An automatic carbon suction feedback control method based on negative pressure dynamic adjustment

Abstract

The present application relates to the technical field of automobile carbon tank production equipment control, in particular to an automatic carbon suction feedback control method based on negative pressure dynamic adjustment, which adds pressure transmitters at the inlet and outlet of the vacuum filter element on the basis of the original PID closed loop, identifies the linear seepage resistance coefficient and the inertial resistance coefficient of the filter element with the Darcy-Forchheimer seepage model online at each production beat, calculates the feedforward compensation amount according to the normalized drift of the linear seepage resistance coefficient, divides the filter element life cycle into the steady state section, the transition section and the steep increase section to schedule the PID parameters in sections, and completes the remaining life prediction and beat adaptive bottoming out by synchronously combining the resistance growth slope extrapolation. The method can significantly suppress the steady state deviation of the suction nozzle end negative pressure caused by dust deposition in the whole life cycle of the filter element, keep the activated carbon filling density of each carbon tank stable, and reduce the off-line weighing rework rate and the unqualified rate of vehicle evaporative emission detection.

Description

Technical Field

[0001] This invention relates to the field of control technology for automotive carbon canister production equipment, specifically to an automatic carbon absorption feedback control method based on negative pressure dynamic adjustment. Background Technology

[0002] The automated carbon filling workstation is a key, cycle-driven process piece of equipment on the automotive carbon canister production line. The carbon canister is a core component of the automotive evaporative emission control system (EVAP). The activated carbon particles inside adsorb fuel vapors evaporated from the fuel tank, preventing them from directly escaping into the atmosphere. The density and uniformity of the activated carbon packing directly affect the vehicle's evaporative emission performance and compliance with emission regulations. The core function of the automated carbon filling workstation is to precisely fill the inner cavity of the carbon canister shell with activated carbon particles at the specified density on a cycle-driven production line.

[0003] During the operation of this workstation, a vacuum pump is needed to establish a stable negative pressure at the suction nozzle end to draw activated carbon particles from the storage bin into the carbon canister shell and fill them into place. A vacuum filter element is connected in series in the particle flow path to prevent particulate dust from entering the pump body and protect the pump chamber. As the production cycle accumulates, activated carbon dust will continuously deposit on the surface of the filter element, causing the filter element's flow resistance to show a non-linear growth trend of being stable in the early stage and then rising sharply in the later stage.

[0004] Currently, most automated carbon filling workstations in the industry use a single, fixed-parameter PID closed-loop to regulate the vacuum pump output. This PID parameter is typically set once during production line commissioning based on the clean filter cartridge operating conditions and then permanently fixed until the entire filter cartridge is replaced. As the filter cartridge resistance gradually increases with dust accumulation, the pressure transmission relationship from the vacuum pump end to the nozzle end changes accordingly. The original PID loop can only rely on the negative pressure deviation at the nozzle end to slowly correct through the integral term, making it difficult to keep up with the non-linear changes in resistance. The actual suction negative pressure at the nozzle end will gradually deviate from the process setting value with each production batch. The actual consequences of this in the production line scenario are as follows: the cumulative suction volume obtained by each carbon canister during the suction time drifts, the weight and density of activated carbon filled into the carbon canister fluctuate considerably between batches, and some carbon canisters are judged to be out of tolerance at the weighing station after they come off the line and are rejected for rework; more importantly, some carbon canisters with a filling density on the edge of the tolerance flow into the vehicle assembly process and will show excessive hydrocarbon emissions during the evaporative emission testing process after the vehicle comes off the line, causing vehicle-level rework or scrapping, and in severe cases, even triggering emission regulation compliance risks, causing significant quality costs and regulatory compliance pressure on the OEM.

[0005] For example, Chinese invention patent CN115213185B discloses a negative pressure suction system, primarily used in oral medical droplet suction scenarios. This system incorporates humidity and / or concentration sensors within the suction pipe, and, through a feedback control circuit and a thyristor speed control circuit, automatically adjusts the suction motor speed based on changes in droplet humidity or concentration in the suction airflow, thereby reducing motor idling and energy consumption. Simultaneously, it incorporates an electrostatic filter, a photocatalyst processor, an activated carbon filter, an ultraviolet sterilizer, a hypochlorous acid injector, and an ozone generator within the airflow channel to purify and disinfect the suction gas. While this solution can adjust the suction intensity based on changes in medium concentration, its control is primarily based on airflow humidity or concentration, and it does not address the changes in negative pressure transmission characteristics caused by dust deposition on the vacuum filter element during industrial carbon filling. Therefore, given the continuous changes in filter element resistance with production cycle in automatic carbon filling equipment for automotive carbon canisters, it remains difficult to guarantee long-term stability of the negative pressure at the suction nozzle and the activated carbon filling amount. Summary of the Invention

[0006] To address the problem that using a single fixed-parameter PID closed-loop to adjust the vacuum pump output is difficult to keep up with the nonlinear changes in filter element resistance, this invention proposes an automatic carbon suction feedback control method based on negative pressure dynamic adjustment. This method is applied to an automatic carbon suction control system, which includes a storage silo, a nozzle docking mechanism, a frequency-controlled vacuum pump, a vacuum filter element, a proportional vacuum valve, a carbon canister workpiece fixture, a nozzle end pressure transmitter, and a programmable logic controller (PLC). The control system further includes pressure transmitters located at the inlet and outlet sides of the vacuum filter element, and a flow acquisition unit. The method includes the following steps: S1, the PLC synchronously samples the inlet... The inlet-side pressure transmitter and outlet-side pressure transmitter obtain the pressure difference signal before and after the filter element, and combine it with the flow signal of the flow acquisition unit to identify the linear seepage resistance coefficient and inertial resistance coefficient online; S2, the programmable logic controller superimposes pseudo-random binary excitation during the suction process and monitors the health status of the identifier, generates a feedforward control quantity based on the linear seepage resistance coefficient and the inertial resistance coefficient, schedules PID parameters to generate a feedback control quantity based on the normalized drift of the linear seepage resistance coefficient relative to the cleanliness reference, combines the feedforward control quantity and the feedback control quantity into a total control quantity sent to the vacuum pump frequency converter or proportional vacuum valve, and outputs maintenance prompts or cycle extension commands based on the normalized drift.

[0007] Compared to existing automatic carbon filling workstations that typically rely solely on the negative pressure at the nozzle for single-loop regulation and struggle to detect the clogging and aging of vacuum filters in a timely manner, this invention can identify changes in filter resistance online through the pressure difference between the filter inlet and outlet and the flow signal. It directly incorporates the filter status into feedforward compensation, PID feedback scheduling, and maintenance decisions, enabling the vacuum system to maintain a stable target negative pressure at the nozzle even during long-term production processes where the filter gradually accumulates carbon and resistance increases. This reduces carbon filling volume fluctuations, cycle time anomalies, and decreased product consistency caused by insufficient negative pressure. Simultaneously, pseudo-random excitation and health monitoring by the identifier improve the reliability of online identification, preventing the controller from making incorrect compensations based on distorted parameters. This allows for adaptive adjustment of suction performance, filter status early warning, and cycle time backup without shutting down the system or disassembling the filter, improving production continuity, equipment maintenance planning, and the stability of carbon canister filling quality.

[0008] Furthermore, the flow acquisition unit includes a speed feedback interface that communicates with the vacuum pump frequency converter and a pump flow characteristic curve stored in the programmable logic controller (PLC). The PLC obtains the volumetric flow rate through the filter element based on the speed feedback. Alternatively, the flow acquisition unit includes a proportional vacuum valve opening acquisition interface and an opening flow lookup table module formed by pre-calibration. The PLC obtains the volumetric flow rate through the filter element based on the valve opening.

[0009] Furthermore, the online identification is based on the Darcy-Forchheimer seepage relationship between the pressure difference between the inlet and outlet of the vacuum filter element and the volumetric flow rate. The programmable logic controller uses a recursive least squares identifier with a forgetting factor to refresh the linear seepage resistance coefficient and the inertial resistance coefficient according to the production cycle, and uses the clean reference resistance parameters and positive definite covariance matrix calibrated during the production line commissioning period as the initial identification state.

[0010] Compared to simply treating filter pressure drop as a fixed loss or judging blockage based solely on empirical thresholds, this invention distinguishes between linear seepage resistance and inertial resistance based on the Darcy-Forchheimer seepage relationship. It also employs a recursive least squares method with a forgetting factor to continuously refresh parameters according to the production cycle, which can more accurately reflect the true resistance characteristics of the filter under different flow conditions and aging stages. The clean baseline parameters and positive definite covariance matrix serve as the initial state, which helps to shorten the identification and convergence process after commissioning, improve the reliability of early control compensation, and transform negative pressure control from experience-based adjustment to adaptive control based on the physical state of the filter.

[0011] Furthermore, the programmable logic controller adds a pseudo-random binary sequence perturbation with an amplitude symmetrically distributed around the current output to the vacuum pump control quantity during the steady-state suction phase, and incorporates the sampled data of the nozzle docking, suction start, suction end and nozzle release transient phases into the identification dataset to expand the spectral coverage of the identification input signal.

[0012] Compared to relying entirely on natural operating condition fluctuations for parameter identification during production, which can easily lead to insufficient excitation and unidentifiable parameters, this invention superimposes symmetrically distributed pseudo-random binary disturbances in the steady-state suction section and expands the identification sample using transient data such as nozzle docking, suction start-up, suction end, and release. This allows the system to obtain richer flow-pressure difference response information without significantly affecting the carbon filling cycle and negative pressure stability, thereby improving the convergence speed of resistance parameter identification and its ability to resist operating condition uniformity, and reducing control errors caused by stagnation or drift of identification results during long-term stable production.

[0013] Furthermore, the programmable logic controller is also used to monitor the condition number of the regression matrix and the trace of the covariance matrix; when the condition number of the regression matrix continues to be higher than a preset upper limit, the update of the inertial drag coefficient is paused and the linear seepage drag coefficient is refreshed; when the trace of the covariance matrix is ​​lower than a preset lower limit, the covariance matrix is ​​reset to a positive definite diagonal matrix.

[0014] Compared to online identifiers that continue updating even when data is ill-conditioned or covariance degrades, potentially generating abnormal parameters and causing control oscillations, this invention determines the health status of the identifier by monitoring the condition number of the regression matrix and the trace of the covariance matrix. When the excitation is insufficient or the parameter correlation is too strong, the update of the inertial drag coefficient is paused, and when the covariance is too small and the identifier loses its adaptability, it is reset to a positive definite diagonal matrix. This improves the numerical stability and fault tolerance of the online identification process, avoids the chain reaction of individual abnormal sampling or indistinguishable operating conditions on negative pressure control, and enables the workstation to maintain reliable operation under complex production disturbances.

[0015] Furthermore, when the condition number of the regression matrix exceeds the limit and meets the preset continuous condition, the programmable logic controller switches to single-parameter identification mode, keeps the inertial drag coefficient at the initial calibration value and only refreshes the linear seepage drag coefficient, and sends an insufficient excitation flag bit to the upper system via the fieldbus; when the condition number of the regression matrix recovers and meets the preset stability condition, it resumes dual-parameter identification mode.

[0016] Furthermore, the feedforward control quantity is generated by the feedforward compensation unit. The feedforward compensation unit is based on the pressure balance relationship between the target negative pressure at the nozzle end, the filter element pressure drop, and the downstream fixed flow channel pressure drop. It adopts a linearized compensation structure around the nominal operating point to map the increment of the linear seepage resistance coefficient relative to the clean reference and the inertial pressure drop contribution corresponding to the inertial resistance coefficient into the actuator control quantity increment through the viscous pressure drop compensation weight coefficient and the inertial pressure drop compensation weight coefficient, respectively.

[0017] Compared to the lag control method that relies solely on PID feedback to correct for negative pressure deviations, this invention, based on the pressure balance relationship between the target negative pressure at the nozzle end, the filter element pressure drop, and the downstream fixed flow channel pressure drop, converts the incremental contribution of filter element resistance and inertial pressure drop into the incremental control quantity of the actuator in advance. This allows the vacuum pump or proportional vacuum valve to pre-compensate before the filter element resistance changes cause a significant deviation in the negative pressure at the nozzle end. This method helps reduce negative pressure fluctuations during the initial suction stage and the filter element aging stage, reduces the adjustment burden of feedback control, and improves the response speed, filling consistency, and cycle stability of the carbon filling process.

[0018] Furthermore, the programmable logic controller maps the normalized drift amount to a set of PID parameters for a steady-state segment, a transition segment, or a steep increase segment. In the transition segment, the proportional gain decreases relative to the steady-state segment and the integral time constant increases. In the steep increase segment, the proportional gain continues to decrease relative to the transition segment, the integral time constant continues to increase, and the derivative gain increases with the increase of the normalized drift amount and is constrained by a preset upper limit.

[0019] Furthermore, the programmable logic controller performs linear interpolation on adjacent PID parameter groups at the boundary of adjacent working sections according to the production cycle to avoid glitches in the total control quantity; the programmable logic controller generates the feedback control quantity based on the deviation between the negative pressure at the nozzle end measured by the nozzle end pressure transmitter and the target negative pressure, and adds the feedback control quantity to the feedforward control quantity at the actuator inlet before issuing it.

[0020] Furthermore, it also includes a manufacturing execution system and a quality control dimming system connected via fieldbus; when the normalized drift amount enters the steep increase phase, the programmable logic controller performs linear fitting on the linear seepage resistance coefficient sequence within a preset window to obtain the growth trend and form a remaining life estimate. When the remaining life estimate meets the warning conditions, it sends a maintenance reminder, and when the maintenance delay reaches the preset maintenance threshold, it extends the suction time and provides feedback on the cycle extension information according to the degree of excess.

[0021] The technical effects of this invention are as follows: This invention focuses on online identification of vacuum filter element resistance, combining inlet and outlet pressure difference, flow estimation, PRBS excitation, and health monitoring to obtain real-time filter element aging status. This data is then used for feedforward compensation, PID segmented scheduling, remaining life prediction, and cycle time fallback control. Compared to traditional carbon filling equipment that relies solely on closed-loop negative pressure at the nozzle or periodic maintenance, this invention maintains stable negative pressure and consistent filling in a real production line environment where filter element clogging gradually develops. It also improves maintenance predictability, reduces downtime risks, and is suitable for the long-term stable operation of automated carbon canister filling workstations. Attached Figure Description

[0022] Figure 1This is a flowchart of an automatic carbon absorption feedback control method based on negative pressure dynamic adjustment in an embodiment of the present invention; Figure 2 This is a flowchart of the decision branch for injecting stimuli and protecting the convergence of the identifier in an embodiment of the present invention; Figure 3 This is a schematic diagram of a closed-loop control circuit in an embodiment of the present invention, in which the feedforward control quantity and the feedback control quantity are superimposed. Detailed Implementation

[0023] In this embodiment, the automatic carbon filling workstation includes mechanical execution units such as a storage bin, a suction nozzle docking mechanism, a frequency-controlled vacuum pump, a vacuum filter, a silencer, a proportional vacuum valve, and a carbon canister workpiece fixture, all uniformly scheduled by a programmable logic controller (PLC). Compared to common automatic carbon filling workstations, this embodiment makes the following hardware additions to the control feedback loop: A pressure transmitter is installed on the inlet side and the outlet side of the vacuum filter element, respectively. The two transmitters are sampled synchronously inside the PLC and the difference is calculated to obtain the pressure difference signal before and after the filter element. The flow signal can be obtained by converting the speed feedback from the vacuum pump frequency converter into the flow characteristic curve of the pump from the factory, or it can be obtained by looking up the opening degree of the proportional vacuum valve after one calibration. The original pressure transmitter at the nozzle end is retained, and its measured value is still used as the feedback source for the feedback control loop. The PLC connects to the host Manufacturing Execution System (MES) and the workstation quality control dimming system via a fieldbus (such as PROFINET) to send maintenance prompts and cycle time extension information. The remaining actuators, fixtures, loading / unloading mechanisms, etc., retain their original configurations.

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] An example of an automatic carbon absorption feedback control method based on negative pressure dynamic adjustment: like Figure 1 As shown, an automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to the present invention includes: S1. Online identification of filter element resistance coefficient.

[0026] This embodiment identifies the pressure differential-flow rate relationship of the vacuum filter element and its downstream flow channel online at each production cycle. The physical model used is the Darcy-Forchheimer form, which has been experimentally verified in the field of porous media seepage. ; In the formula, This indicates the production cycle number, which increments by 1 for each completed carbon canister filling. For the first The measured pressure difference between the inlet and outlet of the vacuum filter element under cycle time, in kPa; The volumetric flow rate through the filter element under the same cycle time is expressed in L / min. The linear seepage resistance coefficient corresponds to the viscous laminar flow pressure drop, with units of kPa·min / L, and increases approximately linearly with the deposition thickness; The coefficient is the inertial drag coefficient, corresponding to the inertial turbulent pressure drop, with units of kPa·min² / L², which increases rapidly in a power-law manner at the end of the deposition process; This is the zero-mean bounded measurement noise term, in kPa.

[0027] parameter , The system employs the mature Recursive Least Squares (RLS) method with a forgetting factor for real-time updates. The standard iterative process of this algorithm is well-known to those skilled in the art and will not be elaborated upon in this embodiment. In this embodiment, the forgetting factor is set to 0.97, but it can be adjusted within the range of 0.95 to 0.99 to give new data a higher weight relative to historical data. The initial value uses the initial linear resistance coefficient calibrated during the production line commissioning period. With initial inertial drag coefficient The initial covariance matrix is ​​a relatively large positive definite diagonal matrix to accelerate convergence.

[0028] An identification update is performed once per beat. The output of this step is the identification update for the current beat. , The estimated value is stored in the controller's internal state variables for subsequent steps to retrieve.

[0029] S2, Inject stimulus and protect the identifier convergence.

[0030] To avoid an ill-conditioned regression matrix caused by an excessively narrow input signal spectrum under steady-state carbon absorption conditions, and To address the issue of inseparability, this embodiment employs the following three safeguards simultaneously during the S2 identification process: The first step: continuous excitation signal injection. During the steady-state suction phase of each production cycle, the PLC superimposes a pseudo-random binary sequence (PRBS) disturbance onto the vacuum pump control quantity. The disturbance amplitude is ±3% of the rated output, and the basic period is 100 milliseconds (recommended range). In other embodiments, the disturbance amplitude range can also be ±2% to ±5% of the rated output, and the basic period range can be 50 to 200 ms. Simultaneously, high-speed sampling data from the four start-stop transient phases—nozzle docking, suction start, suction end, and nozzle release—are synchronously used for identification. The transient data contains rich spectral components from low to mid-frequency, which can improve the condition number of the regression matrix.

[0031] The second item: Regression matrix condition number monitoring and covariance reset. The condition number of the regression matrix is ​​calculated in real time for each identification period. When the condition number exceeds a preset upper limit (typically 10) within 5 consecutive sampling periods, the condition number is reset. 4 Pause Updates are only allowed. Continue refreshing; simultaneously monitor the trace of the covariance matrix, when it falls below 10. -6 When the estimator enters a sleep state, the covariance matrix is ​​reset to a larger positive definite diagonal matrix, restoring the estimator's sensitivity to new data.

[0032] Third item: Single parameter identification rollback branch. When the condition number exceeds the threshold and fails to recover within 20 consecutive production cycles, the control system determines that the input stimulus is severely insufficient and automatically switches to single parameter identification mode: online refresh only. , Fixed to the initial calibration value Simultaneously, an insufficient excitation flag is sent to the upper-level system via the fieldbus, prompting process personnel to appropriately increase the PRBS disturbance amplitude or check the sensor sampling status. When the condition number recovers to below the threshold and remains stable at 50 cycles, the dual-parameter identification mode is automatically restored.

[0033] This step can output stable and reliable identification parameters. , And the health status flag of the identifier. For example... Figure 2 As shown in the figure, starting from the high-speed sampling data of the PRBS disturbance and start-stop transient segment entering the identification cycle, the following steps are sequentially represented: regression matrix condition number monitoring, covariance matrix trace monitoring, covariance matrix reset, single-parameter identification mode, insufficient excitation flag, and two-parameter identification mode recovery. Each judgment branch eventually returns to the health status flag output of the identifier.

[0034] S3, calculate the feedforward compensation control quantity.

[0035] The derivation of the feedforward compensation law starts from the pressure balance of the vacuum circuit under steady-state conditions: the target negative pressure at the nozzle end is equal to the sum of the pressure drop across the vacuum filter and the pressure drop in the downstream fixed flow channel. Substituting the seepage equation obtained in step S1 into the pressure balance equation, a quadratic algebraic equation in one variable for steady-state flow rate can be obtained. After solving the equation, the precise feedforward control quantity can be obtained by inverting the flow rate-control quantity characteristic curve of the vacuum pump. The process of solving this precise solution involves algebraic operations, which will not be elaborated in this embodiment.

[0036] Considering the limited real-time computing resources of the PLC within each cycle, this embodiment does not solve the above nonlinear equations online, but instead uses its first-order Taylor expansion near the nominal operating point as the engineering feedforward law: ; In the formula, This is a feedforward control variable, and the unit is the percentage of the output frequency of the vacuum pump inverter or the percentage of the opening degree of the proportional vacuum valve. The nominal output value corresponding to achieving the target negative pressure under clean filter conditions is determined by a one-time calibration during the production line commissioning period. , These are the initial linear drag coefficient and steady-state volumetric flow rate under the calibration conditions, respectively. The viscosity pressure drop compensation weighting coefficient is represented by the proportion of the viscosity pressure drop term in the total system pressure drop under clean baseline, and is mapped to the control quantity increment through the vacuum pump flow rate-control quantity slope coefficient. In this embodiment, the value range is 0.8 to 1.0. The inertial pressure drop compensation weighting coefficient has the physical meaning of the relative pressure drop growth ratio corresponding to a unit increase in inertial resistance. In this embodiment, the value ranges from 0.3 to 0.6.

[0037] , The specific values ​​were determined through bench calibration, and the calibration conditions were designed as follows: Flow scan range The negative pressure setting is set to 30%–120% of the target negative pressure value, with a sampling point taken every 10%. The filter has four levels: 50%, 75%, 100%, and 125%. At least 200 samples are continuously collected at each working point. The artificial deposition counterweight simulation method is adopted, and adjustable flow valves can be connected in series before and after the filter to simulate different deposition levels covering five deposition states. Nonlinear least squares fitting is adopted, requiring the coefficient of determination of the goodness of fit to be no less than 0.95 and the root mean square of the residuals to be less than 0.2 kPa.

[0038] When migrating to a different workstation model, simply repeat the above calibration process to obtain the new equipment. , The numerical values ​​and the feedforward structure themselves remain unchanged.

[0039] After each cycle of identification and update is completed, the PLC immediately calculates the feedforward control quantity according to the above formula. The output of this step is the feedforward control quantity. .

[0040] S4. Segmented scheduling of PID parameters and superimposed output.

[0041] Definition Identification relative to cleanliness benchmark The normalized drift is the scheduling variable: ; It represents the relative increase in viscous resistance caused by filter cartridge deposits relative to the initial clean state, and is a dimensionless quantity. This embodiment follows... The values ​​divide the filter cartridge lifecycle into three segments, corresponding to three sets of PID parameters: First section (steady-state section) ): Using the standard parameter set calibrated when the production line was put into operation, the baseline proportional gain is denoted as . The baseline integration time is The reference differential gain is , Take 0 or a smaller value.

[0042] Second paragraph (transition paragraph) In this embodiment, the proportional gain is set to 0.75 times the reference value; the integral time constant is extended to 1.5 times the reference value; and the derivative gain remains the reference value. This section suppresses potential oscillations caused by the nonlinear growth of resistance by moderately weakening the feedback strength.

[0043] The third section (sharp increase section) In this embodiment, the proportional gain is set to 0.5 times the reference value; the integral time constant is extended to 2.5 times the reference value; the differential gain corresponding to the normalized drift is calculated using the following formula: ; ; In the formula, The differential gain of the steeply increasing segment; The differential time constant is the base value, which can be 0.075 seconds in this embodiment, and can be set to 0.05 to 0.10 seconds in other embodiments; The preset coefficient, its value varies. The portion exceeding 1.5 increases linearly, with an upper limit of 2.0 set to avoid noise amplification. This is used to make the differential action linearly increase with the increase in deposition, thereby compensating for the phase margin lost due to the slowing of the dynamic response of the controlled object in the section of steep increase in resistance.

[0044] The final values ​​of the three sets of parameters for the steeply increasing segment within the aforementioned interval are determined jointly by the following three methods: Root locus analysis: Based on the maximum value obtained from identification The transfer function of the controlled object in the state is used to construct the closed-loop root locus, requiring that the damping ratio of the dominant pole in the closed loop is not less than 0.7; Frequency domain margin analysis: The open-loop Bode plot phase margin is required to be no less than 45° and the magnitude margin is required to be no less than 6dB. Benchtop trial and error record: In Step response tests were applied under three artificial deposition conditions: 2.0 and 2.5. The overshoot was required to be no more than 10% and the settling time was required to be no more than 3 times the standard cycle time. The specific values ​​were then determined by backtracking.

[0045] To avoid glitch in the control quantity at the stage boundary, step parameter switching is not used between each stage. Instead, linear interpolation with a cycle length of 7 is used for transition. Within the transition interval, the parameters change linearly and smoothly according to the cycle number.

[0046] The selected PID parameter set in this step is applied to the negative pressure deviation at the nozzle end to obtain the feedback control quantity. and the feedforward control quantity output by S3 The values ​​are added directly at the actuator inlet to obtain the total control quantity sent to the vacuum pump inverter or proportional vacuum valve for this cycle. Because the feedforward channel actively raises the reference output before pressure errors occur, the feedback loop can focus on suppressing fast transient disturbances, eliminating the need to rely on the integral term to slowly eliminate steady-state deviations caused by filter aging. This structurally eliminates integral saturation and steady-state deviation accumulation. The output of this step is the total control quantity sent to the actuator for the current cycle.

[0047] like Figure 3 As shown in the figure, the negative pressure deviation is formed by the target negative pressure at the nozzle end and the feedback from the nozzle end pressure transmitter. The feedback law outputs the feedback control quantity according to the PID parameter group. The linear seepage resistance coefficient and inertial resistance coefficient identified by S1 enter the feedforward law of S3 to generate the feedforward control quantity. The two control quantities are added at the actuator inlet to form the total control quantity and sent to the vacuum pump frequency converter or proportional vacuum valve.

[0048] S5. Remaining life prediction and beat fallback.

[0049] In addition to triggering the parameter switching of the S4 steep increase segment, the boundary condition simultaneously initiates the filter cartridge remaining life prediction branch. The specific implementation steps are as follows: Firstly, the PLC can monitor the last 100 production cycles. Perform least-squares linear fitting on the sequence to obtain... The current growth slope, in other embodiments, can also be a production cycle time range of 50–200 cycles; then extrapolation is made from this. Reaching the preset maintenance threshold The number of remaining beats required, of which Typical values ​​are 2.0 to 2.5; finally, when the remaining cycle count is less than the warning threshold, such as 200 pieces, the control system sends a maintenance reminder to the workstation quality control dimming system via the fieldbus, and the maintenance personnel stop the machine to replace the filter element according to the plan.

[0050] If maintenance fails If completed previously, the adaptive fallback branch will be activated: the suction time of a single carbon canister will be adjusted from the calibrated value. This means extending the standard suction time required to achieve the cumulative suction volume needed for the process while the filter element is clean. The extension is given by the following formula: ; In the formula, This is the extension of the suction duration in this beat; The standard suction time is when the filter element is clean. and The definition is the same as before. This means that when the identified resistance exceeds the maintenance threshold, the suction time is linearly extended proportionally to maintain the cumulative suction volume obtained by each carbon canister—that is, the integral of negative pressure over time—at the required process level, thereby ensuring filling quality. Simultaneously, the PLC feeds back the cycle time extension information to the upper-level MES via the fieldbus for capacity scheduling. This branch ensures that even in extreme cases of maintenance delay, the filling quality remains within the process's allowable range. This step outputs a maintenance prompt signal, a remaining lifespan estimate, and a cycle time extension command.

[0051] S6. Closed-loop execution timing of hierarchical control laws.

[0052] The execution sequence of each step within each production cycle is as follows: First, at the start of the cycle, the PLC reads the pressure difference and flow data across the filter element from the previous cycle, and initiates S1 and S2 identification and update to obtain... , Then, immediately after identification, the S3 feedforward law is invoked to calculate the feedforward control quantity; at this time, the pressing... Select the current PID parameter group in S4, calculate the feedback control quantity using the feedback law; further execute the distribution, distributing the sum of the feedforward and feedback control quantities to the vacuum pump inverter or proportional vacuum valve; finally execute step S5 to refresh the life estimate and determine whether a maintenance prompt or cycle time extension is triggered.

Claims

1. An automatic carbon absorption feedback control method based on negative pressure dynamic adjustment, the method being applied to an automatic carbon absorption control system, the control system comprising a storage silo, a nozzle docking mechanism, a frequency-controlled vacuum pump, a vacuum filter element, a proportional vacuum valve, a carbon canister workpiece clamp, a nozzle end pressure transmitter, and a programmable logic controller, characterized in that, The control system also includes pressure transmitters and flow acquisition units respectively located on the inlet and outlet sides of the vacuum filter element; The method includes the following steps: S1. The programmable logic controller synchronously samples the inlet-side pressure transmitter and the outlet-side pressure transmitter to obtain the pressure difference signal before and after the filter element, and combines it with the flow signal of the flow acquisition unit to identify the linear seepage resistance coefficient and the inertial resistance coefficient online. S2. During the suction process, the programmable logic controller superimposes pseudo-random binary excitation and monitors the health status of the identifier. It generates a feedforward control quantity based on the linear seepage resistance coefficient and the inertial resistance coefficient. It schedules PID parameters to generate a feedback control quantity based on the normalized drift of the linear seepage resistance coefficient relative to the cleanliness reference. It combines the feedforward control quantity and the feedback control quantity into a total control quantity sent to the vacuum pump frequency converter or proportional vacuum valve. Based on the normalized drift, it outputs maintenance prompts or cycle extension commands.

2. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 1, characterized in that, The flow acquisition unit includes a speed feedback interface that communicates with the vacuum pump frequency converter and a pump flow characteristic curve stored in the programmable logic controller. The programmable logic controller obtains the volumetric flow rate through the filter element based on the speed feedback. Alternatively, the flow acquisition unit includes a proportional vacuum valve opening acquisition interface and an opening flow lookup module formed by pre-calibration, wherein the programmable logic controller obtains the volumetric flow rate through the filter element according to the valve opening.

3. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 1, characterized in that, The online identification is based on the Darcy-Forchheimer seepage relationship between the pressure difference between the inlet and outlet of the vacuum filter element and the volumetric flow rate. The programmable logic controller uses a recursive least squares identifier with a forgetting factor to refresh the linear seepage resistance coefficient and the inertial resistance coefficient according to the production cycle, and uses the clean reference resistance parameters and positive definite covariance matrix calibrated during the production line commissioning period as the initial identification state.

4. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 3, characterized in that, The programmable logic controller adds a pseudo-random binary sequence perturbation with an amplitude symmetrically distributed around the current output to the vacuum pump control quantity during the steady-state suction phase, and incorporates the sampled data of the nozzle docking, suction start, suction end and nozzle release transient phases into the identification dataset to expand the spectral coverage of the identification input signal.

5. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 4, characterized in that, The programmable logic controller is also used to monitor the condition number of the regression matrix and the trace of the covariance matrix; When the condition number of the regression matrix continues to exceed the preset upper limit, the update of the inertial drag coefficient is paused and the linear seepage drag coefficient is refreshed. When the trace of the covariance matrix is ​​below a preset lower limit, the covariance matrix is ​​reset to a positive definite diagonal matrix.

6. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 5, characterized in that, When the condition number of the regression matrix exceeds the limit and the preset continuous condition is met, the programmable logic controller switches to the single-parameter identification mode, keeps the inertial drag coefficient at the initial calibration value and only refreshes the linear seepage drag coefficient, and sends the excitation insufficient flag bit to the upper system through the fieldbus. The two-parameter identification mode is restored when the condition number of the regression matrix is ​​recovered and the preset stability condition is met.

7. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 1, characterized in that, The feedforward control quantity is generated by the feedforward compensation unit. The feedforward compensation unit is based on the pressure balance relationship between the target negative pressure at the nozzle end, the filter element pressure drop, and the downstream fixed flow channel pressure drop. It adopts a linearized compensation structure around the nominal operating point and maps the increment of the linear seepage resistance coefficient relative to the clean reference and the inertial pressure drop contribution corresponding to the inertial resistance coefficient to the actuator control quantity increment through the viscous pressure drop compensation weight coefficient and the inertial pressure drop compensation weight coefficient, respectively.

8. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 1, characterized in that, The programmable logic controller maps the normalized drift to a set of PID parameters for a steady-state segment, a transition segment, or a steep increase segment. In the transition section, the proportional gain decreases compared to the steady-state section, and the integral time constant increases. During the steep increase phase, the proportional gain continues to decrease relative to the transition phase, the integral time constant continues to increase, and the differential gain increases with the increase of the normalization drift and is constrained by a preset upper limit.

9. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 8, characterized in that, The programmable logic controller performs linear interpolation of adjacent PID parameter groups at the boundary of adjacent working sections according to the production cycle to avoid glitches in the total control quantity. The programmable logic controller generates the feedback control quantity based on the deviation between the negative pressure at the nozzle end measured by the nozzle end pressure transmitter and the target negative pressure, and then adds the feedback control quantity to the feedforward control quantity at the actuator inlet before issuing it.

10. The automatic carbon absorption feedback control method based on negative pressure dynamic adjustment according to claim 1, characterized in that, It also includes manufacturing execution systems and quality control dimming systems connected via fieldbus; When the normalized drift amount enters the steep increase phase, the programmable logic controller performs linear fitting on the linear seepage resistance coefficient sequence within a preset window to obtain the growth trend and form a remaining lifetime estimate. When the remaining lifetime estimate meets the warning conditions, a maintenance prompt is sent. When the maintenance delay reaches the preset maintenance threshold, the pumping time is extended according to the degree of excess and feedback cycle extension information is provided.

Images