A speed feed perfume dosing method and system

Abstract

The present application relates to the technical field of automatic control, and especially relates to a speed feed-forward flavor metering and adding method and system. The method comprises the following steps: constructing a temperature-feed-forward mapping discretization memory grid; when the host speed suddenly changes, a feed-forward compensation coefficient is extracted based on transient absolute temperature addressing, a target feed-forward pump speed is deduced, and a command is issued to jump to the next stage, and a PID integral accumulation anti-saturation is synchronously forced to freeze; after returning to a steady state, a transition period PID closed-loop residual error is extracted and quantified as a correction weight, and a targeted physical overwrite is performed on the coefficients in a specific grid. The present application can solve or at least alleviate the problems of transient response lag, integral saturation divergence, and metering mismatch caused by long-period temperature drift and mechanical wear in flavor adding, and provides a speed feed-forward flavor metering and adding method and system.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology, and in particular to a method and system for metering and adding spices with a speed feedforward. Background Technology

[0002] In the modern production system of the tobacco industry, the filter rod forming process, as a core link to ensure the intrinsic quality and sensory comfort of cigarettes, places extremely high demands on the precise metering of additives, especially flavorings. With the continued deepening of the intelligent manufacturing trend, production lines are evolving towards ultra-high speed, flexibility, and high automation. This not only requires the flavoring addition system to maintain extremely high repeatability under steady-state conditions, but also demands excellent dynamic response and anti-disturbance capabilities during transient abrupt changes such as equipment start-up and process speed adjustment. The uniformity of flavoring addition directly affects the consistency between finished cigarettes. In high-speed continuous operation, any minute fluctuation in physical flow or response lag will induce significant aroma concentration deviations in the filter rod, leading to the generation of a large amount of substandard waste.

[0003] To achieve precise addition of flavorings, various metering and control schemes have been developed in the existing technology field. One typical technical approach is online monitoring and closed-loop zeroing, such as the patent technology with publication number CN103278222B. This technology establishes a correlation model between the real-time change in the weight of the flavoring container and the cumulative value of the flow meter, and then uses a statistical process control (SPC) algorithm to evaluate the operating accuracy of the flow meter online, and performs zeroing calibration after the prediction error exceeds the limit. This type of scheme can detect measurement drift under steady-state operating conditions, and improves long-term reliability through back-end mathematical compensation. Another technical approach starts from the physical delivery structure of flavoring addition and optimizes its structural parameters. For example, the scheme with publication number CN121312875A uses a combination of a heated and insulated box and a precision metering pump. This design idea attempts to achieve the addition of high-viscosity flavorings by controlling the stability of the fluid physical environment and utilizing the forced delivery characteristics of the metering pump.

[0004] Analyzing the working principle of the aforementioned existing technologies and their coupling relationship with complex industrial conditions, it is not difficult to find that these control logics exhibit inherent defects in filter rod machines with large-span dynamic production cycles of thousands of rods per minute: 1. Transient Feedback Lag and Integral Saturation Divergence: Traditional feedback-based monitoring and calibration are essentially compensation methods, with the adjustment action inherently lagging behind the occurrence of physical errors. When the filter rod mill speed undergoes a sudden, step-like change, the pipeline fluid exhibits physical static friction and inertia. Simply relying on PID feedback not only leads to a severe lag in the initial response but also causes a destructive geometrical surge in integral energy during the transition period. Once the target baseline is exceeded, it can easily trigger a vicious integral saturation backlash, resulting in severe overshoot and divergence in the pipeline flow rate.

[0005] 2. Mismatch of Static Feedforward Model to Nonlinear Temperature Drift: While physical heating can improve fluid flow to some extent, the dynamic viscosity of spice fluids is highly sensitive to minute thermodynamic absolute temperature drifts. Current feedforward control often uses a static mathematical straight line slope uniformly allocated across all operating conditions, which cannot accurately offset the drastic changes in fluid dynamic resistance caused by wide-range environmental temperature drift, easily leading to excessive physical injection or insufficient compensation.

[0006] 3. Lack of dynamic adaptive repair for mechanical wear: With long-term, high-load service, the mechanical flow components of precision gear pumps will experience irreversible physical wear, leading to reduced volumetric efficiency and increased internal leakage. A rigid control model cannot detect this physical degradation and make underlying corrections, inevitably resulting in a sharp deterioration in long-term steady-state accuracy over time.

[0007] Therefore, an intelligent adaptive composite control system that integrates multi-source high-frequency physical sensing, eliminates the persistent problem of transient integral saturation, and has the ability to self-repair and evolve mechanical wear across the entire life cycle is a key technical challenge that urgently needs to be solved in the current automation of tobacco machinery. Summary of the Invention

[0008] To achieve the above-mentioned objectives, this invention provides a speed-feedback flavor metering and addition method and system, which aims to solve or at least mitigate the metering mismatch problems caused by transient response lag, integral saturation divergence, long-cycle temperature drift, and mechanical wear during flavor addition.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for metering and adding flavorings with a rate-feedback mechanism, comprising: The operating speed signal of the filter rod machine host and the real-time temperature signal of the fragrance fluid in the fragrance delivery pipeline are acquired in real time. When a sudden change is detected in the running speed signal, the feedforward control process is triggered; The transient temperature signal at the moment the feedforward control process is triggered is obtained, and the transient temperature signal is used as the addressing coordinate to locate the matching specific small temperature range grid in the pre-constructed temperature-feedforward mapping memory grid. Extract the basic feedforward compensation coefficients independently stored within the specific micro-temperature range grid, and combine them with the jump target value of the running speed signal to calculate the target feedforward pump speed setpoint. By blocking conventional control, the target feedforward pump speed setpoint is preferentially output to the spice adding pump to perform transient step-over action; after the action is completed, the control authority is smoothly transferred to the PID feedback control loop for steady-state correction.

[0010] To further realize the present invention, the following technical solutions may be preferred: Preferably, before acquiring the transient temperature signal that triggers the feedforward control process, the method further includes the step of constructing the temperature-feedforward mapping memory grid: The global ambient temperature domain in which the system operates is divided into multiple continuous and non-overlapping micro-temperature interval grids. Based on the fluid viscous resistance characteristics, a corresponding basic feedforward compensation coefficient is pre-set for each of the micro temperature range grids to form a discretized feedforward memory array.

[0011] Preferably, when a sudden change is detected in the running speed signal, the feedforward control process is triggered, including a dynamic dead-zone anti-jitter step: The instantaneous rate of change of the running speed signal is calculated at high frequency; Determine whether the instantaneous rate of change exceeds a preset steady-state dead zone threshold; If no breakthrough is achieved, it is determined to be mechanical gear resonance or electrical high-frequency noise, and the PID feedback control loop is maintained and the feedforward is shielded; if a breakthrough is achieved, it is confirmed that a sudden change in the actual production cycle has occurred, and the feedforward control process is triggered.

[0012] Preferably, the method of blocking conventional control and prioritizing the output of the target feedforward pump speed setpoint to the spice addition pump to perform a transient step-over action includes: Within the microsecond-level period of triggering the feedforward control process, the integral accumulation operation of the PID feedback control loop is forcibly frozen to prevent integral saturation. The target feedforward pump speed setpoint is directly issued to drive the servo pump to overcome physical static inertia and achieve strong fluid injection.

[0013] Preferably, the method further includes a residual extraction step based on historical closed-loop control data: Continuously monitor the actual fragrance flow rate at the end of the pipeline. When it is determined that the fluctuation of the actual fragrance flow rate has subsided and converged to the target demand, the system is confirmed to have returned to steady state. By retrospectively tracing back from the feedforward trigger to the system returning to steady state, the integral regulation load of the PID feedback control loop accumulated during this dynamic transition period to correct the residual error of the feedforward action is extracted.

[0014] Preferably, the method further includes an empirical conversion step based on the integral adjustment load: The extracted integral adjustment load is converted into a corresponding empirical correction weight according to a preset mapping relationship; If the integral regulation load indicates that the PID feedback control loop has performed a large amount of positive additional compensation, then a positive empirical correction weight is generated; if it indicates negative suppression regulation, then a negative empirical correction weight is generated.

[0015] Preferably, the method further includes a closed-loop self-learning targeted overwriting evolution step: Carrying the generated empirical correction weights, reverse tracing is performed to re-lock the specific micro-temperature range grid that was located when the feedforward control process was triggered; The empirically corrected weights are nested and integrated into the specific micro-temperature range grid, and the values ​​of the basic feedforward compensation coefficients stored inside are permanently overwritten to update the system's feedforward performance in the corresponding temperature range.

[0016] Preferably, the step of nesting and fusing the empirically corrected weights into the specific micro-temperature range grid includes: When the empirical correction weight is positive, it is determined that the pump body of the current system has mechanical wear and aging, resulting in insufficient original feedforward delivery capacity. The basic feedforward compensation coefficient in the specific micro temperature range grid is increased proportionally; otherwise, it is decreased proportionally, so as to realize the full life cycle adaptive optimization of the feedforward parameters.

[0017] A system for implementing the above method includes: The multi-source state sensing module includes a speed encoder, a digital temperature sensor attached to the wall of the spice main pipe, and a mass flow meter at the end of the pipe. The physical execution module includes a servo driver with a communication connection and a spice addition pump controlled by it; An adaptive composite controller is communicatively connected to the multi-source state perception module and the physical execution module, respectively, and is used to run a nested temperature-sensing addressing algorithm and a closed-loop self-learning algorithm, and to issue action commands.

[0018] Preferably, the adaptive composite controller is nested in the logical architecture as follows: Discretized memory grid cells are used to store array nodes divided according to temperature gradients and their corresponding feedforward compensation coefficients; The temperature-sensing addressing feedforward unit is used to accurately locate and force the extraction of coefficients in the discretized memory grid unit when a sudden velocity change signal is extracted. The residual evaluation and self-evolution unit is used to take over the historical integral data of the underlying PID after the system returns to steady state, convert it into the empirical correction weights, and feed them back to the discretized memory grid unit for coefficient overwriting.

[0019] The beneficial effects of this invention are: First, this invention divides the global ambient temperature physical domain into a nonlinear topological discretized microgrid and uses ferroelectric materials for persistent and independent storage of the basic feedforward coefficients. Under severe disturbances caused by wide-range ambient temperature drift (e.g., a natural rise from 20°C to 35°C) leading to a sudden change in fluid dynamic viscosity, the microprocessor can achieve seamless switching between coordinate addressing and parameter thermal states. Compared to the high rate of severe temperature drift in traditional control systems, this invention tightly constrains the steady-state absolute deviation across the entire physical temperature range within a very small tolerance band.

[0020] Faced with extreme physical conditions of rapid equipment speed-up (e.g., jumping from 3000 pieces / minute to 8000 pieces / minute), this system forcibly issues a peak torque command at the moment the feedforward is triggered to break down the static friction of the pipeline. Simultaneously, the underlying algorithm cuts off the PID integral accumulation path, fundamentally eliminating the destructive energy accumulation caused by integral saturation. This invention takes over the historical integral load data of the PID during the steady-state transition period, quantifies it into directional empirical correction weights, and nests a Markov chain forgetting factor algorithm to perform permanent targeted overwrite updates on the feedforward coefficients within a specific temperature grid. Attached Figure Description

[0021] Figure 1 This is the overall architecture and electrical topology interconnection logic diagram of the present invention.

[0022] Figure 2 This is a schematic diagram of the nonlinear spatial topology partitioning and discretization addressing principle of the temperature-feedforward mapping memory grid of the present invention.

[0023] Figure 3 This is a schematic diagram of the timing control principle of the feedforward command priority skipping and PID integral freeze anti-saturation timing control under transient operating conditions of the present invention.

[0024] Figure 4 This is a comparison chart of the fluid dynamic response and overshoot suppression effect between the implementation group and the control group under a large step acceleration condition.

[0025] Figure 5 This is a comparison chart of the steady-state measurement accuracy evolution trajectories of the implementation group and the control group under wide-range environmental temperature drift disturbance.

[0026] Figure 6 This is a comparison chart of the added precision and self-repair effect between the implementation group and the control group after mechanical wear caused by long-term high-load operation. Detailed Implementation

[0027] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0028] 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. Example 1

[0029] Overall architecture and electrical topology interconnection logic diagram ( Figure 1 The three-level hardware-software coupling physical boundary of the multi-source state sensing module, the adaptive composite controller, and the physical execution module is defined. The multi-source state sensing module is distributed at the characteristic nodes of the main power transmission chain of the filter rod forming machine and the fragrance conveying fluid pipeline network, and captures the changes of environmental thermodynamic parameters and the motion vector characteristics of the mechanical actuators at a high frequency on a microsecond time scale.

[0030] A high-resolution incremental photoelectric encoder with a sensing network configuration is rigidly mounted on the flange end face of the filter rod host's main drive output shaft. The encoder's internal high-frequency differential signal generation circuit continuously outputs a quadrature pulse sequence, which is fed into the controller's high-speed counting and acquisition terminal via an industrial twisted-pair cable with a double-layer tinned copper wire braided shield. A hardware-level Schmitt trigger performs waveform reshaping and precise voltage amplitude filtering on the pulse sequence before it enters the logic unit, eliminating distortion glitches caused by long-distance line attenuation or common-mode electromagnetic coupling from the high-frequency inverter. This establishes a high signal-to-noise ratio physical benchmark for capturing the host's production cycle time and linear velocity jump signals at the electrical hardware level.

[0031] The fluid sensing network deploys a multi-point distributed high-sensitivity digital temperature sensor physical array at the pipeline end. The high thermal conductivity probes of the sensors penetrate the physical structure through deep holes, reaching the geometric center of the flow field inside the pipeline and penetrating the fluid thermal boundary layer on the inner side of the pipe wall to eliminate temperature gradient hysteresis. The sensor array physically spans the outlet of the spice storage tank, the inlet flow channel of the precision gear pump for fragrance addition, and the pipeline end extremely close to the terminal atomizing nozzle. The probe exterior and the stainless steel pipeline surface are tightly covered with multiple layers of closed-cell polyurethane polymer thermal insulation material and an outer metal protective sleeve to isolate the space convection and thermal radiation disturbances caused by the circulating airflow of the workshop's air conditioning system and the alternating day and night temperature differences. High-precision mass flow meters based on the Coriolis force physical principle are installed in series on the upstream straight pipe section near the nozzle array. The flowmeter measuring tube maintains high-frequency harmonic mechanical vibration under the drive of an alternating electromagnetic field. It captures the minute changes in the phase difference of the Coriolis force caused by physical inertia when the fluid flows through the vibrating tube section. It outputs real instantaneous mass flow rate data that is not limited by random fluctuations in fluid density, tiny bubble inclusions, and sudden changes in back pressure during pipeline operation, providing a high-fidelity physical reference for residual evaluation of the entire control closed loop.

[0032] The adaptive composite controller is built inside the motherboard of an industrial-grade microprocessor system with an integrated hard real-time task scheduling kernel. The software architecture incorporates containerized memory management and independent physical process deployment mechanisms, encapsulating digital filtering and denoising, multi-dimensional feature addressing, and evolutionary mathematical calculations into mutually isolated, standardized, and independent control components. The microprocessor is deeply deconstructed on the logical memory address plane into discretized memory grid units, temperature-sensing addressing feedforward units, and residual evaluation and self-evolution units. Each functional unit occupies a strictly isolated independent preemption priority in the operating system's hardware interrupt vector table.

[0033] Schematic diagram of nonlinear spatial topology partitioning and discretization addressing principle of temperature-feedforward mapping memory grid ( Figure 2 This demonstrates a multidimensional structured data matrix statically allocated within a non-volatile random access memory using discretized memory grid cells. The inherent power-loss-free physical latching characteristics of ferroelectric materials and their near-infinite physical erase / write cycles on the nanosecond scale ensure the persistent security of high-frequency self-learning evolution data. Within each cell, the physical domain of the global ambient temperature encountered by the system during actual operation is divided into multiple continuous, non-overlapping micro-temperature range grids. Figure 2The distribution characteristics of the horizontal and vertical axis curves map the nonlinear topology partitioning strategy of the mesh space. In low-temperature physical states, the dynamic viscosity of the spice fluid exhibits a significant exponential increase with decreasing temperature, and the interlayer shear resistance of the fluid shows extreme sensitivity to minute temperature drifts. The system increases the mesh space distribution density in the low-temperature sensitive physical region, defining small, dense mesh regions; in the high-temperature region where fluid viscosity changes gradually and rheological properties approximate Newtonian fluids, a widened, independent mesh is defined. Nonlinear mesh partitioning uses limited hardware addressing space to perform high-resolution discretization coverage of important fluid temperature drift segments. Each mesh physical node is allocated an independent memory addressing pointer, independently storing the unique basic feedforward compensation coefficients for that temperature region, supporting parallel storage of template parameters for various viscosity-temperature nonlinear spice formulations and seamless switching between thermal states.

[0034] The temperature-sensing addressing feedforward unit is mounted on the highest-priority external interrupt vector table of the hardware processor. Within the nanosecond-level physical cycle of the host speed change pulse signal extraction, this unit forcibly interrupts regular data display and background communication tasks. It uses the transient absolute physical temperature data captured by the temperature sensor array to find the corresponding physical node address in the discretized memory grid storage space, forcibly preempts and binds the basic feedforward compensation coefficient, derives dynamic mathematical actuation instructions, and bypasses the conventional underlying network communication stack to directly send them to the registers.

[0035] The residual evaluation and self-evolution unit runs in a low-priority background system shadow process, taking over the underlying PID feedback loop to generate historical integral physical data for correction during the fluid dynamic transition period. The residual evaluation and self-evolution unit quantifies the physical regulation load depth into mathematically directional empirical correction weights, which are then targeted and fed back to specific addresses in the discretized memory grid to perform permanent physical overwrite updates of the coefficient values.

[0036] The precision physical execution layer consists of a high-performance servo drive supporting industrial real-time fieldbus communication protocols and a precision external meshing fragrance dispensing pump. The dispensing pump features minimal backlash and end-face clearance in its physical mating gears. A deep tungsten carbide infiltration layer on the surface of the stainless steel flow components within the pump body reduces the resistance to shear friction of the viscous fluid. A linear dynamic mapping equation is established between the pump shaft's mechanical rotation angle and the fluid output volume. The servo drive's internal hardware circuitry integrates a three-loop decoupled logic—current, speed, and position—operating at a high-frequency carrier frequency. Upon receiving a transient feedforward strong injection order-breaking digital command, the servo drive allocates internal power devices to instantly generate peak electromagnetic torque, driving the motor rotor to forcefully overcome the static friction inertia of the fluid within the pipeline network. The delivery pipeline hardware system is equipped with an active pulse physical damper to smooth out fluid hydraulic pulsations caused by the high-frequency mechanical cycle of the gear pump. Example 2

[0037] The system rigorously integrates three underlying control processes: benchmark perception and database construction for the operating environment, composite nested control for transient operating conditions, and reverse self-learning and overwriting of steady-state residual frequency domain extraction.

[0038] After the equipment hardware powers on and establishes a handshake using the underlying electrical communication protocol, the control flow enters the stage of building a physical perception database based on the operating environment benchmark. The adaptive composite controller hardware interface captures the physical pulse signals of the filter rod machine's operating speed at high frequency. Addressing the inherent physical backlash and high-frequency vibration characteristics of the mechanical main drive chain and the flexible belt, the microprocessor logic layer invokes a time-window moving average filtering algorithm and a first-order low-pass digital filtering algorithm to perform digital signal purification and extraction. Within a continuous millisecond-level fixed time window, the controller mathematically accumulates the received effective orthogonal pulse counts and performs physical ratio conversions, thoroughly removing high-frequency non-stationary mechanical disturbance noise components from the complex spectrum and extracting smooth instantaneous linear velocity characteristic parameters that characterize the actual physical operating rhythm of the production line.

[0039] The microprocessor synchronously and at high frequency reads discrete ambient temperature data from a multi-point digital temperature sensor array. The system's computational logic loads a fluid dynamics friction resistance mathematical model, along with the equivalent pipe inner diameter physical parameters and absolute pipe wall roughness data from the actual field conditions. Figure 2 The absolute nominal dynamic viscosity parameter value of the spice fluid corresponding to the nominal center point of a specific temperature grid is solidified and written into the feedforward compensation mathematical coefficient of the initial basis after dividing and meshing each small nonlinear temperature range.

[0040] The control flow transitions to a high-density monitoring state for continuous production processes. The logic operation layer calculates the absolute value of the instantaneous rate of change signal over multiple cycles at high frequency, and performs anti-jitter judgment logic to determine if the absolute value of the instantaneous rate of change exceeds the dead zone. The logic mathematical operation layer compares the instantaneous absolute velocity parameter value within the current high-speed physical sampling period with the root mean square velocity parameter value within the historical period sequence using a multi-level time numerical difference method. Combining the physical acceleration amplitude parameter described by the first derivative and the physical acceleration rate of change described by the second derivative, a composite criterion is used to accurately determine whether the transmission system is currently in a uniform motion segment, a uniform acceleration segment, a rapid deceleration segment, or a complex micro-amplitude oscillation segment. If the absolute value of the instantaneous rate of change does not exceed the set steady-state dead zone boundary threshold, the algorithm determines that the current pulse frequency exhibits slight jitter, which may be due to natural inherent resonance of the mechanical transmission gears or high-frequency parasitic noise interference from the electrical system. The system maintains the basic steady-state closed-loop correction motion of the underlying PID feedback control loop, and forcibly shields the trigger-triggered feedforward interception action at the underlying algorithm level. When the microprocessor determines that the host is currently in a long-period micro-amplitude oscillation segment, the logic control layer dynamically increases the physical limit of the steady-state dead zone threshold judgment boundary, thereby expanding the anti-interference and anti-shake signal tolerance of the control system.

[0041] When the absolute value of the instantaneous rate of change exceeds the set steady-state dead zone safety threshold and exhibits continuous unidirectional step changes, the logic operation layer determines that a sudden change in the actual process physical cycle has occurred, and the hardware interrupt port triggers the feedforward control instruction flow. The system forcibly intercepts and latches the current transient fluid absolute physical temperature data fed back in real time by the digital temperature sensor array. The temperature-sensing addressing feedforward operation unit uses this frozen physical data as the memory address address coordinate and enters the temperature-feedforward mapping memory grid structured data array to perform coordinate physical boundary matching. The addressing comparison unit locks the corresponding specific small temperature range grid and retrieves the internally independently stored basic feedforward compensation coefficients. The basic feedforward compensation coefficients extracted by the operation are used as the physical weight of the product gain, and are subjected to composite algebraic cross-operation with the difference between the host running speed and the target instruction for the jump. The independent components of fluid momentum inertia compensation, which are dynamically adjusted according to the host's second-order physical acceleration, are simultaneously superimposed to deduce the target feedforward pump speed physical setpoint. It receives and blocks all data streams sent by conventional PID feedback control, and pushes the target feedforward pump speed signal value as the highest priority transient over-order digital signal of absolute scheduling directly into the underlying bus and sends it to the servo drive physical actuator.

[0042] Schematic diagram of feedforward command priority and step-by-step issuance and PID integral freeze anti-saturation timing control under transient conditions ( Figure 3 This demonstrates the underlying anti-divergence protection measures during the extreme instruction issuance time. Within the microsecond-level execution cycle after triggering feedforward control, the controller performs calculations on the PID control core. Figure 3 The gray shaded area indicates the protective action of forcibly freezing the integral accumulation calculation of the PID feedback control loop. During sudden changes in production speed, a drastically increased transient tracking absolute error occurs between the actual physical flow rate in the pipeline and the process-set target flow rate. Keeping the mathematical integrator on directly leads to a destructive and rapid accumulation of energy within the PID algorithm due to the huge deviation. Disabling the integral action completely blocks the energy accumulation path from the algorithm's core, preventing integral saturation. This completely eradicates the persistent problem of severe overshoot and divergence in pipeline flow caused by the integrator releasing accumulated energy during smooth handover of subsequent control authority. The target feedforward pump speed step-up injection command drives the servo motor rotor to unleash explosive acceleration electromagnetic peak torque, allowing the high-viscosity fragrance fluid to instantly overcome the static friction of the long pipeline physical area, achieving extremely rapid injection.

[0043] After the transient over-order strong injection physical lifecycle is completed, the system instruction releases the frozen state of the PID kernel integral accumulation calculation unit. Figure 3The lower half of the waveform demonstrates the smooth transfer of control authority using cross-fade-in / fade-out timing nested physical logic. The hardware weight coefficient of the feedforward control output decreases linearly with the system's internal master clock cycle, and the weight allocation ratio of the PID feedback control loop in the total output control command signal gradually increases. The PID loop, which fully takes over the physical correction task, is configured with adaptive nonlinear adjustment function for internal control parameters across large, medium, and small ranges. When the absolute value of the flow rate deviation is in the large deviation physical range, the system significantly increases the proportional gain parameter to seek the fastest deviation physical contraction convergence trajectory; after the flow rate enters the medium deviation physical range, the proportional gain is gradually reduced and the integral action accumulation principle is slowly released to eliminate long-tail physical static errors; when the flow rate approaches the small deviation limit process tolerance band, the sensitivity of the differential action operator to high-frequency small disturbances is enhanced to suppress small fluid residual dynamic physical fluctuations.

[0044] When the absolute value of the difference between the flow rate value of the Coriolis mass flow meter at the end of the pipeline and the target process requirement remains stably within a preset stringent tolerance bandwidth for dozens of consecutive sampling cycles, the system state physical machine confirms that it has returned to physical steady state. The residual evaluation and self-evolutionary operation unit ends its dormant polling mechanism and starts the residual physical extraction and reverse overwrite evolution program based on the historical data of steady-state closed-loop control. The reverse backtracking time window is precisely defined physically as the entire dynamic physical transition period from the moment the feedforward trigger interception hard command takes effect to the moment the system confirms that it has returned to steady state.

[0045] Steady-state closed-loop residual extraction, integral regulation load quantization, and the self-evolutionary principle of targeted overwriting of specific grids deconstruct the system's underlying self-healing learning path. The system extracts the accumulated integral regulation load energy from the PID feedback control loop to correct residual physical errors in feedforward actions. High-frequency intensive sampling is performed on the discrete absolute deviation between the physical value of the PID control output and the theoretically expected value of the steady-state reference throughout the entire transient period, and time-weighted integral mathematical operations on the absolute physical error are executed. The time-weighted integral operation model multiplies and integrates the magnitude of the physical deviation with the duration of the deviation on the physical time axis, quantifying the accumulated physical regulation load actually borne by the underlying closed-loop feedback loop from the energy transfer dimension of physical mechanical regulation work. The extraction module transforms the integral regulation load energy, stripped of high-frequency mechanical noise, into corresponding mathematical empirical correction weights according to a preset nonlinear sensitivity function mapping relationship. When the integral regulation load physical quantity is characterized by the PID feedback control loop performing a large number of positive additional compensation physical operations, it is determined that the original feedforward physical model's compensation output physical capability is insufficient, and the system generates positive empirical correction weight values. When the physical quantity of the integral-regulated load is characterized as a negative suppression of the physical action of regulation, the original feedforward model is judged to have overshoot, and a negative empirical correction weight value is generated.

[0046] The underlying memory addressing logic uses temporarily stored frozen physical temperature data to reverse spatial physical tracing, recalculating the memory physical starting address of the specific tiny temperature range grid initially locked when the feedforward was triggered. The overwrite mechanism introduces a forgetting factor iterative mathematical algorithm with Markov chain physical characteristics. It sets a threshold for the proportional allocation of newly generated empirical correction weight values ​​and historical solidified physical data in the update calculation formula. The empirical correction weights, combined with historical features, are nested and fused into the specific physical grid through the forgetting factor algorithm, permanently overwriting and non-volatilely persisting the original basic feedforward compensation coefficients in the internal physical storage.

[0047] When the generated empirical correction weights exhibit a long-term, continuous, positive cumulative physical trend, the microprocessor, combined with the servo driver load electromagnetic torque data, eliminates physical interference from pipeline crystallization and blockage, determining that the current system's pump body suffers from mechanical wear and aging, leading to insufficient original feedforward delivery capacity. The controller proportionally increases the basic feedforward compensation coefficient within a specific micro-temperature range grid to compensate for internal physical deficiencies in delivery capacity. When the empirical correction weights are negative, the basic feedforward compensation coefficient is proportionally decreased. A limited micro-penetration mathematical update step size ensures the smooth convergence of the control physical model's evolutionary trajectory.

[0048] The system synchronously executes fault-prevention logic for abnormal operating conditions in the background of physical coefficient overwriting. When the empirical correction weight value of a single calculation exceeds the bottom line of the model's safe evolution ratio, or when the feedback physical data shows a cliff-like nonlinear abrupt jump that does not conform to the fluid dynamics decay law, the system forcibly executes a physical rejection overwrite interception command, keeping the original matrix coefficients in a read-only frozen state, and preventing the erroneous operating condition data from physically contaminating the adaptive evolution direction. Example 3

[0049] A control system was deployed in a high-speed tobacco filter rod molding and processing digital benchmark workshop equipped with high-precision constant temperature and humidity centralized physical control facilities to conduct verification tests on engineering efficiency and resistance to physical interference. The test fluid medium was strictly selected as a peppermint essential oil-based fragrance emulsion mixture with an extremely steep rheological viscosity-temperature curve and extremely sensitive viscosity physical properties to thermodynamic absolute temperature.

[0050] The molding machine operates at a target physical speed of 3000 pieces / minute in a low-speed ramp-up state and 8000 pieces / minute in a high-speed cruise limit state, performing extremely high-frequency, large-span physical step-like hard switching actions. A control group machine with a traditional "main machine speed static linear proportional follower superimposed with terminal PID mass flow closed-loop feedback" control architecture is configured synchronously and in parallel on-site. The control group rigidly uses a uniform static feedforward mathematical straight line slope across the entire operating condition physical domain, lacking a temperature grid nonlinear discretization physical addressing algorithm, an integral freeze anti-saturation physical protection mechanism, and residual extraction targeted overwriting self-learning evolutionary capabilities.

[0051] The engineering verification plan is divided into two phases on the physical timeline: the initial operational dynamic extreme value physical response testing period and the long-term mechanical limit physical aging simulation period. The core process physical performance parameters and multi-dimensional indicators recorded by the high-frequency acquisition of the high-speed sensor network in the field are shown in the table below.

[0052] Test physical indicators Description of extreme physical conditions control group, traditional control Implementation Group Adaptive Feedforward Optimization range Dynamic transition physical time (s) From 3000 to 8000 in a rapid physical step climb. 4.2 0.8 80.95% Maximum overshoot peak (%) From 3000 to 8000 in a rapid physical step climb. 12.5 2.1 83.20% Steady-state addition of limit accuracy (%) 25℃ standard constant temperature physical boundary environment ±1.2 ±0.3 75.00% Peak value of physical deviation of flow lag (ml) Extreme physical abrupt change in velocity occurs in the initial physical instant. 15.8 1.2 92.41% Temperature drift affects absolute deviation (%) The absolute temperature of the fluid naturally rises from 20°C to 35°C. 8.5 0.6 92.94% Steady-state accuracy (%) after wear compensation After 500 hours of continuous high-load physical operation at full capacity ±2.8 ±0.4 85.71% Discard / defect rate (in ten thousand pieces) Statistics on elimination of speed regulation and hard switching of process physical state 45 3 93.33% The figure shows a comparison of the fluid dynamic response and overshoot suppression effect between the implementation group and the control group under the condition of large step acceleration. Figure 4 To observe the suppressive force of the underlying physical architecture on fluid dynamics nonlinearity, the graph shows the horizontal axis representing millisecond-level physical time and the vertical axis representing the physical flow rate of the instantaneous mass of the spice. Starting from the physical time of the step acceleration command, the physical curve (dashed line) of the control group exhibits severe physical response hysteresis. Due to the slope mismatch of the static feed, the static physical friction of the high-viscosity fluid in the long pipeline is extremely large, causing the dashed line to collapse deeply from the beginning, with a physical hysteresis error as high as 15.8 ml. The control architecture of the control group relies entirely on the feedback deviation of the flow meter to drive the PID calculation, continuously accumulating the energy of the physical error during the hysteresis period. When the liquid momentum crosses the target baseline, a vicious physical integral saturation backlash occurs, with the physical peak of vicious overshoot reaching an astonishing 12.5%. Only after 4.2 seconds of high-frequency damped oscillation does it enter the steady-state tolerance zone.

[0053] The physical curve (solid line trajectory) of the implementation group exhibits a steep straight line pattern indicating full-speed physical convergence. Utilizing high-speed addressing commands via a nonlinear discretized temperature feature grid and dynamic physical acceleration inertial compensation components, the system blocks conventional control, prioritizing the output of the target feed pump speed setpoint to the spice addition pump, executing transient step-up actions, and transmitting the strong injection physical command to the static friction of the fluid pipeline. When the solid line spikes, the system microprocessor executes a limit protection measure to forcibly freeze the spice signal source in the integral accumulation operation of the PID feedback control loop, cutting off the energy accumulation path of the large transient tracking physical deviation within the integrator from the algorithm's underlying layer. The anti-saturation mechanism takes effect, and in conjunction with the timing-based physical transition fade-in / fade-out logic, the maximum flow rate adjustment physical quantity of the implementation group curve is controlled within a small oscillation range of 2.1%. The physical transition time for the entire flow field's steady-state convergence is significantly shortened to 0.8 seconds, and the peak hysteresis deviation is also drastically reduced to 1.2 ml.

[0054] Under wide-range environmental temperature drift disturbances, the steady-state metrological accuracy evolution trajectories of the implementation group and the control group were compared (see...). Figure 5As can be seen from the figure, in the steady-state physical defense line diagram under the boundary of long-term environmental absolute temperature disturbance, the control group curve (dashed line) exhibits a strong unidirectional divergent upward trend as the thermodynamic temperature on the horizontal axis increases. When the fragrance solution is at 35℃, its physical dynamic viscosity decreases nonlinearly, and the physical pipe loss resistance along the pipeline also decreases significantly. However, the static feedforward model of the control group solidifies several physical parameters, and the drive motor is still driven according to the physical conditions of low temperature, high viscosity, and high friction, requiring a large torque output. As a result, the fluid is severely over-injected, causing an 8.5% malignant physical temperature drift deviation. The physical curve of the implementation group (solid line) is different. Throughout the entire physical temperature change range, it almost always remains close to the zero axis with minimal fluctuations. This is because the multi-point weighted digital temperature sensor can collect real-time absolute physical temperature at high frequency, and then enter the multi-layer, multi-dimensional, multi-point microprocessor for high-frequency coordinate addressing. The system can accurately extract the small basic feedforward physical compensation coefficient suitable for the high temperature, low viscosity and low friction range, and control the system-level flow physical oscillation amplitude caused by the wide temperature drift of the environment within the excellent process tolerance range of 0.6%, thus completely avoiding the problem of thermodynamic drift quality exceeding the limit.

[0055] The physical curve (solid line) of the implementation group maintains a near-zero, extremely low-amplitude physical fluctuation band across a wide range of physical temperature variations. Multi-point weighted digital temperature sensors acquire real-time absolute physical temperatures at high frequency, while a multi-core microprocessor performs coordinate addressing at high frequency. The system accurately acquires and matches the minute basic feedforward physical compensation coefficients for high-temperature, low-viscosity, and low-friction zones. This effectively prevents system-level flow oscillations caused by wide-range environmental temperature drift from exceeding the excellent 0.6% process tolerance, and thermodynamically eliminates all thermodynamic drift and quality exceedances.

[0056] Let's look at the comparison chart of the addition accuracy and self-repair effect between the implementation group and the control group after mechanical wear caused by long-term high-load operation. Figure 6 This diagram illustrates the self-healing evolution of the system's core during the long-term mechanical aging stage after 500 hours of full-load, high-pressure extreme physical operation. The horizontal axis represents cumulative physical operating hours, and the vertical axis represents the positive and negative tolerance boundaries of the steady-state accuracy enhancement domain. The physical feedback curve of the control group exhibits a distinctly divergent funnel-shaped geometry, gradually collapsing. After being scoured and rubbed by the high-pressure physical fluid, irreversible mechanical wear and aging occurred in the physical tooth clearance and end-face clearance of the pump body's stainless steel flow gear components. The clearance increased, and the internal leakage of fluid flowing back from the high-pressure physical zone to the low-pressure physical zone of the positive displacement pump intensified geometrically. Due to the lack of an endogenous sensing pathway for the physical state of the mechanical actuators, the steady-state accuracy enhancement of the control group deteriorated sharply from the initial ±1.2% at the factory, eventually reaching a completely uncontrolled ±2.8%.

[0057] The implementation team's physical convergence curve performed exceptionally well. During 500 hours of physical operation, it demonstrated robust adaptive repair capabilities. After each steady-state process speed adjustment, a significant positive integral adjustment load energy, representing the internal leakage loss of volumetric efficiency, was used. The system backend continuously generated positive empirical physical correction weights, employing a physical forgetting factor smoothing penetration mechanism. From Figure 6 The curve on the right shows multiple step-like physical targeted overwrite and lifting actions at a rate of a few per thousand. The system continuously performs permanent overwrites, updating the physical coefficients of the basic feedforward compensation stored within a specific small temperature range grid. Under extremely harsh conditions with no human intervention, the control model achieves closed-loop self-learning and autonomous iterative evolution based on real physical residuals, bridging the physical decay loophole of volumetric efficiency at the mechanical and physical level at the virtual level of the electrical algorithm. After 500 hours of extreme mechanical and physical wear, the physical accuracy of the fragrance fluid addition in the implementation group system remarkably remained at a high-precision control level of +0.4%. On the entire production lifecycle timeline, the physical residual loss of discarded filter rods caused by speed adjustment and process state switching was reduced from 450,000 in the control group to 30,000 in the implementation group. The vast physical system data and comparative graphs demonstrate the fundamental engineering value of the system in eliminating nonlinear response hysteresis, resisting physical drift of parameters, and providing long-term targeted repair of mechanical wear attenuation.

[0058] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for metering and adding flavorings with a rate-feedforward capability, characterized in that, include: The operating speed signal of the filter rod machine host and the real-time temperature signal of the fragrance fluid in the fragrance delivery pipeline are acquired in real time. When a sudden change is detected in the running speed signal, the feedforward control process is triggered; The transient temperature signal at the moment the feedforward control process is triggered is obtained, and the transient temperature signal is used as the addressing coordinate to locate the matching specific small temperature range grid in the pre-constructed temperature-feedforward mapping memory grid. Extract the basic feedforward compensation coefficients independently stored within the specific micro-temperature range grid, and combine them with the jump target value of the running speed signal to calculate the target feedforward pump speed setpoint. By blocking conventional control, the target feedforward pump speed setpoint is preferentially output to the spice adding pump to perform transient step-over action; after the action is completed, the control authority is smoothly transferred to the PID feedback control loop for steady-state correction.

2. The method according to claim 1, characterized in that, Before acquiring the transient temperature signal at the moment the feedforward control process is triggered, the method further includes the step of constructing the temperature-feedforward mapping memory grid: The global ambient temperature domain in which the system operates is divided into multiple continuous and non-overlapping micro-temperature interval grids. Based on the fluid viscous resistance characteristics, a corresponding basic feedforward compensation coefficient is pre-set for each of the micro temperature range grids to form a discretized feedforward memory array.

3. The method according to claim 1, characterized in that, When a sudden change is detected in the operating speed signal, a feedforward control process is triggered, including a dynamic dead-zone debouncing step: The instantaneous rate of change of the running speed signal is calculated at high frequency; Determine whether the instantaneous rate of change exceeds a preset steady-state dead zone threshold; If no breakthrough is achieved, it is determined to be mechanical gear resonance or electrical high-frequency noise, and the PID feedback control loop is maintained and the feedforward is shielded; if a breakthrough is achieved, it is confirmed that a sudden change in the actual production cycle has occurred, and the feedforward control process is triggered.

4. The method according to claim 1, characterized in that, The aforementioned blocking of conventional control prioritizes outputting the target feedforward pump speed setpoint to the spice addition pump to perform transient over-level actions, including: Within the microsecond-level period of triggering the feedforward control process, the integral accumulation operation of the PID feedback control loop is forcibly frozen to prevent integral saturation. The target feedforward pump speed setpoint is directly issued to drive the servo pump to overcome physical static inertia and achieve strong fluid injection.

5. The method according to claim 1, characterized in that, The method also includes a residual extraction step based on historical closed-loop control data: Continuously monitor the actual fragrance flow rate at the end of the pipeline. When it is determined that the fluctuation of the actual fragrance flow rate has subsided and converged to the target demand, the system is confirmed to have returned to steady state. By retrospectively tracing back from the feedforward trigger to the system returning to steady state, the integral regulation load of the PID feedback control loop accumulated during this dynamic transition period to correct the residual error of the feedforward action is extracted.

6. The method according to claim 5, characterized in that, The method further includes an empirical conversion step based on the integral adjustment load: The extracted integral adjustment load is converted into a corresponding empirical correction weight according to a preset mapping relationship; If the integral regulation load represents a large amount of positive additional compensation performed by the PID feedback control loop, then a positive empirical correction weight is generated. If the characterization is negative inhibitory regulation, then negative empirical correction weights are generated.

7. The method according to claim 6, characterized in that, The method also includes a closed-loop self-learning targeted overwrite evolution step: Carrying the generated empirical correction weights, reverse tracing is performed to re-lock the specific micro-temperature range grid that was located when the feedforward control process was triggered; The empirically corrected weights are nested and integrated into the specific micro-temperature range grid, and the values ​​of the basic feedforward compensation coefficients stored inside are permanently overwritten to update the system's feedforward performance in the corresponding temperature range.

8. The method according to claim 7, characterized in that, The step of nesting and fusing the empirically corrected weights into the grid of the specific small temperature range includes: When the empirical correction weight is positive, it is determined that the pump body of the current system has mechanical wear and aging, resulting in insufficient original feedforward delivery capacity. The basic feedforward compensation coefficient in the specific micro temperature range grid is increased proportionally; otherwise, it is decreased proportionally, so as to realize the full life cycle adaptive optimization of the feedforward parameters.

9. A system for implementing the method according to any one of claims 1 to 8, characterized in that, include: The multi-source state sensing module includes a speed encoder, a digital temperature sensor attached to the wall of the spice main pipe, and a mass flow meter at the end of the pipe. The physical execution module includes a servo driver with a communication connection and a spice addition pump controlled by it; An adaptive composite controller is communicatively connected to the multi-source state perception module and the physical execution module, respectively, and is used to run a nested temperature-sensing addressing algorithm and a closed-loop self-learning algorithm, and to issue action commands.

10. The system according to claim 9, characterized in that, The adaptive composite controller is nested in the following logical architecture: Discretized memory grid cells are used to store array nodes divided according to temperature gradients and their corresponding feedforward compensation coefficients; The temperature-sensing addressing feedforward unit is used to accurately locate and force the extraction of coefficients in the discretized memory grid unit when a sudden velocity change signal is extracted. The residual evaluation and self-evolution unit is used to take over the historical integral data of the underlying PID after the system returns to steady state, convert it into the empirical correction weights, and feed them back to the discretized memory grid unit for coefficient overwriting.

Images