A water injection filling pressure disturbance identification compensation method

Abstract

This invention relates to the field of pressure disturbance identification and control technology, and in particular to a method for identifying and compensating pressure disturbances during water injection filling. The method collects and deconstructs the time series of supply pressure during a continuous filling cycle, generating three types of pressure disturbance characteristic quantities. It calculates the filling weight deviation and cross-validates it with the three types of pressure disturbance characteristic quantities to determine the source of the deviation. Based on the deviation source identification results, it obtains the comprehensive filling control quantity for the next filling cycle. Before executing the comprehensive filling control quantity, it judges the fluid dynamic state of the supply pipeline, reconstructs the initial pressure state of the supply pipeline based on the judgment result, or executes segmented peristaltic pump filling control. After filling is completed, it performs closed-loop parameter updates. This invention can reduce the risk of undercompensation, overcompensation, and continuous filling volume drift caused by occasional disturbances being misjudged as continuous trends, and improves the consistency of filling volume during continuous filling.

Description

Technical Field

[0001] This invention relates to the field of pressure disturbance identification and control technology, and in particular to a method for identifying and compensating for pressure disturbances during water injection filling. Background Technology

[0002] Fill volume control in liquid injections is a crucial technical aspect of aseptic production, directly impacting the consistency of finished product fill volume, batch stability, and subsequent quality testing results. During actual continuous filling, the liquid level in the storage tank continuously decreases as the medication is consumed, causing a slow shift in the supply pressure baseline. This results in inconsistent actual fill volumes from the peristaltic pump under the same speed, rotation time, or control pulse.

[0003] However, existing post-filling weighing methods can only reflect the filling deviation with a lag after filling is completed. It is difficult to identify whether the deviation is caused by the pressure trend migration due to the drop in liquid level, the instantaneous disturbance in a single cycle, or the residual pressure state that has not been fully recovered between adjacent filling cycles. If the peristaltic pump parameters of the next bottle are directly corrected based on the weighing result of the previous bottle, it is easy to mistake occasional disturbances as continuous trend compensation, thereby causing undercompensation, overcompensation, or even continuous filling drift. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a method, system, and storage medium for identifying and compensating for pressure disturbances during water injection filling. This method generates three types of pressure disturbance characteristics by collecting and deconstructing the time series of supply pressure during a continuous filling cycle. Based on the cross-validation results of the filling weight deviation and the three types of pressure disturbance characteristics, the source of the deviation is identified, and a comprehensive filling control quantity for the next filling cycle is obtained. Before executing the comprehensive filling control quantity, the fluid dynamics state of the supply pipeline is determined. Based on the determination result, the initial pressure state of the supply pipeline is reconstructed or segmented peristaltic pump filling control is executed. After filling is completed, closed-loop parameter updates are performed, thereby reducing the risks of undercompensation, overcompensation, and continuous fill volume drift, and improving the consistency of continuous filling volume.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for identifying and compensating for pressure disturbances during the filling of water-based injections, executed by a control device, comprising:

[0007] Collect and deconstruct the time series of liquid supply pressure during the continuous filling cycle, and generate three types of pressure disturbance characteristic quantities based on the time series of liquid supply pressure during the continuous filling cycle;

[0008] Calculate the filling weight deviation during the filling cycle, and cross-validate the filling weight deviation with three types of pressure disturbance characteristics to determine the deviation source identification results;

[0009] The comprehensive filling control quantity for the next filling cycle is obtained based on the results of the deviation source identification.

[0010] Before executing the comprehensive filling control quantity, the fluid dynamics state of the liquid supply pipeline is judged; if the filling start-up conditions are not met, the initial pressure state of the liquid supply pipeline is rebuilt; if the filling start-up conditions are met, segmented peristaltic pump filling control is executed according to the comprehensive filling control quantity; after filling is completed, closed-loop parameters are updated.

[0011] Preferably, three types of pressure disturbance characteristic quantities are generated, including:

[0012] The continuous filling cycle liquid supply pressure time sequence is divided according to the start and stop sequence of the peristaltic pump. First, the liquid supply pressure time sequence of the static section before filling is obtained, then the liquid supply pressure time sequence of the filling execution section is obtained, and finally the liquid supply pressure time sequence of the recovery section after filling is obtained.

[0013] First, the baseline migration of the supply pressure is generated based on the time series of the supply pressure during the pre-filling settling period. Then, the disturbance amplitude within the cycle is generated based on the time series of the supply pressure during the filling execution period. Finally, the residual verification amount during the cycle is generated based on the time series of the supply pressure during the recovery period after filling of the adjacent filling cycle, combined with the time series of the supply pressure during the pre-filling settling period of the next filling cycle.

[0014] The three generated results are used as three types of pressure perturbation features.

[0015] Preferably, generating the supply pressure baseline shift includes:

[0016] Calculate the median pressure of the liquid supply pressure time series during the settling period before filling;

[0017] Based on the absolute deviation of the pressure value at each sampling point relative to the median pressure, the median of all absolute deviations is taken as the pressure fluctuation scale.

[0018] Abnormal sampling point pressure values ​​with absolute deviations exceeding three times the pressure fluctuation scale are removed, and the remaining sampling point pressure values ​​are averaged to obtain the baseline value of the liquid supply pressure for the current filling cycle.

[0019] The liquid supply pressure baseline shift is obtained by subtracting the liquid supply pressure baseline value of the previous filling cycle from the liquid supply pressure baseline value of the current filling cycle.

[0020] Preferably, cross-validation includes:

[0021] A historical filling weight deviation sequence is generated based on the filling weight deviation of consecutive filling cycles;

[0022] A moving average is applied to the historical filling weight deviation sequence to obtain the long-term trend sequence of filling weight deviation;

[0023] Perform a first-order difference on the historical filling weight deviation sequence to obtain the first-order difference sequence of filling weight deviation;

[0024] The long-term trend sequence of filling weight deviation and the first-order difference sequence of filling weight deviation were used for cross-validation to obtain test results for three types of sources.

[0025] The results of the deviation source identification were determined based on the test results of the three types of sources.

[0026] Preferably, three types of source test results are obtained, including:

[0027] The time series of supply pressure baseline migration was read from three types of pressure disturbance characteristic quantities. The cross-correlation between the time series of supply pressure baseline migration and the long-term trend series of filling weight deviation was calculated to obtain the test results of supply pressure baseline migration.

[0028] Read the time series of disturbance amplitude within the period, check the event synchronicity of the time series of disturbance amplitude within the period relative to the first difference series of filling weight deviation, check the deviation direction according to the disturbance direction mark within the period, and obtain the disturbance test result within the period;

[0029] Read the residual verification amount during the cycle, and perform cycle-by-cycle mapping verification based on the relative filling weight deviation of the residual verification amount during the cycle to obtain the residual inspection result during the cycle.

[0030] Preferably, the deviation source identification result is determined based on the test results of the three types of sources, including:

[0031] The test results of the three types of sources, along with the characteristic quantities of the three types of pressure disturbances, are input into the deviation source probability identification model to obtain the contribution probability of each type of deviation source.

[0032] By comparing contribution probabilities, the dominant source of bias corresponding to the probability of maximum contribution can be determined.

[0033] The dominant source of bias, along with its contribution probability, is included in the bias source identification results.

[0034] Preferably, the comprehensive filling control quantity for the next filling cycle is obtained based on the deviation source identification results, including:

[0035] Read the contribution probability of the liquid supply pressure baseline migration from the deviation source identification results, and generate an adaptive compensation vector for the liquid supply pressure baseline migration.

[0036] Read the probability of intra-period disturbance contribution from the deviation source identification results and generate an intra-period disturbance isolation compensation vector;

[0037] Read the probability of periodic residual contribution from the deviation source identification results and generate a periodic residual decoupling compensation vector;

[0038] Set the compensation vector that has not met the corresponding compensation trigger condition to zero;

[0039] The comprehensive filling control quantity is obtained based on the three compensation vectors.

[0040] Preferably, the comprehensive filling control quantity is obtained based on three compensation vectors, including:

[0041] Read the contribution probability of each deviation source in the deviation source identification results;

[0042] Transform each compensation vector into the control dimension of the baseline filling control quantity;

[0043] Using the corresponding contribution probability as the fusion weight, the transformed compensation vector is weighted and superimposed onto the benchmark filling control quantity to obtain the comprehensive filling control quantity;

[0044] Based on the adjustment range of the comprehensive filling control quantity relative to the benchmark filling control quantity according to the compensation saturation threshold, the comprehensive filling control quantity for the next filling cycle is output.

[0045] Preferably, determining and reconstructing the fluid dynamics state of the supply pipeline includes:

[0046] Before the start of the next filling cycle, the fluid dynamics of the supply pipeline is deemed to meet the filling start-up conditions only if the average liquid supply pressure in the static section before filling is within the allowable starting pressure range, the standard deviation of the liquid supply pressure time series in the static section before filling is less than the pressure stability threshold, and the deviation ratio of the pressure rebound time constant relative to the standard time constant does not exceed the time constant deviation threshold. Otherwise, the fluid dynamics of the supply pipeline is deemed not to meet the filling start-up conditions.

[0047] If the fluid dynamics of the supply line does not meet the filling start-up conditions, the initial pressure state of the supply line is rebuilt; if the fluid dynamics of the supply line meets the filling start-up conditions, segmented peristaltic pump filling control is executed according to the comprehensive filling control quantity.

[0048] After the segmented peristaltic pump filling control is completed, the closed-loop parameters are updated based on the filling weight deviation of the next filling cycle.

[0049] Preferably, segmented peristaltic pump filling control includes:

[0050] The interval before the slope of the liquid supply pressure rise curve after the peristaltic pump starts drops to the steady-state determination slope threshold is defined as the filling start-up section.

[0051] The interval between the end of the filling start-up phase and the issuance of the peristaltic pump deceleration command is defined as the filling steady-state phase.

[0052] The interval from when the peristaltic pump deceleration command is issued until the peristaltic pump stops is defined as the filling end segment;

[0053] The peristaltic pump is started in the filling start-up section, then runs in the filling steady-state section, and finally stops in the filling end section, based on the comprehensive filling control quantity. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 This is a flowchart illustrating the overall process of a water injection filling pressure disturbance identification and compensation method according to the present invention.

[0056] Figure 2 This is a schematic diagram of a three-segment timing partitioning in an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of the anti-pulsation smoothing process in an embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram of the perturbation spectrum feature library in an embodiment of the present invention;

[0059] Figure 5 This is a schematic diagram of the residual verification amount during the week in an embodiment of the present invention;

[0060] Figure 6 This is a schematic diagram of the temporal correlation matrix in an embodiment of the present invention;

[0061] Figure 7 This is a schematic diagram of the deviation source probability identification model in an embodiment of the present invention;

[0062] Figure 8 This is a schematic diagram of the comprehensive filling control quantity in an embodiment of the present invention. Detailed Implementation

[0063] Example 1:

[0064] 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 a part of the embodiments of the present invention, and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0065] S1: Collect and deconstruct the time series of liquid supply pressure during the continuous filling cycle, and generate three types of pressure disturbance characteristic quantities based on the time series of liquid supply pressure during the continuous filling cycle.

[0066] Specifically, during the continuous filling process, each filling action of the container to be filled is taken as a filling cycle. The liquid supply pressure time series of each filling cycle is continuously collected. The liquid supply pressure time series of the continuous filling cycle is subjected to fluid dynamics deconstruction to generate three types of pressure disturbance characteristic quantities. The characteristic domain separation relationship between the three types of pressure disturbance characteristic quantities and the source of filling weight deviation is established.

[0067] Where k is the current filling cycle number; when calculating the liquid supply pressure baseline migration, k is a positive integer not less than 2, and the liquid supply pressure baseline value of the first filling cycle is used as the initial baseline value for calculating the liquid supply pressure baseline migration in subsequent filling cycles.

[0068] The fluid dynamics deconstruction refers to dividing the liquid supply pressure time series into a pre-filling resting period, a filling execution period, and a post-filling recovery period according to the peristaltic pump start-up and shutdown sequence. Then, it involves extracting baseline migration features, intra-cycle disturbance features, and inter-cycle residual features from each of these periods. The feature domain separation relationship refers to mapping the gradually changing pressure trend, intra-cycle transient pressure disturbance, and inter-cycle residual pressure to different sources of deviation, allowing these three types of deviation sources to participate in subsequent cross-validation and compensation calculations.

[0069] The three types of pressure disturbance characteristics correspond to the following categories: supply pressure baseline migration characteristics, intra-cycle disturbance characteristics, and inter-cycle residual characteristics. Specifically, the supply pressure baseline migration characteristics correspond to the source of slowly changing pressure trends, the intra-cycle disturbance characteristics correspond to the source of transient disturbances within a single cycle, and the inter-cycle residual characteristics correspond to the source of residual pressure at the start of a cross-cycle period. The supply pressure baseline migration characteristics characterize the slow change in the supply pressure baseline, the intra-cycle disturbance characteristics characterize the transient pressure disturbances within a single filling cycle, and the inter-cycle residual characteristics characterize the residual pressure impact transmitted from the end of the previous filling cycle to the start of the next filling cycle.

[0070] Further, step S1 includes:

[0071] S11: Divide the k-th filling cycle into three segments according to the start-stop sequence of the peristaltic pump, generating a pre-filling resting segment, a filling execution segment, and a post-filling recovery segment.

[0072] like Figure 2 The diagram shows a three-segment time sequence division method for the filling cycle. Based on the peristaltic pump start and stop command timestamps, the system divides the pressure sampling sequence into a pre-filling resting segment, a filling execution segment, and a post-filling recovery segment. This distinguishes the pressure baseline, execution disturbances, and recovery residues, avoiding mixed calculations of different pressure mechanisms.

[0073] The three-stage time sequence partitioning is used to avoid mixing pressure samples under different stress mechanisms in the same calculation.

[0074] At the start of the k-th filling cycle, the control device reads the peristaltic pump start command timestamp, the peristaltic pump stop command timestamp, and the pressure acquisition device sampling timestamp. The peristaltic pump start command timestamp and the peristaltic pump stop command timestamp are generated by the control device's filling control program, while the pressure acquisition device sampling timestamp is generated by the pressure acquisition device's sampling clock. The control device aligns the peristaltic pump start command timestamp, the peristaltic pump stop command timestamp, and the pressure acquisition device sampling timestamp with the same system clock to obtain the timing reference for the k-th filling cycle.

[0075] The control device defines the preset settling time before the peristaltic pump start command timestamp as the pre-filling settling period, the time interval between the peristaltic pump start command timestamp and the peristaltic pump stop command timestamp as the filling execution period, and the preset recovery time after the peristaltic pump stop command timestamp as the post-filling recovery period. The preset settling time is determined based on the minimum time required for the pressure fluctuation in the supply line to decay to a steady state; for example, the preset settling time is 0.5 seconds to 2 seconds. The preset recovery time is determined based on the characteristic time constant of the supply line pressure returning from a dynamic response to a quasi-static state; for example, the preset recovery time is 1 second to 3 seconds.

[0076] Based on the timing reference of the k-th filling cycle, the control device extracts the time series of the supply pressure during the pre-filling resting period, the supply pressure during the filling execution period, and the supply pressure during the post-filling recovery period from the continuous supply pressure sampling data output by the pressure acquisition device. The supply pressure time series during the pre-filling resting period is used in S12 to calculate the supply pressure baseline shift; the supply pressure time series during the filling execution period is used in S13 to calculate the disturbance amplitude and disturbance direction marker within the cycle; and the supply pressure time series during the post-filling recovery period is used in S14 to obtain the residual pressure at the end of the recovery period and the pressure rebound time constant. After the start of the next filling cycle, the control device generates the residual verification amount during the cycle based on the deviation of the initial pressure of the pre-filling resting period relative to the supply pressure baseline value for the next filling cycle.

[0077] S12: Perform anti-pulsation smoothing on the liquid supply pressure time series during the static period before filling, calculate the liquid supply pressure baseline value and the liquid supply pressure baseline migration, and convert the liquid supply pressure baseline migration into the flow deviation prediction value.

[0078] like Figure 3 The diagram shows the anti-pulsation smoothing process for pressure data during the static period before filling. The system first calculates the median pressure and pressure fluctuation scale, removing pulsation peaks, sampling spikes, and instantaneous anomalies. Then, it averages the effective sampling points to obtain a stable baseline value for the supply pressure, providing a basis for subsequent baseline migration and flow deviation prediction.

[0079] The anti-pulsation smoothing process refers to the process of removing pulsation peaks, sampling spikes, or instantaneous abnormal sampling points based on the median pressure and pressure fluctuation scale of the liquid supply pressure time series during the static period before filling, and averaging the remaining sampling points to obtain the baseline value of the liquid supply pressure.

[0080] The control device takes the liquid supply pressure time series output from S11 during the pre-filling settling period as input, calculates the median pressure of the liquid supply pressure time series during the pre-filling settling period, then calculates the absolute deviation between the pressure value of each sampling point in the pre-filling settling period liquid supply pressure time series and the median pressure, and takes the median of all absolute deviations as the pressure fluctuation scale. The pressure fluctuation scale is a pressure dispersion index calculated based on the absolute median difference, used to quantify the dispersion of the pressure series and identify abnormal sampling points; abnormal sampling points with an absolute deviation greater than 3 times the pressure fluctuation scale are removed; the arithmetic mean of the pressure values ​​of the remaining sampling points is taken, and the baseline value of the liquid supply pressure for the k-th filling cycle is output. The pressure fluctuation scale comes from the internal statistical results of the liquid supply pressure time series during the pre-filling settling period, and does not require manually setting a fixed noise boundary.

[0081] The control device reads the liquid supply pressure baseline value generated in S12 during the (k-1)th filling cycle, calculates the difference between the liquid supply pressure baseline value of the kth filling cycle and the liquid supply pressure baseline value of the (k-1)th filling cycle, and outputs the liquid supply pressure baseline shift amount of the kth filling cycle.

[0082] The supply pressure baseline shift is the result of subtracting the supply pressure baseline value of the (k-1)th filling cycle from the supply pressure baseline value of the (k-1)th filling cycle. A positive shift indicates an increase in supply pressure, while a negative shift indicates a decrease. The supply pressure baseline value of the (k-1)th filling cycle is written into the historical pressure record table after the (k-1)th filling cycle ends. The historical pressure record table is a key-value table with the filling cycle number as the key and the supply pressure baseline value as the value.

[0083] The control device calculates the time interval between the k-th filling cycle and the (k-1)-th filling cycle, divides the liquid supply pressure baseline shift by the time interval, and outputs the liquid supply pressure baseline shift rate. The time interval is obtained by subtracting the timestamps of the peristaltic pump start commands from those of two adjacent filling cycles. The liquid supply pressure baseline shift rate is used in step S31 to determine the update rate of the adaptive compensation vector for the liquid supply pressure baseline shift. The unit of the liquid supply pressure baseline shift rate is pressure units per second.

[0084] The control device utilizes the relationship between liquid level, pressure, and flow rate to convert the baseline shift of the supply pressure into a predicted flow rate deviation. This relationship is a quantitative mapping between the baseline shift of the supply pressure and the actual change in the peristaltic pump's outlet flow rate under current supply pipeline geometric parameters, drug density, drug viscosity, and peristaltic pump volumetric efficiency. During the initial startup of the filling system, the control device maintains a constant peristaltic pump control at multiple known storage tank levels, records the baseline supply pressure value and the filling mass per unit time, obtains the corresponding actual outlet flow rate based on the drug density, and calibrates the coefficients of the quantitative mapping relationship according to the change in actual outlet flow rate corresponding to the change in the baseline supply pressure value.

[0085] The direction of the predicted flow deviation is represented by the direction of deviation of the actual flow rate from the target flow rate. A positive predicted flow deviation value indicates that the predicted actual flow rate is higher than the target flow rate, while a negative predicted flow deviation value indicates that the predicted actual flow rate is lower than the target flow rate. The predicted flow deviation value is used by S31 to generate an adaptive compensation vector for the supply pressure baseline migration. When the predicted flow deviation value is expressed as mass flow rate deviation, the unit is grams per second or kilograms per second; when expressed as volumetric flow rate deviation, the unit is milliliters per second or liters per minute. In the same batch of calculations, the unit type of the predicted flow deviation value is consistent with the unit of the dependent variable in the peristaltic pump flow rate-speed characteristic curve.

[0086] S13: Perform high-frequency disturbance extraction on the liquid supply pressure time series of the filling execution section, calculate the disturbance amplitude, disturbance direction mark and disturbance main frequency within the period, and output the preliminary diagnosis result of the disturbance source based on the disturbance main frequency within the period.

[0087] like Figure 4 The diagram shows a schematic of the disturbance spectrum feature library, illustrating the establishment and matching logic of the library. After high-pass filtering and fast Fourier transform, the pressure signal from the filling execution section is used to extract the dominant disturbance frequency within the cycle. This frequency is then compared with typical spectral features such as peristaltic pump roller compression pulsation, mechanical vibration of the liquid supply pipeline, air bubble entry, and partial blockage of the filling needle. Here, f1 to f8 represent the boundary frequencies of the corresponding disturbance source frequency range, n represents the harmonic order, and Δf1 to Δf4 represent the bandwidth of the corresponding disturbance source.

[0088] The high-frequency disturbance extraction refers to separating the pressure fluctuation component that changes faster than the baseline of the supply pressure from the liquid supply pressure time series of the filling execution section, and using it to identify transient pressure disturbances within the cycle caused by peristaltic pump roller compression pulsation, mechanical vibration of the supply pipeline, air bubble entry, or local blockage of the filling needle.

[0089] The control device takes the liquid supply pressure time sequence of the filling execution section output by S11 as input, performs high-pass filtering, and outputs a high-frequency pressure signal for the filling execution section. The cutoff frequency of the high-pass filter is determined based on the rotation frequency of the peristaltic pump rollers. For example, the cutoff frequency is taken as 0.5 times the frequency corresponding to the rated speed of the peristaltic pump. The rotation frequency of the peristaltic pump rollers is calculated by converting the speed setting value sent by the control device to the peristaltic pump driver and the number of peristaltic pump rollers.

[0090] The control device takes the high-frequency pressure signal from the filling execution section as input and calculates the maximum positive and maximum negative amplitude values ​​of the high-frequency pressure signal. The control device takes the larger of the absolute values ​​of the maximum positive and maximum negative amplitude values ​​as the intra-cycle disturbance amplitude for the k-th filling cycle, with the unit of the intra-cycle disturbance amplitude being consistent with the unit of the liquid supply pressure. The control device simultaneously records the pressure change direction corresponding to the generation of the larger absolute amplitude value, using it as the intra-cycle disturbance direction marker for the k-th filling cycle. The intra-cycle disturbance amplitude value and the intra-cycle disturbance direction marker are written into a historical pressure disturbance record table for S21 to construct a time-series correlation matrix.

[0091] The control device performs a Fast Fourier Transform (FFT) on the high-frequency pressure signal of the filling execution section to obtain the disturbance spectrum distribution within the period. The frequency with the highest power in the disturbance spectrum distribution within the period is selected as the dominant disturbance frequency within the period. The input of the FFT is the high-frequency pressure signal of the filling execution section arranged in the order of sampling time, and the output is a two-dimensional array composed of frequency and power.

[0092] The control device matches the dominant frequency of the disturbance within a cycle with typical spectral patterns in the disturbance spectrum feature library, and outputs a preliminary diagnostic result of the disturbance source. The data structure of the disturbance spectrum feature library is a key-value table, where the key is the disturbance source category and the value is a feature vector composed of frequency range, octave relationship, and bandwidth. The disturbance spectrum feature library is established through equipment calibration, historical abnormal sample statistics, or preset experience frequency bands. Disturbance source categories include peristaltic pump roller compression pulsation, mechanical vibration of the liquid supply line, air bubble entry, and partial blockage of the filling needle.

[0093] S14: Based on the end pressure of the recovery section after filling, the baseline value of the liquid supply pressure, and the starting pressure of the static section before filling in the next filling cycle, calculate the residual pressure at the end of the recovery section and the residual verification amount during the cycle, and update the pressure rebound time constant according to the pressure decay curve of the recovery section after filling.

[0094] like Figure 5 The diagram shown illustrates the residual verification amount during the cycle, illustrating the calculation relationship of the residual verification amount during the cycle. The system compares the starting pressure of the settling section before filling in the next filling cycle with its baseline value of the supply pressure to determine whether the residual pressure of the recovery section in the previous cycle has been transmitted to the starting state of the next cycle. At the same time, it combines the residual pressure at the end of the recovery section and the pressure rebound time constant to assess the cross-cycle impact.

[0095] The control device reads the end sampled value of the liquid supply pressure time series of the post-filling recovery section output by S11 and generates the end pressure of the post-filling recovery section. The control device reads the beginning sampled value of the liquid supply pressure time series of the pre-filling resting section intercepted in S11 for the (k+1)th filling cycle and generates the beginning pressure of the pre-filling resting section for the (k+1)th filling cycle.

[0096] The control device calculates the difference between the initial pressure of the pre-filling settling section in the (k+1)th filling cycle and the baseline value of the liquid supply pressure in the (k+1)th filling cycle, and outputs the inter-cycle residual verification amount transferred from the kth filling cycle to the (k+1)th filling cycle. This inter-cycle residual verification amount is not used as a real-time calculation result of the internal pressure disturbance in the kth filling cycle, but is generated at the beginning of the (k+1)th filling cycle. It is specifically used to verify whether the residual pressure after the end of the kth filling cycle has been transferred to the beginning state of the (k+1)th filling cycle, reflecting the inter-cycle transfer relationship from the kth cycle to the (k+1)th cycle.

[0097] The unit of the residual verification quantity during the cycle is the same as the unit of the liquid supply pressure. It is used to verify whether the residual pressure at the end of the k-th filling cycle has been transmitted to the starting state of the (k+1)-th filling cycle. When the residual verification quantity during the cycle is positive, it indicates that the starting pressure of the standing section before filling in the (k+1)-th filling cycle is higher than the baseline value of the liquid supply pressure in the (k+1)-th filling cycle, and the control device outputs a positive residual verification mark. When the residual verification quantity during the cycle is negative, it indicates that the starting pressure of the standing section before filling in the (k+1)-th filling cycle is lower than the baseline value of the liquid supply pressure in the (k+1)-th filling cycle, and the control device outputs a negative residual verification mark. When the absolute value of the residual verification quantity during the cycle is less than the residual judgment threshold, a residual negligible mark is output.

[0098] The unit of the residual determination threshold is consistent with the unit of the liquid supply pressure. The residual determination threshold is determined based on the standard deviation of the measurement noise of the pressure acquisition device and the statistical fluctuation range of the liquid supply pressure baseline value. Specifically, it is the larger of three times the standard deviation of the no-load measurement noise of the pressure acquisition device and 1% of the liquid supply pressure baseline value. For example, the residual determination threshold is 1% to 3% of the liquid supply pressure baseline value.

[0099] The control device also calculates the difference between the pressure at the end of the recovery section after filling in the kth filling cycle and the baseline value of the supply pressure in the kth filling cycle, generating the residual pressure at the end of the recovery section for the kth filling cycle. When using the residual pressure at the end of the recovery section to predict the starting pressure state of the (k+1)th filling cycle, the control device corrects the baseline value of the supply pressure in the (k+1)th filling cycle based on the shift in the supply pressure baseline before calculating the deviation in the starting pressure. A positive residual pressure at the end of the recovery section indicates that the pressure at the end of the recovery section after filling in the kth filling cycle is higher than the baseline value of the supply pressure in the kth filling cycle; a negative residual pressure indicates that the pressure at the end of the recovery section after filling in the kth filling cycle is lower than the baseline value of the supply pressure in the kth filling cycle.

[0100] The residual pressure at the end of the recovery period is used to predict the initial pressure deviation of the (k+1)th filling cycle after the end of the kth filling cycle, while the residual verification value during the cycle is used to verify the prediction result at the beginning of the (k+1)th filling cycle. Both characterize the source of the residual during the cycle, but they do not substitute for each other, and are recorded separately in the historical pressure disturbance record table.

[0101] The control device performs exponential decay fitting on the time series of the liquid supply pressure in the recovery section after filling to obtain the pressure rebound time constant. The pressure rebound time constant characterizes the decay rate at which the pressure in the recovery section deviates from the baseline value and returns to near the baseline. The exponential decay fitting uses each sampling time in the liquid supply pressure time series as the independent variable and the difference between the corresponding sampled liquid supply pressure value and the baseline value as the objective variable. The least squares method is used to solve for the parameters of the exponential decay curve, and the time constant corresponding to the decay rate is used as the pressure rebound time constant.

[0102] All sampling times and pressure rebound time constants are expressed in seconds. The difference between the sampled supply pressure value and the baseline supply pressure value is expressed in the same unit as the supply pressure. The pressure rebound time constant is used by S33 to predict the influence of the residual pressure at the end of the recovery section on the starting pressure state of the next filling cycle.

[0103] The control device writes the residual verification amount during the cycle, the residual pressure at the end of the recovery section, the positive residual verification mark, the negative residual verification mark, the residual negligible mark, and the pressure rebound time constant into the historical pressure disturbance record table. Among them, the residual verification amount during the cycle is calculated at the beginning of the (k+1)th filling cycle and written into the historical pressure disturbance record table according to the residual source cycle number k, and associated with its affecting cycle number k+1.

[0104] The data structure of the historical pressure disturbance record table is a table with the filling cycle number as the primary key and fields including the baseline migration of the liquid supply pressure, the disturbance amplitude within the cycle, the disturbance direction marker within the cycle, the residual verification amount during the cycle, the residual pressure at the end of the recovery section, the dominant frequency of the disturbance within the cycle, the preliminary diagnosis result of the disturbance source, and the pressure rebound time constant.

[0105] S2: Calculate the filling weight deviation during the filling cycle, and cross-validate the filling weight deviation with three types of pressure disturbance characteristics to determine the source of deviation.

[0106] Specifically, the filling weight deviation is calculated based on the pre-filling weight, post-filling weight, and target filling weight of the k-th filling cycle. The filling weight deviation is then cross-validated with the baseline migration of the liquid supply pressure generated by S1, the disturbance amplitude within the cycle, the disturbance direction marker within the cycle, the residual verification amount during the cycle, and the residual pressure at the end of the recovery section. The results of deviation source identification and contribution probability distribution are then output.

[0107] Further, step S2 includes:

[0108] S21: Calculate the filling weight deviation based on the pre-filling weight, post-filling weight, and target filling weight, and construct a historical filling weight deviation sequence and time-series correlation matrix.

[0109] like Figure 6 The diagram shows a temporal correlation matrix, illustrating the relationship between filling weight deviation and pressure disturbance characteristics. Each row corresponds to a filling cycle, and each column corresponds to the baseline migration of the liquid supply pressure, the disturbance amplitude within the cycle, the disturbance direction marker within the cycle, the residual verification amount during the cycle, and the residual pressure at the end of the recovery segment, used to achieve intra-cycle alignment analysis.

[0110] Historical filling weight deviation sequences and time-series correlation matrices are used to trace the correspondence between filling weight deviations and three types of pressure disturbance characteristics during the same period. The time-series correlation matrix is ​​used to perform time-series alignment of filling weight deviations and pressure disturbance characteristics under the same filling cycle number.

[0111] The weighing device outputs the container weight before filling at the start of the k-th filling cycle and the container weight after filling at the end of the k-th filling cycle. The control device reads the target filling weight from the production batch record. The target filling weight is obtained by multiplying the target filling volume by the drug density, and the unit of the target filling weight is consistent with the unit output by the weighing device. The drug density is provided by the production batch process document or measured by a densitometer before the start of the batch, and the unit of the drug density matches the units of the target filling volume and the target filling weight.

[0112] The control device calculates the filling weight deviation for the k-th filling cycle according to the following expression:

[0113]

[0114] in, This represents the filling weight deviation in the k-th filling cycle. Let be the weight of the container after filling in the kth filling cycle. Let the container weight before filling be the weight of the container in the k-th filling cycle. The target filling weight. , , and All units are consistent with the output units of the weighing device. In the expression, the actual filled weight is obtained by subtracting the container weight before filling from the container weight after filling, and the filling weight deviation is obtained by subtracting the target filling weight from the actual filled weight. A positive value indicates that the actual filled weight is higher than the target filling weight, and a negative value indicates that the actual filled weight is lower than the target filling weight.

[0115] The control device maintains a historical filling weight deviation sequence with a length of N. N is the historical window length, determined jointly by the peristaltic pump control response time and the rate of decrease in the reservoir level. Specifically, it is the larger of the number of filling cycles required for the reservoir level change to exceed 1% of the initial batch level height and 15, but not exceeding 30. For example, N is between 15 and 30. The reservoir level change and the initial batch level height use the same length unit. The data structure of the historical filling weight deviation sequence is a one-dimensional array arranged in ascending order of filling cycle number.

[0116] The control device takes the historical filling weight deviation sequence as input and performs a first-order difference operation to obtain the first-order difference sequence of filling weight deviation; the control device also takes the historical filling weight deviation sequence as input and performs moving average smoothing to obtain the long-term trend sequence of filling weight deviation. The long-term trend sequence of filling weight deviation is a trend sequence obtained by smoothing the historical filling weight deviation sequence with a moving average, and is used to characterize the slow changing trend of filling weight deviation.

[0117] The window length for moving average smoothing is determined based on the historical window length, specifically one-third of the historical window length rounded down. For example, the window length for moving average smoothing is 5 to 10 filling cycles. The control device then uses the historical pressure disturbance record table as input to extract the baseline migration of the liquid supply pressure, the disturbance amplitude within the cycle, the disturbance direction marker within the cycle, the residual verification amount during the cycle, and the residual pressure at the end of the recovery segment according to the same filling cycle number, and constructs a time-series correlation matrix.

[0118] Each row of the time-series correlation matrix corresponds to a filling cycle, and each column corresponds to the baseline migration of the liquid supply pressure, the disturbance amplitude within the cycle, the disturbance direction marker within the cycle, the residual verification amount during the cycle, and the residual pressure at the end of the recovery segment.

[0119] S22: Perform causal consistency test on the baseline migration of the supply pressure, perform event synchronicity test on the disturbance amplitude and disturbance direction marker within the cycle, perform cycle-by-cycle mapping verification on the residual verification amount during the cycle and the residual pressure at the end of the recovery segment, and output the test results for the three types of sources.

[0120] The control device performs cross-correlation analysis on the time series of the liquid supply pressure baseline migration and the long-term trend series of the filling weight deviation. Before performing the cross-correlation analysis, the control device performs mean removal and standard deviation normalization on the time series of the liquid supply pressure baseline migration and the long-term trend series of the filling weight deviation, respectively. Then, it calculates the normalized correlation coefficients at different time lags and takes the largest normalized correlation coefficient as the baseline trend correlation coefficient.

[0121] The time lag search range is determined by the filling cycle time; for example, the time lag search range is 0 to 3 filling cycles. When the baseline trend correlation coefficient is greater than 0.80 and the maximum normalized correlation coefficient appears within the 0 to 1 filling cycle lag range, the control device outputs a baseline migration test result for the liquid supply pressure.

[0122] The control device performs an event synchronization test on the time series of disturbance amplitudes within a period and the first-order difference series of filling weight deviations. The control device first calculates the peak value judgment threshold of the disturbance amplitude within a period. For example, the peak value judgment threshold of the disturbance amplitude within a period is the average disturbance amplitude within the historical period plus three times the standard deviation of the disturbance amplitude within the historical period; for example, when the average disturbance amplitude within the historical period is 0.12 MPa and the standard deviation is 0.03 MPa, the peak value judgment threshold of the disturbance amplitude within a period is 0.21 MPa.

[0123] Next, it is determined whether the filling cycle in which the disturbance amplitude exceeds the peak value judgment threshold of the disturbance amplitude within the cycle is consistent with the abrupt change cycle of the first-order difference value of the filling weight deviation. For example, the abrupt change cycle of the filling weight deviation is the filling cycle in which the absolute value of the first-order difference value of the filling weight deviation is greater than the average absolute value of the first-order difference value of the historical filling weight deviation plus 3 times the standard deviation of the absolute value of the first-order difference value of the historical filling weight deviation; for example, when the average absolute value of the first-order difference value of the historical filling weight deviation is 0.08g and the standard deviation is 0.04g, the abrupt change judgment threshold of the filling weight deviation is 0.20g.

[0124] When the two types of cycles are consistent and the direction of the disturbance direction marker within the cycle matches the direction of the first-order difference value of the filling weight deviation, the control device outputs a result indicating that the disturbance test within the cycle has passed. The disturbance amplitude within the cycle is used to determine whether the disturbance intensity exceeds the threshold, and the disturbance direction marker within the cycle is used to determine whether the disturbance direction matches the direction of change of the filling weight deviation.

[0125] The control device reads the periodic residual verification amount affecting the k-th filling cycle from the historical pressure disturbance record table, and performs cycle-by-cycle mapping verification between the periodic residual verification amount and the filling weight deviation of the k-th filling cycle. Based on the mapping relationship between the periodic residual verification amount obtained during the equipment calibration phase and the filling weight deviation of the affected filling cycle, the control device determines whether the deviation direction corresponding to the periodic residual verification amount is consistent with the filling weight deviation direction of the k-th filling cycle, and counts the number of consecutive consistent deviations.

[0126] The unit of the residual verification amount during the week is consistent with the unit of the liquid supply pressure, and the unit of the filling weight deviation is consistent with the unit output by the weighing device. When the number of consecutive consistent values ​​exceeds the number of residual mapping verifications, the control device outputs a result indicating that the residual inspection during the week has passed.

[0127] The number of residual mapping verifications is determined by the residual verification samples obtained during the equipment calibration phase. The minimum number of consecutive consistent cycles that can stably characterize the deviation direction of the residual verification quantity during the period is taken and remains unchanged within the same production batch.

[0128] The control device writes the results of the baseline migration test for the liquid supply pressure, the disturbance test within the cycle, and the residue test within the cycle into the source test result table. The data structure of the source test result table is a table with the filling cycle number as the key and three Boolean values ​​and corresponding test intensity values ​​as fields. The test intensity values ​​are the baseline trend correlation coefficient, the peak disturbance exceeding the limit within the cycle, and the number of consecutive consistent residue mappings.

[0129] S23: Based on the source test result table and the three types of pressure disturbance characteristic quantities, construct a deviation source probability identification model and calculate the contribution probability distribution.

[0130] like Figure 7 The diagram shows a schematic of the deviation source probability identification model, illustrating its structure. The model takes three types of pressure disturbance characteristics and source verification results as input. The logic inference layer updates the posterior probability of each deviation source, ultimately outputting the probability of supply pressure baseline migration, intra-cycle disturbance, and intra-cycle residual contribution, providing a basis for compensation decisions.

[0131] The control device defines a probabilistic identification model for deviation sources. This model is a probabilistic inference process consisting of an input layer, a logic inference layer, and an output layer, and is not limited to a neural network structure. The input layer receives the baseline shift of the supply pressure output by S12, the intra-cycle disturbance amplitude and intra-cycle disturbance direction marker output by S13, the intra-cycle residual verification quantity and the residual pressure at the end of the recovery segment output by S14, and the source verification result table output by S22. The logic inference layer updates the prior probabilities of the three types of deviation sources based on the source verification result table.

[0132] The output layer outputs the probability of contribution of the supply pressure baseline migration, the probability of contribution of intra-cycle disturbance, and the probability of contribution of residual during the cycle. Among them, the probability of contribution of supply pressure baseline migration corresponds to the source of supply pressure baseline migration deviation, the probability of contribution of intra-cycle disturbance corresponds to the source of intra-cycle disturbance deviation, and the probability of contribution of residual during the cycle corresponds to the source of residual deviation during the cycle.

[0133] The control device sets the prior probabilities of three types of deviation sources based on historical production data of the same model of equipment; if historical production data is unavailable, the initial prior probabilities of the three types of deviation sources are all set to 1 / 3.

[0134] The control device updates the posterior probabilities of the three types of deviation sources based on the results of the baseline migration test of the supply pressure, the per-cycle disturbance test, and the per-cycle residue test. When the test passes, the likelihood value of the corresponding deviation source is the result of linear normalization of the corresponding test intensity value; when the test fails, the likelihood value of the corresponding deviation source is set to 0.1.

[0135] The upper and lower bounds of linear normalization are determined by the minimum and maximum inspection intensities of the most recent N filling cycles. When the maximum and minimum inspection intensities of the most recent N filling cycles are equal, the control device uses a preset likelihood value to participate in the posterior probability update. The control device multiplies the prior probability by the likelihood value and sums and normalizes the three types of products to obtain the probability of contribution of the liquid supply pressure baseline migration, the probability of contribution of intra-cycle disturbance, and the probability of contribution of contribution of intra-cycle residue. The sum of the three types of contribution probabilities is 1.

[0136] The preset likelihood value is determined by testing samples from historical production data of the same model of equipment and remains unchanged within the same production batch.

[0137] The control device writes the contribution probability of the supply pressure baseline migration, the contribution probability of the intra-cycle disturbance, and the contribution probability of the intra-cycle residual to the contribution probability record table. The data structure of the contribution probability record table is a table with the filling cycle number as the key and the contribution probability of the supply pressure baseline migration, the contribution probability of the intra-cycle disturbance, and the contribution probability of the intra-cycle residual as the values. The contribution probability record table is used in S24 to output the deviation source identification results and in S31 to S34 to generate compensation vectors.

[0138] S24: Output the source identification results and identification confidence index based on the contribution probability distribution.

[0139] The control device reads the contribution probability of the liquid supply pressure baseline migration, the contribution probability of the disturbance within the cycle, and the contribution probability of the residual during the cycle for the k-th filling cycle from the contribution probability record table, and selects the deviation source corresponding to the largest contribution probability as the dominant deviation source. When the contribution probability of a source other than the dominant deviation source is greater than the secondary deviation source threshold, the control device marks the corresponding deviation source as a secondary deviation source. The secondary deviation source threshold is determined based on the minimum identifiable contribution ratio of secondary sources to filling weight deviation in historical composite disturbance batches; for example, the secondary deviation source threshold is set to 0.2.

[0140] The control device calculates the ratio between the probability of the dominant source of deviation and the probability of the second largest source of deviation, and outputs an identification confidence index. When the probability of the second largest source of deviation is less than a preset minimum value, the control device uses the preset minimum value in the ratio calculation to avoid an undefined ratio. When the identification confidence index is greater than a confidence threshold, the control device outputs a high confidence flag; when the identification confidence index is less than or equal to the confidence threshold, the control device outputs a low confidence flag. The confidence threshold is determined based on the lower quartile of the ratio of the probability of the dominant source to the probability of the second largest source in historical correctly identified samples; for example, the confidence threshold is set to 2.

[0141] The preset minimum value is the minimum positive contribution probability that the deviation source probability identification model is allowed to participate in the calculation of the identification confidence index, and its value is determined according to the data precision of the contribution probability record table.

[0142] The control device outputs the primary deviation source, secondary deviation source, high confidence mark or low confidence mark, as well as the probability of contribution of the liquid supply pressure baseline migration, the probability of contribution of the disturbance during the cycle and the probability of the residual contribution during the cycle as deviation source identification results to S3.

[0143] S3: Based on the results of the deviation source identification, obtain the comprehensive filling control quantity for the next filling cycle.

[0144] Specifically, based on the deviation source identification results and contribution probability distribution of the S2 output, an adaptive compensation vector for the liquid supply pressure baseline migration, an intra-cycle disturbance isolation compensation vector, and an inter-cycle residual decoupling compensation vector are generated, and then fused to generate the comprehensive filling control quantity for the next filling cycle. Before fusion, all three types of compensation vectors are converted into positive equivalent filling control quantity directions; the positive equivalent filling control quantity refers to the control quantity whose value, after being mapped by the control direction, corresponds to an increase in actual filling volume when its value increases, and a decrease in its value when its value decreases when its value decreases.

[0145] Further, step S3 includes:

[0146] S31: Generate an adaptive compensation vector for the migration of the liquid supply pressure baseline based on the probability of contribution of the liquid supply pressure baseline migration, the predicted value of the flow deviation, and the long-term trend sequence of the filling weight deviation.

[0147] The supply pressure baseline shift originates from a drop in the liquid level in the storage tank or a slow drift in the supply pressure reference, and its impact is continuous. Compensating for this shift requires both utilizing the output flow deviation prediction value to reflect the physical transmission relationship and using the long-term trend sequence of filling weight deviation to resist drift in the transmission relationship parameters. This step employs a fusion of physical prediction and statistical trend analysis to ensure that the adaptive compensation vector for the supply pressure baseline shift can be smoothly updated as the liquid level in the storage tank decreases.

[0148] When the probability of contribution of the supply pressure baseline migration output by S24 is greater than the trend source threshold, the control device initiates the supply pressure baseline migration compensation calculation. The trend source threshold is determined based on the probability distribution of contribution when the supply pressure baseline migration is identified as the dominant source in historical production batches, specifically by taking the lower quartile of the probability distribution. For example, the trend source threshold is 0.5.

[0149] The control device takes the predicted output flow rate deviation as input and calculates the direct pressure compensation based on the peristaltic pump flow rate-speed characteristic curve. The peristaltic pump flow rate-speed characteristic curve is obtained during the equipment commissioning phase through standard weighing and filling experiments at different speeds. The data structure is a calibration table with the peristaltic pump speed as the independent variable and the filling mass per unit time as the dependent variable; where the unit of peristaltic pump speed is revolutions per minute, and the unit of filling mass per unit time is grams per second or kilograms per second.

[0150] If the predicted flow rate deviation is expressed as a volumetric flow rate deviation, the control device first converts it into a mass flow rate deviation based on the drug solution density, and then matches it with the peristaltic pump flow rate-speed characteristic curve. Based on the current control mode, the control device converts the predicted flow rate deviation into peristaltic pump speed increments, peristaltic pump running time increments, or pulse count increments, generating a direct pressure compensation value.

[0151] Specifically, when the current control mode is speed control, the control device converts the predicted flow deviation value into speed increment based on the peristaltic pump flow-speed characteristic curve; when the current control mode is running time control, the control device combines the target filling weight, the baseline filling flow rate, and the baseline running time to convert the predicted flow deviation value into running time increment; when the current control mode is pulse number control, the control device combines the unit pulse filling weight or the unit pulse filling volume to convert the predicted flow deviation value into pulse number increment.

[0152] The direction of the direct pressure compensation is opposite to the direction of the flow deviation prediction. When the flow deviation prediction indicates that the actual flow rate is higher than the target flow rate, the direct pressure compensation is to reduce the peristaltic pump speed, shorten the peristaltic pump running time, or reduce the number of pulses. When the flow deviation prediction indicates that the actual flow rate is lower than the target flow rate, the direct pressure compensation is to increase the peristaltic pump speed, extend the peristaltic pump running time, or increase the number of pulses. The units of the converted direct pressure compensation correspond to the current control mode, namely revolutions per minute, seconds, or pulses.

[0153] The control device takes the long-term trend sequence of filling weight deviation output by S21 as input, fits a linear trend model, and extrapolates the predicted value of the trend filling weight deviation for the (k+1)th filling cycle. The predicted value of the trend filling weight deviation refers to the predicted deviation value for the (k+1)th filling cycle obtained by extrapolation based on the long-term trend sequence of filling weight deviation.

[0154] The linear trend model uses the filling cycle number as the independent variable and the long-term trend sequence of filling weight deviation as the objective variable, employing the least squares method to solve for the slope and intercept. The unit of the predicted trend filling weight deviation is consistent with the unit output by the weighing device. The control device divides the predicted trend filling weight deviation by the drug density to obtain the predicted filling volume deviation, and the unit of the predicted filling volume deviation is consistent with the volume unit corresponding to the drug density.

[0155] The control device then combines the peristaltic pump flow-speed characteristic curve and the current control mode to convert the predicted filling volume deviation or its corresponding mass deviation into peristaltic pump speed increment, peristaltic pump running time increment, or pulse number increment, generating an indirect trend compensation quantity. Specifically, when converting to running time increment, the control device uses the baseline filling mass flow rate or baseline filling volume flow rate as the conversion benchmark; when converting to speed increment, the control device uses the baseline filling running time and peristaltic pump flow-speed characteristic curve as the conversion benchmark; and when converting to pulse number increment, the control device uses the unit pulse filling weight or unit pulse filling volume as the conversion benchmark.

[0156] The control device determines the fusion weights of direct pressure compensation and indirect trend compensation based on the identification confidence index. A higher identification confidence index results in a larger weight for direct pressure compensation, while a lower index results in a larger weight for indirect trend compensation. The fusion yields an adaptive compensation vector for the supply pressure baseline migration. The update rate of this vector is determined by the output supply pressure baseline migration rate; a higher migration rate results in a higher update rate, and vice versa.

[0157] S32: Based on the probability of disturbance contribution within the cycle and the filling weight deviation, generate a disturbance isolation compensation vector within the cycle and restrict the transmission of disturbances within the cycle to subsequent filling cycles.

[0158] Intra-cycle disturbances typically originate from air bubble ingress, short-term increased pulsation of the peristaltic pump rollers, or mechanical vibration of the supply line, exhibiting sporadic and directional instability. If the filling weight deviation corresponding to the intra-cycle disturbance is fully propagated to the next filling cycle, the next filling cycle may still be compensated for after the disturbance has subsided, resulting in a reverse error. This step limits the compensation propagation range of intra-cycle disturbances through attenuation isolation, direction reversal identification, and fading judgment, ensuring that the intra-cycle disturbance isolation compensation vector does not enter the historical update data of the supply pressure baseline migration adaptive compensation vector.

[0159] When the probability of the intra-period disturbance contribution output by S24 is greater than the transient source threshold, the control device initiates the intra-period disturbance isolation compensation calculation. The transient source threshold is determined based on the probability distribution of the contribution of disturbances identified as the dominant source in historical disturbance samples. For example, the transient source threshold is set to 0.4.

[0160] The control device determines the disturbance attenuation factor based on the probability of disturbance contribution within a period. The disturbance attenuation factor is inversely related to the probability of disturbance contribution within a period, and the inverse relationship adopts a linear decreasing model. The calculation formula is: the disturbance attenuation factor equals the maximum attenuation factor minus the product of the attenuation coefficient and the probability of disturbance contribution within a period, where the maximum attenuation factor is the upper limit of the disturbance attenuation factor, and the attenuation coefficient controls the attenuation rate.

[0161] The above formula guarantees that: the greater the probability of the disturbance contribution within a period, the smaller the disturbance attenuation factor; and the smaller the probability of the disturbance contribution within a period, the larger the disturbance attenuation factor. The maximum attenuation factor and the attenuation coefficient are determined based on the number of historical transient disturbance decay periods. For example, the maximum attenuation factor is taken as 0.3 and the attenuation coefficient is taken as 0.2, so that the disturbance attenuation factor ranges from 0.1 to 0.3.

[0162] The control device multiplies the output filling weight deviation by a disturbance attenuation factor to obtain the attenuated weight deviation to be compensated. The unit of the weight deviation to be compensated is consistent with the output unit of the weighing device. Based on the current control mode, and in conjunction with the peristaltic pump flow-speed characteristic curve, the reference filling flow rate, and the reference filling control quantity, the control device converts the attenuated weight deviation to be compensated into the peristaltic pump speed increment, peristaltic pump running time increment, or pulse number increment, generating a disturbance isolation compensation vector within the cycle.

[0163] Specifically, when converting to runtime increments, the control device uses the baseline filling flow rate as the conversion benchmark; when converting to speed increments, the control device uses the baseline filling runtime and the peristaltic pump flow-speed characteristic curve as the conversion benchmark; and when converting to pulse count increments, the control device uses the unit pulse filling weight or unit pulse filling volume as the conversion benchmark. Before output, the disturbance isolation compensation vector within the cycle is uniformly represented according to the positive equivalent filling control quantity direction defined in S3.

[0164] The control device determines whether the disturbance amplitude exceeds the disturbance amplitude threshold for two consecutive filling cycles, and whether the signs of the filling weight deviations for two consecutive filling cycles are opposite. For example, the disturbance amplitude threshold is set to 0.20 MPa. When the disturbance amplitude is greater than 0.20 MPa for two consecutive filling cycles, and the signs of the filling weight deviations for two consecutive filling cycles are opposite, the control device determines that a disturbance direction reversal has occurred and sets the in-cycle disturbance isolation compensation vector to zero.

[0165] The control device determines whether the disturbance amplitude within three consecutive filling cycles is below the disturbance amplitude threshold and whether the absolute value of the filling weight deviation shows a decreasing trend. For example, the disturbance amplitude threshold is set to 0.20 MPa. When the disturbance amplitude within three consecutive filling cycles is less than 0.20 MPa and the absolute value of the filling weight deviation shows a decreasing trend, the control device determines that the disturbance within the cycle has subsided and allows the next filling cycle to regenerate the disturbance isolation compensation vector within the cycle.

[0166] The intra-cycle disturbance isolation compensation vector only applies to the next immediately following filling cycle each time, and its application period is limited to the second half of the filling execution segment. It is not cumulatively transferred across cycles and is not written into the historical update data of the liquid supply pressure baseline migration adaptive compensation vector. Whether the intra-cycle disturbance isolation compensation vector is generated again in subsequent filling cycles is determined based on the intra-cycle disturbance test results obtained again in the corresponding cycle.

[0167] S33: Based on the residual contribution probability during the cycle, the residual pressure at the end of the recovery section, and the pressure rebound time constant, generate the residual decoupling compensation vector during the cycle, and perform pre-correction of the initial pressure state before the start of the next filling cycle or at the beginning of the start.

[0168] The initial pressure state pre-correction refers to adjusting the initial control quantity of the peristaltic pump, the pre-filling rest time, or the low-speed pre-filling action based on the predicted initial pressure deviation before formal filling in the k+1th filling cycle or at the beginning of filling.

[0169] The residual pressure at the end of the recovery section is not a random disturbance within the current filling cycle, but rather a residual pressure state formed at the end of the recovery section after incomplete recovery in the previous filling cycle. If the deviation of the initial pressure corresponding to the residual pressure at the end of the recovery section is not addressed before or at the beginning of the next filling cycle, an excessively high or low initial pressure in the liquid supply line will change the actual liquid output of the peristaltic pump in the initial stage, causing the filling volume deviation to be amplified in the next filling cycle. This step uses a combination of prediction and actual measurement correction to handle cross-cycle pressure coupling.

[0170] When the probability of residual contribution during the cycle output by S24 is greater than the residual source threshold, the control device initiates the decoupling compensation calculation for residual during the cycle. The residual source threshold is determined based on the contribution probability distribution of residuals identified as the dominant source in the residual pressure samples at the end of the historical recovery segment and the residual verification samples during the cycle. For example, the residual source threshold is set to 0.3.

[0171] The control device uses the generated residual pressure at the end of the recovery section and the output pressure rebound time constant as inputs to predict the initial pressure deviation of the pre-filling resting section in the (k+1)th filling cycle. During prediction, the control device calculates the residual pressure's remaining influence based on the time interval between the end of the recovery section after filling in the kth filling cycle and the start of the pre-filling resting section in the (k+1)th filling cycle, combined with the pressure rebound time constant. This remaining influence is then used as the initial pressure deviation for the (k+1)th filling cycle. The initial pressure deviation is converted into the initial flow rate deviation through the level, pressure, and flow rate transmission relationship.

[0172] The control device calculates the pre-correction control quantity for the peristaltic pump startup phase based on the initial flow deviation. If the predicted initial pressure deviation is positive, the control device reduces the peristaltic pump filling control quantity at the beginning of the next filling cycle, or extends the pre-filling settling time. If the predicted initial pressure deviation is negative, the control device increases the peristaltic pump filling control quantity at the beginning of the next filling cycle, or controls the peristaltic pump to perform a low-speed pre-filling action.

[0173] The rotational speed of the low-speed pre-charge fluid action is taken as 10% to 30% of the rated speed of the peristaltic pump, and the duration is determined by the initial pressure deviation and the low-speed pre-charge fluid pressure recovery rate. The unit of the initial pressure deviation is the same as the unit of the supply pressure, the unit of the low-speed pre-charge fluid pressure recovery rate is pressure units per second, and the calculated duration is in seconds.

[0174] Before the end of the pre-filling resting period and the start of the filling execution period in the (k+1)th filling cycle, the control device measures the actual liquid supply pressure during the pre-filling resting period and compares it with the pressure value corresponding to the predicted initial pressure deviation to obtain the prediction error. The unit of the prediction error is consistent with the unit of the liquid supply pressure. Based on the prediction error, the control device adjusts the real-time control quantity of the peristaltic pump at the beginning of the filling execution period and outputs the residual decoupling compensation vector during the cycle. Before being output, the residual decoupling compensation vector during the cycle is uniformly represented according to the positive equivalent filling control quantity direction defined in S3 and only preferentially acts on the start-up stage of the filling execution period.

[0175] If any compensation vector fails to meet the corresponding compensation trigger condition, the control device sets the compensation vector to zero.

[0176] S34: Integrate the adaptive compensation vector for baseline migration of liquid supply pressure, the intra-cycle disturbance isolation compensation vector, and the inter-cycle residual decoupling compensation vector to generate the comprehensive filling control quantity for the (k+1)th filling cycle, and perform priority arbitration.

[0177] like Figure 8 The diagram shown illustrates the comprehensive filling control quantity, illustrating its generation process. The system unifies the supply pressure baseline migration adaptive compensation vector, the intra-cycle disturbance isolation compensation vector, and the inter-cycle residual decoupling compensation vector to the same control dimension, and then weights and fuses them based on their contribution probabilities. Finally, it combines these with the baseline filling control quantity to form the control setting for the next cycle.

[0178] The three types of compensation vectors originate from different physical mechanisms. If they are directly superimposed without considering the contribution probability and the timing of their action, over-adjustment may occur during the start-up phase of the filling process, or unnecessary transient compensation may be introduced during the steady-state phase of filling. This step uses the contribution probability to determine the fusion weight and uses the dominant deviation source to determine the timing of the action, so that the compensation amount matches the disturbance source in the time dimension.

[0179] The control unit reads the baseline filling control quantity from the equipment process parameter table. The baseline filling control quantity is the peristaltic pump speed, peristaltic pump running time, or number of pulses required to achieve the target filling weight under pressure disturbance-free calibration conditions. The control unit takes the adaptive compensation vector for supply pressure baseline migration output by S31, the intra-cycle disturbance isolation compensation vector output by S32, and the inter-cycle residual decoupling compensation vector output by S33 as inputs, and first converts the three types of compensation vectors to the same control dimension as the baseline filling control quantity.

[0180] If the baseline filling control quantity is the peristaltic pump running time, then all three types of compensation vectors are converted into running time increments in seconds; if the baseline filling control quantity is the peristaltic pump speed, then all three types of compensation vectors are converted into speed increments in revolutions per minute; if the baseline filling control quantity is the number of pulses, then all three types of compensation vectors are converted into pulse number increments in pulses. After the conversion, the control device generates the comprehensive filling control quantity for the (k+1)th filling cycle according to the following expression:

[0181]

[0182] in, This represents the comprehensive filling control quantity for the (k+1)th filling cycle. As the baseline filling control quantity, This is an adaptive compensation vector for the baseline migration of the supply pressure. This is the disturbance isolation compensation vector within the period. This is the residual decoupling compensation vector during the cycle. , and Used only as formula symbols, the full names of the fluid supply pressure baseline migration adaptive compensation vector, the intra-cycle disturbance isolation compensation vector, and the inter-cycle residual decoupling compensation vector are used in the text description. , , , and They use the same control dimension, the same positive and negative directions, and the same units, which are seconds, revolutions per minute, or pulses. , and These are the fusion weight coefficients for the adaptive compensation vector for the supply pressure baseline migration, the intra-cycle disturbance isolation compensation vector, and the inter-cycle residual decoupling compensation vector, respectively. The initial fusion weight coefficients are obtained by normalizing the corresponding contribution probabilities.

[0183] Since the sum of the three contribution probabilities output by S23 is 1, the initial fusion weight coefficients are equal to the corresponding contribution probabilities when priority arbitration, conservative compensation strategy, or closed-loop self-verification correction is not executed. When priority arbitration, conservative compensation strategy, or closed-loop self-verification parameter updates are subsequently executed, the control device adjusts the initial fusion weight coefficients according to the corresponding weight correction coefficients and renormalizes the three types of fusion weight coefficients after adjustment.

[0184] The weight correction coefficient is an adjustment applied to the initial fusion weight coefficient based on the multi-layer closed-loop self-verification result output by S44.

[0185] The design principle of the expression is as follows: the baseline filling control quantity provides the basic liquid discharge capacity under undisturbed conditions; the three types of compensation vectors respectively correct the direction of the gradual trend, transient disturbance and cross-cycle residue; the weighted summation represents the incremental contribution of the three types of physical sources to the same peristaltic pump control quantity; and the fusion weight coefficient makes the source with the greater probability of contribution account for a higher proportion in the comprehensive filling control quantity.

[0186] For example, if the baseline filling control quantity is the peristaltic pump running time of 2.000 seconds, the adaptive compensation vector for the supply pressure baseline migration is 0.060 seconds, the intra-cycle disturbance isolation compensation vector is -0.010 seconds, the intra-cycle residual decoupling compensation vector is 0.020 seconds, the contribution probability of the supply pressure baseline migration is 0.60, the intra-cycle disturbance contribution probability is 0.15, and the intra-cycle residual contribution probability is 0.25, then the comprehensive filling control quantity for the (k+1)th filling cycle is 2.000 + 0.60 × 0.060 + 0.15 × (-0.010) + 0.25 × 0.020 = 2.0395 seconds. The control device uses 2.0395 seconds as the setpoint for the peristaltic pump running time in the (k+1)th filling cycle.

[0187] The control device performs priority arbitration. When the probability of residual contribution during the cycle is the maximum among the three types of contribution probabilities, the residual decoupling compensation vector during the cycle is preferentially applied to the start-up stage of the filling execution segment, while the adaptive compensation vector for the liquid supply pressure baseline migration is delayed until the middle and later stages of the filling execution segment. When the probability of liquid supply pressure baseline migration contribution is the maximum among the three types of contribution probabilities, the adaptive compensation vector for liquid supply pressure baseline migration runs through the entire filling execution segment, and the intra-cycle disturbance isolation compensation vector and the intra-cycle residual decoupling compensation vector are superimposed as fine-tuning quantities. When the intra-cycle disturbance contribution probability is the maximum among the three types of contribution probabilities, the control device reduces the single-cycle compensation amplitude of the intra-cycle disturbance isolation compensation vector and applies conservative compensation correction to the initial fusion weight coefficients of the three types of compensation vectors, adopting a conservative compensation strategy.

[0188] The conservative compensation strategy refers to reducing the single-cycle compensation amplitude, limiting the disturbance isolation compensation vector within a cycle to only apply to the immediately following filling cycle, and prohibiting the compensation result from being written into the historical update data of the supply pressure baseline migration adaptive compensation vector. The reduction ratio of the single-cycle compensation amplitude is determined based on historical transient disturbance compensation overadjustment samples; for example, the reduction ratio is 20% to 50%. When the probability of disturbance contribution within a cycle is lower than the transient source threshold for multiple consecutive filling cycles, and the absolute value of the filling weight deviation shows a decreasing trend, the control device releases the conservative compensation strategy.

[0189] The control device applies a saturation limit to the overall filling control quantity. The adjustment range of the overall filling control quantity relative to the baseline filling control quantity must not exceed the compensation saturation threshold. The compensation saturation threshold is determined based on the target filling weight allowable error and the peristaltic pump flow rate-speed characteristic curve. Specifically, the target filling weight allowable error is converted into the peristaltic pump control quantity allowable variation range. For example, the compensation saturation threshold is taken as ±15% of the baseline filling control quantity. The overall filling control quantity after saturation limitation is output to S4.

[0190] S4: Determine the fluid dynamics state of the liquid supply line before executing the comprehensive filling control quantity; if the filling start-up conditions are not met, rebuild the initial pressure state of the liquid supply line; if the filling start-up conditions are met, execute segmented peristaltic pump filling control according to the comprehensive filling control quantity; after filling is completed, execute closed-loop parameter update.

[0191] Specifically, before the start of the (k+1)th filling cycle, the fluid dynamics state of the liquid supply pipeline is judged, and the initial pressure state reconstruction strategy of the liquid supply pipeline or the segmented peristaltic pump filling control is executed according to the judgment result. After filling is completed, multi-layer closed-loop self-verification is performed, and the parameters of the deviation source probability identification model and the compensation algorithm coefficients are updated according to the self-verification results.

[0192] The comprehensive filling control output of S3 can only be accurately converted into the actual filling volume when the initial state of the liquid supply line is stable. If the initial pressure of the liquid supply line is still in a state of excessively high residual pressure, excessively low residual pressure, or continuous oscillation, directly starting the peristaltic pump will cause the comprehensive filling control to act on incorrect initial conditions, resulting in compensation failure. This step judges the fluid dynamic state of the liquid supply line before filling and performs initial pressure state reconstruction of the liquid supply line when the filling start conditions are not met; it performs segmented control during filling and updates the identification and compensation parameters in reverse with the weighing results after filling, forming an online adaptive closed loop.

[0193] Further, step S4 includes:

[0194] S41: Before the start of the (k+1)th filling cycle, collect the liquid supply pressure of the static section before filling and determine whether the fluid dynamics of the liquid supply pipeline meets the filling start-up conditions.

[0195] The fluid dynamics state of the supply pipeline refers to the initial pressure conditions and pressure recovery characteristics of the supply pipeline before the start of the filling cycle, which includes three aspects: initial pressure amplitude, pressure time-domain stability, and pressure rebound time constant. The initial pressure amplitude determines the actual liquid output in the initial stage of peristaltic pump startup; pressure time-domain stability reflects whether the supply pipeline is still oscillating; and the pressure rebound time constant reflects whether the viscoelasticity and fluid viscosity of the supply pipeline deviate from the calibrated state. These three aspects together determine whether the comprehensive filling control quantity can be directly executed.

[0196] The control device collects the time series of liquid supply pressure during the pre-filling settling period in the (k+1)th filling cycle and calculates the average liquid supply pressure during this period. Using the obtained baseline value of the liquid supply pressure in the (k+1)th filling cycle as the center, the control device constructs an allowable initial pressure range. The upper limit of the allowable initial pressure range is the baseline value of the liquid supply pressure plus the positive pressure tolerance, and the lower limit is the baseline value of the liquid supply pressure minus the negative pressure tolerance.

[0197] The positive and negative pressure tolerances are calculated by inversely converting the preset filling error threshold using the relationship between liquid level, pressure, and flow rate. The preset filling error threshold is determined based on the container specifications and is taken as a positive value. The preset filling error threshold for small-sized glass bottles is 1% to 2% of the target filling weight, and the preset filling error threshold for 50 ml to 100 ml liquid bottles is 2% of the target filling weight. The unit of the preset filling error threshold is consistent with the unit of the target filling weight.

[0198] The control device calculates the standard deviation of the liquid supply pressure time series during the pre-filling settling period and determines whether the standard deviation is less than a pressure stability threshold. The pressure stability threshold is determined based on the standard deviation of the no-load noise from the pressure acquisition device and the standard deviation of the pressure during the normal, stable pre-filling settling period; specifically, it is the larger of the two values ​​multiplied by 2. The unit of the pressure stability threshold is consistent with the unit of the liquid supply pressure; for example, the pressure stability threshold corresponds to 0.5% to 1% of the baseline value of the liquid supply pressure.

[0199] The control device reads the output pressure rebound time constant and compares it with the standard time constant initially calibrated by the system. The standard time constant is obtained during the equipment installation and commissioning phase through a pressure decay experiment in the recovery section after filling under standard chemical solutions, standard temperatures, and new supply pipeline conditions. If the deviation of the pressure rebound time constant from the standard time constant exceeds a time constant deviation threshold, the control device determines that the viscoelastic state of the supply pipeline has deviated. The time constant deviation threshold is determined based on the fluctuation range of the pressure rebound time constant of historical qualified batches; for example, the time constant deviation threshold is set to 20%.

[0200] When the average liquid supply pressure in the static section before filling is within the allowable starting pressure range, the standard deviation of the liquid supply pressure time series in the static section before filling is less than the pressure stability threshold, and the deviation ratio of the pressure rebound time constant does not exceed the time constant deviation threshold, the control device outputs a filling start allow flag; otherwise, it outputs a liquid supply pipeline starting pressure state reconstruction flag.

[0201] S42: When S41 outputs the liquid supply line initial pressure state reconstruction flag, perform liquid supply line initial pressure state reconstruction, oscillation suppression, or parameter recalibration.

[0202] Reconstructing the initial pressure of the supply line is not a post-filling correction of deviations in filling weight, but rather a process of bringing the initial pressure of the supply line back within a controllable range before the peristaltic pump begins filling. Positive residue, negative residue, pressure oscillation, and rebound parameter drift correspond to different physical causes and require different reconstruction actions. Uniformly extending the settling period or uniformly increasing the peristaltic pump speed may exacerbate another type of anomaly.

[0203] When the average supply pressure during the pre-filling settling period exceeds the upper limit of the allowable initial pressure range, the control device determines that positive residue is causing the initial pressure to be too high. The control device first extends the pre-filling settling time to the maximum settling duration. The maximum settling duration is determined based on the allowable waiting time of the production cycle; for example, the maximum settling duration is 3 to 5 seconds. If, after extending the pre-filling settling time, the average supply pressure during the pre-filling settling period still exceeds the upper limit of the allowable initial pressure range, the control device opens the pressure relief valve in the supply line to perform a small pressure relief. The pressure relief flow rate is determined based on the rated flow rate of the peristaltic pump, and the pressure relief flow rate uses the same unit as the rated flow rate of the peristaltic pump. For example, the pressure relief flow rate is 1% to 3% of the rated flow rate of the peristaltic pump.

[0204] When the average supply pressure during the pre-filling settling period is lower than the lower limit of the allowable starting pressure range, the control device determines that negative residue has caused the starting pressure to be too low. The control device controls the peristaltic pump to perform a low-speed pre-filling operation at 10% to 30% of its rated speed, and monitors the supply pressure recovery rate in real time during the low-speed pre-filling operation. When the supply pressure recovery rate is lower than the rate threshold, the control device stops the low-speed pre-filling operation and re-enters the pre-filling settling period. The rate threshold is determined based on the normal distribution of the pressure recovery rate at the end of the low-speed pre-filling stage, and the rate threshold and pressure recovery rate use the same unit, which is pressure units per second. For example, the rate threshold is taken as 10% of the average pressure recovery rate at the end of the normal low-speed pre-filling stage.

[0205] When the standard deviation of the liquid supply pressure time series during the pre-filling settling period is not less than the pressure stability threshold, the control device performs a fast Fourier transform on the liquid supply pressure time series during the pre-filling settling period to identify the oscillation frequency.

[0206] The dominant oscillation frequency is the frequency of unstable oscillation of the liquid supply pressure in the pre-filling static section, which is different from the periodic disturbance frequency corresponding to the high-frequency pressure signal in the filling execution section in S13. The control device controls the peristaltic pump to perform micro-amplitude reciprocating motion based on the dominant oscillation frequency. The displacement amplitude of the micro-amplitude reciprocating motion is determined based on the volumetric elastic coefficient of the liquid supply pipeline. For example, the single discharge volume corresponding to the micro-amplitude reciprocating motion does not exceed 0.5% of the target filling volume. During the micro-amplitude reciprocating motion, the control device continuously collects the liquid supply pressure time series in the pre-filling static section. When the standard deviation of the liquid supply pressure time series decreases below the pressure stability threshold, the micro-amplitude reciprocating motion is stopped.

[0207] When the deviation of the pressure rebound time constant exceeds the time constant deviation threshold, the control device performs a probing filling. The probing filling refers to a low-risk test filling action performed without using the normal full-volume filling control quantity to re-acquire the pressure decay curve and verify the pressure rebound time constant. The probing filling uses 80% of the baseline filling control quantity as the peristaltic pump control quantity, and in the recovery phase after filling, the pressure sampling frequency is increased to above 200 Hz to re-acquire the liquid supply pressure decay curve and refit the pressure rebound time constant.

[0208] If the deviation of the refitted pressure rebound time constant from the standard time constant exceeds the maintenance alarm threshold, the control device outputs a maintenance warning signal for the fluid supply line. The maintenance alarm threshold is determined based on the boundary of the pressure rebound time constant change during the fluid supply line aging test; for example, the maintenance alarm threshold is set to 30%.

[0209] S43: When S41 outputs the filling start permission flag, the segmented peristaltic pump filling control is executed according to the comprehensive filling control quantity output by S34, and the liquid supply pressure status is monitored in real time.

[0210] The peristaltic pump filling process is not a single process with constant pressure. The startup phase involves fluid inertia establishment, the steady-state phase involves flow maintenance, and the ending phase involves pressure rebound triggering. If the overall filling control quantity is applied evenly throughout the entire filling process, the timing of the residual decoupling compensation vector and the adaptive compensation vector for the supply pressure baseline migration cannot be distinguished. This step distributes the overall filling control quantity to different execution sub-segments.

[0211] The control device divides the filling execution phase into a filling start-up phase, a filling steady-state phase, and a filling end phase. The filling start-up phase continues until the slope of the liquid supply pressure rise curve falls below the steady-state determination slope threshold. Both the liquid supply pressure rise curve slope and the steady-state determination slope threshold are measured in pressure units per second. The steady-state determination slope threshold is determined based on the slope distribution of the pressure rise curve during the normal filling start-up phase; for example, the steady-state determination slope threshold is taken as 10% of the maximum slope during normal start-up. The filling end phase begins when the peristaltic pump deceleration command is issued and continues until the peristaltic pump speed drops to below 5% of the rated speed.

[0212] During the filling start-up phase, the control device preferentially applies the residual decoupling compensation vector during the cycle and controls the peristaltic pump to start according to an S-shaped acceleration curve. The maximum acceleration of the S-shaped acceleration curve is determined based on the allowable pressure surge of the liquid supply line. The unit of the maximum acceleration corresponds to the peristaltic pump drive control mode; when the controlled object is the peristaltic pump speed, the unit of the maximum acceleration is revolutions per minute per second; when the controlled object is the pulse frequency, the unit of the maximum acceleration is Hertz per second; when the controlled object is the equivalent filling flow rate, the unit of the maximum acceleration is flow rate per second.

[0213] The unit of the allowable pressure surge in the liquid supply pipeline is consistent with the unit of the liquid supply pressure, and the allowable pressure surge in the liquid supply pipeline is determined by the width of the allowable initial pressure range. The pressure acquisition device collects the liquid supply pressure in real time. When the slope of the liquid supply pressure rise curve is lower than the steady-state determination slope threshold, the control device determines that the filling start-up phase has ended.

[0214] During the steady-state filling phase, the control device applies an adaptive compensation vector for the supply pressure baseline shift and controls the peristaltic pump to operate at a constant speed or in pulse width modulation mode. The pressure acquisition device continuously collects the supply pressure. When the supply pressure deviates from the expected steady-state pressure value by more than a steady-state deviation threshold, the control device triggers an abnormal disturbance recording. The expected steady-state pressure value is obtained by adding the current supply pressure baseline value to the peristaltic pump's steady-state operating pressure increment. The steady-state deviation threshold is determined based on the standard deviation of normal steady-state pressure fluctuations; for example, the steady-state deviation threshold is three times the standard deviation of normal steady-state pressure fluctuations.

[0215] At the end of the filling process, the control device stops the peristaltic pump according to an S-shaped deceleration curve. The maximum deceleration of the S-shaped deceleration curve is determined based on the upper limit of the pressure rebound amplitude in the liquid supply line. The unit of the maximum deceleration corresponds to the peristaltic pump drive control mode; when the controlled object is the peristaltic pump speed, the unit of the maximum deceleration is revolutions per minute per second; when the controlled object is the pulse frequency, the unit of the maximum deceleration is Hertz per second; when the controlled object is the equivalent filling flow rate, the unit of the maximum deceleration is flow rate per second.

[0216] In the latter half of the steady-state filling phase, the control device applies the periodic disturbance isolation compensation vector output by S32; when the periodic disturbance isolation compensation vector is zero, the control device maintains the current filling control quantity.

[0217] The unit of the upper limit of the pressure rebound amplitude in the liquid supply line is consistent with the unit of the liquid supply pressure, and the upper limit of the pressure rebound amplitude in the liquid supply line is determined by the residual judgment threshold. When the peristaltic pump speed drops to less than 5% of the rated speed, the control device shuts off the peristaltic pump drive power, the filling execution section ends, and the process enters the post-filling recovery section.

[0218] S44: After filling is completed, perform multi-layer closed-loop self-verification based on the filling weight deviation of the (k+1)th filling cycle, and update the deviation source probability identification model parameters and compensation algorithm coefficients based on the multi-layer closed-loop self-verification results.

[0219] The multi-layer closed-loop self-verification includes absolute accuracy self-verification, relative improvement degree self-verification, and compensation direction consistency self-verification.

[0220] Closed-loop self-checking is used to determine whether the source of deviation in S2 is correctly identified, whether the compensation range in S3 is appropriate, and whether the initial state reconstruction in S4 is effective. Simply judging whether the filling weight deviation falls within the allowable range cannot distinguish between different situations such as correct compensation direction but insufficient magnitude, over-adjustment of compensation, and model parameter drift. This step uses three levels of judgment—absolute accuracy, relative improvement, and consistency of compensation direction—to write the post-filling weighing results back into the parameter update process.

[0221] The weighing device acquires the container weight before and after filling for the (k+1)th filling cycle, and the control device obtains the filling weight deviation for the (k+1)th filling cycle. The control device determines whether the absolute value of the filling weight deviation for the (k+1)th filling cycle is less than a preset filling volume error threshold and outputs the absolute accuracy self-verification result. The preset filling volume error threshold is determined based on the container specifications and is a positive value. The preset filling volume error threshold for small-sized glass bottles is 1% to 2% of the target filling weight, and the preset filling volume error threshold for 50 ml to 100 ml liquid bottles is 2% of the target filling weight. The unit of the preset filling volume error threshold is consistent with the unit of the target filling weight.

[0222] The control device first determines whether the absolute value of the filling weight deviation in the k-th filling cycle is greater than the minimum evaluable deviation. The unit of the minimum evaluable deviation is consistent with the output unit of the weighing device and is determined based on the repeatability error of the weighing device and the allowable error of the target filling weight.

[0223] When the absolute value of the filling weight deviation in the kth filling cycle is greater than the minimum evaluable deviation, the control device calculates the difference between the absolute value of the filling weight deviation in the kth filling cycle and the absolute value of the filling weight deviation in the (k+1)th filling cycle, then divides the difference by the absolute value of the filling weight deviation in the kth filling cycle to obtain the filling weight deviation improvement degree, and determines whether the filling weight deviation improvement degree is greater than the improvement degree threshold, and outputs the relative improvement degree self-verification result.

[0224] When the absolute value of the filling weight deviation in the k-th filling cycle is not greater than the minimum evaluable deviation, the control device skips the relative improvement calculation and outputs a relative improvement rating without evaluation. The improvement threshold is determined based on the improvement distribution of historically successful compensation samples; for example, the improvement threshold is set to 0.3.

[0225] The control device first determines whether the direction of increase of the comprehensive filling control quantity in the control dimension of the (k+1)th filling cycle is consistent with the direction of increase of the actual filling volume. When they are consistent, the control device directly uses the comprehensive filling control quantity as the positive equivalent filling control quantity; when they are inconsistent, the control device converts the comprehensive filling control quantity into the positive equivalent filling control quantity according to a preset control direction mapping relationship. The positive equivalent filling control quantity and each compensation vector in S3 share the same positive and negative directions; when its value increases, the actual filling volume increases, and when its value decreases, the actual filling volume decreases.

[0226] The control device determines whether the adjustment direction of the positive equivalent filling control quantity in the (k+1)th filling cycle relative to the reference filling control quantity is opposite in sign to the filling weight deviation in the kth filling cycle, and whether the absolute value of the filling weight deviation in the (k+1)th filling cycle is less than the absolute value of the filling weight deviation in the kth filling cycle. When both conditions are met, the control device outputs a self-verification result of successful compensation direction consistency; otherwise, it outputs a self-verification result of failed compensation direction consistency.

[0227] When the absolute accuracy self-verification result passes, the compensation direction consistency self-verification result passes, and the relative improvement degree self-verification result passes or requires no evaluation mark, the control device increases the prior probability corresponding to the current dominant deviation source and increases the baseline value of the fusion weight coefficient of the compensation vector corresponding to the current dominant deviation source. The prior probability update adopts an exponential weighted moving average method, and the decay factor is determined based on the stability of historical batches. For example, the decay factor is 0.1.

[0228] When the absolute accuracy self-verification result fails, the relative improvement degree self-verification result passes, and the compensation direction consistency self-verification result passes, the control device determines that the compensation direction is correct but the compensation magnitude is insufficient, and increases the generation gain coefficient of the compensation vector corresponding to the current dominant deviation source. The increase in the generation gain coefficient is determined based on the difference between the filling weight deviation improvement degree and the improvement degree threshold. The larger the difference, the smaller the increase; the smaller the difference, the larger the increase. The compensation vector after the generation gain coefficient is updated is still limited by the compensation saturation threshold.

[0229] When the self-verification result of the compensation direction consistency fails, the control device determines whether the maximum difference between the contribution probability of the supply pressure baseline migration, the contribution probability of the intra-cycle disturbance, and the contribution probability of the intra-cycle residual contribution is less than the probability approach threshold. The probability approach threshold is determined based on the contribution probability difference when miscompensation occurs in historical multi-source mixed samples. For example, the probability approach threshold is 0.15.

[0230] If the maximum difference is less than the probability approach threshold, the control device reduces the compensation saturation threshold of the comprehensive filling control quantity; if the maximum difference is not less than the probability approach threshold, the control device reduces the prior probability corresponding to the current dominant deviation source and increases the degree of conservative compensation after the low confidence marker is triggered.

[0231] When the absolute accuracy self-calibration pass rate of M consecutive filling cycles is lower than the pass rate threshold, the control device triggers a global parameter recalibration process. M is determined based on the production cycle time and weighing feedback delay; for example, M is 5. The pass rate threshold is determined based on the allowable short-term fluctuation range of qualified batches; for example, the pass rate threshold is 60%.

[0232] The global parameter recalibration process refers to the process of recalibrating the pressure transmission relationship and compensation parameters when the closed-loop self-calibration results of multiple consecutive filling cycles fail to meet the requirements. The global parameter recalibration process includes pausing normal filling, executing a calibration filling sequence under known liquid level gradient and standard temperature conditions, re-acquiring the baseline value of the liquid supply pressure, instantaneous flow rate and filling weight data, and refitting the liquid level, pressure, flow rate transmission relationship coefficients and pressure rebound time constant.

[0233] When the residual of the refitted liquid level, pressure, and flow transmission relationship is less than 5%, and the deviation of the refitted pressure rebound time constant from the standard time constant does not exceed 30%, the control device updates the corresponding model parameters and resumes normal filling; otherwise, the control device remains in the suspended normal filling state and outputs a liquid supply pipeline maintenance warning signal.

[0234] For example, the preset calibration residual threshold is 5%, and the maintenance alarm threshold is 20%. That is, when the residual of the refitted liquid level, pressure, and flow transmission relationship is less than 5%, and the deviation of the refitted pressure rebound time constant from the standard time constant does not exceed 20%, the control device updates the corresponding model parameters and resumes normal filling; otherwise, the control device remains in the paused normal filling state and outputs a liquid supply pipeline maintenance warning signal.

Claims

1. A method for identifying and compensating for pressure disturbances during the filling of water-based injections, characterized in that, Performed by a control device, including: Collect and deconstruct the time series of liquid supply pressure during the continuous filling cycle, and generate three types of pressure disturbance characteristic quantities based on the time series of liquid supply pressure during the continuous filling cycle; Calculate the filling weight deviation during the filling cycle, and cross-validate the filling weight deviation with three types of pressure disturbance characteristics to determine the source of deviation identification results; The comprehensive filling control quantity for the next filling cycle is obtained based on the results of the deviation source identification. Before executing the comprehensive filling control quantity, the fluid dynamics state of the liquid supply pipeline is judged; if the filling start-up conditions are not met, the initial pressure state of the liquid supply pipeline is rebuilt; if the filling start-up conditions are met, segmented peristaltic pump filling control is executed according to the comprehensive filling control quantity; after filling is completed, closed-loop parameters are updated.

2. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 1, characterized in that, The generation of three types of pressure disturbance characteristic quantities includes: The continuous filling cycle liquid supply pressure time sequence is divided according to the start and stop sequence of the peristaltic pump. First, the liquid supply pressure time sequence of the static section before filling is obtained, then the liquid supply pressure time sequence of the filling execution section is obtained, and finally the liquid supply pressure time sequence of the recovery section after filling is obtained. First, the baseline migration of the supply pressure is generated based on the time series of the supply pressure during the pre-filling settling period. Then, the disturbance amplitude within the period is generated based on the time series of the supply pressure during the filling execution period. Finally, the residual verification amount during the period is generated based on the time series of the supply pressure during the recovery period after filling of the adjacent filling period, combined with the time series of the supply pressure during the pre-filling settling period of the next filling period. The three generated results are used as the three types of pressure perturbation feature quantities.

3. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 2, characterized in that, The generation of the liquid supply pressure baseline shift includes: Calculate the median pressure of the liquid supply pressure time series during the pre-filling settling period; Based on the absolute deviation of the pressure value at each sampling point relative to the median pressure, the median of all absolute deviations is taken as the pressure fluctuation scale. Abnormal sampling point pressure values ​​with absolute deviations exceeding three times the pressure fluctuation scale are removed, and the remaining sampling point pressure values ​​are averaged to obtain the baseline value of the liquid supply pressure for the current filling cycle. The liquid supply pressure baseline shift is obtained by subtracting the liquid supply pressure baseline value of the previous filling cycle from the liquid supply pressure baseline value of the current filling cycle.

4. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 1, characterized in that, The cross-validation includes: A historical filling weight deviation sequence is generated based on the filling weight deviation of the continuous filling cycle; A moving average is applied to the historical filling weight deviation sequence to obtain a long-term trend sequence of filling weight deviation; Perform a first-order difference on the historical filling weight deviation sequence to obtain a first-order difference sequence of filling weight deviation; The long-term trend sequence of the filling weight deviation and the first-order difference sequence of the filling weight deviation are used for the cross-validation to obtain the test results of three types of sources; The deviation source identification result is determined based on the test results of the three types of sources.

5. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 4, characterized in that, The obtained test results for the three types of sources include: Read the time series of supply pressure baseline migration from the three types of pressure disturbance characteristic quantities, calculate the cross-correlation between the time series of supply pressure baseline migration and the long-term trend series of filling weight deviation, and obtain the test result of supply pressure baseline migration. Read the time series of disturbance amplitude within the period, check the event synchronicity of the time series of disturbance amplitude within the period relative to the first difference series of the filling weight deviation, check the deviation direction according to the disturbance direction mark within the period, and obtain the disturbance test result within the period; Read the residual verification amount during the cycle, and perform cycle-by-cycle mapping verification based on the residual verification amount during the cycle relative to the filling weight deviation of the affected filling cycle to obtain the residual inspection result during the cycle.

6. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 4, characterized in that, The determination of the deviation source identification result based on the test results of the three types of sources includes: The test results of the three types of sources, together with the three types of pressure disturbance characteristics, are input into the deviation source probability identification model to obtain the contribution probability of each type of deviation source. By comparing the contribution probabilities, the dominant source of deviation corresponding to the maximum contribution probability is determined; The dominant source of deviation, along with its contribution probability, is written into the deviation source identification result.

7. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 1, characterized in that, The process of obtaining the comprehensive filling control quantity for the next filling cycle based on the deviation source identification results includes: Read the contribution probability of the liquid supply pressure baseline migration from the deviation source identification result, and generate an adaptive compensation vector for the liquid supply pressure baseline migration. Read the intra-period disturbance contribution probability from the deviation source identification result and generate an intra-period disturbance isolation compensation vector; Read the periodic residual contribution probability from the deviation source identification result and generate a periodic residual decoupling compensation vector; Set the compensation vector that has not met the corresponding compensation trigger condition to zero; The comprehensive filling control quantity is obtained based on the three compensation vectors.

8. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 7, characterized in that, The process of obtaining the comprehensive filling control quantity based on three compensation vectors includes: Read the contribution probability of each deviation source in the deviation source identification results; Each of the compensation vectors is converted to the control dimension of the baseline filling control quantity; Using the corresponding contribution probability as the fusion weight, the converted compensation vector is weighted and superimposed onto the benchmark filling control quantity to obtain the comprehensive filling control quantity; Based on the compensation saturation threshold limiting the adjustment range of the comprehensive filling control quantity relative to the benchmark filling control quantity, the comprehensive filling control quantity for the next filling cycle is output.

9. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 1, characterized in that, The determination of the fluid dynamic state of the supply pipeline includes: Before the start of the next filling cycle, determine whether the average liquid supply pressure of the static section before filling is within the allowable starting pressure range. Continue to determine whether the standard deviation of the liquid supply pressure time series during the static period before filling is less than the pressure stability threshold. Next, determine whether the deviation of the pressure rebound time constant from the standard time constant exceeds the time constant deviation threshold; When all three judgment results meet the filling start conditions, the segmented peristaltic pump filling control is executed according to the comprehensive filling control quantity; If any judgment result does not meet the filling start condition, the initial pressure state of the liquid supply line shall be rebuilt. Once completed, the closed-loop parameters are updated based on the filling weight deviation of the next filling cycle.

10. The method for identifying and compensating for pressure disturbances during water injection filling according to claim 9, characterized in that, The segmented peristaltic pump filling control includes: The interval before the slope of the liquid supply pressure rise curve after the peristaltic pump starts drops to the steady-state determination slope threshold is defined as the filling start-up section. The interval between the end of the filling start-up phase and the issuance of the peristaltic pump deceleration command is defined as the filling steady-state phase. The interval from the issuance of the peristaltic pump deceleration command to the stopping of the peristaltic pump is defined as the filling end segment; The peristaltic pump is started in the filling start-up section according to the comprehensive filling control quantity, then runs in the filling steady-state section, and finally stops in the filling end section.

Images