Tablet press pressure control method and system

Abstract

The invention discloses a pressure control method and system for a tablet press, and relates to the technical field of pressure control, and the method comprises the steps: directly representing the physical state of a punch through a pressure rise rate and a peak holding duration by adopting a pressure curve characteristic-wear state-dynamic compensation cascade control logic; when the pressure transmission efficiency is reduced due to the abrasion of the punch, the rising rate is obviously reduced due to the increase of mechanical resistance, and the peak holding time is shortened due to the reduction of elastic deformation, the system calculates a compensation factor phi in real time according to the rising rate, and the original target pressure value is increased by 1 / phi times, so that the actual bearing pressure of the material is consistent with a set value; and the fatal defect that the feedback distortion of the sensor cannot be identified by the system is fundamentally solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pressure control, in particular to a tablet press pressure control method and system. BACKGROUND

[0002] In the continuous production of high-speed rotary tablet press, the punch is subjected to periodic impact load for a long time, and micro-wear is gradually formed on the surface of the workpiece. The degradation leads to an increase in the impedance of the pressure transmission path, and the actual pressure acting on the powder is significantly attenuated. Especially for hard-sensitivity drugs such as sustained-release tablets, the deviation between the target pressure and the actual forming pressure will directly cause abnormal tablet disintegration time or uneven content. Although the existing system based on the calibrated pressure closed-loop control can maintain the nominal pressure stable, it cannot sense the energy loss caused by the wear of the punch.

[0003] Some solutions use coordinated control of multiple punches, such as patent CN111890722A, which compares the pressure feedback values of each punch through a central controller and dynamically compensates, but the nonlinear characteristics of wear make the fixed compensation coefficient invalid. Or introduce a machine learning model to predict the amount of wear based on historical data and adjust the pressure set value, but the model relies on a large amount of offline calibration data and is difficult to adapt to changes in material characteristics on site. In addition, both types of solutions do not solve the pressure transmission attenuation, i.e. the sensor readings of the worn punch still show the standard pressure.

[0004] In recent years, some research has analyzed the rising slope characteristics of the pressure curve to determine the health of the punch, or used adjacent punch pressure mutual calibration to achieve self-diagnosis, but these solutions still rely on additional sensors or complex algorithms, and have real-time defects under high-speed working conditions. In essence, when the performance of the actuator itself deteriorates, such as punch wear, the traditional set value-feedback value closed-loop paradigm cannot guarantee the terminal forming quality, and it is necessary to build a tablet press pressure control method with the real forced state of the powder as the control object. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] The present application provides a tablet press pressure control method and system to solve the problem of pressure transmission attenuation caused by punch wear. The existing closed-loop control cannot sense the actual pressure loss due to sensor feedback distortion, leading to loss of control of tablet quality.

[0007] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a tablet press pressure control method, comprising, Step S1, real-time acquisition of the time sequence signal of the pressure sensor of each punch; Step S2, extracting a feature parameter set representing the punch wear state from the time sequence signal, the feature parameter set at least containing the change rate of the pressure rising stage and the pressure peak duration; Step S3, calculating a dynamic wear compensation factor based on the feature parameter set; Step S4, correcting the target pressure value of the corresponding punch according to the dynamic wear compensation factor.

[0008] As a preferred scheme of the tablet press pressure control method, in step S2, the feature parameter set is extracted by: calculating the pressure increment change rate per unit time in the pressure rising stage; in the pressure peak interval, identifying the duration that the pressure value is maintained above the set threshold as the pressure peak duration.

[0009] As a preferred scheme of the tablet press pressure control method, in step S2, the step of calculating the pressure increment change rate per unit time in the pressure rising stage includes: setting a weighted sliding window: , wherein, is the triangular weight of the i-th sampling point, dimensionless, is the internal index of the window, is the window length, the number of sampling points; calculating the weighted mean value, the formula is: , wherein, is the weighted time mean value of the i-th window, unit s, is the weighted pressure mean value of the i-th window, unit MPa, is the timestamp of the i-th sampling point in the window, is the pressure of the i-th sampling point in the window; calculating the window slope: , wherein, is the pressure increment change rate per unit time of the i-th window, unit ; extracting the rising stage feature value: , , , , ​​​​wherein, is the first is the validity indication of the window, unit Boolean, is the global pressure peak of this stroke, unit MPa, is the pressure rise rate index, unit .

[0010] As a preferred scheme of the tablet press pressure control method, in step S3, the calculation of the dynamic wear compensation factor includes: inputting the feature parameter set into a predefined parameter mapping model to output a compensation factor scalar; The parameter mapping model is obtained by training historical wear data and is independent of the tablet press running speed parameter.

[0011] As a preferred scheme of the tablet press pressure control method, in step S3, the step of inputting the feature parameter set into a predefined parameter mapping model to output a compensation factor scalar includes: performing feature normalization, and the normalization formula is: , wherein, denotes the normalized pressure rise rate, dimensionless, denotes the pressure rise rate index, unit , is the historical average of the pressure rise rate, unit , is the historical standard deviation of the pressure rise rate, unit , denotes the normalized pressure peak duration, dimensionless, is the pressure peak duration, unit ms, is the historical average of the pressure peak duration, unit ms, is the historical standard deviation of the pressure peak duration, unit ms, and unit conversion is performed before normalization; secondary mapping transformation, formula: , wherein, is the feature mapping intermediate quantity, dimensionless, is the linear term weight coefficient, dimensionless, is the quadratic term weight coefficient, dimensionless, is the bias term, dimensionless; generate a compensation factor: , wherein, is a dynamic wear compensation factor, dimensionless, is an amplification factor, typically 0.05-0.15, is a logic function output; the amplification factor typically ranges from 0.05 to 0.15 and the maximum compensation increment is an empirical reference value determined based on the fatigue characteristics of the punch material and the safety threshold of the hydraulic system, and the specific parameters can be adjusted by the skilled person according to the actual working conditions; a confidence threshold clipping is performed: , wherein, is a dynamic wear compensation factor after clipping, is a maximum allowed increment of the compensation factor, typically 0.10, dimensionless.

[0012] As a preferred scheme of the tablet press pressure control method described in the present application, in step S4, the corrected target pressure value comprises: dividing the original target pressure value by the dynamic wear compensation factor to generate a compensation pressure command; driving the punch through the hydraulic actuator to implement the compensation pressure command, while constraining the compensation pressure value to not exceed the preset floating range of the original target value.

[0013] As a preferred scheme of the tablet press pressure control method described in the present application, wherein calculating the change rate of the pressure rising stage comprises: performing sliding window segmentation on the pressure time sequence signal; using linear regression to fit the pressure-time curve in the window; using the absolute value of the slope of the fitted curve as the change rate output value.

[0014] As a preferred scheme of the tablet press pressure control method described in the present application, wherein identifying the duration of the pressure peak value comprises: setting the 90%-100% interval of the pressure peak value as the effective holding section; cumulatively counting the number of consecutive sampling points in the effective holding section and converting it into a time length.

[0015] In a second aspect, the present application provides a tablet press pressure control system, comprising, a signal processing unit integrated in the punch base, configured to perform pressure signal acquisition and feature parameter extraction; a central compensation calculation unit connected to each signal processing unit, configured to calculate a dynamic wear compensation factor and generate a compensation command; an electro-hydraulic servo drive module receiving the compensation command and regulating the punch pressure.

[0016] As a preferred solution of the tablet press pressure control system, wherein: the signal processing unit comprises: An anti-aliasing filter circuit limits the frequency band of the original pressure signal; A real-time feature extractor with a sliding window processing unit and a linear regression operator.

[0017] The present application has the following advantages: The present application addresses the problem of pressure transmission attenuation caused by punch wear in the tablet press. Through deep mining of pressure curve features at the method level and innovation of distributed architecture at the system level, the sensing blind area in pressure control is broken. Traditional closed-loop control relies on the feedback value of the punch sensor, but wear causes the sensor reading to deviate significantly from the actual pressure acting on the material. The present application uses a cascade control logic of pressure curve features-wear state-dynamic compensation, which directly represents the physical state of the punch using the pressure rise rate and peak value holding time. When punch wear causes a decrease in pressure transmission efficiency, the rise rate decreases significantly due to increased mechanical resistance, and the peak value holding time shortens due to reduced elastic deformation. The system calculates a compensation factor φ in real time based on this, which increases the original target pressure value by 1 / φ times, making the actual pressure of the material consistent with the set value, and fundamentally solving the fatal flaw that the sensor feedback distortion cannot be identified by the system.

[0018] In addition, the parameter mapping model of the present application is independent of the speed of the tablet press, avoiding compensation failure due to changes in production rhythm. The compensated pressure value is constrained to not exceed the floating range of the original target value, preventing overcompensation from causing equipment overload. The system does not need to be recalibrated when changing material varieties or adjusting capacity, significantly reducing maintenance frequency. Through the distributed hardware architecture, the feature extraction task is sunk to the signal processing unit of the punch, and the central compensation calculation unit only performs lightweight mapping operations, avoiding the delay bottleneck of data upload / instruction issuance under high-speed working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0020] Figure 1 The flowchart of the tablet press pressure control method in embodiment 1.

[0021] Figure 2 The framework diagram of the tablet press pressure control system in embodiment 1. DETAILED DESCRIPTION

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1, referring to Figure 1 and Figure 2 This embodiment provides a method for controlling the pressure of a tablet press, including the following steps: Step S1: Real-time acquisition of the timing signals of the pressure sensors of each punch; Step S2: Extract a set of characteristic parameters representing the wear state of the punch from the time-series signal. The set of characteristic parameters includes at least the rate of change of the pressure rise phase and the duration of the pressure peak. Step S2 extracts the feature parameter set, including: Calculate the rate of change of pressure increment per unit time during the pressure rise phase; Within the pressure peak range, the duration for which the pressure value remains above a set threshold is identified as the pressure peak duration. Step S2, which involves calculating the rate of change of pressure increment per unit time during the pressure rise phase, includes: Setting up a weighted sliding window: , in, For the first The triangular weights of each sampling point are dimensionless. For internal window indexing, The window length is the number of sampling points. The formula for calculating the weighted mean is: , in, For the first The weighted time mean of the window, in seconds. For the first Weighted average pressure of the window, in MPa. For the first in the window Each sampling point timestamp For the first in the window Pressure at each sampling point; Calculate the window slope: , in, For the first The rate of change of pressure increment per unit time within the window, in units of ; Extracting feature values ​​during the ascent phase: , , in, For the first Window validity indicator, in Boolean form. This represents the peak global pressure during this stroke, expressed in MPa. This is an indicator of the rate of change of pressure increment per unit time during the pressure rise phase, with units of... ; Specifically, triangular weighting gives higher influence to the central sample of the window, which can suppress random noise without increasing computational latency; weighted mean and The intercept coupling was removed, making the slope It only reflects the instantaneous upward speed; validity judgment Excluding the buffer zone near the peak, a single quantitative indicator is provided for changes in the rate of ascent due to wear. ; The calculation of the rate of change during the pressure rise phase specifically includes: Sliding window segmentation is applied to the pressure time-series signal; Linear regression was used to fit the stress-time curve within the window; The absolute value of the slope of the fitted curve is used as the output value of the rate of change. The duration of the pressure peak is identified as follows: Set the effective holding range as the pressure peak range of 90%-100%; Accumulate the number of consecutive sampling points within the effective retention segment and convert it into time length; Step S3: Calculate the dynamic wear compensation factor based on the feature parameter set; Step S3 involves calculating the dynamic wear compensation factor, including: Input the feature parameter set into a predefined parameter mapping model and output a compensation factor scalar; The parameter mapping model is trained using historical wear data and is independent of the tablet press operating speed parameters; In step S3, the feature parameter set is input into the pre-defined parameter mapping model, and the step of outputting the compensation factor scalar is: The feature normalization is performed, and the normalization formula is: , wherein, represents the normalized pressure rise rate, dimensionless, represents the pressure increment change rate per unit time in the pressure rise stage, unit: , is the historical average of the pressure rise rate, unit: , is the historical standard deviation of the pressure rise rate, unit: , represents the normalized pressure peak duration, dimensionless, is the pressure peak duration, unit ms, is the historical average of the pressure peak duration, unit ms, is the historical standard deviation of the pressure peak duration, unit ms, and unit conversion is performed before normalization; The secondary mapping transformation formula is: , wherein, is the feature mapping intermediate quantity, dimensionless, is the linear term weight coefficient, dimensionless, is the quadratic term weight coefficient, dimensionless, is the bias term, dimensionless; The compensation factor is generated: , wherein, is the dynamic wear compensation factor, dimensionless, is the amplification coefficient, typical value 0.05-0.15, is the logic function output; the typical value range of the amplification coefficient 0.05-0.15 and the maximum compensation increment are the empirical reference values determined based on the fatigue characteristics of the punch material and the safety threshold of the hydraulic system, and the specific parameters can be adjusted according to the actual working conditions by those skilled in the art; The confidence threshold clipping is performed: , wherein, is the clipped dynamic wear compensation factor, is the maximum allowed increment of the compensation factor, typical value 0.10, dimensionless; Specifically, zero-mean and unit-variance normalization of the original features can weaken the influence of dimension difference on mapping sensitivity; the quadratic mapping introduces cross terms , capturing the coupling effect between the rising rate and the peak duration, making up for the insufficient expression of the complex wear pattern by the linear model; the logical function output is scaled to the interval, ensuring that the adjustment range of the compensation factor relative to the original pressure command is limited and continuous, and then using threshold clipping to suppress extreme abnormal points, controlling the stable fluctuation of the compensation range; Step S4, correcting the target pressure value of the corresponding punch according to the dynamic wear compensation factor; The correction of the target pressure value in step S4 includes: Divide the original target pressure value by the dynamic wear compensation factor to generate a compensation pressure command; Drive the punch through the hydraulic actuator to achieve the compensation pressure command, while constraining the compensation pressure value not to exceed the preset floating range of the original target value.

[0026] The embodiment also provides a tablet press pressure control system, including: A signal processing unit integrated in the punch base, configured to perform pressure signal acquisition and feature parameter extraction; A central compensation calculation unit connected to each signal processing unit, configured to calculate a dynamic wear compensation factor and generate a compensation command; An electro-hydraulic servo drive module receiving the compensation command and regulating the punch pressure; The signal processing unit includes: An anti-aliasing filter circuit for band limiting the original pressure signal; A real-time feature extractor with a sliding window processing unit and a linear regression operator.

[0027] In summary, the embodiment aims at the problem of pressure transmission attenuation caused by punch wear of the tablet press, and breaks the sensing blind area in pressure control through deep excavation of pressure curve features at the method level and innovation of distributed architecture at the system level. The traditional closed-loop control relies on the feedback value of the punch sensor, but wear makes the sensor reading and the actual pressure acting on the material deviate seriously. The invention adopts the cascade control logic of pressure curve features-wear state-dynamic compensation, and uses the pressure rise rate and peak value retention time to directly represent the physical state of the punch. When the punch wear causes the pressure transmission efficiency to decrease, the rise rate decreases significantly due to the increase of mechanical resistance, and the peak value retention time shortens due to the decrease of elastic deformation. The system calculates the compensation factor φ in real time according to this, and increases the original target pressure value by 1 / φ times, so that the actual pressure borne by the material is consistent with the set value, and the fatal defect that the sensor feedback distortion cannot be identified by the system is fundamentally solved. The parameter mapping model is independent of the speed of the tablet press, avoiding compensation failure caused by changes in production rhythm; the pressure value after constraint compensation does not exceed the floating range of the original target value, preventing over-compensation from causing equipment overload; the system does not need to be recalibrated when changing the material variety or adjusting the capacity, significantly reducing the maintenance frequency; through the distributed hardware architecture, the feature extraction task is sunk to the signal processing unit of the punch locally, and the central compensation calculation unit only performs light mapping operation, avoiding the delay bottleneck of data upload / instruction issuance under high-speed working conditions.

[0028] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method of tablet press pressure control, characterized by, Comprising, Step S1, collecting the time sequence signal of the pressure sensor of each punch in real time; Step S2, extracting a feature parameter set representing the wear state of the punch from the time sequence signal, the feature parameter set at least including the change rate of the pressure rising stage and the pressure peak duration; Step S3, calculating a dynamic wear compensation factor based on the feature parameter set; Step S4, correcting the target pressure value of the corresponding punch according to the dynamic wear compensation factor.

2. A method of tablet press pressure control as claimed in claim 1, characterized in that, The feature parameter set extraction in step S2 includes: Calculate the pressure increment change rate per unit time in the pressure rising stage; In the pressure peak interval, identify the duration that the pressure value maintains above the set threshold as the pressure peak duration.

3. A method of tablet press pressure control as claimed in claim 2, wherein, In step S2, the step of calculating the pressure increment change rate per unit time in the pressure rising stage includes: Set a weighted sliding window: ， wherein, is the triangle weight for the th sampling point, dimensionless, is the window interior index, is the window length, number of sampling points; Calculate the weighted mean, the formula is: ， wherein, is the first weighted time mean of the window, in s, is the first weighted pressure mean of the window, in MPa, is the time stamp of the first sampling point in the window, is the pressure of the first sampling point in the window; Calculate the window slope: ， wherein, is the first window pressure change rate, in units of ; Extract the rising stage feature value: ， ， wherein, is the first is the validity indication of the window, in Boolean quantity, is the global pressure peak of this stroke, in MPa, is the pressure rise phase pressure increment change rate index per unit time, in .

4. A method of tablet press pressure control as defined in claim 1, wherein, The calculation of the dynamic wear compensation factor in step S3 includes: Input the feature parameter set into a predefined parameter mapping model to output a compensation factor scalar; The parameter mapping model is obtained by training historical wear data and is independent of the tablet press running speed parameter.

5. A tablet press pressure control method as claimed in claim 4, characterized in that, In step S3, the step of inputting the feature parameter set into the predefined parameter mapping model to output the compensation factor scalar is: Perform feature normalization, the normalization formula is: ， wherein, denotes the normalized pressure rise rate, dimensionless, denotes the pressure rise phase pressure increment change rate per unit time, unit , is the pressure rise rate historical mean value, unit , is the pressure rise rate historical standard deviation, unit , denotes the normalized pressure peak duration, dimensionless, is the pressure peak duration, unit ms, is the pressure peak duration historical mean value, unit ms, is the pressure peak duration historical standard deviation, unit ms, unit conversion is performed before normalization; Secondary mapping transformation, the formula is: ， wherein, is a feature mapping intermediate quantity, dimensionless, is a linear term weight coefficient, dimensionless, is a quadratic term weight coefficient, dimensionless, is a bias term, dimensionless; Generate a compensation factor: ， wherein, is a dynamic wear compensation factor, dimensionless, is an amplification factor, typical value 0.05-0.15, is a logic function output; the amplification factor typical value range 0.05-0.15 and maximum compensation increment is an empirical reference value based on the fatigue properties of the punch material and the safety threshold of the hydraulic system, and the specific parameters can be adjusted by the person skilled in the art according to the actual working conditions; Perform confidence threshold clipping: ， wherein, is the dynamic wear compensation factor after cropping, is the maximum allowed increment of the compensation factor, typical value 0.10, dimensionless.

6. A method of tablet press pressure control as defined in claim 1, wherein, The correction of the target pressure value in step S4 includes: Divide the original target pressure value by the dynamic wear compensation factor to generate a compensation pressure instruction; Drive the punch through the hydraulic actuator to realize the compensation pressure instruction, while constraining the compensation pressure value not to exceed the preset floating range of the original target value.

7. A method of tablet press pressure control as defined in claim 2, wherein, The calculation of the change rate of the pressure rising stage includes: Divide the pressure time sequence signal into sliding windows; Fit the pressure-time curve in the window using linear regression; Take the absolute value of the slope of the fitted curve as the change rate output value.

8. A tablet press pressure control method as claimed in claim 2, characterized in that, The identification of the pressure peak duration includes: Set the 90%-100% interval of the pressure peak as the effective maintenance segment; Cumulative number of consecutive sampling points in the effective maintenance segment and convert it to time length.

9. A tablet press pressure control system based on the tablet press pressure control method of any one of claims 1 to 8, characterized by, Comprising, The signal processing unit integrated in the punch base is configured to perform pressure signal collection and feature parameter extraction; The central compensation calculation unit is connected to each signal processing unit and is configured to calculate the dynamic wear compensation factor and generate a compensation instruction; The electro-hydraulic servo drive module receives the compensation instruction and controls the punch pressure.

10. A tablet press pressure control system as claimed in claim 9, wherein, The signal processing unit includes: An anti-aliasing filter circuit for band limiting the original pressure signal; A real-time feature extractor with a sliding window processing unit and a linear regression operator.

Images