Full-automatic filling and tail sealing method and system based on image recognition analysis

Abstract

The invention discloses a full-automatic filling and tail sealing method and system based on image recognition analysis, and relates to the technical field of image analysis, and the method comprises the following steps: extracting fluid dynamic control parameters and bubble group dynamic characteristic parameters; according to fluid dynamic control parameters, the annular microelectrode array is controlled to generate a directional electric field, and the liquid flowing state is regulated and controlled; ultrasonic action parameters are determined according to the bubble group dynamic characteristic parameters, and a frequency-adjustable ultrasonic generator is controlled to output ultrasonic modulation pulses; detecting the temperature distribution of the sealing area of the container, and applying differential heat sealing parameters to different temperature areas; and establishing a multi-layer perception neural network optimization controller, and optimizing a control parameter combination by taking the quality index of the filling and end-capping process as an optimization target. The filling method combining image recognition and electric field regulation and control is adopted, accurate control over the liquid flowing state is achieved, the problem that in the traditional filling process, liquid drawing and flow cutoff are not accurate is solved, and the filling precision and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly to a fully automatic filling and sealing method and system based on image recognition and analysis. Background Art

[0002] In the field of filling and sealing, automated filling equipment has been widely used, but still faces many technical challenges. Traditional filling equipment usually adopts a metering pump or a volumetric metering device and operates relying on a mechanical control system. The filling accuracy and efficiency are limited to a certain extent. During the production process, due to the influence of liquid surface tension, viscosity differences, and environmental factors, problems such as uneven filling, liquid overflow, or insufficient filling amount are likely to occur. At the same time, the existing filling systems have limited ability to handle bubbles, and the bubbles formed during the filling process are difficult to effectively eliminate, affecting product quality.

[0003] In the sealing process, the existing technology mainly relies on a constant temperature heating device for sealing, and it is difficult to meet the sealing requirements of containers with different materials and shapes. The problem of uneven temperature distribution in the sealing area of the container during the sealing process is widespread, which easily leads to inconsistent sealing strength. Most of the filling and sealing equipment on the market currently adopts an independent control system, lacking an overall coordination mechanism and intelligent analysis ability. The adjustment of production parameters mainly relies on manual experience, and it is difficult to achieve efficient parameter optimization and adaptive adjustment, restricting the further development of the filling and sealing process. Summary of the Invention

[0004] The present invention provides a fully automatic filling and sealing method and system based on image recognition and analysis, which is used to solve the technical problems of low liquid filling accuracy, difficult bubble elimination, and uneven sealing temperature.

[0005] In view of this, the first aspect of the present invention provides a fully automatic filling and sealing method based on image recognition and analysis, including: Constructing a three-dimensional dynamic model of the fluid, and extracting fluid dynamic control parameters and dynamic characteristic parameters of the bubble group through image recognition; According to the fluid dynamic control parameters, controlling a ring-shaped microelectrode array arranged around the filling nozzle to generate a directional electric field and regulating the liquid flow state; Determining ultrasonic action parameters according to the dynamic characteristic parameters of the bubble group and the physical property parameters of the container liquid, and controlling an adjustable frequency ultrasonic generator to output ultrasonic modulation pulses to the bubble aggregation area; Detecting the temperature distribution in the sealing area of the container, and controlling a locally adjustable hot head according to the detection result to apply differential heat sealing parameters to different temperature areas; Establishing a multi-layer perceptron neural network optimization controller, receiving a set of system state parameters, and optimizing the control parameter combination with the quality index of the filling and sealing process as the optimization goal to complete the adaptive adjustment of the parameters in the whole filling and sealing process.

[0006] Optionally, the fluid dynamic control parameters include the surface stress tensor field, the three-dimensional flow velocity vector distribution, and the liquid surface motion state parameters. The dynamic characteristic parameters of the bubble group include the bubble volume concentration gradient, the bubble equivalent particle size distribution, and the three-dimensional coordinates of the bubble aggregation region.

[0007] Optionally, generating a directional electric field by the annular microelectrode array arranged around the filling nozzle includes: Arranging a plurality of microelectrodes around the filling nozzle to form an annular microelectrode array; Generating a control parameter set for each microelectrode based on the fluid dynamic control parameters; Applying the control parameter set to the annular microelectrode array to generate a directional electric field; Forming an electric stress distribution at the liquid-gas interface through the directional electric field to adjust the liquid flow state; At the end of filling, based on the liquid surface motion state parameters, forming a convergent electric field gradient pointing to the center of the filling nozzle by adjusting the microelectrode voltage distribution to control the liquid cut-off.

[0008] Optionally, controlling the tunable frequency ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation region includes: Calculating the transmission loss coefficient and the liquid attenuation coefficient of the ultrasonic wave according to the physical property parameters of the container liquid; Determining the bubble resonance frequency range based on the bubble equivalent particle size distribution; Adjusting the beam focusing parameters of the tunable frequency ultrasonic generator based on the bubble volume concentration gradient and the three-dimensional coordinates of the bubble aggregation region; Configuring the tunable frequency ultrasonic generator as a pulse output mode, and determining the pulse period and pulse intensity according to the transmission loss coefficient, the liquid attenuation coefficient, and the bubble resonance frequency range; Outputting ultrasonic modulation pulses to the bubble aggregation region through the tunable frequency ultrasonic generator.

[0009] Optionally, applying differential heat sealing parameters to different temperature regions includes: Scanning the container sealing region by using thermal imaging technology to obtain the temperature distribution data of the container sealing region; Performing temperature gradient analysis according to the temperature distribution data to identify the temperature abnormal regions in the container sealing region; Based on the temperature abnormal regions, dividing the container sealing region into multiple independent control blocks, and calculating the temperature characteristic parameters of each control block; Extracting the standard heat sealing parameters corresponding to the container material; According to the temperature characteristic parameters of each control block and the standard heat sealing parameters, generating differential heat sealing parameters for each control block through parameter compensation calculation; Apply corresponding differential heat-sealing parameters to each control block through a locally adjustable heat-sealing head to perform the heat-sealing operation.

[0010] Optionally, establishing a multi-layer perceptron neural network optimization controller includes: Construct a multi-layer perceptron neural network optimization controller, which includes an input layer, a hidden layer, and an output layer. The input layer receives the system state parameter set, and the output layer generates an optimal control parameter combination; Taking the filling and sealing process quality index evaluation system as the optimization goal, train the multi-layer perceptron neural network optimization controller; Configure a dual-mode control mechanism for the multi-layer perceptron neural network optimization controller; Through the parameter sensitivity analysis mechanism of the multi-layer perceptron neural network optimization controller, optimize the control parameter combination to complete the adaptive adjustment of the parameters in the whole filling and sealing process.

[0011] Optionally, identifying the temperature abnormal area in the container sealing area includes: Convert the temperature distribution data into a two-dimensional temperature matrix; Introduce a multi-scale thermal anomaly detection mechanism, perform multi-scale decomposition on the two-dimensional temperature matrix through wavelet transform, establish a thermal feature subspace, and extract abnormal feature points; Calculate the temperature gradient vector between adjacent points in the two-dimensional temperature matrix to obtain a temperature gradient field, and mark the temperature abnormal candidate points according to a preset temperature gradient threshold; Merge the abnormal feature points with the temperature abnormal candidate points to obtain an enhanced temperature abnormal candidate point set; Perform connectivity analysis on the enhanced temperature abnormal candidate point set to obtain the temperature abnormal area.

[0012] The second aspect of the present invention provides a fully automatic filling and sealing system based on image recognition and analysis, including: A parameter analysis module, used to construct a three-dimensional dynamic model of the fluid, and extract fluid dynamic control parameters and bubble group dynamic characteristic parameters through image recognition; An electric field regulation module, used to control the annular microelectrode array arranged around the filling nozzle to generate a directional electric field according to the fluid dynamic control parameters to regulate the liquid flow state; An ultrasonic modulation module, used to determine the ultrasonic action parameters according to the bubble group dynamic characteristic parameters and the container liquid physical property parameters, and control the frequency-adjustable ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation area; A heat-sealing adjustment module, used to detect the temperature distribution in the container sealing area, and control the locally adjustable heat-sealing head according to the detection result to apply differential heat-sealing parameters to different temperature areas; A control optimization module is used to establish a multi-layer perceptron neural network optimization controller, receive a set of system state parameters, optimize the control parameter combination with the quality index of the filling and sealing process as the optimization goal, and complete the adaptive adjustment of the parameters in the whole filling and sealing process.

[0013] The beneficial effects of the present invention are as follows: The present invention adopts a filling method combining image recognition and electric field regulation, realizing precise control of the liquid flow state, effectively solving the problems of liquid drawing and inaccurate flow interruption in the traditional filling process, and improving the filling accuracy and efficiency; The directional ultrasonic technology based on the dynamic characteristics of bubbles can specifically eliminate the bubbles in the container, greatly reducing the bubble residue rate in the product, and at the same time avoiding damage to the active ingredients of the liquid, ensuring the product quality; The differential heat-sealing parameter application system solves the problems such as poor sealing reliability caused by uneven container sealing temperature in the traditional sealing process, improving the consistency of the sealing strength and the product shelf life; The multi-layer perceptron neural network optimization control system realizes the intelligent collaborative adjustment of the parameters in the whole filling and sealing process, breaking through the limitations of independent control in traditional links, enabling the system to have the adaptive ability to cope with complex and changeable production scenarios, and greatly improving the production stability and product consistency. Description of the Drawings

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a flowchart of a full-automatic filling and sealing method based on image recognition and analysis.

[0016] Figure 2 It is a flowchart of the ultrasonic modulation pulse output of a full-automatic filling and sealing method based on image recognition and analysis.

[0017] Figure 3 It is a flowchart of determining the temperature abnormal area of a full-automatic filling and sealing method based on image recognition and analysis. Detailed Embodiments

[0018] To make the objectives, features, and advantages of the present invention more apparent and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0019] Example 1. Refer to Figures 1 to 3 , which is the first embodiment of the present invention. This embodiment provides a fully automatic filling and sealing method based on image recognition and analysis. The flowchart of this method is as shown in Figure 1 and includes: S1: Use a dual-frequency flash stereoscopic imaging system to construct a three-dimensional dynamic model of the fluid, and extract fluid dynamic control parameters and bubble group dynamic characteristic parameters through image recognition.

[0020] Among them, the fluid dynamic control parameters include the surface stress tensor field, three-dimensional flow velocity vector distribution, and liquid surface motion state parameters. The bubble group dynamic characteristic parameters include the bubble volume concentration gradient, bubble equivalent particle size distribution, and three-dimensional coordinates of the bubble aggregation region.

[0021] First, turn on the dual-frequency flash stereoscopic imaging system, take multi-angle pictures of the fluid at different times, and obtain multiple sets of fluid image data. Preprocess the multiple sets of fluid image data: enhance the image through histogram equalization to improve the distinguishability between the fluid and bubbles; use Gaussian filtering to remove interference such as thermal noise to ensure clear images.

[0022] Next, use an image matching algorithm to process images from different perspectives, identify the corresponding relationships of the same features; combine the principle of stereoscopic vision to convert two-dimensional image information into three-dimensional space information and construct a three-dimensional dynamic model of the fluid. For fluid physical property data, calibrate according to standard methods to reduce measurement errors; for multi-source spatial coordinate data, establish a conversion model to unify the coordinate system.

[0023] Subsequently, based on the constructed three-dimensional dynamic model of the fluid, extract various parameters: analyze the surface information of the fluid, calculate the surface stress tensor field according to the theory of fluid mechanics; use particle image velocimetry technology to track the movement trajectories of tracer particles to obtain the three-dimensional flow velocity vector distribution; monitor the changes in the liquid surface height and tilt angle, and correct considering factors such as liquid surface reflection and fluctuations to obtain the liquid surface motion state parameters; count the proportion of bubble volume in different regions and its spatial changes to obtain the bubble volume concentration gradient; segment and identify the bubbles, measure the sizes and perform statistical analysis to obtain the bubble equivalent particle size distribution; mark the relatively dense regions of bubbles, and combine three-dimensional information to determine the three-dimensional coordinates of the bubble aggregation region.

[0024] Finally, the obtained fluid dynamic control parameters and bubble group dynamic characteristic parameters are used to precisely control the generation of a directional electric field by the annular microelectrode array, so that the liquid flow state meets the filling requirements, and appropriate ultrasonic action parameters are determined to eliminate bubbles, improving product quality and production efficiency.

[0025] S2: According to the fluid dynamic control parameters, control the annular microelectrode array arranged around the filling nozzle to generate a directional electric field, regulate the liquid flow state, and suppress the liquid drawing phenomenon at the end of filling.

[0026] In a specific embodiment of the present invention, step S2 specifically includes: S2.1: Arrange a plurality of independently adjustable microelectrodes in a circumferential direction around the filling nozzle to form an annular microelectrode array.

[0027] Among them, the number of microelectrodes is 8 - 16, evenly distributed. Each microelectrode is made of an anti-corrosion alloy, with a working voltage range of 0 - 3000V, an electrode size of 1.0mm × 3mm, and a spacing from the filling channel maintained at 3.0 - 5.0mm to ensure sufficient field strength without disturbing fluid flow. By arranging the annular microelectrode array around the filling nozzle, the liquid flow state can be flexibly and precisely regulated, breaking through the limitation of single liquid control means in traditional filling technology.

[0028] S2.2: Based on the fluid dynamic control parameters, generate a control parameter set for each microelectrode.

[0029] First, perform eigen-decomposition on the surface stress tensor field to determine the maximum principal stress direction and stress gradient distribution; then, decompose the three-dimensional velocity vector into radial and tangential components, determine the vorticity distribution by calculating the curl of the velocity field, and define the region with vorticity greater than the critical threshold as the vortex core region, where the critical threshold is 1.5 times the average vorticity or 30% of the maximum vorticity, taking the larger value of the two; then, combine the fluctuation frequency information in the liquid surface motion state parameters to optimize the microelectrode response timing; subsequently, generate the microelectrode voltage distribution according to the stress gradient distribution and superimpose the electric field rotation compensation in the vortex core region; in addition, configure the phase difference between adjacent microelectrodes in the vortex core region, and keep the phase synchronous in the non-vortex core region; finally, calculate the hydrodynamic response time according to the Reynolds number and Weber number of the fluid, and generate the microelectrode trigger timing accordingly; integrate the voltage distribution, phase difference, and trigger timing to output the microelectrode control parameter set. Generating the microelectrode control parameter set based on the fluid dynamic control parameters can flexibly adjust the electric field according to the real-time state of the fluid, effectively improving the adaptability to complex fluid flow states and the controllability of liquid flow during filling.

[0030] S2.3: Apply the control parameter set to the annular microelectrode array to generate a directional electric field with spatial non-uniformity and time-varying characteristics.

[0031] First, the voltage distribution parameters are converted into driving voltage signals of each microelectrode, and then output to the corresponding microelectrode through the high-voltage signal amplification module; then, the multi-channel signal generator is configured according to the phase difference parameters to realize the phase difference control of adjacent microelectrodes; then, according to the trigger timing parameters, the driving circuit of each microelectrode is activated in time-sharing through the central control unit; at the same time, the actual output voltage and phase signal of each microelectrode are monitored in real time to verify the spatial non-uniformity and time-varying characteristics; when the electric field parameter deviation exceeds the limit, hierarchical compensation control is performed, including adjacent electrode coordinated compensation, trigger timing redistribution, and control parameter regeneration request. The deviation exceeds the limit here specifically refers to: the driving voltage deviates from the set value by ±5%; the actual phase difference deviates from the set value by ±3°; the electric field intensity distribution deviates from the preset mode by more than 10%.

[0032] Among them, spatial non-uniformity verification refers to confirming that the electric field forms a non-uniform field (including intensity gradient distribution and / or direction gradient distribution) that conforms to the preset distribution pattern within the array coverage area, and time-varying characteristic verification refers to confirming that the changes of electric field parameters over time conform to the preset dynamic change laws.

[0033] S2.4: The electric stress distribution is formed at the liquid-gas interface through a directional electric field to regulate the liquid flow state.

[0034] First, the electric stress distribution on the liquid-gas interface is calculated based on the electric field intensity distribution generated by the annular microelectrode array, which is proportional to the square of the electric field intensity. Then, the calculated electric stress distribution is converted into an equivalent surface tension correction and superimposed with the surface stress tensor field extracted from S1 to obtain the corrected surface force distribution. Then, based on the liquid surface motion state parameters and three-dimensional flow velocity vector distribution extracted from S1, the liquid surface stability index (defined as the ratio of the local curvature change rate of the liquid surface to the average curvature) is calculated. When the stability index exceeds the preset liquid surface fluctuation allowable threshold of 0.25, the electric field adjustment is performed: The target electric field distribution is calculated according to the liquid surface state, and the voltage adjustment parameters and phase adjustment parameters of the microelectrode are generated, wherein the preset liquid surface fluctuation allowable threshold is determined through pre-experimental calibration according to the fluid characteristics and filling accuracy requirements; then, the voltage adjustment parameters and phase adjustment parameters are applied to the microelectrode array to reconstruct the electric field distribution; the reconstructed electric field generates corrected electric stress at the liquid-gas interface, changes the local surface tension balance of the liquid surface, induces surface flow, and thus regulates the overall liquid flow state; finally, the changes in the liquid surface motion state parameters are continuously monitored, and the electric field control strategy is updated in real time to form a closed-loop control to maintain the stability of the filling process.

[0035] It is worth noting that the strict verification of the spatial non-uniformity and time-varying characteristics of the electric field and the hierarchical compensation control mechanism ensure the stable operation of the electric field, reduce filling anomalies caused by electric field deviations, and improve the stability and reliability of the entire filling process.

[0036] S2.5: At the end of filling, based on the liquid surface motion state parameters, a convergent electric field gradient pointing to the center of the filling nozzle is formed by adjusting the microelectrode voltage distribution to control liquid breakage.

[0037] First, by monitoring the liquid volume and flow rate, when the preset conditions are met, it is determined that the end stage of filling is entered. The preset conditions include that the liquid volume reaches 95% of the target value and the flow rate is reduced to below 0.1 m / s. Then, based on the liquid surface motion state parameters, the microelectrode voltage distribution is reconfigured: the voltage of the peripheral electrodes is set higher than that of the electrodes in the central region to form an electric field gradient pointing to the center of the filling nozzle. Next, the phases of the microelectrodes are adjusted to be in-phase to ensure the consistency of the electric field direction. Subsequently, the change in the liquid neck diameter is monitored. When the liquid neck diameter drops to the breakage trigger diameter (usually 10% of the inner diameter of the filling nozzle), an electric field pulse sequence is applied to accelerate liquid breakage. In addition, after liquid breakage, a weakened electric field (intensity reduced to 30% of the peak value) is briefly maintained to eliminate residual droplets and prevent liquid drawing. Finally, the electric field is smoothly turned off according to the exponential decay curve to complete the filling process. The special electric field adjustment means designed for the end stage of filling effectively solves the problem of liquid drawing that has long troubled the filling process, improves the filling accuracy, reduces material loss, and improves product consistency.

[0038] S3: Determine the ultrasonic action parameters according to the dynamic characteristic parameters of the bubble group and the physical property parameters of the container liquid, and control the frequency-adjustable ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation region to eliminate the bubbles generated during the filling process.

[0039] In a specific embodiment of the present invention, the flowchart of the ultrasonic modulation pulse output is as Figure 2 , and specifically includes: S3.1: Calculate the transmission loss coefficient and liquid attenuation coefficient of ultrasonic waves according to the physical property parameters of the container liquid.

[0040] Among them, the physical property parameters of the container liquid include the container structure characteristics and the physical property parameters of the liquid. The container structure characteristics include the acoustic impedance of the container material and the container wall thickness, and the physical property parameters of the liquid include the liquid viscosity, liquid density, and sound velocity.

[0041] First, based on the liquid density and sound velocity, calculate the liquid acoustic impedance. Then, combined with the acoustic impedance of the container material, calculate the interface sound energy transmission coefficient and reflection coefficient according to acoustic theory. Next, based on the container wall thickness, ultrasonic frequency, and sound velocity in the container material, calculate the phase difference of ultrasonic wave propagation in the container wall according to wave theory, and obtain the correction factor through transformation. In addition, considering that there are multiple reflections of ultrasonic waves at the interface between the liquid and the container wall, based on the reflection coefficient and transmission coefficient, calculate the total transmitted energy by summing an infinite geometric series. Finally, calculate the transmission loss value by comprehensively considering the total transmitted energy and the correction factor.

[0042] Meanwhile, based on the classical viscous attenuation theory, the liquid attenuation coefficient is calculated using the liquid viscosity, liquid density, sound speed, and ultrasonic frequency.

[0043] S3.2: Determine the bubble resonance frequency range based on the equivalent bubble size distribution.

[0044] First, divide the equivalent bubble size distribution data into multiple size intervals at a preset interval, and count the proportion of the number of bubbles in each interval; then, for the bubbles in each size interval, apply the Minnaert resonance formula to calculate the theoretical resonance frequency; next, consider the influence of liquid viscosity on bubble resonance, determine the viscous damping correction coefficient according to the liquid viscosity, and correct the theoretical resonance frequency; subsequently, calculate the frequency weighted average value according to the proportion of the number of bubbles, and determine the frequency interval covering the main bubble population (the cumulative proportion of the number exceeds 85%); finally, adjust the frequency interval boundary according to the acoustic response characteristics (including vibration amplitude, phase response, and damping characteristics) of bubbles with different equivalent sizes under the action of the sound field to obtain the bubble resonance frequency range. For example, when the vibration amplitude of a bubble with a certain equivalent size changes violently near a certain frequency, the interval boundary near that frequency can be appropriately widened; if there is an obvious mutation in the phase response, the boundary can also be adjusted accordingly.

[0045] Among them, the preset interval is the unit for dividing the equivalent bubble size determined according to the resolution accuracy of the image recognition system and the bubble size distribution range.

[0046] It should be noted that by dividing the bubble size in detail, considering the correction of the resonance frequency by liquid viscosity, and adjusting the frequency interval boundary in combination with the acoustic response characteristics, the determined frequency range can better cover the main bubble population. This method can make the ultrasonic wave more effectively excite bubble resonance, improve the bubble elimination efficiency to a certain extent, and improve the situation that it is difficult for traditional methods to accurately act on bubbles of different particle sizes.

[0047] S3.3: Adjust the beam focusing parameters of the tunable ultrasonic generator based on the bubble volume concentration gradient and the three-dimensional coordinates of the bubble aggregation region, so that the acoustic energy is directionally focused on the bubble aggregation region.

[0048] First, based on the bubble volume concentration gradient data, identify the regions in the container where the bubble concentration is higher than the preset concentration threshold, and determine the boundaries of the bubble aggregation regions that need to be processed. Subsequently, based on the three-dimensional coordinates of the bubble aggregation regions, use the ray tracing algorithm to calculate the propagation paths of sound waves from the generator to the discrete sampling points within the bubble aggregation regions. Then, correct the propagation paths in combination with the container structure and the liquid sound speed to determine the spatial pointing angle of the generator. Next, according to the geometric characteristics of the bubble aggregation regions, adjust the phased array element parameters of the generator, control the phase difference between the elements, and control the shape of the sound wavefront. In addition, by adjusting the phase difference and amplitude of each element, form a focused sound field that effectively covers the bubble aggregation regions. Finally, evaluate the focusing effect, and trigger automatic calibration when the focusing effect is lower than the set standard. Real-time monitor the changes in the bubble aggregation state, and dynamically adjust the beam focusing parameters at a preset frequency to ensure that the acoustic energy is always directionally focused on the current bubble aggregation region.

[0049] Among them, the beam focusing parameters include phased array element parameters, the phase difference between elements, and element amplitudes; the preset concentration threshold is determined according to the type of filling liquid and the filling rate; the set standard refers to that the uniformity of the acoustic energy distribution of the focused sound field in the target region is not less than 80%; the preset frequency refers to the time interval for updating the beam focusing parameters, which depends on the rate of change of the bubble aggregation state.

[0050] Importantly, adjust the beam focusing parameters based on the bubble volume concentration gradient and the three-dimensional coordinates of the bubble aggregation regions to achieve the directional focusing of acoustic energy on the bubble aggregation regions. This focusing method reduces the ineffective dispersion of ultrasonic energy and enhances the effect on bubbles in specific regions. At the same time, real-time monitoring and dynamic adjustment ensure that during the filling process, even if the bubble state changes, the bubbles can be continuously processed, improving the stability of the bubble elimination process and optimizing the problems of energy dispersion and poor treatment effect on bubbles in specific regions in traditional methods.

[0051] S3.4: Configure the tunable ultrasonic generator into a pulse output mode, and determine the pulse period and pulse intensity according to the transmission loss coefficient, liquid attenuation coefficient, and bubble resonance frequency range.

[0052] First, based on the transmission loss coefficient and the liquid attenuation coefficient, calculate the average attenuation value of the ultrasonic energy from the generator to each discrete sampling point within the bubble aggregation region; then, according to the bubble resonance frequency range, set the ultrasonic frequency and modulation bandwidth of the generator to ensure that the output ultrasonic frequency can effectively excite bubble resonance; next, determine the initial pulse intensity based on the average attenuation value and the container structure characteristics, so that the acoustic energy is sufficient to trigger bubble resonance when it reaches the bubble aggregation region; in addition, by analyzing the distribution of acoustic energy in the bubble aggregation region under different pulse duty cycles, adjust the pulse duty cycle to ensure the efficient utilization of acoustic energy; subsequently, combined with the dynamic characteristic parameters of the bubble group, design a pulse repetition frequency that matches the bubble resonance characteristics; at the same time, based on the energy distribution of ultrasonic waves in the bubble aggregation region and the bubble resonance characteristics, determine the pulse duration that can ensure the best resonance effect; finally, collect the data of different pulse parameter combinations and the corresponding bubble elimination efficiency, and organize them into a parameter optimization table; compare the real-time bubble elimination situation with the data in the parameter optimization table, and use an adaptive control algorithm to adjust the pulse period and pulse intensity, laying a foundation for the subsequent efficient elimination of bubbles.

[0053] S3.5: Output ultrasonic modulation pulses from the tunable ultrasonic generator to the bubble aggregation region to make the bubbles reach the resonance state and promote the floating and elimination of the bubbles.

[0054] Specifically, use shielded cables to connect the generator with the power supply and the control system. After the equipment starts, execute the self-check program, and feedback the result to the control system after the self-check is completed; subsequently, the control system sends the pulse parameters determined in S3.4 to the generator, and the generator receives and stores the parameters; then, the generator uses digital signal processing technology to generate and modulate the electrical pulse signal according to the stored pulse parameters, including frequency and duty cycle modulation, so that the generated pulse characteristics match the bubble resonance frequency range; then, amplify the power of the modulated electrical pulse signal and convert the amplified electrical signal into ultrasonic waves; in addition, according to the beam focusing parameters determined in S3.3, direct and focus the ultrasonic waves to the bubble aggregation region to improve the effect of exciting bubble resonance; finally, the focused ultrasonic waves act on the bubbles, prompting the bubbles to overcome the resistance and float to the liquid surface and be eliminated.

[0055] S3.6: Adaptively adjust the ultrasonic frequency, pulse period and pulse intensity according to the bubble elimination situation, and at the same time control the pulse intensity below the liquid component protection threshold.

[0056] Among them, the liquid component protection threshold refers to the highest pulse intensity value that will not cause the effective components of the filled liquid to decompose and denature due to thermal, mechanical effects, etc., thereby affecting the product quality. This threshold is determined by comprehensively analyzing the components of the filled liquid, combined with thermal stability tests, mechanical sensitivity tests, and reaction experiments under simulated ultrasonic action.

[0057] S4: Use thermal imaging technology to detect the temperature distribution in the sealing area of the container, control the local adjustable heat sealing head based on the detection results, and apply differentiated heat sealing parameters to different temperature areas.

[0058] In a specific embodiment of the present invention, step S4 specifically includes: S4.1: Use thermal imaging technology to scan the container sealing area and obtain temperature distribution data of the container sealing area.

[0059] S4.2: Perform temperature gradient analysis based on the temperature distribution data to identify abnormal temperature areas in the container sealing area.

[0060] Specifically, the temperature abnormality area determination flow chart is as follows: Figure 3 As shown, the method includes: pre-processing the temperature distribution data and converting it into a two-dimensional temperature matrix, wherein each element of the two-dimensional temperature matrix represents the temperature value at the corresponding position of the container sealing area; introducing a multi-scale thermal anomaly detection mechanism, performing multi-scale decomposition of the two-dimensional temperature matrix by wavelet transform, establishing thermal feature subspaces in different frequency domains, and extracting abnormal feature points respectively, and the steps are as follows: selecting an adaptive wavelet basis function according to the data characteristics and analysis requirements of the two-dimensional temperature matrix, and determining the number of decomposition layers of the wavelet transform; performing multi-layer wavelet decomposition of the two-dimensional temperature matrix by using the selected wavelet basis function, obtaining a low-frequency submatrix, a horizontal high-frequency submatrix, a vertical high-frequency submatrix, and a diagonal high-frequency submatrix in different frequency domains, and constructing a thermal feature subspace; calculating statistical features for each thermal feature subspace, setting a threshold according to the extracted feature vector, and marking points whose feature values exceed the threshold as abnormal feature candidate points; comprehensively considering the correlation and consistency of different thermal feature subspaces, determining points that present abnormal features in multiple subspaces as final abnormal feature points, and removing isolated abnormal points generated by noise.

[0061] Subsequently, the temperature gradient vectors between adjacent points in the two-dimensional temperature matrix are calculated to obtain the temperature gradient field in the container sealing area; a preset temperature gradient threshold is set, and points whose gradient values exceed the threshold are marked as temperature anomaly candidate points; the abnormal feature points are merged with the temperature gradient anomaly point set through a feature fusion algorithm to form an enhanced temperature anomaly candidate point set; a connectivity analysis is performed on the enhanced temperature anomaly candidate point set, and points with adjacent spatial positions and similar temperature characteristics are classified into the same temperature anomaly area; the area size, average temperature value and temperature standard deviation of each temperature anomaly area are calculated, and areas with an area smaller than the preset area threshold or a temperature standard deviation lower than the preset standard deviation threshold are eliminated to obtain the final temperature anomaly area.

[0062] Among them, the preset temperature gradient threshold, the preset area threshold and the preset standard deviation threshold are determined based on statistical analysis of historical data, characteristics of the container sealing area and production practice, normal temperature fluctuation range and expert experience.

[0063] S4.3: Based on the temperature anomaly region, divide the container sealing region into multiple independent control blocks, and calculate the temperature characteristic parameters of each control block.

[0064] Specifically, specifically, taking the boundary contour of the temperature anomaly region as the demarcation, divide the container sealing region into a temperature anomaly region and a temperature normal region; according to the preset control accuracy requirements, determine the minimum control unit size of the container sealing region; further divide the temperature anomaly region according to the minimum control unit size to generate temperature anomaly control blocks; divide the temperature normal region into several temperature normal control blocks according to the preset region division rules, and the region division rules include the maximum size limit of the control unit and the boundary continuity requirement; assign a unique identification code to each control block, and establish a control block index table; calculate the temperature characteristic parameters of each control block, including the block average temperature value, the highest temperature value, the lowest temperature value, the temperature standard deviation, and the temperature uniformity index; associate the temperature characteristic parameters of all control blocks with the control block index table to form a block-based temperature characteristic data structure of the container sealing region. This step breaks through the limitation of treating the entire sealing region uniformly in the traditional way, and can formulate personalized processing strategies for regions with different temperature characteristics, improving the pertinence and effectiveness of heat sealing parameter adjustment.

[0065] S4.4: Extract the standard heat sealing parameters corresponding to the container material from the preset material-temperature parameter database.

[0066] Among them, the standard heat sealing parameters include the reference heat sealing temperature, pressure, and time parameters.

[0067] S4.5: Generate differential heat sealing parameters for each control block through parameter compensation calculation based on the temperature characteristic parameters of each control block and the standard heat sealing parameters.

[0068] Specifically, a mapping relationship function between temperature characteristic parameters and the adjustment amount of heat-sealing parameters is established. This function describes the influence of temperature differences on heat-sealing temperature, pressure, and time parameters. For each control block, calculate the difference between the average temperature value of the block and the reference heat-sealing temperature, and based on the block temperature standard deviation and temperature uniformity index, determine the parameter adjustment direction and amplitude. According to the calculation results, determine the adjustment amounts of heat-sealing temperature, pressure, and time parameters respectively. When the temperature difference is positive, lower the heat-sealing temperature; when the temperature standard deviation is large, increase the heat-sealing pressure; when the temperature uniformity is low, extend the heat-sealing time. Add the reference heat-sealing parameters to the corresponding parameter adjustment amounts to generate the differential heat-sealing temperature, pressure, and time parameters for the control block. Perform smoothing processing on the differential heat-sealing parameters of adjacent control blocks to ensure continuous spatial variation of the parameters and avoid uneven sealing quality. This step can better adapt to the temperature changes at different positions in the container sealing area, improving the adaptability of the heat-sealing process to the actual condition of the container compared with the traditional method of fixed heat-sealing parameters, and effectively ensuring the consistency of sealing quality.

[0069] S4.6: Apply the corresponding differential heat-sealing parameters to each control block through a locally adjustable heat-sealing head with multi-region independent control function to perform the heat-sealing operation.

[0070] Specifically, map the spatial position information of the control block to the execution unit of the locally adjustable heat-sealing head to establish a one-to-one correspondence between the control block and the execution unit. Transmit the differential heat-sealing parameters of each control block to the control system of the locally adjustable heat-sealing head. Each execution unit of the locally adjustable heat-sealing head independently adjusts its own heat-sealing temperature, pressure, and action time according to the received differential heat-sealing parameters. Start the locally adjustable heat-sealing head to synchronously perform the heat-sealing operation on each control block in the container sealing area to form a complete seal.

[0071] S4.7: During the heat-sealing process, use thermal imaging technology to monitor the temperature changes of each control block in real time, and feedback the monitoring data to the control system to dynamically adjust the differential heat-sealing parameters to ensure the sealing quality.

[0072] S5: Establish a multi-layer perceptron neural network optimization controller, receive the system state parameter set, and optimize the control parameter combination with the quality index of the filling and sealing process as the optimization goal to complete the adaptive adjustment of the parameters in the whole filling and sealing process.

[0073] In a specific embodiment of the present invention, step S5 specifically includes: S5.1: Construct a multi-layer perceptron neural network optimization controller. The multi-layer perceptron neural network optimization controller includes an input layer, a hidden layer, and an output layer. The input layer receives the system state parameter set, and the output layer generates the optimal control parameter combination.

[0074] Specifically, a deep multi-layer perceptron neural network structure with memory units is constructed, including 1 input layer, an experimentally determined appropriate number of hidden layers (such as 4), and 1 output layer. At least one hidden layer is an LSTM structure for processing sequence data; a multi-channel feature mapping module is configured in the input layer to receive fluid dynamic control parameters, bubble swarm dynamic characteristic parameters, container temperature distribution data, and equipment status parameters to form a system state parameter set; a residual connection structure is set between the hidden layers to mitigate the problem of gradient disappearance caused by the increase in network depth; an attention mechanism module is embedded in the hidden layers to perform dynamic weight allocation for key features; a multi-task learning structure is designed in the output layer to simultaneously generate a combination of control parameters, that is, the optimal combination of the control parameters of the annular microelectrode array, the ultrasonic action parameters, and the heat-sealing parameters; the batch normalization processing technology is adopted to accelerate network training and improve the generalization ability of the model; the dropout mechanism is introduced to prevent overfitting, and an appropriate dropout rate (such as 0.3) is set.

[0075] S5.2: Based on the historical filling and sealing process data, taking the filling and sealing process quality index evaluation system as the optimization goal, train a multi-layer perceptron neural network optimization controller.

[0076] Specifically, collect historical filling and sealing process data, covering the system state parameter set, the corresponding control parameter combination, and the quality index data, perform cleaning and annotation, and divide them into a training set, a validation set, and a test set according to the ratio of 70%, 15%, and 15%; construct a multi-dimensional filling and sealing process quality index evaluation system, including filling accuracy index, liquid level stability index, bubble residue rate index, and sealing strength index; transform the multi-dimensional filling and sealing process quality index evaluation system into a weighted combined loss function, and the weight coefficients of each index are determined according to product characteristics, production process requirements, and expert experience.

[0077] To enrich the training data and enhance the adaptability of the model, a generative adversarial network (GAN) is used to expand the training data. The generator learns the characteristics of real data and generates a system state parameter set similar to the real data distribution; at the same time, a filling and sealing knowledge graph is constructed to extract and graphically represent the relationship between parameters and quality indicators. The model uses graph embedding technology to quickly grasp the key data connections and accelerate convergence.

[0078] In the model training session, input the training set into the multi-layer perceptron neural network optimization controller, and use the stochastic gradient descent method for iterative training. Obtain the predicted value through forward propagation and calculate the loss, update the weights and biases through backpropagation, and use the validation set to regularly evaluate the model performance to dynamically adjust the hyperparameters.

[0079] After the model training is completed, input the test set data into the model, calculate the prediction accuracy of each quality index, verify the generalization ability of the model on unknown data, and ensure that an effective control parameter combination can be generated.

[0080] S5.3: Configure a dual-mode control mechanism for the multi-layer perceptron neural network optimized controller. The dual-mode control mechanism includes an offline learning mode and an online adaptive mode.

[0081] Specifically, in the offline learning mode, the multi-layer perceptron neural network optimized controller learns and trains based on historical filling and sealing process data. Using algorithms such as stochastic gradient descent, it adjusts the model weights and biases until the loss function converges to a satisfactory range. In the online adaptive mode, the system state parameters during the filling and sealing process are monitored in real time. Once it detects that the parameters deviate from the normal range, it immediately generates a new control parameter combination according to the trained model and makes real-time adjustments to the filling and sealing process.

[0082] At the same time, design a mode switching mechanism to switch between the offline learning mode and the online adaptive mode according to production requirements and model status. For example, when introducing new products or new processes, switch to the offline learning mode. If the model has consecutive prediction errors exceeding the preset threshold in the online adaptive mode, switch to the offline learning mode for retraining.

[0083] S5.4: Through the parameter sensitivity analysis mechanism of the multi-layer perceptron neural network optimized controller, optimize the control parameter combination to complete the adaptive adjustment of the parameters throughout the filling and sealing process.

[0084] Due to the complex and changeable actual production conditions with many uncertainties, it is difficult for the control parameter combination generated by the controller trained only by S5.2 to always maintain the optimal state. Therefore, a parameter sensitivity analysis mechanism is needed to deeply analyze the influence degree of each parameter on the filling and sealing process and optimize the control parameter combination accordingly.

[0085] Specifically, first determine the parameters to be analyzed, including control parameters such as filling speed, sealing temperature, pressure, and influencing factors such as material properties and environmental conditions. Use the global sensitivity analysis method to evaluate the influence degree of each parameter on the quality index. Set a multi-objective optimization function that comprehensively considers quality index, production efficiency, and cost factors, and use intelligent optimization algorithms to search for the optimal control parameter combination in the parameter space. Install various sensors on the production line to collect system state parameters and quality index data in real time and transmit them to the controller for processing. The controller dynamically adjusts the control parameters according to the parameter sensitivity analysis results and real-time monitoring data, adopts different adjustment strategies at different production stages, and continuously optimizes the parameter analysis and control parameter combination according to the deviation between the actual production results and the optimization objectives to achieve the adaptive adjustment of the parameters throughout the filling and sealing process.

[0086] By constructing a multi-layer perceptron neural network optimization controller, the integration and collaborative optimization of the system state parameters in each link of filling and sealing are realized. With the help of a generative adversarial network to expand data and a knowledge graph to assist training, the adaptability of the model to complex working conditions and the accuracy of parameter optimization are improved. Through a dual-mode control mechanism and a mode switching strategy, the automation level of the filling and sealing process and the stability of product quality are enhanced. Through a parameter sensitivity analysis mechanism, the adaptive adjustment of parameters in the whole process of filling and sealing is realized, and the ability of the system to cope with complex and changeable production scenarios is enhanced.

[0087] Furthermore, this embodiment also provides a full-automatic filling and sealing system based on image recognition and analysis, including: a parameter parsing module for constructing a three-dimensional dynamic model of the fluid and extracting fluid dynamic control parameters and bubble group dynamic characteristic parameters through image recognition; an electric field regulation module for controlling a ring-shaped microelectrode array arranged around the filling nozzle to generate a directional electric field according to the fluid dynamic control parameters and regulating the liquid flow state; an ultrasonic modulation module for determining ultrasonic action parameters according to the bubble group dynamic characteristic parameters and the physical properties of the liquid in the container and controlling an adjustable-frequency ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation region; a heat-sealing adjustment module for detecting the temperature distribution in the container sealing area and controlling a locally adjustable heat-sealing head according to the detection result to apply differential heat-sealing parameters to different temperature regions; a control optimization module for establishing a multi-layer perceptron neural network optimization controller, receiving a set of system state parameters, taking the quality index of the filling and sealing process as the optimization goal, optimizing the control parameter combination, and completing the adaptive adjustment of parameters in the whole process of filling and sealing.

[0088] In summary, the filling method of the present invention combines image recognition and electric field regulation, realizes the precise control of the liquid flow state, solves the problems of liquid drawing and inaccurate interruption of liquid flow in the traditional filling process, and improves the filling accuracy and efficiency; the directional ultrasonic technology based on the dynamic characteristics of bubbles can specifically eliminate bubbles in the container, reduce the bubble residue rate of the product, and avoid damage to the effective components of the liquid while ensuring the product quality; the differential heat-sealing parameter application system solves the problems such as poor sealing reliability caused by uneven container sealing temperature in the traditional sealing process, improves the consistency of the sealing strength and the product shelf life; the multi-layer perceptron neural network optimization control system realizes the intelligent collaborative adjustment of parameters in the whole process of filling and sealing, breaks through the limitations of independent control in traditional links, enables the system to have the adaptive ability to cope with complex and changeable production scenarios, and improves the production stability and product consistency.

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fully automatic filling and sealing method based on image recognition and analysis, characterized in that, Including: Construct a three-dimensional dynamic model of the fluid, and extract fluid dynamic control parameters and dynamic characteristic parameters of the bubble group through image recognition; According to the fluid dynamic control parameters, control the annular microelectrode array arranged around the filling nozzle to generate a directional electric field and regulate the liquid flow state; Determine the ultrasonic action parameters according to the dynamic characteristic parameters of the bubble group and the physical property parameters of the container liquid, and control the frequency-tunable ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation region; Detect the temperature distribution in the container sealing area, and control the locally adjustable heat sealing head according to the detection results to apply differential heat sealing parameters to different temperature regions; Establish a multi-layer perceptron neural network optimization controller, receive the system state parameter set, optimize the control parameter combination with the quality index of the filling and sealing process as the optimization goal, and complete the adaptive adjustment of the parameters in the whole filling and sealing process.

2. The fully automatic filling and sealing method based on image recognition and analysis according to claim 1, wherein The fluid dynamic control parameters include the surface stress tensor field, three-dimensional flow velocity vector distribution, and liquid surface motion state parameters. The dynamic characteristic parameters of the bubble group include the bubble volume concentration gradient, bubble equivalent particle size distribution, and three-dimensional coordinates of the bubble aggregation region.

3. The fully automatic filling and sealing method based on image recognition analysis according to claim 1, characterized in that, The control of the annular microelectrode array arranged around the filling nozzle to generate a directional electric field includes: Arrange multiple microelectrodes around the filling nozzle to form an annular microelectrode array; Generate a control parameter set for each microelectrode based on the fluid dynamic control parameters; Apply the control parameter set to the annular microelectrode array to generate a directional electric field; Form an electric stress distribution at the liquid-gas interface through the directional electric field to regulate the liquid flow state; At the end of filling, based on the liquid surface motion state parameters, form a convergent electric field gradient pointing to the center of the filling nozzle by adjusting the microelectrode voltage distribution to control liquid cutoff.

4. The fully automatic filling and sealing method based on image recognition analysis according to claim 1, wherein, The control of the frequency-tunable ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation region includes: Calculate the transmission loss coefficient and liquid attenuation coefficient of ultrasonic waves according to the physical property parameters of the container liquid; Determine the bubble resonance frequency range based on the bubble equivalent particle size distribution; Adjust the beam focusing parameters of the frequency-tunable ultrasonic generator based on the bubble volume concentration gradient and the three-dimensional coordinates of the bubble aggregation region; Configure the frequency-tunable ultrasonic generator as a pulse output mode, and determine the pulse period and pulse intensity according to the transmission loss coefficient, liquid attenuation coefficient, and bubble resonance frequency range; Output ultrasonic modulation pulses for the bubble aggregation region through the frequency-tunable ultrasonic generator.

5. The fully automatic filling and sealing method based on image recognition analysis according to claim 1, wherein, The application of differential heat sealing parameters to different temperature regions includes: Use thermal imaging technology to scan the container sealing area to obtain temperature distribution data of the container sealing area; Conduct temperature gradient analysis according to the temperature distribution data to identify the temperature abnormal regions in the container sealing area; Based on the temperature abnormal regions, divide the container sealing area into multiple independent control blocks and calculate the temperature characteristic parameters of each control block; Extract the standard heat sealing parameters corresponding to the container material; Generate differential heat sealing parameters for each control block through parameter compensation calculation according to the temperature characteristic parameters of each control block and the standard heat sealing parameters; Apply the corresponding differential heat sealing parameters to each control block through the locally adjustable heat sealing head to perform heat sealing operations.

6. The fully automatic filling and sealing method based on image recognition analysis according to claim 1, wherein, The establishment of the multi-layer perceptron neural network optimization controller includes: Construct a multi-layer perceptron neural network optimization controller, which includes an input layer, a hidden layer, and an output layer. The input layer receives a set of system state parameters, and the output layer generates an optimal control parameter combination; Taking the filling and sealing process quality index evaluation system as the optimization goal, train the multi-layer perceptron neural network optimization controller; Configure a dual-mode control mechanism for the multi-layer perceptron neural network optimization controller; Through the parameter sensitivity analysis mechanism of the multi-layer perceptron neural network optimization controller, optimize the control parameter combination to complete the adaptive adjustment of the parameters in the whole filling and sealing process.

7. The fully automatic filling and sealing method based on image recognition and analysis according to claim 5, characterized in that The temperature abnormal area for identifying the container sealing area includes: Convert the temperature distribution data into a two-dimensional temperature matrix; Introduce a multi-scale thermal anomaly detection mechanism, perform multi-scale decomposition on the two-dimensional temperature matrix through wavelet transform, establish a thermal feature subspace, and extract abnormal feature points; Calculate the temperature gradient vector between adjacent points in the two-dimensional temperature matrix to obtain a temperature gradient field, and mark temperature abnormal candidate points according to a preset temperature gradient threshold; Merge the abnormal feature points with the temperature abnormal candidate points to obtain an enhanced temperature abnormal candidate point set; Perform connectivity analysis on the enhanced temperature abnormal candidate point set to obtain the temperature abnormal area.

8. A fully automatic filling and sealing system based on image recognition and analysis, characterized in that, Include: A parameter analysis module for constructing a three-dimensional dynamic model of the fluid, and extracting fluid dynamic control parameters and bubble group dynamic characteristic parameters through image recognition; An electric field regulation module for controlling a ring-shaped microelectrode array arranged around the nozzle to generate a directional electric field according to the fluid dynamic control parameters, and regulating the liquid flow state; An ultrasonic modulation module for determining ultrasonic action parameters according to the bubble group dynamic characteristic parameters and container liquid physical property parameters, and controlling an adjustable frequency ultrasonic generator to output ultrasonic modulation pulses for the bubble aggregation area; A heat sealing adjustment module for detecting the temperature distribution in the container sealing area, and controlling a locally adjustable heat sealing head according to the detection result to apply differential heat sealing parameters to different temperature areas; A control optimization module for establishing a multi-layer perceptron neural network optimization controller, receiving a set of system state parameters, taking the filling and sealing process quality index as the optimization goal, optimizing the control parameter combination, and completing the adaptive adjustment of the parameters in the whole filling and sealing process.

Images