands. Simultaneously, thickness data is used as feedback to correct prediction deviations. Feedback refers to the control term that adjusts the control strategy in reverse based on the difference between the actual system output (here, the measured material thickness) and the predicted value. Its core function is to correct the error in the feedforward prediction and ensure control accuracy. Prediction deviation refers to the difference between the thickness prediction value output by the feedforward model and the time-aligned measured material thickness value. This difference directly reflects the accuracy of the feedforward prediction and is the core basis for correction. The specific process is as follows: The algorithm compares the measured material thickness with the feedforward predicted value in real time and calculates the absolute and relative deviations between the two. If the deviation is within the preset allowable range (calibrated to ±0.01mm), the current prediction model parameters are maintained. If the deviation exceeds the allowable range, an adaptive correction mechanism is activated, adjusting the correlation coefficient of the feedforward prediction model according to the direction and magnitude of the deviation (e.g., if the predicted value is 0.02mm smaller than the measured value, the weight of the rolling force on the thickness change is increased). At the same time, a proportional-integral-derivative (PID) control logic is introduced, with the integral term accumulating historical deviations and the derivative term predicting the trend of deviation changes. The prediction results are corrected through the collaborative operation of the three to ensure that the subsequent thickness change prediction can accurately match the actual working conditions. After collaborative computation involving feedforward prediction and feedback correction, the algorithm outputs the roll gap adjustment amount to the servo motor drive module. The roll gap adjustment amount refers to the specific distance that needs to be increased or decreased between the rolls to achieve the target material thickness, measured in millimeters. A positive value indicates an increased gap, while a negative value indicates a decreased gap. The servo motor drive module is the intermediate unit connecting the control algorithm and the actuator. It is responsible for converting the digital adjustment amount output by the algorithm into analog control signals (such as voltage or current signals) that the servo motor can recognize, driving the servo motor to perform precise mechanical actions. The specific process is as follows: The thickness control algorithm calculates the required roll gap adjustment based on the difference between the corrected thickness prediction and the target thickness (preset in the algorithm parameters). For example, if the target thickness is 5mm and the corrected predicted thickness is 4.95mm, then an adjustment of +0.05mm is output (to increase the gap and reduce the rolling extrusion intensity). The adjustment is converted into an analog control signal by the digital-to-analog converter module and then transmitted to the servo motor drive module through a dedicated control cable. The drive module amplifies and calibrates the phase of the control signal to ensure that the signal can drive the servo motor to rotate with preset precision, thereby driving the rolls to perform gap adjustment actions, realizing real-time closed-loop control of the material thickness, and ensuring that the rolled material thickness always meets the process requirements.

[0029] The present invention also provides an integrated control system for implementing a PLC-free rolling mill dedicated controller, characterized in that: the PLC-free rolling mill dedicated controller is mounted on the control board of the integrated control system, wherein the control board is provided with multiple connection interfaces corresponding to the PLC-free rolling mill dedicated controller, and a name identifier corresponding to each connection interface.

[0030] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A rolling mill-specific controller without a PLC architecture, characterized in that, include: The mill-specific controller unit (1) collects rolling force data and material thickness data through pressure sensors and thickness sensors, and receives orthogonal pulse signals from the encoder at the end of the roll shaft through the roll rotary encoder. The pulse signals are converted into rolling length in real time by measuring the number of roll rotations, providing a rolling length observation benchmark for dynamic compensation. The dynamic phase compensation unit (2) uses the rolling length observation benchmark to calculate the cumulative amount of plastic deformation in real time. The increase in rolling length is positively correlated with the cumulative amount of plastic deformation. When the cumulative amount of plastic deformation reaches a preset threshold, a phase delay trigger signal is generated. When the phase delay trigger signal is received, the dynamic compensation engine is started. The dynamic compensation engine establishes a basic phase difference compensation amount based on the fixed sampling rate difference between the pressure sensor and the thickness sensor. An initial timestamp offset is applied to the rolling force data. The dynamic compensation coefficient is increased according to the increase in the cumulative amount of plastic deformation by a preset step function. An incremental time delay is injected into the material thickness data. Finally, a linear interpolation algorithm is used to dynamically generate compensation data segments between adjacent sampling points of the thickness sensor. The physical response timing of the peak moment of the rolling force data and the valley moment of the material thickness data are forcibly aligned by real-time reconstruction of the data stream. The integrated control output unit (3) inputs the rolling force data and material thickness data after the physical response timing is forcibly aligned into the thickness control algorithm, and generates the roll gap adjustment command to the actuator.

2. The rolling mill controller without PLC architecture according to claim 1, characterized in that: The generation of the rolling length observation benchmark includes: The orthogonal pulse signal from the encoder at the end of the roll shaft is received by the roll rotary encoder. The rising and falling edges of the orthogonal pulse signal are counted by a fourfold frequency using a dual-channel counter to obtain the cumulative number of pulses. The cumulative number of pulses is multiplied by the preset material travel length corresponding to each rotation of the roll, and the rolling length is output to the dynamic phase compensation unit (2).

3. The rolling mill controller without PLC architecture according to claim 1, characterized in that: The calculation of the cumulative amount of plastic deformation includes: In the dynamic phase compensation unit (2), a mapping relationship between rolling length and cumulative plastic deformation is established. The mapping relationship is realized by calling the material deformation coefficient lookup table through the embedded processor. The material deformation coefficient is dynamically adjusted according to the instantaneous value of rolling force, so that the increase in cumulative plastic deformation corresponding to each unit increase in rolling length increases stepwise with the increase in rolling force.

4. The rolling mill controller without PLC architecture according to claim 3, characterized in that: The generation and response of the phase delay trigger signal includes: When the cumulative amount of plastic deformation exceeds a preset threshold, the processor comparator module outputs a high-level trigger signal to the enable port of the dynamic compensation engine. After receiving the high-level trigger signal, the dynamic compensation engine activates a three-stage pipeline processing architecture. The first-stage pipeline reads the basic phase difference compensation amount, the second-stage pipeline calculates the step function increment, and the third-stage pipeline initializes the interpolation buffer.

5. The rolling mill controller without PLC architecture according to claim 4, characterized in that: The establishment of the basic phase difference compensation amount includes: The dynamic compensation engine pre-stores the integer multiple relationship between the sampling period of the pressure sensor and the sampling period of the thickness sensor. After dividing the sampling clock signal of the thickness sensor, it compares the phase of the sampling clock signal with that of the sampling clock of the pressure sensor to generate an initial timestamp offset and writes it into the index value of the circular buffer.

6. The rolling mill controller without PLC architecture according to claim 5, characterized in that: The improvement of the dynamic compensation coefficient includes: The incremental value of the compensation coefficient is output by querying the step function mapping table based on the cumulative increase of plastic deformation. The initial timestamp offset is then superimposed with the incremental value of the compensation coefficient to generate a composite offset, which is then reloaded into the ring buffer index register so that the overall delay of the rolling force data stream on the time axis is equal to the duration of the composite offset.

7. The rolling mill controller without PLC architecture according to claim 6, characterized in that: The injection of incremental delay includes: Material thickness data is written to the front end of the circular buffer according to the sampling timestamp, and data is read from the back end of the circular buffer according to the composite offset index value. When the composite offset index value crosses the adjacent sampling point, a transition data segment generated by the interpolation algorithm is inserted into the read data stream to realize the continuous translation of the material thickness data time axis.

8. The rolling mill controller without PLC architecture according to claim 7, characterized in that: The generation of the compensation data segment includes: A linear interpolation algorithm is used to process adjacent sampling point data of the thickness sensor. The value of the previous sampling point is used as a reference, and the value is linearly and gradually changed to the value of the next sampling point according to the time ratio. Interpolation data points with equal time intervals are generated within the original sampling interval to fill the data gaps caused by the index value crossing the sampling point.

9. The rolling mill controller without PLC architecture according to claim 8, characterized in that: The physical response timing forced alignment includes: The system monitors the local maxima and minima of the differential curve of rolling force data and the differential curve of material thickness data in real time. When the time deviation between the local maxima and minima exceeds a preset tolerance, the system adaptively adjusts the incremental value of the compensation coefficient in the composite offset until the difference between the peak rolling force timestamp and the valley thickness timestamp stabilizes within the tolerance range. The generation of the roll gap adjustment command includes: The timing-aligned rolling force data and material thickness data are input into the data fusion layer of the thickness control algorithm. The rolling force data is used as a feedforward term to predict the thickness change trend, and the thickness data is used as a feedback term to correct the prediction deviation. The roll gap adjustment amount is then output to the servo motor drive module.

10. An integrated control system for implementing a rolling mill-specific controller with a PLC-free architecture as described in any one of claims 1-9, characterized in that: The PLC-free rolling mill dedicated controller is installed on the control board of the integrated control system. The control board has multiple connection interfaces corresponding to the PLC-free rolling mill dedicated controller, and a name identifier for each connection interface.

Images