Intelligent control method for multi-process coordination management of injection stretch blow molding machine

Abstract

The present application relates to the technical field of industrial process control, in particular to an intelligent control method for multi-process collaborative management of injection stretch blow molding machines, comprising: in the mold cooling main process for medical consumable production, when the monitored main process variable reaches the preset threshold value, based on the material property model associated with the rheological melt index, the operation instruction of the preparatory process executor is adaptively generated, and the preparatory process parallel to the main process is started; using the model prediction double-loop control loop, through real-time control of the inner loop, according to the dynamic prediction of the synchronization deviation of the two process procedures, the preparatory process executor is continuously adjusted to realize synchronization, and at the same time, through online learning of the outer loop, according to historical error feedback, the prediction model inside the loop is continuously self-corrected. Through the establishment of a model prediction double-loop self-correcting control loop, the present application realizes predictive collaborative control of multiple asynchronous executed process procedures.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial process control, in particular to an intelligent control method for multi-process collaborative management of injection stretch blow molding machines. BACKGROUND

[0002] In the automated production of medical consumables, a set of equipment usually contains multiple independent functional units driven by different actuators. However, the traditional program controller usually adopts a fixed operation time sequence to drive each actuator to complete its own process stage in turn. A common problem of this control mode is that when a certain actuator (for example, a temperature control loop for maintaining the temperature of a mold) is executing a long-time program segment, other independent actuators (for example, a driving device for material preparation) must be in an unnecessary waiting state. This bottleneck caused by the fixed control time sequence seriously limits the system-level operation efficiency of the entire set of equipment.

[0003] To solve this problem, the prior art has made some improvements. One is to realize independent driving of each execution unit in hardware, which provides a physical basis for asynchronous operation. The other is to add a simple state monitoring function to the equipment to realize remote monitoring of the process parameters during equipment operation. However, the existing adjustment scheme is still limited to adjusting the process variables within a single loop, and has not been able to realize cross-loop control instruction adjustment and time sequence coordination based on the feedback signals of key process variables.

[0004] Therefore, the present application provides an intelligent control method for multi-process collaborative management of injection stretch blow molding machines. SUMMARY

[0005] The purpose of the present application is to provide an intelligent control method for multi-process collaborative management of injection stretch blow molding machines, which realizes predictive collaborative control of multiple asynchronous process procedures by establishing a double-loop self-correcting control loop based on model prediction.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] An intelligent control method for multi-process collaborative management of injection stretch blow molding machines, comprising:

[0008] In the mold cooling main process of medical consumable production, the main process variable is continuously monitored;

[0009] When the main process variable reaches a preset threshold, the operation instruction of the preparatory process actuator is generated based on the material property model associated with the flow melt index, and the preparatory process parallel to the main process is started;

[0010] Based on the model predictive double-loop control loop, continuously compare the output results of the linear regression model for predicting the main process and the kinematic model for calculating the preparatory process, generate the process synchronization deviation;

[0011] The model predictive double-loop control loop analyzes the correlation between the historical production data containing the history of rheological melt index change and the process synchronization deviation through an adaptive filtering algorithm, online corrects the parameters of the linear regression model and the kinematic model, and continuously generates and outputs the adjusted control instructions to the preparatory process executor according to the process synchronization deviation.

[0012] Preferably, the process of continuously monitoring the main process variable includes: collecting temperature data of the mold cavity surface in real time through a non-contact infrared sensing array deployed in a clean environment; and processing the time series of the temperature data through a cooling time prediction model to generate the main process variable containing a cooling remaining time prediction value.

[0013] Preferably, the cooling time prediction model includes:

[0014] The temperature data is input as a thermal boundary condition into a finite element analysis model based on the heat conduction equation; the finite element analysis model is solved to obtain the transient temperature field distribution of the mold cavity surface; and according to the transient temperature field distribution, the remaining time required for the entire mold cavity to reach the glass transition temperature is calculated as the cooling remaining time prediction value contained in the main process variable.

[0015] Preferably, the material property model based on the correlation of rheological melt index includes:

[0016] A database containing multiple known rheological melt indexes and corresponding optimal screw rotation speeds and optimal back pressures for producing qualified medical consumables is pre-established through experimental calibration; based on the database, a curve fitting algorithm is used to generate a feedforward model for calculating the screw rotation speed and the back pressure, which can interpolate and calculate the corresponding screw rotation speed set value and back pressure set value according to the input of any rheological melt index.

[0017] Preferably, the process of generating the operation instructions of the preparatory process executor includes:

[0018] The rheological melt index is input into a feedforward model to interpolate a corresponding screw rotation speed setting value and a back pressure setting value, the operation instruction being a set of control parameters including the screw rotation speed setting value and the back pressure setting value; a monitoring threshold value associated with the rheological melt index is generated, the local controller extracting a time series of real-time power data for a batch production of qualified medical consumables; an average value and a standard deviation of the time series are calculated; and upper and lower limits of the monitoring threshold value are set as the average value plus or minus a preset multiple of the standard deviation.

[0019] Preferably, the generation process of the procedure synchronization deviation includes:

[0020] Based on the time rate of change of the main process variable, a linear regression model is established to dynamically predict the remaining completion time of the main process; based on real-time position and speed feedback of the preparatory process actuator, a kinematics model is established to calculate the estimated completion time of the preparatory process in real time; the remaining completion time and the estimated completion time are compared to generate the procedure synchronization deviation.

[0021] Preferably, the model predictive double-loop control circuit includes an inner loop for real-time control and an outer loop for online learning:

[0022] The inner loop generates a control instruction according to the current value, historical cumulative value and change trend of the procedure synchronization deviation, and outputs the control instruction to the preparatory process actuator through a PID controller; the outer loop analyzes the correlation between historical procedure synchronization deviations and historical production data including historical rheological melt index changes through an adaptive filtering algorithm, and online corrects the linear regression model parameters for predicting the remaining completion time of the main process and the kinematics model parameters for calculating the estimated completion time of the preparatory process.

[0023] Preferably, the method further includes a monitoring and warning process performed concurrently with the adjustment process of the model predictive double-loop control circuit:

[0024] The real-time power generated by the preparatory process actuator is compared with the monitoring threshold value as a real-time process variable to generate a warning signal; a fast Fourier transform is applied to time series data of the real-time power to generate a power spectrum diagram; a harmonic component at a preset frequency associated with electromagnetic noise of a driving motor of the preparatory process actuator is identified in the power spectrum diagram; and the amplitude of the harmonic component is compared with the monitoring threshold value to generate the warning signal and write it into a production process electronic record; the warning signal is used to output an interrupt instruction to the local controller to interrupt the execution of the main process.

[0025] Preferably, the production process electronic record includes the control instruction, the warning signal, the operator identity and the related time stamp:

[0026] The cloud server receives the production process electronic record, calculates a digital digest of the production process electronic record, binds the digital digest with a trusted timestamp provided by a trusted third-party timestamp service, and stores the electronic record, the digital digest, and the trusted timestamp in a database with tamper-proof characteristics for integrity checking of historical control behavior of the local controller.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] 1. The present application realizes predictive cooperation and synchronization of multiple processes and improves the operation efficiency of the control method. The model prediction double-loop control circuit can predict the remaining time and estimated time of two independent processes in real time and dynamically, and based on the generated process synchronization deviation, continuously outputs adjustment instructions through a real-time control inner loop to make the end time of the two processes accurately synchronized. This reduces the inherent waiting time in traditional control due to serial execution of processes, thereby improving the overall production efficiency.

[0029] 2. The present application constructs a cross-loop linkage adjustment mechanism to realize cooperative management of multiple independent processes. When a key process variable in one process reaches a preset threshold, it can event-drivenly trigger the start of another independent process, and during parallel execution, the time states of the two processes are continuously associated and compared to generate a process synchronization deviation for adjusting one of the processes. This cross-loop information interaction and closed-loop adjustment realizes cooperative management of multiple independent processes.

[0030] 3. The present application gives the control method the ability of self-correction to ensure the accuracy of long-term operation. The model prediction double-loop control circuit of the present application includes an online learning outer loop. The outer loop can update its internal prediction model parameters online and continuously according to the correlation between historical production data and historical process synchronization deviations. This self-correcting ability enables the control method of the present application to adaptively compensate for system characteristic drift caused by factors such as equipment wear and environmental changes, maintaining control accuracy and synchronization efficiency in long-term operation.

[0031] 4. A model-based diagnostic function linked to product quality was established, improving the level of automation. This invention achieves adaptive setting of control parameters by establishing a material property model linked to the rheological melt index. Through signal processing techniques such as FFT spectrum analysis, in-depth analysis of real-time process variables is performed to identify and warn of process anomalies that may lead to defects in medical consumables. This model-based diagnostic capability improves the automation level and process stability of the production process. Attached Figure Description

[0032] Figure 1 This is a flowchart of an intelligent control method for multi-process collaborative management of an injection stretch blow molding machine according to the present invention;

[0033] Figure 2 This is a logic block diagram of the model prediction dual-loop control loop in an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the fault early warning principle based on power spectrum analysis according to an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Other embodiments obtained by those skilled in the art based on the ideas in this specification without creative effort all fall within the protection scope of this invention.

[0036] For clarity, the term "rheological melt index" as used in this invention refers to the melt index measured by a standardized method recognized in the art (e.g., ASTM D1238 or ISO 1133) and used in this invention to characterize the macroscopic rheological properties of medical polymer materials. The use of this term is intended to emphasize that, specifically, the material property model based on the correlated rheological melt index, and the model's predictive dual-loop control loop, described in this invention, utilizes this index based on an understanding of the deep flow characteristics of the material, rather than simply using a factory parameter.

[0037] Reference Figures 1 to 3The embodiment provides an intelligent control method for multi-process collaborative management of an injection stretch blow molding machine for medical consumable production, which is implemented in an environment including a local controller, a main process actuator and a preparatory process actuator. In the embodiment, the main process is a process that plays a leading role and serves as a time reference in the whole control cycle, and specifically refers to a mold cooling process driven by the main process actuator; and the preparatory process is a process that is executed in parallel with the main process and prepares for the next production cycle, and specifically refers to a plasticized sol process driven by the preparatory process actuator.

[0038] Embodiment 1

[0039] With reference to Figure 1 An intelligent control method for multi-process collaborative management of an injection stretch blow molding machine, comprising:

[0040] In the mold cooling main process for medical consumable production, a main process variable is continuously monitored;

[0041] When the main process variable reaches a preset threshold, operation instructions of a preparatory process actuator are generated based on a material property model associated with a rheological melt index, and a preparatory process that is parallel to the main process is started;

[0042] Based on a model predictive double-loop control loop, output results of a linear regression model for predicting the main process and a kinematics model for calculating the preparatory process are continuously compared to generate a process synchronization deviation;

[0043] The model predictive double-loop control loop analyzes the correlation between historical production data containing historical rheological melt index changes and the process synchronization deviation through an adaptive filtering algorithm, corrects the parameters of the linear regression model and the kinematics model online, and continuously generates and outputs adjusted control instructions to the preparatory process actuator according to the process synchronization deviation.

[0044] Further, the process of continuously monitoring the main process variable includes: acquiring temperature data of a mold cavity surface in real time through a non-contact infrared sensing array deployed in a clean environment; and processing a time sequence of the temperature data through a cooling time prediction model to generate the main process variable containing a cooling remaining time prediction value.

[0045] Specifically, the non-contact infrared sensing array is deployed on the injection stretch blow molding machine, and the position of the array can ensure that the temperature data of multiple predetermined monitoring areas of the mold cavity surface is acquired in real time without obstruction after the mold is opened. The sensing array adopts a non-contact measurement method and is installed in a manner meeting the requirements of the clean environment, so as to avoid causing any pollution to the production process of the medical consumables.

[0046] The cooling time prediction model is an algorithm module fixed in the main controller memory. The model continuously receives the temperature data stream uploaded by the non-contact infrared sensing array at a fixed sampling frequency, and organizes these data into a time series of temperature data. A set of reference cooling curves is stored in the model, which is determined by experiments under standard process conditions for specific specifications of medical consumables and their raw materials. In real-time operation, the model compares the temperature decay trend presented by the current collected temperature data time series with the reference cooling curves composed of multiple sets of "time-temperature" data points stored internally, and uses the least squares method to perform exponential decay model fitting, and solves the equation to calculate the time required for the current mold cavity surface temperature to reach the preset demolding temperature threshold.

[0047] The cooling time prediction model processes the time series of temperature data to generate a dynamically updated numerical value. This value is the main process variable containing the cooling remaining time prediction value, which is transmitted to the central task scheduler as the core time reference for determining when the mold cooling process ends and safely starting the next main process (such as the mold opening process).

[0048] The preset threshold for starting the parallel preparation process can be determined by experiment calibration, setting the threshold to a fixed time value slightly larger than the standard execution time of the preparation process; in a preferred embodiment, the threshold can also be dynamically set to a preset proportion of the initial prediction value of the main process variable, for example 55%, to adapt to process fluctuations.

[0049] This embodiment realizes adaptive cooling control by real-time monitoring of mold cavity temperature and dynamic prediction of cooling time, effectively avoiding efficiency loss and product defects caused by improper cooling time, while non-contact measurement ensures a clean production environment, significantly improving the yield, consistency and overall beat stability of medical consumable production.

[0050] Further, the cooling time prediction model includes: inputting the temperature data as a thermal boundary condition into a finite element analysis model based on the heat conduction equation; solving the finite element analysis model to obtain the transient temperature field distribution of the mold cavity surface; and calculating the remaining time required for the entire mold cavity to reach the glass transition temperature based on the transient temperature field distribution, as the cooling remaining time prediction value contained in the main process variable.

[0051] Specifically, the finite element analysis model based on the heat conduction equation is a pre-constructed digital mold thermal simulation model. The model is formed by meshing the three-dimensional geometric structure of the mold to form a collection of a large number of finite element units, and the units are given thermal physical properties consistent with the actual mold material, such as thermal conductivity, specific heat capacity, etc. At the same time, the model also presets the glass transition temperature of the raw material of the target medical consumable as a key threshold.

[0052] As a more preferred embodiment, in order to greatly improve the online calculation speed while ensuring physical precision, a hybrid prediction model based on physical information neural network can be used to replace the traditional finite element solver. In the offline training stage, the hybrid model not only learns a large amount of “surface temperature-internal temperature field” data generated by finite element simulation, but also adds the heat conduction partial differential equation itself as a loss function to the training process of the neural network.

[0053] During the operation of the molding machine, the real-time temperature data collected by the non-contact infrared sensor array is used as the input boundary condition of the hybrid model. Since the model has embedded the physical law of heat conduction in the training, it can quickly and physically solve the transient temperature field distribution inside the product at the current time.

[0054] The local controller continuously analyzes the transient temperature field distribution, especially the temperature change inside the product forming area. Through extrapolation algorithm, the controller predicts the temperature change trend in the future period of time according to the current temperature drop rate, and calculates the time required for all nodes inside the product to have a temperature lower than the preset glass transition temperature. The calculation result is determined as the cooling remaining time prediction value and is updated in real time.

[0055] As a more preferred embodiment, in order to further improve the prediction accuracy and stability of the main process variable (i.e. cooling remaining time prediction value), the mold cooling process further includes a precise temperature tracking control process, which includes: presetting an optimal cooling curve representing the change of the target temperature of the mold with time; monitoring the actual mold temperature measured by the sensor array in real time during the cooling process; comparing the actual mold temperature with the corresponding target temperature on the optimal cooling curve at the corresponding time to generate a temperature deviation; based on the temperature deviation, continuously adjusting the opening of the proportional control valve connected to the mold cooling water circuit through the PID controller, so that the actual mold temperature tracks the optimal cooling curve.

[0056] Specifically, by offline finite element simulation or standardized experiment, an optimal mold temperature cooling curve varying with time is preset for specific products and raw materials. The curve aims to maximize the cooling rate while controlling the internal stress, warping deformation and other quality indicators of the final product within the qualified range.

[0057] In the actual production cooling process, the local controller compares the actual temperature of the mold cavity key area collected by the non-contact infrared sensor array in real time with the target temperature corresponding to the current time on the optimal cooling curve, and generates a temperature deviation.

[0058] Subsequently, a controller with a feedforward feedback composite control algorithm inside the local controller calculates a regulating output in real time. The controller includes two parts working together:

[0059] Feedforward control part: This part directly reads the preset optimal cooling curve and calculates a basic valve opening instruction based on the future rate of change of the curve (i.e. the target cooling rate) through a pre-established physical model (which describes the relationship between cooling water flow and mold cooling rate). This instruction can be adjusted in advance and actively to compensate for the thermal inertia and pure delay of the mold system;

[0060] PID feedback control part: This part calculates a compensatory regulating amount based on the size, historical accumulation and change trend of the deviation between the actual temperature and the target temperature. This regulating amount can fine-tune on the basis of feedforward control to reduce residual errors caused by model inaccuracies or external disturbances.

[0061] The final regulating output is the superposition of the calculation results of the above two parts, which is used to continuously adjust the opening of a proportional control valve connected to the main cooling water circuit of the mold, thereby accurately controlling the flow of cooling water.

[0062] Through this feedforward and feedback combined control strategy, the actual cooling process of the mold can be stabilized and effectively followed along the preset optimal cooling curve, significantly reducing the quality defects such as warping and shrinkage marks of the product caused by uneven or too fast cooling, and achieving the dual improvement of production efficiency and product structure stability.

[0063] The embodiment introduces finite element analysis based on physical equations to improve the prediction from surface temperature to accurate simulation of the real temperature field inside the product, which can accurately determine whether the product has been completely solidified to the core, especially for complex medical consumables with uneven wall thickness, to ensure that the product is demolded after reaching the glass transition temperature, thereby fundamentally avoiding quality defects such as shrinkage marks and warping caused by incomplete cooling inside, and achieving the dual improvement of the limit optimization of cooling time and the stability of product structure.

[0064] Further, the material property model based on the associated rheological melt index comprises: a method of establishing a database comprising a plurality of known rheological melt indices and corresponding optimal screw rotation speeds and optimal back pressures for producing qualified medical consumables in advance through experimental calibration; based on the database, a feedforward model for calculating the screw rotation speed and the back pressure is generated by using a curve fitting algorithm, which can calculate the corresponding screw rotation speed set value and back pressure set value according to the input of any rheological melt index.

[0065] Specifically, the method of experimental calibration comprises the following steps: a plurality of batches of medical-grade raw materials with different known rheological melt indices are subjected to optimization experiments of process parameters respectively; in the experiments, the process parameters including the screw rotation speed and the back pressure are iteratively adjusted, and the samples produced after each adjustment are subjected to quality inspection; when a set of process parameters can stably produce qualified products meeting the predetermined medical consumable standards in terms of dimensional accuracy, internal stress and surface finish, the set of process parameters is determined as the optimal process parameters corresponding to the current rheological melt index; and a set of corresponding data of the rheological melt index and the optimal process parameters in the database is established accordingly.

[0066] The feedforward model is a continuous function model generated by mathematical processing of discrete data points in the database; the local controller uses a curve fitting algorithm such as polynomial curve fitting or spline interpolation to establish two independent mathematical relationships for the screw rotation speed and the back pressure respectively, and the two relationships describe the nonlinear correspondence between the rheological melt index and the optimal screw rotation speed, and the rheological melt index and the optimal back pressure.

[0067] The technical solutions of the present application will be further described below through a specific example. As an example, part of the database content obtained through the foregoing standardized experiments can be shown in the following table:

[0068] Table 1: Example of Correspondence between Rheological Melt Index and Process Parameters

[0069]

[0070] Based on the above table and at least ten groups of all data points in the database covering the commonly used rheological melt index range, the local controller generates a feedforward model using a curve fitting algorithm based on the least square method principle; the model is mathematically represented as a functional relationship, in which the screw speed is determined by the weighted sum of multiple powers of the rheological melt index plus a constant, the weight coefficient is determined by the fitting algorithm through numerical calculation, and the target is to minimize the sum of squared errors between the model prediction value and the actual value of all database data points.

[0071] When the production task starts, the operator inputs the rheological melt index of the current batch of raw materials into the human-computer interaction interface, and the feedforward model automatically obtains a theoretically optimal screw speed setting value and an optimal back pressure setting value through interpolation calculation using the established functional relationship, and uses them as the initial control parameters of the plastic sol process.

[0072] As a more preferred embodiment, to further improve the accuracy of the process to cope with the fluctuations in the actual properties of the raw materials, the method further includes a dynamic correction mechanism for the input rheological melt index. During the production process, the local controller continuously collects real-time power data generated by the motor driven by the preparatory process executor, and calculates its statistical characteristics, such as the average power value, in a complete plasticizing cycle; compare the statistical characteristics with the reference characteristics of the historical power data corresponding to the current input rheological melt index in the database to generate a power deviation signal; the power deviation signal is used to dynamically correct the input rheological melt index to generate a corrected rheological melt index reflecting the true flowability of the raw materials; this corrected rheological melt index is used in the feedforward model to calculate more accurate screw speed setting value and back pressure setting value.

[0073] The present embodiment solves the problem of production instability caused by the flowability difference of different batches of raw materials and the batch fluctuation by establishing a direct correlation model between rheological melt index and core plasticizing parameters and combining real-time feedback correction. The operator only needs to input an easily obtained rheological melt index, and the local controller can automatically set and dynamically maintain the optimized process parameters, greatly reducing the dependence on human experience, shortening the material change debugging time, ensuring the consistency of the melt glue quality, and improving the stability and efficiency of medical consumable production.

[0074] Further, the process of generating the operation instruction of the preparatory process executor comprises: inputting the rheological melt index into a feedforward model, interpolating to calculate corresponding screw rotation speed setting value and back pressure setting value, the operation instruction being a set of control parameters comprising the screw rotation speed setting value and the back pressure setting value; generating a monitoring threshold value associated with the rheological melt index, the local controller extracting a time sequence of real-time power data of batch production for producing qualified medical consumables; calculating the average value and the standard deviation of the time sequence; and setting the upper and lower limits of the monitoring threshold value as the average value plus or minus a preset multiple of the standard deviation.

[0075] Specifically, at the beginning of the production task, an operator inputs the rheological melt index of the currently used raw material through a human-computer interaction interface. The rheological melt index is automatically transmitted by the main controller to the feedforward model. The feedforward model performs interpolation operation according to the internal function relationship, and outputs two determined values, i.e. the screw rotation speed setting value and the back pressure setting value. The two setting values are encapsulated into a data packet to form a complete set of operation instructions, and are issued by the main controller to the local controller responsible for driving the screw rotation and forward and backward movement, as the initial execution basis of the sol gel process.

[0076] The establishment of the monitoring threshold value is based on a calibration process. The process first performs one or more standard productions, i.e. using confirmed qualified raw materials to produce under the best process parameters. During this process, the local controller continuously monitors the real-time power consumption of the screw driving motor through the built-in current sensor at a high sampling frequency, and records these data to form a time sequence of real-time power data covering the entire sol gel period. The calculation unit in the local controller statistically analyzes all data points in the time sequence to calculate the arithmetic mean and the standard deviation of the entire sequence. The selection of the preset multiple is based on the industrial quality control standard, and its typical value range is between two and a half to three and a half; using three times the standard deviation corresponds to about 99.7% of the process data within the normal range, which is a common setting that can effectively distinguish between normal fluctuations and real anomalies.

[0077] The local controller calculates the upper limit and the lower limit of the numerical interval by calculating “the average value plus three times the standard deviation” and “the average value minus three times the standard deviation” according to a preset multiple (for example, the value 3) stored in a configuration file, and the upper limit and the lower limit of the interval are set as the monitoring threshold value, which is used to judge in real time whether the power consumption of the sol gel process is within the normal statistical range during subsequent batch production; for the monitoring of specific harmonic components, the threshold value is also set based on the average value and the standard deviation of the harmonic amplitude calibrated in multiple normal production periods, so as to effectively distinguish between normal fluctuations and abnormal signals.

[0078] The embodiment realizes the automatic setting of plasticizing parameters by combining the rheological melt index with the feedforward model, and further establishes a statistical monitoring threshold using historical power data, which provides a quantitative and dynamic monitoring standard for the sol process, can detect process deviation caused by abnormal raw materials or equipment state changes in real time, and effectively ensures the stability of the medical consumable production process and the consistency of product quality.

[0079] Further, the generation process of the process synchronization deviation includes: based on the time change rate of the main process variable, a linear regression model is established to dynamically predict the remaining completion time of the main process; based on the real-time position and speed feedback of the preparatory process executor, a kinematics model is established to calculate the estimated completion time of the preparatory process in real time; the remaining completion time and the estimated completion time are compared to generate the process synchronization deviation.

[0080] Specifically, the linear regression model is an algorithm deployed in the main controller for fine prediction of the end time point of the main process. A combination prediction strategy is used to balance global accuracy and real-time response: first, an initial cooling time benchmark prediction value is obtained through a global solution based on the initial thermal field distribution by a finite element analysis model; then, during the cooling process, the linear regression model analyzes the change trend, i.e. the time change rate, of the sequence within a very short time window to locally and linearly dynamically correct the benchmark prediction value obtained by the finite element model, thereby generating a more smooth and current real cooling trend of the remaining completion time of the main process.

[0081] At the same time, the local controller obtains the real-time rotational speed and axial position of the screw rod of the preparatory process executor (i.e. the sol assembly) through the encoder and position sensor installed on the driving motor of the preparatory process executor. The kinematics model calculates the time required for the screw rod to move from the current position to the target end position based on the preset sol amount (i.e. the target end position of the screw rod to be retreated), combined with the real-time position and speed feedback, by dividing the remaining travel distance by the current speed through kinematics formula, which is defined as the estimated completion time of the preparatory process.

[0082] Finally, the main controller synchronously obtains the values output by the two models at a high frequency in a separate comparison unit: one is the remaining completion time of the main process, and the other is the estimated completion time of the preparatory process. The controller performs subtraction operation on these two time values, and the difference is generated as the process synchronization deviation in real time and dynamically. The process synchronization deviation value accurately quantifies the leading, lagging or synchronous state of the two parallel processes in the completion time.

[0083] The embodiment realizes accurate and real-time prediction of the completion time of two parallel processes by establishing a double-path dynamic prediction model, rather than simple state synchronization. This prediction-based comparison can generate a predictive process synchronization deviation signal, which not only reflects the current synchronization state, but also predicts the future completion time difference. This provides a core basis for the controller to implement higher-order, prediction-based collaborative control, thereby maximizing the use of process gaps.

[0084] Further, with reference to Figure 2 , the model prediction double-loop control circuit includes an inner loop for real-time control and an outer loop for online learning: the inner loop generates a control instruction according to the current value, historical cumulative value and change trend of the process synchronization deviation, and outputs the control instruction to the preparatory process actuator through a PID controller; the outer loop analyzes the correlation between historical process synchronization deviation and historical production data containing historical rheological melt index change through an adaptive filtering algorithm, and online corrects the linear regression model parameters for predicting the remaining completion time of the main process and the kinematics model parameters for calculating the estimated completion time of the preparatory process.

[0085] Specifically, the model prediction double-loop control circuit is a hierarchical control structure. The inner loop at the core of the structure is a high-speed response real-time feedback controller. The input of the inner loop is the real-time process synchronization deviation value generated by comparing the two time prediction values output by the upstream linear regression model and kinematics model through a comparator; a standard proportional-integral-derivative (PID) controller continuously processes the process synchronization deviation signal, calculates a real-time adjustment amount according to the current size (proportion P), historical cumulative amount (integral I) and future change trend (derivative D) of the process synchronization deviation signal; the adjustment amount is the control instruction, which is sent to the local controller of the preparatory process actuator (sol gel assembly) to fine-tune the running speed, such as screw speed; if the process synchronization deviation shows that the sol process is too fast, the control instruction will slightly reduce the screw speed, and vice versa.

[0086] The outer loop is an optimization learning loop running on a longer time scale, which continuously collects and stores key data in multiple production cycles to form the historical production data, including but not limited to actual process synchronization deviation, rheological melt index of raw materials, environmental temperature and finally calculated cooling time of each cycle; these historical data are periodically analyzed by an adaptive filtering algorithm, such as a recursive least squares algorithm, which identifies the potential correlation and systematic errors between historical process synchronization deviation and specific production conditions (such as a certain rheological melt index range) through mathematical operations.

[0087] Based on the analysis result, the outer loop updates the model prediction parameters inside the MPC loop in real time. Specifically, the update process follows a standard iterative algorithm, i.e. according to the prediction error of the last period, multiplied by a gain coefficient calculated according to historical data, to make a small, compensatory adjustment to the current model parameters, so that the model is more accurate in future prediction.

[0088] Through this online updating process, the outer loop can continuously and adaptively correct the model parameters inside the linear regression model and the kinematic model to compensate for the system characteristic drift caused by equipment wear or environmental changes, etc.

[0089] The embodiment combines fast response and online learning by constructing an inner-outer double-loop control structure. The inner loop PID controller ensures rapid and stable correction of the instantaneous deviation that occurs in each production cycle; the adaptive algorithm of the outer loop gives the method the ability to optimize long-term, enabling it to learn and compensate for systematic errors introduced by factors such as equipment wear or raw material batch changes, thereby ensuring immediate production synergy while continuously improving the long-term stability and prediction accuracy of the entire control method.

[0090] Further, with reference to Figure 1 , Figure 3 , the method further comprises a monitoring and warning process executed concurrently with the adjustment process of the MPC loop: concurrently comparing the real-time power generated by the preparatory process actuator as a real-time process variable with a monitoring threshold to generate a warning signal; applying a fast Fourier transform to the time series data of the real-time power to generate a power spectrum; identifying the harmonic component at the preset frequency associated with the electromagnetic noise of the motor driving the preparatory process actuator in the power spectrum; and comparing the amplitude of the harmonic component with the monitoring threshold to generate the warning signal and write it into the production process electronic record; the warning signal is used to output an interrupt instruction to the local controller to interrupt the execution of the main process.

[0091] Specifically, the local controller continuously acquires the real-time current and voltage of the motor driving the preparatory process actuator (sol component) and multiplies them to obtain the real-time power; these power readings are continuously recorded to form a time series data of the real-time power over time; the signal processing unit in the local controller intercepts a latest time series data segment (e.g. containing 1024 sampling points) at fixed time intervals and applies a fast Fourier transform algorithm to the data segment; the transform converts the power signal in the time domain to a frequency domain representation, and the result is the power spectrum showing the energy distribution of the power signal at different frequencies.

[0092] In the power spectrum, the local controller focuses on one or more preset frequencies, which are characteristic frequencies highly related to the health status or failure mode of the specific equipment, and are calibrated by systematically analyzing the power signal spectrum of the equipment under various working conditions such as normal operation and simulated slight faults (such as bearing wear) during equipment debugging. For example, for an asynchronous motor driven by a power frequency power supply, the preset frequencies can be set to twice the power frequency (100 Hz) and the characteristic frequencies related to the rotor speed of the motor, which are usually characteristic frequencies generated by the modulation effect of mechanical load fluctuations on motor current (such as sideband frequencies near the fundamental frequency). These frequency points are extremely sensitive to slight fluctuations in load caused by mechanical abnormalities (such as bearing wear and screw scraping), and the local controller automatically identifies and extracts the amplitude of the harmonic components at these preset frequency points.

[0093] Subsequently, the local controller compares the extracted harmonic component amplitudes with the monitoring thresholds established by the aforementioned method in real time. Once the amplitude of any harmonic component exceeds its corresponding monitoring threshold range, the local controller immediately determines that it is an abnormal state and generates a digitized warning signal. After the warning signal is attached with a timestamp and specific out-of-limit value, it is written into the production process electronic record in the memory; at the same time, the signal triggers a high-level interrupt to send the interrupt instruction to the main controller, instructing it to immediately suspend or terminate the main process currently being executed, so that the equipment enters a safe standby state.

[0094] As a more preferred embodiment, to further improve the long-term prediction ability of the equipment state, the method further includes a predictive maintenance warning generation logic, specifically: store the statistical characteristics (such as peak value or average value) of the harmonic component amplitude in each cycle within a plurality of consecutive production cycles in chronological order to form a health degree time series; periodically perform trend analysis on the health degree time series to calculate the change slope over time; when the change slope exceeds the preset degradation rate threshold, generate a predictive maintenance warning that is different from the warning signal used to interrupt production, indicating that the equipment has a gradual degradation risk.

[0095] Specifically, the extracted harmonic amplitude single feature (e.g. peak value or average value) in each production cycle is stored in a first-in-first-out queue in chronological order to form a dynamically updated health degree time series; a health degree evaluation module is periodically started and applies least square method for linear regression analysis on the data points of the health degree time series to calculate a change slope representing long-term change trend; the change slope is compared with a pre-set degradation rate threshold according to historical data of the equipment, and when the change slope exceeds the threshold, a predictive maintenance alarm that is distinguished from the warning signal for interrupting production is generated, and a notification is sent to the maintenance system through a human-computer interaction interface or a network interface to indicate that the equipment has a progressive degradation risk.

[0096] The alarm does not interrupt the current production, but is used to prompt the operator or maintenance department that the equipment has a progressive degradation risk, providing a basis for decision-making for planned maintenance.

[0097] As shown in Figure 3 The method combines the aforementioned judgment based on the instantaneous amplitude of harmonic components with the judgment based on the long-term change trend thereof to form a hierarchical logical decision-making process, thereby achieving comprehensive monitoring of immediate failures and progressive degradation risks.

[0098] The embodiment deepens the monitoring from the macroscopic average power consumption to the microscopic vibration feature level through spectral analysis of the driving power. This method can detect potential equipment failures caused by mechanical wear or raw material blockage at an extremely early stage, and achieve long-term prediction of the health status of the equipment through trend analysis; by immediately interrupting production and recording the event, not only is the production of batches of substandard products effectively avoided, ensuring the highest quality standards of medical consumables, but also precise data support is provided for subsequent fault diagnosis and quality traceability.

[0099] Further, the production process electronic record includes the control instruction, the warning signal, the operator identity and the related time stamp; the cloud server receives the production process electronic record and calculates a digital digest of the production process electronic record; binds the digital digest with a trusted time stamp provided by a trusted third-party time stamp service agency; and stores the electronic record, the digital digest and the trusted time stamp in a database with tamper-proof characteristics for integrity verification of the historical control behavior of the local controller; wherein the integrity verification is achieved by recalculating the digital digest of the stored electronic record when needed and comparing it with the original digital digest stored in the database and bound by the time stamp.

[0100] Specifically, when the warning signal is generated, the local controller automatically compiles a structured production process electronic record in the local memory immediately. The record, as a data set, contains the complete parameters of the control instructions issued to the preparatory process actuators at the time of the exception, the specific content of the warning signal triggering the exception (such as the harmonic amplitude exceeding the limit), the current operator identity obtained through login information, and the relevant timestamp provided by the system clock to the millisecond.

[0101] After the record is generated, the local controller encrypts and uploads the production process electronic record to a remote cloud server at a preset address through its network interface. After receiving the record, the cloud server performs a one-way hash operation (such as the SHA-256 algorithm) on all binary contents of the record to generate a fixed-length and unique string, which is the digital digest. The characteristics of the digest are that any minor changes in the original record will result in a completely different digest.

[0102] The cloud server binds the digital digest with a trusted timestamp provided by an authoritative third-party timestamp service, which can prove that the digital digest has existed at a certain time point and its content has not been tampered with. Finally, the cloud server stores the original electronic record, its corresponding digital digest, and the bound trusted timestamp in an internal database with strict access control and anti-tampering or only append write features, completing the evidence storage.

[0103] When subsequent quality traceability or audit is required, anyone can perform the same hash operation on the original electronic record retrieved from the server database and compare the result with the stored digital digest. If they are completely consistent, and the trusted timestamp is verified valid by the third-party institution, it can be strongly proven that the historical control behavior record of the local controller is complete and has not been tampered with since the time of storage.

[0104] The embodiment combines local records with centralized trusted third-party services, fundamentally solving the pain points of data tampering and traceability in medical consumable production processes. This not only provides strong technical evidence for production operation compliance audit, but also greatly improves the evidence credibility in quality accident accountability, ensuring the integrity and authenticity of the production process historical control behavior.

[0105] The embodiment realizes parallel starting of the cooling and sol process through real-time monitoring of the main process; by introducing a material property model associated with the raw material rheological melt index, the automatic process parameters are set, and the monitoring baseline dynamically matched with the material properties is established; the model prediction double-loop control circuit predicts the completion time of the parallel process in advance and adjusts online self-learning, accurately coordinates the end time of the two processes, and effectively compresses the overall production cycle; in addition, the concurrent abnormal monitoring mechanism can immediately find the process deviation, realize the improvement of production efficiency, stable guarantee of medical consumable quality and automatic control of production process.

[0106] Embodiment 2:

[0107] The embodiment of the above-mentioned intelligent control method for multi-process coordination management of an injection stretch blow molding machine is deployed on an all-electric injection stretch blow molding machine in a clean production workshop of medical consumables, which is used to produce a specific specification of thin-walled medical-grade PET material pipe blank.

[0108] At the beginning of the production task, the operator inputs the rheological melt index of the batch of PET raw materials into the control interface of the device. The material property model is called, the initial running instruction (including the core parameters such as screw speed and back pressure for sol) is generated according to the input rheological melt index, and the monitoring threshold value associated with the rheological melt index is calculated at the same time. The upper and lower limits of the monitoring threshold value are determined based on the statistical distribution of real-time power data corresponding to the rheological melt index in the historical production data.

[0109] With the start of the production cycle, after the injection molding is completed and the mold is smoothly closed on its double-track structure, the main process, i.e., the mold cooling process, begins. Through a non-contact infrared sensor array, temperature data on the surface of the mold cavity is continuously collected; these data are sent to a finite element analysis model based on the heat conduction equation. The model obtains the transient temperature field distribution of the pipe blank by solving, and identifies that the gate position at the bottom of the pipe blank is the key quality area in the entire cooling process, and its cooling rate determines the overall efficiency. To actively optimize this process, multiple independent auxiliary cooling units are immediately aligned with the key quality area and perform intensity-adjustable cooling to accelerate its cooling.

[0110] The model continuously calculates the remaining time required for the entire pipe blank to reach its glass transition temperature based on the changes in the temperature field, and this time is the main process variable. When the main process variable reaches a preset threshold value (indicating that the cooling has entered the second half), the system determines that the time for parallel starting is mature, and immediately starts the sol process as a preliminary process through the running instruction generated earlier, which is executed by the sol component driven by its independent motor.

[0111] Once the sol process is started, the model predictive double-loop control starts to dynamically coordinate the two parallel processes. At a certain moment, the loop predicts the remaining time of the cooling process and also calculates the estimated completion time of the sol process. By comparing the two, a small process synchronization deviation is generated, for example, indicating that the sol process has a tendency to end a few tenths of a second early. Based on this process synchronization deviation, the loop immediately generates an adjusted control command to dynamically reduce the screw speed by a small amount, thereby accurately reducing the deviation and ensuring that the end points of the two processes are accurately synchronized.

[0112] During the entire period of concurrent execution of the sol process, the real-time power generated by the sol motor is continuously compared with the previously set monitoring threshold through the motor's built-in sensor for sensing current size. During this process, the outer loop learning control also performs more intelligent online self-correction: through real-time power monitoring, it is found that the average power of the current batch of raw materials is slightly higher than the historical baseline value, which indicates that the actual fluidity of the raw materials is slightly lower than the nominal value. The rheological melt index used in internal operation is automatically dynamically corrected, thereby fine-tuning the back pressure setting and ensuring the consistency of the melt glue quality, which is crucial for ensuring the batch-to-batch stability of medical consumables.

[0113] In one production cycle, due to accidental contamination of the raw materials, the screw torque increased dramatically, and the amplitude of the harmonic component related to the motor load in the power spectrum graph suddenly increased, exceeding the upper limit of the absolute monitoring threshold. The local controller immediately determined that it was a serious abnormality within 0.1 seconds and triggered an interrupt command, causing the molding machine to safely pause production. At the same time of triggering the interrupt, an electronic production record was automatically compiled, including the last set of control commands before the interrupt, the specific content of the warning signal (such as the harmonic amplitude exceeding the limit), the current operator's identity information, and the relevant time stamp accurate to milliseconds. This record is transmitted to the cloud server, effectively avoiding the production of batch defects and providing tamper-proof data support for subsequent quality traceability.

[0114] In another scenario, after running for thousands of cycles, the local controller analyzes the long-term trend of a certain harmonic component amplitude and calculates that its change slope has exceeded the preset degradation rate threshold. At this time, a predictive maintenance warning is generated on the device's human-machine interface and the mobile application side, suggesting that the relevant mechanical parts be checked during the next planned maintenance, but production is not interrupted. This function converts potential sudden failures into controllable planned maintenance, greatly ensuring the continuity of the production line.

[0115] In the normal production cycle without any alarm, the main process and the preparation process are synchronously ended under the accurate cooperative control; after confirming that the two processes are synchronously completed, the current control cycle is ended and seamlessly connected to the next production cycle, so that the parallel processing capability of the equipment in the hardware design is fully exerted, and the maximization of the production efficiency and the energy saving effect is realized.

[0116] In the actual application, the production time of the single medical consumable is stably shortened from 14 seconds under the traditional serial control to 11 seconds, the time is saved by 3 seconds, and the production efficiency is improved by more than 21%, which is significantly higher than the theoretical value that can be achieved by only relying on the hardware upgrade; meanwhile, due to the predictive cooperative control, the unnecessary waiting of the equipment is greatly reduced, and the whole machine can realize about 30% of the comprehensive energy saving compared with the traditional control mode. Finally, through the above technical means, the production efficiency is improved, the quality of the medical consumable is stably ensured, and the highly automatic control of the production process is realized.

[0117] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application, therefore, the protection scope of the present application should be limited by the protection scope defined in the claims.

Claims

1. An intelligent control method for multi-process coordination management of an injection stretch blow molding machine, characterized in that, The application relates to a method for controlling the production of medical consumables, comprising: continuously monitoring main process variables in a main process of mold cooling; when the main process variables reach preset threshold values, generating operation instructions of a preparatory process executor based on a material property model associated with a rheological melt index, and starting a preparatory process in parallel with the main process; a model predictive double-loop control circuit continuously compares output results of a linear regression model for predicting the main process and a kinematic model for calculating the preparatory process, generates a process synchronization deviation, and outputs adjusted control instructions to the preparatory process executor according to the process synchronization deviation; after the main process and the preparatory process are synchronized, the current control period ends. The process of continuously monitoring main process variables comprises: collecting temperature data of a mold cavity surface in real time through a non-contact infrared sensor array arranged in a clean environment; and processing a time sequence of the temperature data through a cooling time prediction model to generate the main process variables including a cooling remaining time prediction value.

2. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 1, characterized in that, The cooling time prediction model comprises: inputting the temperature data as a thermal boundary condition into a finite element analysis model based on a heat conduction equation; solving the finite element analysis model to obtain a transient temperature field distribution of the mold cavity surface; and calculating a remaining time required for the entire mold cavity to reach a glass transition temperature according to the transient temperature field distribution as the cooling remaining time prediction value included in the main process variables.

3. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 2, characterized in that, The material property model associated with the rheological melt index comprises: pre-establishing a database including multiple known rheological melt indexes and corresponding optimal screw rotation speeds and optimal back pressures for producing qualified medical consumables through an experimental calibration method; generating a feedforward model for calculating the screw rotation speed and the back pressure based on the database by using a curve fitting algorithm, the feedforward model being capable of interpolating and calculating corresponding screw rotation speed set values and back pressure set values according to an inputted arbitrary rheological melt index.

4. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 1, characterized in that, The process of generating operation instructions of the preparatory process executor comprises: inputting the rheological melt index into the feedforward model to interpolate and calculate corresponding screw rotation speed set values and back pressure set values, the operation instructions being a group of control parameters including the screw rotation speed set values and the back pressure set values; generating a monitoring threshold value associated with the rheological melt index; extracting a time sequence of real-time power data of batch production for producing qualified medical consumables by a local controller; calculating an average value and a standard deviation of the time sequence; and setting upper and lower limits of the monitoring threshold value as the average value plus or minus a preset multiple of the standard deviation.

5. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 1, characterized in that, ​ 6. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 1, characterized in that, The generation process of the process synchronization deviation comprises: based on the time rate of change of the main process variable, establishing a linear regression model to dynamically predict the remaining completion time of the main process; based on the real-time position and speed feedback of the preparatory process executor, establishing a kinematics model to calculate the estimated completion time of the preparatory process in real time; comparing the remaining completion time with the estimated completion time to generate the process synchronization deviation.

7. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 1, characterized in that, The model predictive double-loop control circuit comprises an inner loop for real-time control and an outer loop for online learning: The inner loop generates a control instruction according to the current value, historical cumulative value and change trend of the process synchronization deviation through a PID controller and outputs the control instruction to the preparatory process executor; the outer loop analyzes the correlation between the historical process synchronization deviation and the historical production data containing the change of the rheological melt index through an adaptive filtering algorithm, and online corrects the linear regression model parameters for predicting the remaining completion time of the main process and the kinematics model parameters for calculating the estimated completion time of the preparatory process.

8. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 7, characterized in that, The method further comprises a monitoring and warning process performed concurrently with the adjustment process of the model predictive double-loop control circuit: The real-time power generated by the preparatory process executor is compared with a monitoring threshold as a real-time process variable to generate a warning signal; Fast Fourier transform is applied to the time series data of the real-time power to generate a power spectrum diagram; harmonic components at preset frequencies associated with electromagnetic noise of the motor driving the preparatory process executor are identified in the power spectrum diagram; The amplitude of the harmonic components is compared with the monitoring threshold to generate the warning signal and write into a production process electronic record; the warning signal is used to output an interrupt instruction to the local controller to interrupt the execution of the main process.

9. The intelligent control method for multi-process synergic management of an injection stretch blow molding machine according to claim 8, characterized in that, The production process electronic record comprises the control instruction, the warning signal, operator identity and related time stamp: A cloud server receives the production process electronic record and calculates a digital digest of the production process electronic record; The digital digest is bound with a trusted timestamp provided by a trusted third-party timestamp service agency; The electronic record, digital digest and trusted timestamp are stored together in a database with anti-tampering features for integrity verification of the historical control behavior of the local controller; wherein the integrity verification is achieved by recalculating the digital digest of the stored electronic record when needed and comparing it with the original digital digest stored in the database and bound with the timestamp.

Images