Intelligent optimization method for tapioca pearl forming and shaping and control system of tapioca pearl forming and shaping

Abstract

The invention relates to the technical field of automatic control of food processing, and discloses an intelligent optimization method for forming and shaping tapioca pearls and a control system of the tapioca pearls. The system comprises a forming parameter acquisition module, a feature fusion processing module, an optimization strategy generation module and a shaping execution control module. Historical temperature, pressure and rotating speed data of a tapioca ball material are collected through a multi-source sensor, temperature, pressure and rotating speed characteristic parameters are determined through characteristic fusion processing, the parameters serve as indexes, a target optimization strategy is searched in a forming process configuration space, forming parameters are adjusted according to the strategy, and a target forming and shaping result is obtained through shaping effect judgment. Intelligent control over the tapioca pearl forming and shaping process is achieved, parameters can be dynamically adjusted according to material characteristics and production data, the tapioca pearl quality and production efficiency are improved, and the problems that in the prior art, parameter setting is not accurate, quality is not stable, and production efficiency is low are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of food processing automation control, and in particular to an intelligent optimization method for tapioca pearl forming and shaping and a control system thereof. Background Art

[0002] As a key ingredient in beverages like pearl milk tea, tapioca pearls (Taobao) have a quality that directly impacts consumer taste and market acceptance. Forming and shaping are crucial steps in the production of tapioca pearls, but current traditional tapioca pearl forming and shaping technologies present numerous challenges that need to be addressed.

[0003] From a molding perspective, traditional processes rely heavily on manual experience to set molding parameters such as temperature, pressure, and rotation speed. Due to a lack of precise monitoring methods, operators struggle to capture real-time data changes in the tapioca pearl material during the molding process. Different batches of tapioca pearl material can exhibit subtle variations in their physical and chemical properties, making manually set parameters difficult to adapt to these variations, leading to instability in the tapioca pearl molding process. For example, excessively high temperatures can cause the surface of the tapioca pearls to harden too quickly, while the interior remains under-molded, resulting in a hollow interior. Excessively low temperatures prolong molding time, reducing production efficiency and making the tapioca pearls too hard, impacting taste. Improper pressure settings can result in irregular shapes, such as oval or flat, failing to meet market demand for round, plump tapioca pearls. Improper speed control can lead to uneven force distribution during the molding process, resulting in cracks or uneven surfaces.

[0004] In the shaping process, traditional technologies usually use fixed process rules and lack effective analysis and feedback adjustment mechanisms for the shaping process data. Once parameter fluctuations occur during the shaping process, the shaping process cannot be adjusted in time, resulting in poor shaping of the tapioca pearls. For example, if the temperature fluctuates abnormally during the shaping process, and conventional shaping parameters are still used during shaping, the physical properties of the tapioca pearls, such as elasticity and toughness, may be affected. The tapioca pearls are prone to deterioration, adhesion, and other problems during subsequent storage and use. In addition, traditional tapioca pearl forming and shaping equipment often operate independently, and data between devices cannot be shared and interacted, making it difficult to achieve coordinated optimization of the entire production process and unable to meet the requirements of modern large-scale, high-efficiency production.

[0005] As the market demands for tapioca pearl quality and production efficiency continue to rise, traditional tapioca pearl forming and shaping technologies, which rely on manual experience and fixed process rules, are no longer able to meet this growing market demand. Therefore, there is an urgent need for an intelligent optimization control system and method that can collect and analyze key data from the tapioca pearl forming process in real time, dynamically adjust forming and shaping parameters, and achieve high-quality and efficient tapioca pearl production. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent optimization method for forming and shaping tapioca pearls and a control system thereof, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above-mentioned object, the present invention provides the following technical solution: an intelligent optimization control system for forming and shaping tapioca pearls, the system comprising: A molding parameter acquisition module is used to perform multi-source sensor data acquisition to obtain a historical temperature fluctuation data set, a historical pressure change data set, and a historical speed adjustment data set of the tapioca pearl material monitored by the multi-source sensors within a preset process cycle; a feature fusion processing module, configured to traverse the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis and determine temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters; An optimization strategy generation module is used to perform a dynamic range search in the molding process configuration space using the temperature characteristic parameter, the pressure characteristic parameter, and the speed characteristic parameter as indexes to determine a target optimization strategy; The shaping execution control module is used to adjust the shaping parameters of the tapioca pearl material according to a preset process rule set based on the target optimization strategy, obtain a periodic control data set, and use a shaping effect discriminator to perform state discrimination on the periodic control data set to obtain a target shaping result.

[0008] Preferably, the feature fusion processing module is further used to: Using a time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence; Counting historical temperature fluctuation data between an upper stability threshold and a lower stability threshold in the temperature segment sequence to obtain a calibration historical temperature data set, wherein the upper stability threshold is a critical value obtained by shifting the maximum value in the segment downward by a preset amplitude, and the lower stability threshold is a critical value obtained by shifting the minimum value in the segment upward by a preset amplitude; Calculating the variance of the calibration history temperature data set to obtain the temperature characteristic parameter; Multi-dimensional feature analysis is performed on the historical pressure change data set and the historical speed adjustment data set to obtain the pressure characteristic parameters and the speed characteristic parameters.

[0009] Preferably, the feature fusion processing module is further used to: Arrange the historical temperature fluctuation data set in ascending order according to chronological order, and take the historical temperature fluctuation data in the middle period as the median of the sequence; Calculating the top decile and the bottom decile of the historical temperature fluctuation data set; The first decile is used as the lower threshold, the second decile is used as the upper threshold, and the sequence median is combined to construct the temperature segmented sequence.

[0010] Preferably, the optimization strategy generation module is further used to: Collect multiple sets of sample temperature characteristic parameters, multiple sets of sample pressure characteristic parameters, and multiple sets of sample speed characteristic parameters, as well as corresponding multiple sets of sample optimization strategies as training data; Pre-constructing a three-dimensional feature space, wherein the coordinate origin of the three-dimensional feature space is P, the A axis is the temperature characteristic parameter, the B axis is the pressure characteristic parameter, and the C axis is the speed characteristic parameter; The training data is input into the three-dimensional feature space to obtain multiple groups of sample feature points, and the multiple groups of sample feature points are labeled using the multiple groups of sample optimization strategies to obtain the molding process configuration space.

[0011] Preferably, the optimization strategy generation module is further used to: A plane in the molding process configuration space that passes through the temperature characteristic parameter and is perpendicular to the BC plane is used as a first reference plane; A plane in the molding process configuration space that passes through the pressure characteristic parameter and is perpendicular to the AC plane is used as a second reference plane; A plane in the molding process configuration space that passes through the rotational speed characteristic parameter and is perpendicular to the AB plane is used as a third reference plane; The area enclosed by the BC plane, the AC plane, the AB plane, the first reference plane, the second reference plane, and the third reference plane is used as a configuration sub-area, wherein the configuration sub-area includes multiple groups of configuration sample feature points; A dynamic range search is performed on the multiple groups of configuration sample feature points to determine a target configuration sample feature point, and a sample optimization strategy corresponding to the target configuration sample feature point is used as the target optimization strategy.

[0012] Preferably, the optimization strategy generation module is further used to: Extracting a central configuration sample feature point of the configuration sub-region, and constructing a central neighborhood region with the central configuration sample feature point as a starting point according to a preset dynamic search range, wherein the central neighborhood region is a spherical sub-region constructed with the central configuration sample feature point as a sphere center and the preset dynamic search range as a radius; Counting the number of sample feature points configured in the central neighborhood area, and dividing the statistical value by the volume of the central neighborhood area to obtain a central neighborhood density value; Randomly selecting a configuration sample feature point from the edge of the central neighborhood area as a first search feature point, and constructing a first search neighborhood density value of the first search feature point; Determine whether the first search neighborhood density value is greater than or equal to the central neighborhood density value. If so, update the first search feature point as the starting point and continue the dynamic range search until the preset number of dynamic searches is reached. The search feature point obtained from the last search is used as the target configuration sample feature point.

[0013] Preferably, the optimization strategy generation module is further used to: If not, the search update failure count with an initial value of 0 is updated to 1, and a configuration sample feature point is randomly selected from the edge of the central neighborhood area as the first search feature point for dynamic range search analysis. When the search update failure count is greater than the preset maximum search update failure count, the central configuration sample feature point is used as the target configuration sample feature point.

[0014] Preferably, the present invention further includes an intelligent optimization method for tapioca pearl forming and shaping, the method comprising: Execute multi-source sensor data acquisition to obtain a historical temperature fluctuation data set, a historical pressure change data set, and a historical speed adjustment data set of the tapioca pearl material monitored by the multi-source sensor within a preset process cycle; Traversing the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis to determine temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters; Using the temperature characteristic parameter, the pressure characteristic parameter, and the speed characteristic parameter as indexes, a dynamic range search is performed in the molding process configuration space to determine a target optimization strategy; Based on the target optimization strategy, the molding parameters of the tapioca pearl material are adjusted according to a preset process rule set to obtain a periodic control data set, and the state of the periodic control data set is judged using a shaping effect discriminator to obtain a target molding shaping result.

[0015] Preferably, the step of traversing the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis to determine the temperature characteristic parameters, the pressure characteristic parameters, and the speed characteristic parameters specifically includes: Using a time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence; Counting historical temperature fluctuation data between an upper stability threshold and a lower stability threshold in the temperature segment sequence to obtain a calibration historical temperature data set, wherein the upper stability threshold is a critical value obtained by shifting the maximum value in the segment downward by a preset amplitude, and the lower stability threshold is a critical value obtained by shifting the minimum value in the segment upward by a preset amplitude; Calculating the variance of the calibration history temperature data set to obtain the temperature characteristic parameter; The historical pressure change data set and the historical speed adjustment data set are subjected to multi-dimensional feature analysis using the same logic to obtain the pressure characteristic parameters and the speed characteristic parameters.

[0016] Preferably, the step of using the time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence specifically includes: Arrange the historical temperature fluctuation data set in ascending order according to chronological order, and take the historical temperature fluctuation data in the middle period as the median of the sequence; Calculating the top decile and the bottom decile of the historical temperature fluctuation data set; The first decile is used as the lower threshold, the second decile is used as the upper threshold, and the sequence median is combined to construct the temperature segmented sequence.

[0017] Compared with the prior art, the present invention has the following beneficial effects: At the data acquisition and processing level, the molding parameter acquisition module utilizes multi-source sensors to collect data, capturing historical temperature fluctuation data sets, pressure change data sets, and speed adjustment data sets for tapioca pearl material monitoring within a preset process cycle. This enables comprehensive, real-time monitoring of key parameters in the tapioca pearl molding process. The feature fusion processing module employs various algorithms, including time series segmentation, to deeply analyze this data. For example, it performs dimension splitting, calibration, and variance calculation on the historical temperature fluctuation data set, accurately extracting temperature, pressure, and speed characteristic parameters. This effectively removes noise and interference from the data, improving its accuracy and reliability and providing a solid data foundation for subsequent optimization strategy development.

[0018] In terms of optimization strategy generation, the module constructs a molding process configuration space by collecting multiple sets of sample characteristic parameters and corresponding sample optimization strategies. Using a three-dimensional feature space and a dynamic range search algorithm, the module uses temperature, pressure, and speed characteristic parameters as indexes to precisely locate the target optimization strategy within the molding process configuration space. This big data and algorithm-based optimization strategy generation approach changes the traditional model of manually setting parameters based on experience. It can quickly and accurately generate the most appropriate optimization strategy based on the characteristics of different batches of tapioca pearls and actual data from the molding process, enabling personalized customization of molding parameters and significantly improving the adaptability and stability of the tapioca pearl molding process.

[0019] In the molding and shaping execution control, the shaping execution control module precisely adjusts the molding parameters of the tapioca pearl material according to a target optimization strategy and a set of preset process rules. It also uses a shaping effect discriminator to perform status discrimination on the periodic control data set, ensuring that the tapioca pearls are in an optimal state at every step of the molding and shaping process. This closed-loop control mechanism provides real-time feedback on the molding and shaping effect, promptly identifying and correcting parameter deviations, and effectively avoiding tapioca pearl quality issues caused by improper parameters, such as irregular shape, poor texture, and substandard physical properties. Compared with traditional technologies, the tapioca pearls produced using the present invention have a more regular and rounded shape, a smooth and flawless surface, a uniform internal structure, and a chewy and resilient texture. Key physical performance indicators such as elasticity and toughness have been significantly improved, with the product qualification rate increasing by over [X]%, greatly enhancing the quality and market competitiveness of the tapioca pearls.

[0020] From the perspective of production efficiency, the system's ability to automatically monitor and adjust parameters in real time reduces manual intervention and parameter debugging time, avoiding production interruptions and rework caused by improper parameters. This significantly shortens the tapioca pearl production cycle and increases production efficiency by over [X]%, reducing the company's production costs and improving its economic benefits. Furthermore, the system's intelligent and automated features reduce reliance on manual experience, reduce operator labor intensity, and improve the safety and stability of the production process. This facilitates large-scale, standardized production and promotes the transformation and upgrading of the tapioca pearl production industry towards intelligent and automated production. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the intelligent optimization control system for tapioca pearl forming and shaping according to the present invention; Figure 2 This is a flow chart for obtaining temperature feature parameters of the feature fusion processing module; Figure 3 Flowchart for configuring the space for the molding process; Figure 4 Flowchart determined for the configuration subarea. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figures 1-4 The present invention relates to an intelligent optimization control system for tapioca pearl molding and shaping. The system comprises a molding parameter acquisition module, a feature fusion processing module, an optimization strategy generation module, and a shaping execution control module. The specific implementation steps are as follows: The molding parameter acquisition module performs multi-source sensor data acquisition to obtain the historical temperature fluctuation data set, historical pressure change data set and historical speed adjustment data set of the tapioca pearl material monitored by the multi-source sensors within a preset process cycle.

[0024] The feature fusion processing module traverses the historical temperature fluctuation data set, the historical pressure change data set and the historical speed adjustment data set to perform multi-dimensional feature analysis to determine the temperature characteristic parameters, the pressure characteristic parameters and the speed characteristic parameters.

[0025] The optimization strategy generation module uses temperature characteristic parameters, pressure characteristic parameters and speed characteristic parameters as indexes, conducts dynamic range search in the molding process configuration space, and determines the target optimization strategy.

[0026] The shaping execution control module adjusts the shaping parameters of the tapioca pearl material according to the preset process rule set based on the target optimization strategy to obtain a periodic control data set, and uses the shaping effect discriminator to perform state discrimination on the periodic control data set to obtain the target shaping result.

[0027] Example 1: In the intelligent optimization control system for tapioca pearl forming and shaping, the feature fusion processing module needs to perform a series of specific data processing operations when performing multi-dimensional feature analysis on the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to determine the temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters.

[0028] For the historical temperature fluctuation data set, the feature fusion processing module uses the time series segmentation method to split the data dimension to obtain a temperature segmentation sequence. In this process, the historical temperature fluctuation data set needs to be arranged in ascending chronological order. The chronological arrangement can make the data show a certain regularity, which is convenient for subsequent analysis and processing. After the arrangement is completed, it is necessary to determine the median of the sequence, that is, to select the historical temperature fluctuation data in the middle period as the median of the sequence. The determination of the middle period can be divided according to the time span of the data. For example, the entire time period can be divided into several equal segments, and the data in the middle segment is taken as the data of the middle period.

[0029] Calculate the top and bottom decile of a historical temperature fluctuation data set. Deciles are values ​​that divide the data into ten equal parts after sorting it from smallest to largest. The top decile is the upper limit of the smallest 10 percent of the data, while the bottom decile is the lower limit of the largest 10 percent. Common statistical methods can be used to calculate the top and bottom decile, such as sorting the data and determining the top and bottom decile positions based on the number of data points.

[0030] The top decile is used as the lower threshold, the bottom decile as the upper threshold, and combined with the previously determined sequence median, a temperature segment sequence is constructed. The lower and upper thresholds define a data range, while the sequence median serves as a reference within this range. In this way, the historical temperature fluctuation data set is divided into different segments, each of which forms a temperature segment sequence.

[0031] After obtaining the temperature segment sequence, it is necessary to count the historical temperature fluctuation data between the upper limit of the stability threshold and the lower limit of the stability threshold in the sequence to obtain the calibration historical temperature data set. Among them, the upper limit of the stability threshold is the critical value obtained by offsetting the maximum value in the segment downward by the preset amplitude, and the lower limit of the stability threshold is the critical value obtained by offsetting the minimum value in the segment upward by the preset amplitude. The preset amplitude can be set according to actual process requirements and experience. For example, it can be set to a certain percentage of the maximum or minimum value in the segment. By determining the upper and lower limits of the stability threshold by offsetting, some abnormally fluctuating data can be excluded, making the calibration historical temperature data set more stable and reliable.

[0032] When counting data between the upper and lower stability thresholds, it is necessary to examine each data point in the temperature segment sequence one by one to determine whether it falls within the range. The data points that meet the criteria are collected to form a calibration history temperature data set.

[0033] The variance of the calibration historical temperature data set is calculated to obtain temperature characteristic parameters. Variance is an important indicator of data dispersion and can reflect the fluctuation of the data within the calibration historical temperature data set. The variance is calculated by first calculating the data mean, then taking the average of the squares of the differences between each data point and the mean as the variance. The variance accurately describes the characteristics of the temperature data and provides an important basis for the subsequent optimization strategy generation.

[0034] For the historical pressure change data set and the historical speed adjustment data set, the feature fusion processing module uses the same logic as the historical temperature fluctuation data set to perform multi-dimensional feature analysis to obtain pressure characteristic parameters and speed characteristic parameters. Specifically, for the historical pressure change data set, it is also necessary to split the data dimension using the time series segmentation method, determine the pressure segmentation sequence, and then calculate the upper and lower limits of the stability threshold. The data within this range is counted to obtain a calibrated historical pressure data set, and finally the variance value is calculated to obtain the pressure characteristic parameters. The historical speed adjustment data set is also processed according to the same steps to obtain the speed characteristic parameters.

[0035] When processing historical pressure change data sets using the time series segmentation method, the steps of arranging the chronological order, determining the pressure data in the middle period as the sequence median, and calculating the top and bottom deciles are consistent with the processing of historical temperature fluctuation data sets. Similarly, when determining the upper and lower limits of the stability threshold, a preset amplitude offset is applied based on the maximum and minimum values ​​within the segment. By counting the data that meet the criteria and calculating the variance, a pressure characteristic parameter that reflects the characteristics of the pressure data is obtained.

[0036] The same processing flow was followed for the historical speed regulation data set. The data was segmented using a time series segmentation method, the speed segment sequence was determined, the upper and lower limits of the stability threshold were calculated, the historical speed data set was statistically calibrated, and finally, the speed characteristic parameters were obtained through variance calculation. By applying the same logic to all three data sets, the consistency and accuracy of the feature analysis were ensured, providing reliable characteristic parameters for subsequent optimization strategy generation.

[0037] Throughout the entire processing process, each step must be strictly performed according to established rules and methods to ensure the accuracy and reliability of data processing. For example, when arranging time series, they must be arranged in ascending chronological order to avoid errors in the order that could lead to bias in subsequent analysis. When calculating deciles, the correctness of the calculation method must be ensured to obtain accurate lower and upper thresholds. When determining the upper and lower limits of the stability threshold, the preset amplitude setting needs to be combined with actual process requirements and experience. It must be able to exclude abnormal data while not excessively excluding normal data, which would affect data integrity.

[0038] Through the feature fusion processing module, the multi-dimensional features of the historical temperature fluctuation data set, the historical pressure change data set and the historical speed adjustment data set are analyzed, and finally the temperature characteristic parameters, pressure characteristic parameters and speed characteristic parameters are obtained. These characteristic parameters can accurately reflect the changing characteristics of temperature, pressure and speed during the tapioca pearl forming and shaping process, and provide a solid data foundation for the subsequent optimization strategy generation module to conduct dynamic range search in the forming process configuration space and determine the target optimization strategy.

[0039] Example 2: A key function of the optimization strategy generation module is to construct a molding process configuration space to facilitate the subsequent search for target optimization strategies based on characteristic parameters. The specific process by which this module achieves this function is as follows: The optimization strategy generation module needs to collect multiple sets of sample temperature characteristic parameters, multiple sets of sample pressure characteristic parameters, and multiple sets of sample speed characteristic parameters, and at the same time obtain the corresponding multiple sets of sample optimization strategies, and use these data together as training data. The collection of sample data needs to cover the characteristic parameters under different process conditions during the tapioca pearl molding and shaping process, such as data collected under different temperature ranges, pressure levels, and speed gears, to ensure that the training data is sufficiently representative. The sample optimization strategy is a specific process parameter adjustment plan formulated for each combination of sample characteristic parameters, such as the temperature control range, pressure application timing, or speed adjustment gradient. These strategies are usually formed based on historical production experience or the results of previous process debugging.

[0040] When collecting training data, data accuracy and consistency must be ensured. For example, the acquisition of sample temperature characteristic parameters must be synchronized with the sensor's sampling frequency and calibration cycle to avoid data deviations due to equipment errors. Sample optimization strategies must be recorded clearly corresponding to process stages, such as the initial molding stage and the mid-stage finalization stage, to ensure the accuracy of subsequent feature point labels.

[0041] The optimization strategy generation module pre-constructs a three-dimensional feature space. This three-dimensional feature space is based on the coordinate origin P, where the A-axis is defined as the temperature characteristic parameter axis, the B-axis is defined as the pressure characteristic parameter axis, and the C-axis is defined as the speed characteristic parameter axis. The construction of the three-dimensional space must follow the basic rules of the mathematical coordinate system, and the unit scale of each coordinate axis must be reasonably divided according to the numerical range of the characteristic parameter. For example, if the variance value of the temperature characteristic parameter ranges from 0.1 to 10, the A-axis scale can be increased by 0.5 units to ensure a good visualization of the distribution of data points in space.

[0042] When constructing a three-dimensional feature space, the physical significance of the coordinate origin P must also be considered. The origin P can be set to the theoretical standard process state, that is, the coordinate point when the characteristic parameters of temperature, pressure, and speed are in an ideal equilibrium state. In actual production, the characteristic parameter data points will form different distribution areas around the origin P.

[0043] After constructing the three-dimensional feature space, the optimization strategy generation module inputs the training data into this space. Specifically, each set of sample temperature, pressure, and speed characteristic parameters corresponds to a coordinate point in the three-dimensional space. In other words, each sample data set is converted into a sample feature point in the space. For example, if a set of sample data has a temperature characteristic parameter of 2.5, a pressure characteristic parameter of 3.2, and a speed characteristic parameter of 1.8, then its coordinates in the three-dimensional space are (2.5, 3.2, 1.8), corresponding to a specific sample feature point.

[0044] When inputting training data, ensure the accuracy of data mapping to avoid coordinate errors that could lead to feature point position deviations. For massive amounts of training data, batch import can be used to improve processing efficiency, but this requires simultaneous data verification. For example, using visualization tools to check whether the feature point distribution conforms to process logic to avoid abnormal outliers.

[0045] The optimization strategy generation module uses multiple sets of sample optimization strategies to label multiple sets of sample feature points. This labeling process essentially assigns each feature point a corresponding process optimization strategy attribute, mapping the geometric position in three-dimensional space to the actual process strategy. For example, if the sample optimization strategy corresponding to a sample feature point is "increase temperature by 5°C, maintain pressure at its current value, and reduce speed by 10%," the feature point is labeled as an associated point for that strategy.

[0046] Labeling must follow unified rules to ensure comparability and traceability between labels for different feature points. For example, labels should clearly specify the direction and magnitude of parameter adjustments for the optimization strategy, avoiding ambiguous descriptions. For areas with similar feature point distributions, clustering algorithms can be used to identify common optimization strategy labels, improving the systematic nature of the labeling system.

[0047] Through the above steps, the optimization strategy generation module ultimately obtains the molding process configuration space. This space is a three-dimensional data field containing a large number of sample feature points and their corresponding optimization strategy labels. The coordinate position of each feature point reflects the combined state of the temperature, pressure, and speed characteristic parameters, and the label content indicates the optimal process adjustment solution for that state. The construction of the molding process configuration space provides the data foundation for the subsequent search for target optimization strategies based on real-time characteristic parameters. This enables the system to quickly locate the corresponding optimization strategy based on the monitoring characteristics of the current tapioca pearl material, realizing intelligent adjustment of the molding and shaping processes.

[0048] In practice, the molding process configuration space is not fixed. The optimization strategy generation module regularly iterates the feature points and labels in the space based on newly collected training data to adapt to process parameter drift caused by factors such as changes in raw material properties and equipment wear, ensuring the timeliness and accuracy of the optimization strategy. For example, when changing batches of tapioca pearl raw materials, the system automatically collects the feature parameters of the new raw material under different process conditions and the corresponding optimization strategy. It then inputs the newly added training data into the three-dimensional feature space, relabels the feature points, and updates the molding process configuration space, ensuring that the system always maintains optimal adaptability to current production conditions.

[0049] Example 3: After constructing the molding process configuration space, the optimization strategy generation module needs to further determine the configuration sub-regions in the space based on the temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters, thereby laying the foundation for the dynamic range search target optimization strategy. The specific implementation method is as follows: The optimization strategy generation module needs to determine three reference planes in the molding process configuration space. The first reference plane is a plane passing through the temperature characteristic parameter and perpendicular to the BC plane, called the first reference plane. The temperature characteristic parameter here is a parameter that reflects the temperature fluctuation characteristics obtained by the feature fusion processing module, such as the value obtained by calculating the variance of the historical temperature fluctuation data set. The BC plane is a plane in the three-dimensional feature space formed by the pressure characteristic parameter axis (B axis) and the speed characteristic parameter axis (C axis). The plane perpendicular to this plane must be parallel to the temperature characteristic parameter axis (A axis). Therefore, the first reference plane is a plane in space that is perpendicular to the BC plane and extends along the A axis. Its position is determined by the current temperature characteristic parameter value.

[0050] Determine the second reference plane, a plane in the molding process configuration space that passes through the pressure characteristic parameter and is perpendicular to the AC plane. The pressure characteristic parameter is also obtained by analyzing the characteristics of the historical pressure change data set, such as the variance value. The AC plane is formed by the temperature characteristic parameter axis (A-axis) and the speed characteristic parameter axis (C-axis). The plane perpendicular to this plane is parallel to the pressure characteristic parameter axis (B-axis). Therefore, the second reference plane is a plane perpendicular to the AC plane and extending along the B-axis. Its position in space is determined by the pressure characteristic parameter value.

[0051] Determine the third reference plane, which is a plane in the molding process configuration space that passes through the speed characteristic parameter and is perpendicular to the AB plane. The speed characteristic parameter is the result of feature analysis of the historical speed adjustment data set, such as the variance value. The AB plane is formed by the temperature characteristic parameter axis (A-axis) and the pressure characteristic parameter axis (B-axis). The plane perpendicular to this plane is parallel to the speed characteristic parameter axis (C-axis). Therefore, the third reference plane is a plane perpendicular to the AB plane and extending along the C-axis. Its position is determined by the speed characteristic parameter value.

[0052] After determining these three reference planes, the optimization strategy generation module needs to identify the region enclosed by the BC, AC, and AB planes along with the first, second, and third reference planes, defining this as the configuration subregion. Specifically, the BC, AC, and AB planes are the three coordinate planes of the three-dimensional feature space, while the first, second, and third reference planes are three planes perpendicular to the coordinate planes, determined in space based on the current actual temperature, pressure, and speed characteristic parameters. These six planes collectively enclose a closed region in three-dimensional space, known as the configuration subregion.

[0053] This configuration subregion has a clear spatial range: along the A-axis, its range is defined by the intersection of the first reference plane determined by the temperature characteristic parameter value with the AB plane and the AC plane; along the B-axis, its range is defined by the intersection of the second reference plane determined by the pressure characteristic parameter value with the BC plane and the AB plane; and along the C-axis, its range is defined by the intersection of the third reference plane determined by the speed characteristic parameter value with the BC plane and the AC plane. In this way, the configuration subregion forms a three-dimensional spatial subregion bounded by the three characteristic parameter values, which contains a large number of configuration sample feature points. These configuration sample feature points are sample feature points that were previously input into the three-dimensional feature space through training data and labeled. They are distributed within the configuration subregion, and each point corresponds to a set of sample temperature characteristic parameters, sample pressure characteristic parameters, sample speed characteristic parameters, and a corresponding sample optimization strategy.

[0054] The optimization strategy generation module performs a dynamic range search on multiple sets of configuration sample feature points within the configuration sub-region to identify the target configuration sample feature point and use the sample optimization strategy corresponding to that point as the target optimization strategy. The goal of the dynamic range search is to find the sample feature point in the configuration sub-region that best matches the current feature parameter combination, thereby obtaining the most suitable optimization strategy.

[0055] When performing a dynamic range search, it's important to consider the distribution of the configuration sample's feature points within the configuration subregion. Because training data may come from different process conditions, the distribution of the configuration sample's feature points within the subregion may be uneven, with some areas densely populated and others sparsely populated. Therefore, the search process needs to be able to adaptively adjust the search range, prioritizing areas with dense feature points to increase the probability of finding the optimal strategy.

[0056] The specific implementation process of dynamic range search can be to start from an initial point in the configuration sub-region, gradually expand the search range according to certain rules, calculate the similarity or distance between each feature point in the search range and the current feature parameter, and select the feature point with the highest similarity or the closest distance as the candidate point. After multiple iterative searches, the target configuration sample feature point is finally determined.

[0057] After determining the target configuration sample feature point, the sample optimization strategy corresponding to that point is the target optimization strategy. This strategy includes a specific adjustment plan for the current temperature, pressure, and speed feature parameter combination, such as the temperature adjustment range, the pressure range to be maintained, and the speed change rate.

[0058] It's important to note that the temperature, pressure, and speed characteristic parameters change in real time throughout the entire process. Therefore, each time an optimization strategy search is performed, the positions of the first, second, and third reference planes are redefined based on the current characteristic parameter values, causing the scope of the configuration sub-region to change accordingly. This enables the system to dynamically adjust the search space based on the real-time monitoring of the tapioca pearl material state, ensuring that the target optimization strategy obtained in each search is the most appropriate for the current process conditions.

[0059] The process of determining the configuration sub-region and searching the dynamic range requires a balance between computational efficiency and accuracy. On the one hand, it is necessary to ensure that the configuration sub-region contains a sufficient number of configuration sample feature points to ensure the possibility of finding an effective strategy; on the other hand, it is necessary to avoid excessively large sub-regions, which would lead to a surge in computational complexity and affect the real-time performance of the system. Therefore, in practical applications, it may be necessary to appropriately adjust and optimize the determination of the reference plane and the range of the sub-region based on historical data and process experience to achieve the best processing results.

[0060] In this way, the optimization strategy generation module accurately identifies configuration subregions within the molding process configuration space based on the current characteristic parameters of temperature, pressure, and speed. It then uses dynamic range search to find the target optimization strategy, providing precise parameter adjustment guidance for the subsequent shaping execution control module, thereby achieving intelligent optimization control of the tapioca pearl molding and shaping process. The entire process relies closely on the accurate analysis of the molding process configuration space and characteristic parameters constructed in the early stages, ensuring the rationality and effectiveness of the optimization strategy.

[0061] Example 4: After determining the configuration sub-region, the optimization strategy generation module needs to further locate the target configuration sample feature point through dynamic range search. This embodiment specifically involves the process of constructing a search region with the central configuration sample feature point as the starting point and adjusting the search direction based on the neighborhood density value. The following is a detailed description with reference to a specific example: Assume that a set of sample feature points exists within a configuration subregion, whose temperature, pressure, and speed characteristic parameters correspond to the A-axis, B-axis, and C-axis coordinates in three-dimensional space, respectively. First, the optimization strategy generation module extracts the central configuration sample feature point within the configuration subregion. This central configuration sample feature point can be determined by calculating the mean coordinates of all feature points within the subregion on the three coordinate axes. For example, if there are 100 feature points within the subregion, and their mean A-axis coordinates are 2.5, their mean B-axis coordinates are 3.0, and their mean C-axis coordinates are 1.8, then the coordinates of the central configuration sample feature point are (2.5, 3.0, 1.8).

[0062] Taking the central configuration sample feature point as the starting point, the module needs to construct a central neighborhood area according to the preset dynamic search range. The preset dynamic search range can be set based on process experience, for example, it can be set to a spherical area with a radius of 1.0. In this case, the central neighborhood area is a spherical sub-area with a radius of 1.0 and a center of (2.5, 3.0, 1.8), which contains several configuration sample feature points. Assume that according to statistics, there are 20 feature points in the spherical area, and the volume of the spherical area can be calculated using the three-dimensional space sphere volume formula (although the formula is not involved here, it is necessary to clarify that the volume calculation is based on the preset radius). Dividing the number of feature points 20 by the volume can obtain the central neighborhood density value, which reflects the density of the feature points in the central neighborhood.

[0063] Next, the module randomly selects a configuration sample feature point from the edge of the central neighborhood area as the first search feature point. For example, a feature point with coordinates (3.5, 3.0, 1.8) is selected from the edge of the spherical area, located at a position tangent to the sphere in the positive direction of the A axis. Then, with this first search feature point as the center of the sphere, a neighborhood area corresponding to the first search neighborhood density value is constructed using the preset dynamic search range (radius 1.0) and the number of feature points in this area is counted. Assuming that this neighborhood contains 15 feature points, the volume of the neighborhood is calculated using the same volume calculation method, and the first search neighborhood density value is then calculated.

[0064] At this point, it is necessary to determine whether the density value of the first search neighborhood is greater than or equal to the density value of the central neighborhood. If the density value of the first search neighborhood is greater than or equal to the density value of the central neighborhood, it indicates that the feature points in that search direction are more densely distributed, and there may be more optimal configuration sample feature points. For example, if the central neighborhood density value is 5 per unit volume and the first search neighborhood density value is 6 per unit volume, the module will update the first search feature point as the new starting point and continue the dynamic range search.

[0065] To continue searching, the module uses the new starting point (i.e., the original first search feature point) as the center and randomly selects a feature point on the edge of its neighborhood as the next search feature point. The process of neighborhood construction, density calculation, and comparison is repeated. Assuming the coordinates of the second selected search feature point are (3.5, 3.5, 1.8), the number of feature points counted after neighborhood construction is 18, and the neighborhood density is calculated to be 7 per unit volume. If this value is greater than the neighborhood density of the current starting point, the search continues with that point as the new starting point. This iteration continues until the preset number of dynamic searches is reached, for example, 5 search iterations. Finally, the search feature point obtained from the last search is used as the target configuration sample feature point.

[0066] In the above example, each search iteration determines whether to update the search starting point based on a comparison of neighborhood density values. The core logic is to move toward areas with denser feature points to increase the probability of finding an optimization strategy that better matches the current process features. If the neighborhood density value of the first search feature point selected in a certain time is smaller than the central neighborhood density value, for example, the first search neighborhood density value is 4 per unit volume, which is smaller than the central neighborhood density value of 5 per unit volume, then it means that the distribution of feature points in that search direction is more sparse. At this time, the module will update the search update failure count from an initial value of 0 to 1, and randomly select another configuration sample feature point from the edge of the central neighborhood area as the first search feature point for analysis.

[0067] For example, the second time a feature point with coordinates (2.5, 3.0, 2.8) is selected from the edge of the central neighborhood, the number of feature points counted after neighborhood construction is 12, and the calculated density is 4.5 per unit volume, which is still less than the central neighborhood density. In this case, the number of search update failures accumulates to 2. When the number of search update failures exceeds the preset maximum number of search update failures (such as 3), it means that no area with higher density can be found in the random search direction of the central neighborhood edge. In this case, the module uses the central configuration sample feature point as the target configuration sample feature point to avoid wasting computing resources due to invalid searches.

[0068] Throughout the dynamic range search process, the preset dynamic search range must be determined based on the fluctuation range of the process parameters. For example, if the normal fluctuation variance of the temperature characteristic parameter during the tapioca pearl molding process is 0.5-3.0, the preset search radius can be set to 0.8-1.2 to ensure that the neighborhood area covers a reasonable range of characteristic variations. Furthermore, the preset maximum number of search update failures must balance search efficiency and accuracy. This threshold can typically be adjusted based on the distribution density of feature points in historical data. If the distribution of feature points within the configured sub-area is relatively uniform, the threshold can be appropriately reduced; otherwise, the threshold can be increased to expand the search range.

[0069] Furthermore, the construction of the central neighborhood is not fixed. When process parameters fluctuate significantly, the module can adaptively adjust the preset dynamic search range. For example, if the variance of the current temperature characteristic parameter suddenly increases, indicating intensified temperature fluctuations, the search radius can be temporarily adjusted from 1.0 to 1.5 to ensure that the neighborhood contains more relevant feature points. This dynamic adjustment mechanism improves the search process's adaptability to process changes and ensures the accuracy of the target optimization strategy.

[0070] Through these steps, the optimization strategy generation module uses the central configuration sample feature point as the initial search starting point within the configuration subregion. It iteratively adjusts the search direction by comparing neighborhood density values, ultimately determining the target configuration sample feature point and providing precise coordinate positioning for subsequent target optimization strategy acquisition. This process fully leverages the distribution patterns of feature points within the configuration subregion. This density-guided search improves the optimization strategy's search efficiency and matching accuracy, ensuring intelligent and precise parameter adjustment during the tapioca pearl molding and shaping process.

[0071] Example 5: When the optimization strategy generation module performs a dynamic range search, if the search neighborhood density value does not meet the expected situation, it needs to adjust the search strategy according to a specific logic. Example 5 specifically involves the processing flow when the search fails, which is described in detail below with specific examples: Assume that the coordinates of the central configuration sample feature point of the configuration subregion are (2.5, 3.0, 1.8), the preset dynamic search range is a spherical area with a radius of 1.0, and there are 20 feature points counted in the central neighborhood. The calculated central neighborhood density is 5 per unit volume. The first search feature point is randomly selected from the edge of the central neighborhood with coordinates (3.5, 3.0, 1.8). After constructing a neighborhood with a radius of 1.0, 15 feature points are counted, and the calculated neighborhood density is 4.5 per unit volume. This value is less than the central neighborhood density value. At this time, the search update failure count is updated from 0 to 1.

[0072] The module randomly selects another configuration sample feature point from the edge of the central neighborhood. For example, the second feature point selected is at coordinates (2.5, 3.0, 2.8), which is located at the edge of the neighborhood in the positive direction of the C axis. After constructing the neighborhood, 12 feature points are counted, and the calculated density is 4 per unit volume, which is still less than the 5 per unit volume of the central neighborhood. The number of search and update failures accumulates to 2.

[0073] The third time, a feature point with coordinates (1.5, 3.0, 1.8) was selected from the edge of the central neighborhood, located on the negative edge of the A-axis. After constructing the neighborhood, 14 feature points were counted, with a calculated density of 4.2 per unit volume, still below the central neighborhood density. The number of search update failures increased to 3. At this point, if the preset maximum number of search update failures is 3, the module determines that no higher density area can be found in the current search direction and selects the central configuration sample feature point (2.5, 3.0, 1.8) as the target configuration sample feature point.

[0074] In another scenario, if the preset maximum number of search update failures is 5, after the first three searches fail, feature points are randomly selected. The fourth time, a feature point with coordinates (2.5, 4.0, 1.8) is selected, located at the positive edge of the B axis. After constructing a neighborhood, 16 feature points are counted, and the calculated density is 4.8 per unit volume, which is still less than 5, and the number of failures increases to 4. The fifth time, a feature point with coordinates (2.5, 2.0, 1.8) is selected, located at the negative edge of the B axis. 17 feature points are counted, with a density of 5.1 per unit volume, which is the first time it exceeds the central neighborhood density. This point is then updated as the new search starting point, and subsequent search iterations continue until the preset number of searches is reached.

[0075] The core of this processing mechanism is to avoid invalid searches by setting a search failure threshold. For example, when the distribution of feature points shifts due to the replacement of tapioca pearl raw material batches, the central neighborhood feature points may be sparse in the configuration sub-area. Assuming that the original central neighborhood density value is 5 / unit volume, after the raw materials are replaced, the new feature points are concentrated in the area of ​​B-axis coordinates 4.0-5.0, while the B-axis coordinate of the central configuration sample feature point is still 3.0. At this time, when searching from the edge of the central neighborhood in the positive direction of the B axis, the coordinates of the feature points selected for the first time are (2.5, 4.0, 1.8). After constructing the neighborhood, 25 feature points may be counted, and the calculated density value is 7.5 / unit volume, which is much larger than the central neighborhood density value. The module will immediately update the search starting point and move in that direction to quickly locate the new feature-dense area.

[0076] The preset maximum number of search update failures should be considered in conjunction with process stability. If the tapioca pearl production process fluctuates minimally and the feature point distribution is relatively stable, a threshold of 2-3 failures can be used to improve search efficiency. If process parameters are susceptible to environmental influences (such as temperature and humidity) and the feature point distribution fluctuates significantly, a threshold of 5-7 failures can be used to provide more search attempts. For example, in the summer, when workshop temperatures are high, the temperature characteristic parameters of the tapioca pearl molding process fluctuate more, causing the feature point distribution range to expand. Increasing the threshold in this situation can prevent premature termination of the search due to accidental deviations in the search direction.

[0077] The counting logic for search update failures must strictly adhere to the rule that "each time a randomly selected edge point fails to meet the density requirement, the count increases by 1." For example, in a search, if the density of the first selected edge point fails to meet the requirement, the count will be 1. However, even if the second selected edge point does not find more feature points in the neighborhood, the count will still increase by 1 because it is randomly selected in a different direction until the threshold is reached. This mechanism ensures the randomness and comprehensiveness of the search direction, avoiding omissions caused by searching in a fixed direction.

[0078] When the central configuration sample feature point is ultimately selected as the target point, the module invokes the sample optimization strategy corresponding to that point. For example, the strategy corresponding to the central configuration sample feature point (2.5, 3.0, 1.8) might be "maintain temperature ±0.5°C from the current value, increase pressure by 5%, and reduce speed by 3%." This strategy is based on scenarios matching the central feature in historical process data. While not currently optimal, it serves as a backup plan to ensure uninterrupted production.

[0079] In practical applications, this mechanism complements neighborhood density search. When process parameters are within a normal fluctuation range, density-guided search can quickly locate the optimal strategy. When parameter fluctuations are abnormal or training data is incomplete, the failure handling mechanism prevents the system from falling into invalid search, ensuring the continuity of the control process. For example, if a device sensor experiences a brief anomaly, causing feature parameter data to deviate from the normal range, the search process may frequently trigger failure counts, ultimately transitioning to a central feature point strategy to avoid parameter misadjustments caused by erroneous data.

[0080] Furthermore, the system can periodically update the coordinates of the central configuration sample feature point based on new training data. For example, after every 10 batches of tapioca pearls are produced, the mean of all feature points within the configuration sub-area is recalculated and the center coordinates are updated to better align the search starting point with the current process state. If the mean of the temperature characteristic parameter during a certain production phase increases from 2.5 to 2.8, the A-axis coordinate of the new central configuration sample feature point is adjusted to 2.8. Subsequent searches will use this new center as the starting point, improving search accuracy.

[0081] As the above examples demonstrate, the core of this embodiment lies in using search failure counts and thresholds to determine the optimal feature points. When a dynamic range search fails to find a superior feature point, the centrally configured sample feature points serve as a fallback, ensuring the availability of the optimization strategy. This mechanism combines random search with deterministic strategies, ensuring search accuracy while improving the system's adaptability to abnormal operating conditions, providing dual guarantees for stable control of the tapioca pearl forming and shaping process.

[0082] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0083] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent optimization control system for tapioca pearl forming and shaping, characterized in that: The system comprises: A molding parameter acquisition module is used to perform multi-source sensor data acquisition to obtain a historical temperature fluctuation data set, a historical pressure change data set, and a historical speed adjustment data set of the tapioca pearl material monitored by the multi-source sensors within a preset process cycle; a feature fusion processing module, configured to traverse the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis and determine temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters; An optimization strategy generation module is used to perform a dynamic range search in the molding process configuration space using the temperature characteristic parameter, the pressure characteristic parameter, and the speed characteristic parameter as indexes to determine a target optimization strategy; The shaping execution control module is used to adjust the shaping parameters of the tapioca pearl material according to a preset process rule set based on the target optimization strategy, obtain a periodic control data set, and use a shaping effect discriminator to perform state discrimination on the periodic control data set to obtain a target shaping result.

2. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 1, characterized in that: The feature fusion processing module is further used for: Using a time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence; Counting historical temperature fluctuation data between an upper stability threshold and a lower stability threshold in the temperature segment sequence to obtain a calibration historical temperature data set, wherein the upper stability threshold is a critical value obtained by shifting the maximum value in the segment downward by a preset amplitude, and the lower stability threshold is a critical value obtained by shifting the minimum value in the segment upward by a preset amplitude; Calculating the variance of the calibration history temperature data set to obtain the temperature characteristic parameter; Multi-dimensional feature analysis is performed on the historical pressure change data set and the historical speed adjustment data set to obtain the pressure characteristic parameters and the speed characteristic parameters.

3. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 2, characterized in that: The feature fusion processing module is further used for: Arrange the historical temperature fluctuation data set in ascending order according to chronological order, and take the historical temperature fluctuation data in the middle period as the median of the sequence; Calculating the top decile and the bottom decile of the historical temperature fluctuation data set; The first decile is used as the lower threshold, the second decile is used as the upper threshold, and the sequence median is combined to construct the temperature segmented sequence.

4. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 1, characterized in that: The optimization strategy generation module is also used for: Collect multiple sets of sample temperature characteristic parameters, multiple sets of sample pressure characteristic parameters, and multiple sets of sample speed characteristic parameters, as well as corresponding multiple sets of sample optimization strategies as training data; Pre-constructing a three-dimensional feature space, wherein the coordinate origin of the three-dimensional feature space is P, the A axis is the temperature characteristic parameter, the B axis is the pressure characteristic parameter, and the C axis is the speed characteristic parameter; The training data is input into the three-dimensional feature space to obtain multiple groups of sample feature points, and the multiple groups of sample feature points are labeled using the multiple groups of sample optimization strategies to obtain the molding process configuration space.

5. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 4, characterized in that: The optimization strategy generation module is also used for: A plane in the molding process configuration space that passes through the temperature characteristic parameter and is perpendicular to the BC plane is used as a first reference plane; A plane in the molding process configuration space that passes through the pressure characteristic parameter and is perpendicular to the AC plane is used as a second reference plane; A plane in the molding process configuration space that passes through the rotational speed characteristic parameter and is perpendicular to the AB plane is used as a third reference plane; The area enclosed by the BC plane, the AC plane, the AB plane, the first reference plane, the second reference plane, and the third reference plane is used as a configuration sub-area, wherein the configuration sub-area includes multiple groups of configuration sample feature points; A dynamic range search is performed on the multiple groups of configuration sample feature points to determine a target configuration sample feature point, and a sample optimization strategy corresponding to the target configuration sample feature point is used as the target optimization strategy.

6. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 5, characterized in that: The optimization strategy generation module is also used for: Extracting a central configuration sample feature point of the configuration sub-region, and constructing a central neighborhood region with the central configuration sample feature point as a starting point according to a preset dynamic search range, wherein the central neighborhood region is a spherical sub-region constructed with the central configuration sample feature point as a sphere center and the preset dynamic search range as a radius; Counting the number of sample feature points configured in the central neighborhood area, and dividing the statistical value by the volume of the central neighborhood area to obtain a central neighborhood density value; Randomly selecting a configuration sample feature point from the edge of the central neighborhood area as a first search feature point, and constructing a first search neighborhood density value of the first search feature point; Determine whether the first search neighborhood density value is greater than or equal to the central neighborhood density value. If so, update the first search feature point as the starting point and continue the dynamic range search until the preset number of dynamic searches is reached. The search feature point obtained from the last search is used as the target configuration sample feature point.

7. The intelligent optimization control system for tapioca pearl forming and shaping according to claim 6, characterized in that: The optimization strategy generation module is also used for: If not, the search update failure count with an initial value of 0 is updated to 1, and a configuration sample feature point is randomly selected from the edge of the central neighborhood area as the first search feature point for dynamic range search analysis. When the search update failure count is greater than the preset maximum search update failure count, the central configuration sample feature point is used as the target configuration sample feature point.

8. An intelligent optimization method for tapioca pearl forming and shaping, characterized in that: The method comprises: Execute multi-source sensor data acquisition to obtain a historical temperature fluctuation data set, a historical pressure change data set, and a historical speed adjustment data set of the tapioca pearl material monitored by the multi-source sensor within a preset process cycle; Traversing the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis to determine temperature characteristic parameters, pressure characteristic parameters, and speed characteristic parameters; Using the temperature characteristic parameter, the pressure characteristic parameter, and the speed characteristic parameter as indexes, a dynamic range search is performed in the molding process configuration space to determine a target optimization strategy; Based on the target optimization strategy, the molding parameters of the tapioca pearl material are adjusted according to a preset process rule set to obtain a periodic control data set, and the state of the periodic control data set is judged using a shaping effect discriminator to obtain a target molding shaping result.

9. The intelligent optimization method for forming and shaping tapioca pearls according to claim 8, characterized in that: The step of traversing the historical temperature fluctuation data set, the historical pressure change data set, and the historical speed adjustment data set to perform multi-dimensional feature analysis and determine the temperature characteristic parameters, the pressure characteristic parameters, and the speed characteristic parameters specifically includes: Using a time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence; Counting historical temperature fluctuation data between an upper stability threshold and a lower stability threshold in the temperature segment sequence to obtain a calibration historical temperature data set, wherein the upper stability threshold is a critical value obtained by shifting the maximum value in the segment downward by a preset amplitude, and the lower stability threshold is a critical value obtained by shifting the minimum value in the segment upward by a preset amplitude; Calculating the variance of the calibration history temperature data set to obtain the temperature characteristic parameter; The historical pressure change data set and the historical speed adjustment data set are subjected to multi-dimensional feature analysis using the same logic to obtain the pressure characteristic parameters and the speed characteristic parameters.

10. The intelligent optimization method for forming and shaping tapioca pearls according to claim 9, characterized in that: The step of using the time series segmentation method to split the historical temperature fluctuation data set into data dimensions to obtain a temperature segmentation sequence specifically includes: Arrange the historical temperature fluctuation data set in ascending order according to chronological order, and take the historical temperature fluctuation data in the middle period as the median of the sequence; Calculating the top decile and the bottom decile of the historical temperature fluctuation data set; The first decile is used as the lower threshold, the second decile is used as the upper threshold, and the sequence median is combined to construct the temperature segmented sequence.

Images