Intelligent blow molding and wall thickness closed loop control system for PET plastic bottle

Abstract

The application belongs to the technical field of intelligent manufacturing of plastic packaging containers, and specifically discloses an intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles, which integrates feedforward correction and closed-loop feedback, pre-corrects process parameters by acquiring material characteristic parameters, generates point cloud data through full-circumferential scanning of a non-contact three-dimensional optical measuring device, constructs a partitioned three-dimensional model and accurately calculates the wall thickness, dynamically adjusts blow molding process parameters based on wall thickness deviation data through a mapping relationship database, configures a hierarchical quality control strategy, a multivariate adaptive feedback mechanism and an online incremental learning mechanism. The system realizes adaptive compensation under conditions such as material fluctuation and environmental disturbance, improves the wall thickness control precision and product consistency, reduces manual intervention, guarantees high-speed continuous production, has self-learning and long-term stable regulation and control capabilities, and provides technical support for lightweight and high-precision molding of PET bottles.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology for plastic packaging containers, and relates to an intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles. Background Technology

[0002] Polyethylene terephthalate (PET) plastic bottles are widely used in the packaging of beverages, food, pharmaceuticals, and daily chemical products due to their lightweight, transparency, moderate strength, and good recyclability. In modern high-speed packaging production systems, blow molding is the core process in PET bottle manufacturing, and its molding quality directly determines the physical properties, appearance consistency, and material utilization efficiency of the final product.

[0003] Currently, most mainstream PET bottle blow molding equipment in the industry adopts open-loop or semi-closed-loop control strategies based on preset parameters. Specifically, such systems typically set key process parameters such as mold temperature, tension rod stroke, high-pressure gas pressure, and application timing before blow molding, based on empirical data or offline simulation results, and maintain these parameters relatively constant during production. Although some advanced equipment has introduced local feedback mechanisms for blow pressure or mold cooling status, its control logic is still limited to the static adjustment of a single or a few physical quantities, lacking the ability to perceive and dynamically respond to the actual wall thickness distribution of the bottle in real time.

[0004] The inherent characteristics of the above-mentioned technical solutions at the principle level have gradually revealed their limitations in addressing new challenges. On the one hand, the wall thickness distribution of PET bottles is not solely determined by preset mechanical and pneumatic parameters, but is a complex result influenced by a variety of dynamic factors, including batch differences in raw materials, fluctuations in ambient temperature and humidity, the coupling effect of gas dynamics during pre-blowing and main-blowing stages, and the uniformity of preform heating.

[0005] On the other hand, under high-speed production cycles, these disturbances often act in a nonlinear and time-varying manner, resulting in a significant deviation between the actual wall thickness distribution of the molded bottle and the ideal design. Traditional control systems, lacking closed-loop feedback on the wall thickness state after molding, cannot identify the source and evolution of such deviations, making it difficult for the system to adaptively compensate for drifts in operating conditions such as changes in raw material viscosity or minor wear of the mold.

[0006] On the other hand, the market demand for lightweight bottles and integrated structural functions requires wall thickness control to reach sub-millimeter level or even higher precision, and to be able to implement differentiated control for different areas of the bottle. The control architecture of the existing blow molding system has fundamental deficiencies in information perception dimension and decision response speed. Its open-loop or local feedback mechanism cannot build a complete closed loop from molding result to process execution, thus making it difficult to support truly intelligent control.

[0007] Therefore, how to construct a closed-loop control system that can sense the wall thickness distribution characteristics of PET bottles in real time and dynamically adjust the blow molding process parameters based on this feedback information, so as to maximize material optimization and process self-adaptation while ensuring structural strength, has become a key challenge and a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0008] In view of this, in order to solve the problems mentioned in the background technology, a smart blow molding and wall thickness closed-loop control system for PET plastic bottles is proposed.

[0009] The objective of this invention can be achieved through the following technical solution: This invention provides an intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles, comprising:

[0010] The feedforward correction module obtains the raw material information for the current batch of production, analyzes the systematic wall thickness offset based on the raw material information, and performs feedforward correction based on the systematic wall thickness offset.

[0011] The data acquisition module uses a non-contact three-dimensional optical measurement device to perform a full-circumferential scan of the outer surface of the PET bottle and acquire point cloud data of the bottle surface.

[0012] The model building module constructs a three-dimensional geometric model of the PET bottle based on the point cloud data, and divides the three-dimensional geometric model into multiple structural regions, including the bottle shoulder region, the bottle body region, and the bottle bottom region.

[0013] The wall thickness analysis module compares the point cloud data of the outer surface of the bottle with the preset geometric contour of the inner cavity based on the three-dimensional geometric model to calculate the actual wall thickness value of each structural region.

[0014] The deviation analysis module compares the actual wall thickness of each structural region with the corresponding target wall thickness setting value to generate wall thickness deviation data for each structural region.

[0015] The closed-loop control module dynamically adjusts the blow molding process parameters of the corresponding structural area in the next blow molding cycle based on the wall thickness deviation data, including the timing of the main blowing pressure, the downward speed of the tension rod, and the local mold cooling intensity.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By constructing a multi-dimensional process parameter and wall thickness mapping relationship database, the present invention can accurately correct the main blowing sequence, stretching speed and local cooling intensity while ensuring the safety of parameter adjustment, thereby avoiding equipment failure or molding failure due to parameter exceeding limits.

[0017] 2. This invention integrates trend drift compensation and raw material batch feedforward compensation, identifies systematic operating condition drift through time series deviation analysis, and implements feedforward correction based on raw material physical property parameters. The two work together to significantly improve the system's adaptability to slow-changing disturbances such as raw material fluctuations and mold wear.

[0018] 3. This invention sets up a graded quality control strategy, dynamically adjusts the monitoring frequency and process parameters based on the wall thickness deviation statistics, and automatically initiates diagnostic and conservative control processes when an anomaly occurs. This not only ensures product consistency but also reduces the frequency of manual intervention, thus guaranteeing the stable operation of high-speed continuous production.

[0019] 4. This invention utilizes an online model evolution mechanism, along with a sliding time window and cross-validation to continuously update the mapping relationship database, enabling the control system to possess long-term self-learning capabilities. This effectively addresses equipment performance degradation and process environment evolution, ensuring that the system maintains high-precision wall thickness control performance throughout its entire lifecycle. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the system module connections of the present invention.

[0022] Figure 2 This is a schematic diagram of the overall production structure layout of the present invention.

[0023] Figure 3 This is a schematic diagram of the process parameter correction flow for the closed-loop control module. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1As shown, the present invention provides an intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles, including a feedforward correction module, a data acquisition module, a model building module, a wall thickness analysis module, a deviation analysis module, and a closed-loop control module. The feedforward correction module is connected to the data acquisition module, the data acquisition module is connected to the model building module, the model building module is connected to the wall thickness analysis module, the wall thickness analysis module is connected to the deviation analysis module, and the deviation analysis module is connected to the closed-loop control module.

[0026] It should be noted that you should refer to [link / reference]. Figure 2 As shown, the entire system consists of a blow molding station, a non-contact 3D optical measurement device, and a rejection station arranged sequentially along the conveyor line. After heating, the PET preform is fed into the blow molding station, where it is blow molded using a tension rod and high-pressure gas. After demolding, it travels downstream with the non-contact 3D optical measurement device for a full-surface 3D scan. After measurement, the PET bottle continues along the conveyor line to the rejection station. If it is determined to be defective, a pneumatic pusher actuator pushes it out of the main conveyor path.

[0027] The feedforward correction module obtains the raw material information for the current batch of production, analyzes the systematic wall thickness offset based on the raw material information, and performs feedforward correction based on the systematic wall thickness offset.

[0028] In a preferred embodiment of the present invention, the feedforward correction module is specifically configured as follows: read the radio frequency identification tag on the preform tray to obtain the characteristic parameters of the current batch of raw materials, the characteristic parameters including melt flow index, crystallization rate and thermal shrinkage rate parameters.

[0029] It should be explained that the melt flow index reflects the fluidity of the raw material, the crystallization rate is related to the degree of crystallization during molding, and the thermal shrinkage rate is related to the dimensional deformation after cooling. Differences among these three factors can lead to systematic wall thickness deviations. By acquiring these parameters through RFID tags, they can be input into a pre-trained model to accurately calculate the deviation. This allows for the pre-correction of process parameters before the first blow molding, offsetting disturbances caused by batch-to-batch variations in raw materials. This provides a foundation for closed-loop control, addresses the shortcomings of existing technologies in handling raw material fluctuations, and ensures the accuracy of wall thickness control.

[0030] The characteristic parameters are input into a pre-trained wall thickness influence prediction model to calculate the systematic wall thickness offset of each structural region that may be caused by this batch of raw materials under the baseline process parameters.

[0031] It should be noted that the wall thickness influence prediction model has been trained based on historical data. The input layer consists of melt flow index, crystallization rate, and thermal shrinkage rate, while the output layer is the systematic wall thickness offset of the shoulder, body, and bottom regions under the baseline process. The offset output by the model is converted into a feedforward correction. Preferably, if a higher melt flow index results in a 0.12mm thinner body, the feedforward pressure is increased by 0.15MPa. This feedforward correction and the closed-loop feedback correction are superimposed in the closed-loop control module to form the final execution parameters.

[0032] Furthermore, the training method of the wall thickness influence prediction model is as follows: during the system initialization phase, multiple blow molding experiments are performed, and the parameters other than the target process parameters are fixed. The target parameters are adjusted according to a preset step size. The PET bottles produced by each group of experiments are 3D scanned and the wall thickness is calculated to obtain the actual wall thickness values ​​of each structural region under different parameters. The difference between the actual wall thickness and the target wall thickness setting value is calculated, and a univariate response function of characteristic parameters and wall thickness offset is constructed. All functions are integrated and optimized by combining the process parameter records of historical blow molding batches and the corresponding wall thickness measurement results to complete the pre-training, so as to achieve accurate calculation of systematic wall thickness offset under the benchmark process.

[0033] Based on the systematic wall thickness offset, before the first blow molding of this batch of preforms, the timing of the main blowing pressure, the downward speed of the stretching rod, and the local mold cooling intensity are pre-corrected using a feedforward method.

[0034] It's important to explain that the timing of the main blowing pressure, the downward speed of the stretching rod, and the local mold cooling intensity are the core process parameters for PET bottle blow molding, directly determining the bottle's molding quality and wall thickness uniformity. The timing of the main blowing pressure controls the bottle's expansion rhythm, ensuring uniform stretching in all areas; the downward speed of the stretching rod is related to the stretch ratio, affecting wall thickness distribution and structural strength; and the local mold cooling intensity adjusts the cooling rate, influencing the raw material's crystallization state and thermal shrinkage effect. Based on systematic wall thickness offset pre-correction, disturbances such as batch differences in raw materials can be offset in advance, laying the foundation for closed-loop control, effectively avoiding wall thickness deviations, ensuring product consistency and molding stability, and supporting high-precision blow molding production.

[0035] The feedforward correction is superimposed on the feedback correction output by the current closed-loop control algorithm, and together they constitute the final execution parameters of the current blow molding cycle.

[0036] Preferably, a method of superimposing feedforward correction and feedback correction is as follows: the feedforward correction amount is calculated by the feedforward correction module to obtain the timing of the main blowing pressure, the downward speed of the tension rod, and the local mold cooling intensity; then the closed-loop control module outputs the feedback correction amount based on the wall thickness deviation data; the two types of correction amounts are fused according to the linear superposition rule, the weight of the feedforward correction amount is fixed to a preset value, and the weight of the feedback correction amount is dynamically matched to the degree of wall thickness deviation; after superposition, the result is checked against the equipment limit; if it exceeds the range of the actuator's stroke or pressure output, the total correction amount is scaled proportionally to the compliant range, and finally the final execution parameters of the current blow molding cycle are formed.

[0037] The data acquisition module uses a non-contact three-dimensional optical measurement device to perform a full-circumferential scan of the outer surface of the PET bottle and acquire point cloud data of the bottle surface.

[0038] It should be noted that the installation method of the non-contact three-dimensional optical measurement device is as follows: a ring-shaped support is set downstream of the blow molding station, and eight laser contour sensors are installed at equal intervals along the circumference of the support. Each sensor is fixed to the inside of the support by a rigid connector, and its optical axis points to the central axis of the PET bottle on the conveyor line.

[0039] In a preferred embodiment of the present invention, the acquisition of point cloud data on the surface of the bottle is specifically configured as follows: the non-contact three-dimensional optical measurement device includes multiple laser contour sensors arranged in a ring.

[0040] After the PET bottle is demolded and transferred to the measurement station, the multiple laser contour sensors are triggered to simultaneously emit laser beams, which irradiate the outer surface of the PET bottle.

[0041] It receives laser signals reflected from the outer surface of PET bottles and calculates the coordinates of local cross-sectional contour points corresponding to each laser contour sensor based on the principle of triangulation.

[0042] The coordinates of local cross-sectional contour points collected by all laser contour sensors are stitched together according to their spatial position to generate an original point cloud dataset covering the entire height and circumference of the bottle.

[0043] The original point cloud dataset is resampled and meshed to generate structured point cloud data.

[0044] The model building module constructs a three-dimensional geometric model of the PET bottle based on the point cloud data, and divides the three-dimensional geometric model into multiple structural regions, including the bottle shoulder region, the bottle body region, and the bottle bottom region.

[0045] It's important to explain the reason for dividing the structure into regions: the purpose is to adapt to the molding characteristics and quality control requirements of different parts of the PET bottle. The geometry, stress conditions, and molding process requirements of the shoulder, body, and bottom areas differ, resulting in different target wall thickness settings. This division allows for targeted processing, precise calculation of the actual wall thickness and deviation data for each region, and provides a basis for adjusting the process parameters of the closed-loop control module. This ensures that parameters such as the main blowing pressure and timing are adapted to the molding needs of different regions, while also supporting the precise implementation of graded quality control strategies. This prevents local wall thickness deviations from exceeding limits due to overall control, ultimately ensuring the overall wall thickness uniformity and structural stability of the bottle.

[0046] It should be noted that the shoulder area is defined as the region from the lower edge of the bottle neck to the shoulder inflection point, the body area is the cylindrical segment from the shoulder inflection point to the bottom starting point, and the bottom area is the curved surface region from the bottom starting point to the center of the bottom. The axial boundaries of each structural region are determined by the feature height coordinates of the corresponding bottle shape and are marked with different colors or labels in the model for easy subsequent partitioning.

[0047] It should be added that the system loads the bottle shape parameters of the current production model from the product design database, including key dimensional information such as the total bottle height, shoulder starting height, and bottom ending height. According to the preset spatial boundary definition rules, this module divides the reconstructed 3D geometric model into three structural regions along the bottle body axis: the shoulder region, the body region, and the bottom region.

[0048] In a preferred embodiment of the present invention, the construction of the three-dimensional geometric model of the PET bottle is specifically configured as follows: based on the original geometric parameters of the preform and the theoretical model of the stretch ratio during the blow molding process, an initial internal cavity preset geometric contour is generated.

[0049] The structured point cloud data is compared with the initial internal cavity preset geometric contour by calculating the normal distance to obtain the preliminary wall thickness distribution.

[0050] If the wall thickness in the initial wall thickness distribution exceeds the preset material physical limit or the abrupt change in wall thickness gradient with the adjacent area exceeds the set threshold, then the internal cavity contour is determined to have a modeling error.

[0051] It should be explained that the material's physical limit range is a critical wall thickness range determined based on the inherent properties of PET material and the feasibility of the blow molding process. Specifically, it refers to the minimum wall thickness of PET material that, after blow molding, can ensure the structural strength of the bottle and prevent cracking or embrittlement, as well as the maximum wall thickness that can prevent molding defects caused by material accumulation and uneven cooling. This range is a quantitative standard preset based on the mechanical properties and processing fluidity of PET material, used to determine the rationality of the internal cavity contour modeling. If the calculated local wall thickness exceeds this range, it indicates that the modeling does not conform to the actual molding capacity of the material, and the internal cavity contour needs to be optimized through finite element reverse analysis.

[0052] It should be explained that a sudden change in wall thickness gradient in adjacent areas exceeding a set threshold refers to a rapid change in the wall thickness value of adjacent spatial locations within a short distance in the initial wall thickness distribution of the PET bottle. This change rate exceeds the pre-set quantitative standards based on the processing flowability, molding stability, and bottle structural strength requirements of the PET material. This threshold is a key indicator for determining the rationality of the internal cavity contour modeling. If this condition is triggered, it indicates that the initial modeling does not accurately reflect the actual molding state. It is necessary to iteratively optimize the internal cavity contour node positions through finite element inverse analysis combined with measured external surface deformation data to ensure the accuracy of subsequent wall thickness calculations for each structural region and provide reliable data support for closed-loop control.

[0053] Furthermore, the definition of the adjacent area is as follows: based on the structural partition of the PET bottle three-dimensional geometric model, with the wall thickness sampling points in each structural area as the benchmark, it refers to the local range that is directly adjacent in spatial position in the same structural area, including the area corresponding to the adjacent sampling points that are uniformly distributed circumferentially on the same measurement section, and the local area corresponding to the adjacent measurement sections that are equally spaced axially, to ensure that the judgment of the wall thickness gradient change focuses on the similar areas of continuous molding, which is in line with the actual molding law and wall thickness distribution characteristics of PET bottles.

[0054] For areas with modeling errors, the finite element inverse analysis method is used, combined with measured external surface deformation data, to iteratively optimize the position of the internal cavity contour nodes.

[0055] It's important to explain why error assessment and iterative optimization of the inner cavity contour node positions are necessary: ​​to ensure the accuracy of PET bottle wall thickness calculations and provide reliable data support for closed-loop control. The initial inner cavity contour is generated based on a theoretical model and may deviate from the actual molding state. Without error assessment, the initial wall thickness distribution could exceed the material's physical limits or result in unreasonable abrupt changes in wall thickness gradients, leading to distortion in the actual wall thickness calculation. However, by using finite element reverse analysis combined with measured external surface deformation data for iterative optimization, the inner cavity contour can be made to conform to the actual bottle molding situation, correcting modeling deviations and ensuring the accuracy of subsequent calculations of wall thickness values ​​and deviation data for each structural region. This, in turn, ensures the targeted adjustment of blow molding process parameters, ultimately achieving high-precision control of bottle wall thickness and meeting product quality requirements.

[0056] The optimized inner cavity profile is used as the final preset geometric profile of the inner cavity to accurately calculate the actual wall thickness of each structural region.

[0057] The wall thickness analysis module compares the point cloud data of the outer surface of the bottle with the preset geometric contour of the inner cavity based on the three-dimensional geometric model to calculate the actual wall thickness value of each structural region.

[0058] Preferably, the steps for calculating the actual wall thickness of each structural region are as follows: 1. Spatially align the constructed 3D geometric model of the PET bottle with the optimized final internal cavity preset geometric contour to ensure that the coordinate systems of the bottle shoulder area, bottle body area, and bottle bottom area are completely consistent; 2. Select equally spaced measurement sections along the axial direction in each structural region, and uniformly extract no less than a preset number of wall thickness sampling points along the circumference of each measurement section; 3. For each sampling point, obtain its outer surface spatial coordinates in the 3D geometric model based on the principle of triangulation, and simultaneously determine the corresponding projection point of the point on the final internal cavity preset geometric contour; 4. Calculate the normal distance between the outer surface coordinates of each sampling point and the corresponding internal cavity projection point, which is the actual wall thickness value of the sampling point; 5. Summarize the actual wall thickness values ​​of all sampling points in the same structural region to form complete actual wall thickness distribution data for each structural region, and complete the calculation of the actual wall thickness value of each region.

[0059] The deviation analysis module compares the actual wall thickness of each structural region with the corresponding target wall thickness setting value to generate wall thickness deviation data for each structural region.

[0060] In a preferred embodiment of the present invention, the specific configuration for generating wall thickness deviation data for each structural region is as follows: retrieve the target wall thickness distribution map of the current production model of PET bottle from the product design database. The target wall thickness distribution map defines the target wall thickness setting values ​​for the bottle shoulder region, bottle body region, and bottle bottom region.

[0061] The constructed three-dimensional geometric model is spatially aligned with the target wall thickness distribution map to ensure that the coordinate system of each structural region is consistent.

[0062] Within each structural region, multiple equally spaced measurement sections are selected, and at least a preset number of wall thickness sampling points are uniformly extracted along the circumference of each measurement section.

[0063] Calculate the algebraic difference between the actual wall thickness value and the corresponding target wall thickness setting value at each wall thickness sampling point to form the wall thickness deviation vector of the structural region.

[0064] The wall thickness deviation vector is statistically analyzed to calculate its mean, standard deviation, and maximum absolute deviation value, which are used as the wall thickness deviation data for each structural region.

[0065] Preferably, five measurement sections are selected equidistantly along the axial direction within each structural region. Eight wall thickness sampling points are uniformly extracted circumferentially from each section, resulting in a total of 40 sampling points in the shoulder region, 40 in the body region, and 40 in the bottom region. For each sampling point, the algebraic difference between its actual wall thickness and the target value is calculated, forming a wall thickness deviation vector containing 120 elements. This vector is then statistically processed by region grouping, outputting the mean, standard deviation, and maximum absolute deviation of the wall thickness deviation for each region.

[0066] In a preferred embodiment of the present invention, the wall thickness deviation data is configured with a graded quality control strategy, specifically as follows: when the maximum absolute deviation value of the wall thickness deviation vector of any structural region exceeds the first warning threshold but does not exceed the second alarm threshold, the system records the bottle as an edge qualified product and starts the process parameter fine-tuning process, while increasing the measurement frequency of the next three consecutive PET bottles.

[0067] It's important to explain the reason for configuring a tiered quality control strategy: PET bottle wall thickness deviations vary in severity. Simply judging them as acceptable or unacceptable can easily lead to minor deviations escalating into batch defects, or excessive control can negatively impact production efficiency. Fine-tuning parameters and intensifying monitoring for borderline acceptable products can promptly curb the spread of deviations; removing severely defective products and investigating anomalies reduces waste. This approach ensures product consistency while adapting to the demands of high-speed continuous production, providing a precise basis for closed-loop control.

[0068] Furthermore, the first warning threshold and the second alarm threshold are set based on the following: The target wall thickness of the current production model of PET bottle is used as a benchmark, combined with the physical limits of PET material, the equipment's process control capabilities, and reference to the wall thickness deviation distribution data of stable and qualified products in historical blow molding batches. Simultaneously, the structural strength requirements of the product's usage scenario are considered. The first warning threshold is set within the range where the equipment can correct deviations through fine-tuning parameters. The second alarm threshold is set as a critical value that exceeds the material's safe molding range or cannot be effectively compensated for by the equipment, potentially leading to product failure. This ensures the targeted and feasible nature of the tiered control. The physical limits of PET material include, but are not limited to, the minimum safe wall thickness and the maximum molding wall thickness. The equipment's process control capabilities include, but are not limited to, the effective fine-tuning range of parameters such as main blowing pressure and stretching speed.

[0069] Preferably, the measurement frequency for the next three consecutive PET bottles is increased from once per bottle to three times per bottle.

[0070] When the maximum absolute deviation exceeds the second alarm threshold, the system determines that the bottle is a defective product, triggers the rejection mechanism to remove it from the production line, immediately freezes the current process parameter set, and starts the anomaly diagnosis subroutine.

[0071] The anomaly diagnosis subroutine retrieves the process parameter logs, environmental temperature and humidity records, and preform heating temperature distribution diagrams for the most recent preset number of blow molding cycles to analyze whether there are any abnormal sources such as equipment actuator failure, gas pipeline leakage, or heating lamp power attenuation.

[0072] Preferably, through correlation analysis, if the fluctuation in the position of the tension rod is found to exceed [a certain threshold], [further action can be taken]. If the pressure decreases, it is determined that the tension rod is stuck; if the infrared image shows that the heating temperature of a certain area is lower than the set value... The above indicates that the heating lamp power has decreased.

[0073] If an identifiable source of the anomaly is identified, a maintenance prompt message is generated and pushed to the operation terminal.

[0074] If no identifiable source of anomaly is found, the system will automatically switch to a conservative set of process parameters and implement enhanced monitoring for a preset number of blow molding cycles.

[0075] Preferably, the conservative set of process parameters includes: increasing the main blowing pressure. The stretching speed decreased .

[0076] The closed-loop control module dynamically adjusts the blow molding process parameters of the corresponding structural area in the next blow molding cycle based on the wall thickness deviation data, including the timing of the main blowing pressure, the downward speed of the tension rod, and the local mold cooling intensity.

[0077] For a preferred embodiment of the present invention, please refer to Figure 3 As shown, the specific configuration for dynamically adjusting the blow molding process parameters of the corresponding structural region in the next blow molding cycle is as follows: a mapping relationship database between the blow molding process parameters and the wall thickness deviation of each structural region is established. The mapping relationship database is obtained by training based on the process parameter records and corresponding wall thickness measurement results in historical blow molding batches.

[0078] For any structural region in the current blow molding cycle, the mapping database is queried to determine the process parameter correction amount that matches the wall thickness deviation data.

[0079] The process parameter correction is superimposed on the baseline process parameters used in the current blow molding cycle to generate an updated set of process parameters for the next blow molding cycle.

[0080] The feasibility of the updated process parameter set is verified to determine whether it is within the range of the travel limit of the equipment actuator and the gas pressure output limit.

[0081] Preferably, based on the equipment manufacturing manual, the downward speed of the tension rod must meet 100mm / s-800mm / s, and the main blowing pressure must meet 0.8MPa-4.0MPa.

[0082] If the updated process parameter set passes the feasibility verification, it will be used as the execution parameter for the next blow molding cycle.

[0083] If the feasibility verification fails, the process parameter correction amount is scaled proportionally until the limit constraint conditions are met, and then the final execution parameters are generated.

[0084] Preferably, the specific method of scaling is as follows: clearly define the constraint boundaries such as the travel limit of the equipment actuator and the gas pressure output limit; calculate the over-limit range of the current process parameter correction amount after superimposing the reference parameter; based on the maximum or minimum allowable limit of the equipment, scale the correction amount proportionally according to the ratio between the original correction amount and the over-limit range, ensuring that the process parameter obtained by superimposing the scaled correction amount and the reference parameter is completely within the equipment limit range; and keep the correction direction unchanged during the scaling process, and finally generate the final execution parameters that meet the constraint conditions.

[0085] In a preferred embodiment of the present invention, the specific configuration for establishing the mapping relationship database between blow molding process parameters and wall thickness deviations of each structural region is as follows: During the system initialization phase, multiple sets of blow molding tests are executed. In each set of tests, all process parameters except for one target process parameter are fixed, and the target process parameter is adjusted according to a preset step size.

[0086] Three-dimensional scanning and wall thickness calculation were performed on the PET bottles produced in each blow molding test to obtain the actual wall thickness values ​​of each structural region under different target process parameter values.

[0087] Based on the difference between the actual wall thickness value and the target wall thickness setting value, a univariate response function between the target process parameters and the wall thickness deviation is constructed.

[0088] Repeat the above process to establish individual univariate response functions for the timing of the main blowing pressure, the downward speed of the tension rod, and the local mold cooling intensity, respectively.

[0089] All univariate response functions are integrated into a unified data structure to form a mapping database.

[0090] In a preferred embodiment of the present invention, the closed-loop control module is configured with a multivariable adaptive feedback mechanism, specifically as follows: after each blow molding cycle, the wall thickness deviation data of each structural region generated in the current blow molding cycle is compared with the wall thickness deviation data corresponding to the previous cycle in a time sequence to determine whether the wall thickness deviation shows a trend of drift.

[0091] If the wall thickness deviation of any structural region is detected to change in the same direction for two consecutive cycles and the change exceeds a preset threshold, the dynamic compensation module for the process parameters corresponding to that structural region is activated.

[0092] The dynamic compensation module generates parameter correction suggestions based on parameter adjustment records under similar deviation patterns in the historical mapping relationship database.

[0093] It should be noted that the definition of similar deviation pattern is as follows: based on the data characteristics in the historical mapping relationship database, the core is to match the consistency between the current and historical wall thickness deviations from multiple dimensions: taking the same structural region as a premise, focusing on the temporal trend and magnitude of the wall thickness deviation, while matching the statistical characteristics of the deviation vector, and ensuring that the process environment disturbance factors and raw material characteristic parameters are consistent when the deviation occurs. Historical deviation patterns that meet the above-mentioned dimensional matching conditions are judged as similar deviation patterns to the current deviation.

[0094] The parameter correction suggestions are weighted and fused with the correction amount obtained from the current mapping relation database query, wherein the weight coefficient of trend drift is dynamically adjusted according to the rate of deviation change.

[0095] Preferably, a difference operation is performed on the two-period data to calculate the change in deviation of each structural region. If a certain structural region has two consecutive periods The same sign indicates that the deviation continues to increase or decrease, and If this is not the case, the dynamic compensation unit for that structural region is activated. This unit accesses the historical database, retrieves successful correction records with similar deviation evolution patterns, and extracts the corresponding forward-looking correction strategy. This forward-looking correction suggestion is then fused with the correction amount output by the closed-loop control module according to weights. The weighting coefficients for trend drift are relative to the correction amount obtained from the mapping database. By the rate of change of deviation Real-time calculation, in which This refers to the blow molding cycle time. .

[0096] Finally, the integrated correction will be applied to update the process parameters for the next blow molding cycle.

[0097] It should be noted that the reason for using a weighted fusion of the correction amount obtained from trend drift analysis and the correction amount obtained from the mapping database to determine the final correction is to balance immediate deviation correction and trend disturbance prediction, thereby improving the accuracy and stability of process adjustments. The correction amount obtained from the mapping database is a direct adaptation adjustment to the current wall thickness deviation based on historical data, enabling rapid response to immediate deviations. Meanwhile, the parameter correction suggestions are derived from historical successful records of similar trend deviation patterns, allowing for targeted responses to continuous periodic trend drifts. By dynamically adjusting the weighting coefficients of trend drift, the fused correction amount can both address current deviations and proactively offset slow-changing disturbances, preventing the deviation from continuously expanding. Simultaneously, it adapts to dynamic changes in operating conditions, ensuring long-term consistency of wall thickness control and meeting the adaptive control requirements of the closed-loop control system.

[0098] In a preferred embodiment of the present invention, the mapping relationship database is configured with an online incremental learning and model evolution mechanism, specifically as follows: During normal system operation, the process parameter set used in each blow molding cycle, the corresponding measured wall thickness deviation data, and environmental disturbance factors are continuously collected.

[0099] It should be noted that environmental disturbance factors refer to dynamic variables in the production environment that affect the bottle molding quality and wall thickness distribution during the PET bottle blow molding process. These factors primarily include key environmental parameters such as ambient temperature and humidity, air pressure fluctuations, and air cleanliness in the production workshop. These factors indirectly alter the wall thickness by affecting raw material flowability, mold heat dissipation efficiency, and gas dynamics. They are crucial data supports for the system to build a mapping database, achieve precise adjustment of process parameters, and enable online model evolution. Therefore, they must be collected synchronously with process parameters and wall thickness deviation data to ensure control accuracy.

[0100] Newly collected data samples are labeled and stored in the historical data pool, and model update tasks are triggered periodically.

[0101] In the model update task, a sliding time window mechanism is used to select data from the most recent N effective production cycles and refit the response function between each process parameter and the wall thickness deviation.

[0102] Cross-validate the newly fitted response function with the old function stored in the current database and calculate the prediction error reduction rate.

[0103] It should be noted that the method for calculating the prediction error reduction rate is as follows: Using cross-validation data as a benchmark, select effective samples from the historical data pool, independent of the new and old function training sets. These samples include process parameters, environmental disturbance factors, and corresponding measured wall thickness deviations. Use both the old response function stored in the current database and the newly fitted response function to predict the wall thickness deviation for this batch of data. Calculate the error between the two sets of predicted values ​​and the measured values. Record the old error as... New error Using formulas The prediction error reduction rate was calculated.

[0104] If the prediction error reduction rate is greater than the preset convergence threshold, the old function is replaced with the new response function, and the indexing mechanism is updated.

[0105] It should be noted that the preset convergence threshold is set as follows: It is based on the iterative optimization objective of the wall thickness prediction model, combined with the minimum error reduction rate that effectively improves prediction accuracy from historical model update data, the current production accuracy requirements for wall thickness control, and also takes into account the system's computational resource consumption and update time cost. This threshold is determined through multiple offline simulations. This threshold must satisfy the following condition: only when the prediction error reduction rate of the new response function exceeds this value will the accuracy improvement brought by the model update outweigh the cost of the system update, ensuring that the indexing mechanism and response function updates have practical application value and avoiding ineffective iterations.

[0106] If the prediction error reduction rate is less than or equal to the preset convergence threshold, the original function is retained and the time window length is increased for retraining.

[0107] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications or additions should fall within the protection scope of the present invention.

Claims

1. A closed-loop control system for intelligent blow molding and wall thickness control of PET plastic bottles, characterized in that: include: The feedforward correction module obtains the raw material information of the current batch of production, analyzes the systematic wall thickness offset based on the raw material information, and performs feedforward correction based on the systematic wall thickness offset. The data acquisition module uses a non-contact three-dimensional optical measurement device to perform a full-circumferential scan of the outer surface of the PET bottle and acquire point cloud data of the bottle surface. The model building module constructs a three-dimensional geometric model of the PET bottle based on the point cloud data, and divides the three-dimensional geometric model into multiple structural regions, including the bottle shoulder region, the bottle body region, and the bottle bottom region. The wall thickness analysis module compares the point cloud data of the outer surface of the bottle with the preset geometric contour of the inner cavity based on the three-dimensional geometric model to calculate the actual wall thickness of each structural region. The deviation analysis module compares the actual wall thickness of each structural region with the corresponding target wall thickness setting value to generate wall thickness deviation data for each structural region. The closed-loop control module dynamically adjusts the blow molding process parameters of the corresponding structural area in the next blow molding cycle based on the wall thickness deviation data, including the main blowing pressure action sequence, the downward speed of the tension rod, and the local mold cooling intensity. The specific configuration of the feedforward correction module is as follows: Read the RFID tag on the preform tray to obtain the characteristic parameters of the current batch of raw materials, including melt flow index, crystallization rate and thermal shrinkage rate. The characteristic parameters are input into a pre-trained wall thickness influence prediction model to calculate the systematic wall thickness offset of each structural region that may be caused by this batch of raw materials under the baseline process parameters. Based on the aforementioned systematic wall thickness offset, before the first blow molding of this batch of preforms, the timing of the main blowing pressure, the downward speed of the stretching rod, and the local mold cooling intensity are pre-corrected using a feedforward method. The feedforward correction is superimposed on the feedback correction output by the current closed-loop control algorithm, and together they constitute the final execution parameters of the current blow molding cycle. The specific configuration for constructing the three-dimensional geometric model of the PET bottle is as follows: Based on the original geometric parameters of the preform and the theoretical model of the stretch ratio during the blow molding process, the initial internal cavity preset geometric contour is generated. The structured point cloud data is compared with the initial internal cavity preset geometric contour by calculating the normal distance to obtain the preliminary wall thickness distribution; If the wall thickness value in a local area in the initial wall thickness distribution exceeds the preset material physical limit or the abrupt change in wall thickness gradient with the adjacent area exceeds the set threshold, it is determined that there is a modeling error in the inner cavity contour. For areas with modeling errors, the finite element inverse analysis method is used, combined with measured external surface deformation data, to iteratively optimize the position of internal cavity contour nodes; The optimized inner cavity contour is used as the final preset geometric contour of the inner cavity.

2. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 1, characterized in that: The specific configuration for acquiring the point cloud data of the bottle surface is as follows: The non-contact three-dimensional optical measurement device includes multiple laser profile sensors arranged in a ring. After the PET bottle is demolded and transferred to the measurement station, the multiple laser contour sensors are triggered to simultaneously emit laser beams to irradiate the outer surface of the PET bottle. Receive the laser signal reflected from the outer surface of the PET bottle and calculate the coordinates of the local cross-sectional contour point corresponding to each laser contour sensor; The coordinates of local cross-sectional contour points collected by all laser contour sensors are stitched together according to their spatial position to generate an original point cloud dataset covering the entire height and circumference of the bottle. The original point cloud dataset is resampled and meshed to generate structured point cloud data.

3. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 1, characterized in that: The specific configuration for generating wall thickness deviation data for each structural region is as follows: Retrieve the target wall thickness distribution map of the current production model of PET bottle from the product design database. The target wall thickness distribution map defines the target wall thickness setting values ​​for the bottle shoulder area, bottle body area and bottle bottom area. The constructed three-dimensional geometric model is spatially aligned with the target wall thickness distribution map to ensure that the coordinate system of each structural region is consistent; Within each structural region, select multiple equally spaced measurement sections, and uniformly extract no less than a preset number of wall thickness sampling points along the circumference on each measurement section; Calculate the algebraic difference between the actual wall thickness value and the corresponding target wall thickness setting value at each wall thickness sampling point to form the wall thickness deviation vector of the structural region; The wall thickness deviation vector is statistically analyzed to calculate its mean, standard deviation, and maximum absolute deviation value, which are used as the wall thickness deviation data for each structural region.

4. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 3, characterized in that: The wall thickness deviation data is configured with a graded quality control strategy, as follows: When the maximum absolute deviation value of the wall thickness deviation vector of any structural region exceeds the first warning threshold but does not exceed the second alarm threshold, the system records this bottle as an edge qualified product and starts the process parameter fine-tuning process, while increasing the measurement frequency of the next three consecutive PET bottles. When the maximum absolute deviation exceeds the second alarm threshold, the system determines that the bottle is a defective product, triggers the rejection mechanism to remove it from the production line, immediately freezes the current process parameter set, and starts the anomaly diagnosis subroutine. The anomaly diagnosis subroutine retrieves the process parameter logs, environmental temperature and humidity records, and preform heating temperature distribution diagrams for the most recent preset number of blow molding cycles to analyze whether there are any sources of equipment actuator failure, gas pipeline leakage, or abnormal power attenuation of heating lamps. If an identifiable source of the anomaly is identified, a maintenance prompt message is generated and pushed to the operation terminal. If no identifiable source of anomaly is found, the system will automatically switch to a conservative set of process parameters and implement enhanced monitoring for a preset number of blow molding cycles.

5. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 1, characterized in that: The specific configuration for dynamically adjusting the blow molding process parameters of the corresponding structural region in the next blow molding cycle is as follows: A mapping relationship database between blow molding process parameters and wall thickness deviations of each structural region is established. The mapping relationship database is obtained by training based on process parameter records and corresponding wall thickness measurement results in historical blow molding batches. For any structural region in the current blow molding cycle, the mapping database is queried to determine the process parameter correction amount that matches the wall thickness deviation data. The process parameter correction amount is superimposed on the baseline process parameters used in the current blow molding cycle to generate an updated process parameter set for the next blow molding cycle. The feasibility of the updated process parameter set is verified to determine whether it is within the range of the travel limit of the equipment actuator and the gas pressure output limit. If the updated process parameter set passes the feasibility verification, it will be used as the execution parameter for the next blow molding cycle; If the feasibility verification fails, the process parameter correction amount is scaled proportionally until the limit constraint conditions are met, and then the final execution parameters are generated.

6. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 5, characterized in that: The database establishing the mapping relationship between blow molding process parameters and wall thickness deviations of each structural region is configured as follows: During the system initialization phase, multiple sets of blow molding tests are performed. In each set of tests, all process parameters except for one target process parameter are fixed, and the target process parameter is adjusted according to a preset step size. Three-dimensional scanning and wall thickness calculation were performed on the PET bottles produced by each blow molding test to obtain the actual wall thickness values ​​of each structural region under different target process parameter values. Based on the difference between the actual wall thickness value and the target wall thickness setting value, a univariate response function between the target process parameters and the wall thickness deviation is constructed. Repeat the above process to establish individual univariate response functions for the timing of the main blowing pressure, the downward speed of the tension rod, and the local mold cooling intensity, respectively. All univariate response functions are integrated into a unified data structure to form a mapping database.

7. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 6, characterized in that: The closed-loop control module is configured with a multivariable adaptive feedback mechanism, as detailed below: After each blow molding cycle, the wall thickness deviation data of each structural region generated in the current blow molding cycle is compared with the wall thickness deviation data of the previous cycle in a time sequence to determine whether the wall thickness deviation shows a trend of drift. If the wall thickness deviation of any structural region is detected to change in the same direction for two consecutive cycles and the change exceeds the preset threshold, the dynamic compensation module for the process parameters corresponding to that structural region is activated. The dynamic compensation module generates parameter correction suggestions based on parameter adjustment records in the similar deviation pattern in the historical mapping relationship database. The parameter correction suggestions are weighted and fused with the correction amount obtained from the current query of the mapping relationship database, wherein the weight coefficient of the trend drift is dynamically adjusted according to the rate of change of the deviation; Finally, the integrated correction will be applied to update the process parameters for the next blow molding cycle.

8. The intelligent blow molding and wall thickness closed-loop control system for PET plastic bottles according to claim 6, characterized in that: The mapping database is configured with an online incremental learning and model evolution mechanism, as detailed below: During normal system operation, the system continuously collects the set of process parameters used in each blow molding cycle, the corresponding measured wall thickness deviation data, and environmental disturbance factors. Newly collected data samples are labeled and stored in the historical data pool, and model update tasks are triggered periodically. In the model update task, a sliding time window mechanism is used to select data from the most recent N effective production cycles and refit the response function between each process parameter and the wall thickness deviation. Cross-validate the newly fitted response function with the old function stored in the current database and calculate the prediction error reduction rate. If the prediction error reduction rate is greater than the preset convergence threshold, the old function is replaced with the new response function, and the indexing mechanism is updated. If the prediction error reduction rate is less than or equal to the preset convergence threshold, the original function is retained and the time window length is increased for retraining.

Images