Granulation defect and taste quality synergistic evaluation of ultra-small recombination porridge rice granulation quality control method and system

Abstract

This application discloses a method and system for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and taste quality, relating to the field of adaptive control technology. The method includes: based on the batch data package to be controlled and the benchmark parameter set, extracting granulation defect deviations from dry granule images, and generating a taste quality risk set from cooking, texture, rice soup, and flavor data; further combining the two types of results for synergistic diagnosis, determining the main control target, candidate correction directions, and conflict indicators, and finally generating production parameter control information and executing production parameter control to achieve synergistic control of granulation quality and taste quality of ultra-small recombinant porridge rice. Under conflict constraint control scenarios, this application enables production parameter control to both reduce the main control deviation items and limit the continued deterioration of the secondary control deviation items, thereby improving the stability and executability of granulation quality control of ultra-small recombinant porridge rice.

Description

Technical Field

[0001] This application relates to the field of adaptive control technology, and more specifically, to a method and system for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and taste quality. Background Technology

[0002] Ultra-small reconstituted porridge rice is a granular porridge rice product made from grain raw materials such as broken rice through processes including compounding, conditioning, extrusion cooking, die extrusion, cutting and granulation, and drying and cooling. Compared with natural broken rice, the particle shape of ultra-small reconstituted porridge rice is reconstructed by the extrusion and granulation processes, which can improve the rehydration speed under smaller particle size conditions, adapting to consumer demands such as light cooking, short-time steaming, and rapid porridge formation. As food processing develops towards high-value utilization of resources, nutritional compounding, and convenient cooking, the preparation of ultra-small reconstituted porridge rice using by-product resources such as broken rice is beneficial for improving the problems of irregular shape and insufficient uniformity in steaming and cooking of traditional broken rice, and also for forming a new type of porridge rice product that combines the adaptability to light cooking and nutritional compounding characteristics.

[0003] In the engineered production of ultra-small reconstituted porridge rice, granulation quality and eating quality jointly determine product stability. Granulation quality typically involves particle size, cut end condition, particle integrity, and particle dispersibility; eating quality typically involves cooking and rehydration performance, grain texture, rice soup consistency and stability, as well as aroma, taste, and sensory acceptance. Due to the small particle size, the cutting process after die extrusion is highly sensitive to material gel strength, cutting speed, material moisture content, and emulsifier system, easily leading to granulation defects such as cutting tails, particle breakage, adhesion, and size fluctuations. Furthermore, process and formulation parameters not only affect particle formation but also the texture, rice soup consistency, and flavor acceptance after cooking. Therefore, quality control of ultra-small reconstituted porridge rice is not solely based on granulation indicators but involves comprehensive coordination between the granulation and eating aspects.

[0004] In the current production control of ultra-small reconstituted rice porridge or similar reconstituted rice products, the focus is usually on individually detecting particle appearance defects, individually evaluating cooking quality, or individually adjusting extrusion process parameters. There is a lack of continuous data transmission between the results of granulation defect detection, the results of taste quality evaluation, and the control of production parameters. Even if data such as particle size, cutting tailing, rehydration performance, rice grain hardness, rice soup state, and sensory scores can be obtained during the production process, it is difficult to uniformly convert the above data into deviation information that can be used for control decisions. As a result, the main quality shortcomings of the current batch are not easily identified, and subsequent adjustments to production parameters often rely on manual experience or judgment of single indicators.

[0005] Furthermore, there may be inconsistencies in the control directions between improving granulation defects and improving taste quality. For example, adjusting extrusion strength, moisture content, or emulsifier systems to improve granule forming stability may alter the hardness, cohesiveness, rice soup consistency, and flavor acceptability of cooked rice grains. Conversely, adjusting compound ratios or formulation structures to improve taste performance may affect the structural support capacity of the material during die extrusion and cutting. Existing technologies lack a collaborative diagnostic mechanism for granulation-side deviations and taste-side risks, and also lack methods for identifying control direction conflicts and constraining the adjustment range of production parameters in cases of bilateral deviations. This can easily lead to problems such as a decline in taste quality due to simply reducing granulation defects, or a decrease in granulation stability due to simply improving taste quality. Therefore, there is an urgent need for a method for controlling the granulation quality of ultra-small recombinant porridge rice that can link granulation defect evaluation, taste quality evaluation, and production parameter control to improve the accuracy of batch quality judgment and the stability of production control. Summary of the Invention

[0006] To address the lack of synergistic correlation between granulation defect evaluation, taste quality evaluation, and production parameter control in the existing ultra-small recombinant porridge rice production process, which leads to inaccurate identification of current batch quality shortcomings, tendency for control to be biased towards a single indicator, and difficulty in simultaneously achieving granulation stability and taste quality, this application provides a method and system for controlling the granulation quality of ultra-small recombinant porridge rice through synergistic evaluation of granulation defects and taste quality. By linking granulation-side deviations, taste-side risks, and production parameter control processes, production control can make synergistic judgments and constraints based on the current batch's granulation and taste quality status, thereby improving the accuracy and stability of ultra-small recombinant porridge rice granulation quality control.

[0007] Firstly, this application proposes a method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality, including:

[0008] Establish a batch data package and a set of baseline parameters to be adjusted. The batch data package includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data.

[0009] Effective particle instances are extracted based on dry particle image data and granulation benchmark parameters. Based on the effective particle instances, particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation are generated to form a set of granulation defect deviations.

[0010] Based on the cooking quality data, texture data, rice soup state data, and flavor data, a taste detection feature vector is constructed. This vector is then compared with the taste target feature vector to generate a taste deviation vector. Finally, a taste quality risk set is generated based on the taste deviation vector.

[0011] Based on the set of granulation defect deviations and the set of taste quality risks, the dominant deviations on the granulation side, the dominant risks on the taste side, the collaborative diagnosis types and the main control targets are determined. Based on the set of parameter response coefficients, a set of main control candidate correction directions and conflict identifiers are generated to form a set of collaborative diagnosis results.

[0012] The control type is determined based on the set of collaborative diagnostic results, the set of production parameters, and the set of baseline parameters. The production parameters to be controlled and the control step size are determined from the set of main control candidate correction directions according to the control type. Production parameter control information is generated and production parameter control is executed.

[0013] Secondly, this application proposes a quality control system for ultra-small recombinant porridge rice granulation that coordinates the evaluation of granulation defects and taste quality, and a method for implementing the aforementioned quality control system for ultra-small recombinant porridge rice granulation that coordinates the evaluation of granulation defects and taste quality, including:

[0014] The batch baseline filing module establishes a batch data package and a set of baseline parameters to be adjusted. The batch data package to be adjusted includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data.

[0015] The granulation defect deviation generation module extracts effective particle instances based on dry particle image data and granulation benchmark parameters, and generates particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation based on the effective particle instances, forming a set of granulation defect deviations.

[0016] The food quality risk assessment module constructs a food quality detection feature vector based on cooking quality data, texture data, rice soup state data, and flavor data. It compares the feature vector with the target food quality vector to generate a food quality deviation vector. Based on the food quality deviation vector, it generates a food quality risk set including a food quality risk profile, dominant food quality risk type, and food quality risk level.

[0017] The collaborative diagnosis module determines the dominant deviation on the granulation side, the dominant risk on the taste side, the collaborative diagnosis type, and the main control target based on the set of granulation defect deviations and the set of taste quality risks. It generates a set of main control candidate correction directions and conflict identifiers based on the set of parameter response coefficients, thus forming a set of collaborative diagnosis results.

[0018] The conflict constraint correction module determines the control type based on the collaborative diagnosis result set, the production parameter set, and the baseline parameter set. It then determines the production parameters to be controlled and the control step size from the master control candidate correction direction set according to the control type, generates production parameter control information, and executes production parameter control.

[0019] The method and system for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality proposed in this application have the following beneficial effects:

[0020] By establishing a batch data package and a set of baseline parameters to be controlled, the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data are incorporated into a unified data caliber. This allows the granulation results, cooking performance, grain texture, rice soup state, and flavor acceptance status of ultra-small recombinant porridge rice to be continuously accessed within the same batch dimension. This avoids the problem of fragmented granulation defect detection, taste quality evaluation, and process parameter adjustment in existing production processes, and provides a complete data foundation for subsequent collaborative diagnosis and production parameter control.

[0021] Based on dry particle image data and granulation benchmark parameters, valid particle instances are extracted. A set of granulation defect deviations is formed, encompassing particle size deviation, cutting tailing deviation, particle integrity deviation, and adhesion ratio deviation. This allows for the conversion of appearance defects such as dimensional instability, end tailing, particle breakage, and adhesion aggregation that occur during the cutting and granulation process of ultra-small reconstituted rice into deviation data that can be used in control decisions. This processing method not only improves the objectivity of granulation defect identification results but also provides a clear basis for subsequent judgment of dominant deviations on the granulation side.

[0022] Based on cooking quality data, texture data, rice soup state data, and flavor data, a taste detection feature vector is constructed. This vector is then compared with the target taste feature vector to generate a taste deviation vector. This process transforms taste-related data from different dimensions and testing sources into a unified deviation evaluation standard. By further generating a taste quality risk profile, dominant taste risk types, and taste risk levels, the taste shortcomings of the current batch can be identified from dimensions such as light cooking and rehydration, rice grain texture, rice soup harmony, and flavor acceptability. This allows taste quality evaluation to move beyond single test results and form a set of taste quality risks that can be used for control decisions.

[0023] Based on the set of granulation defect deviations and the set of taste quality risks, the dominant deviations on the granulation side, the dominant risks on the taste side, the collaborative diagnostic types, and the main control objectives are determined. Furthermore, based on the set of parameter response coefficients, a set of candidate correction directions for the main control and conflict markers are generated. This allows identification of whether the current batch is in a state of deviation on the granulation side, the taste side, or both. This collaborative diagnostic process avoids the decline in taste quality caused by simply pursuing a reduction in cutting tailing, and also avoids the deterioration of granulation stability caused by simply improving taste quality, thus achieving a coordinated judgment between the granulation defect control objectives and the taste quality objectives.

[0024] Further, based on the collaborative diagnostic result set, production parameter set, and baseline parameter set, the control type is determined. Then, according to the control type, the production parameters to be controlled and the control step size are determined from the set of primary control candidate correction directions. Production parameter control information is generated and production parameter control is executed. In conflict-constrained control scenarios, this application can constrain candidate control directions based on the remaining allowable deviation of secondary controls. This allows production parameter control to both reduce primary control deviations and limit the further deterioration of secondary control deviations, thereby improving the stability and feasibility of quality control in ultra-small recombinant rice granulation.

[0025] Therefore, this application can form a closed-loop control process from batch data filing, granulation defect deviation generation, taste quality risk assessment, granulation taste co-diagnosis to production parameter control execution. This provides clear data basis and control objectives for adjusting production parameters such as barrel temperature, material moisture addition, screw speed, feeding speed, cutting speed, emulsifier addition, emulsifier type, die specifications, and the ratio of broken rice to quinoa. This is beneficial for improving the uniformity of granule formation of ultra-small recombinant porridge rice, reducing the risk of granulation defects such as cutting tailing and adhesion, and taking into account light cooking and rehydration, rice grain texture, rice soup state and flavor acceptance quality, thereby improving the quality stability in the engineering continuous production of ultra-small recombinant porridge rice. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the ultra-small recombinant porridge rice granulation quality control system for the synergistic evaluation of granulation defects and taste quality in Example 1 of this application.

[0027] Figure 2 This is a flowchart of the method for controlling the granulation quality of ultra-small recombinant porridge rice granules in Example 2 of this application, which involves synergistic evaluation of granulation defects and taste quality.

[0028] Figure 3 This is a flowchart illustrating the method for forming a set of collaborative diagnostic results according to Embodiment 1 of this application;

[0029] Figure 4 This is a flowchart of the method for generating production parameter control information in Embodiment 1 of this application. Detailed Implementation

[0030] The technical solutions in the embodiments of this application will be described in detail, clearly and completely below with reference to the accompanying drawings.

[0031] Example 1:

[0032] like Figure 1As shown in the figure, this embodiment discloses an ultra-small recombinant porridge rice granulation quality control system for synergistic evaluation of granulation defects and taste quality. It includes a batch baseline filing module, a granulation defect deviation generation module, a taste quality risk assessment module, a synergistic diagnosis module, and a conflict constraint correction module. Each module realizes data transmission through wired connection, wireless connection, or a combination of wired and wireless connection.

[0033] The batch baseline filing module establishes a batch data package and a set of baseline parameters to be adjusted. The batch data package to be adjusted includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data.

[0034] The batch baseline documentation module first establishes a batch data package to be adjusted. This batch data package refers to a set of data used to store the current batch's production parameters, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data. The batch data package to be adjusted is created by the production control system.

[0035] When multiple sets of sampling data exist for the same batch to be regulated, the batch baseline filing module generates representative values ​​for the current batch according to preset statistical rules. For barrel temperature, screw speed, feeding speed, and cutting speed, the system generates time averages based on actual feedback values ​​during the extrusion stabilization phase, and uses these time averages as representative values ​​for the current batch's production parameters. For cooking time, rice grain hardness, rice grain elasticity, rice grain cohesion, apparent viscosity of rice soup, rice soup flow time, rice soup turbidity, and settling stratification ratio, the system generates averages based on multiple sets of test data from the same batch, and uses these averages as representative values ​​for the current batch's quality inspection. For rehydration rate and cooking loss, the system accumulates the statistical numerator and denominator from multiple sets of sampling data from the same batch, and then uses the ratio of the accumulated numerator to the accumulated denominator as representative values ​​for the current batch's quality inspection. For the blending ratio of broken rice and quinoa, emulsifier type, emulsifier addition amount, and die specifications, the system uses the actual usage information in the current batch's production record as representative information for the current batch's production parameters. Dry particle image data is not directly used to generate cutting tailing rate, particle integrity rate and adhesion ratio in this module. Dry particle image data is called in the subsequent granulation defect deviation generation process and generates the corresponding granulation defect representative value after particle identification and defect statistics are completed.

[0036] The set of production parameters refers to the set of parameters that reflect the current batch processing conditions and are used to generate subsequent control strategies, including the ratio of broken rice to quinoa, barrel temperature, material moisture content, screw speed, feeding speed, type of emulsifier, amount of emulsifier added, die specifications, and cutting speed.

[0037] The ratio of broken rice to quinoa is collected by the ingredient weighing device before feeding, with the measurement scope being the ratio between the dry weight of broken rice and the dry weight of quinoa. This is subsequently used in the co-diagnosis of granulation defects and taste quality to determine the flavor risk and granulation taste conflict risk related to the ratio. The barrel temperature, screw speed, feeding speed, and cutting speed are collected by the equipment controller during the extrusion stabilization phase, with the measurement scope being the time average of the actual feedback values ​​during the extrusion stabilization phase. This is subsequently used to generate extrusion parameter control information and cutting parameter control information. The amount of moisture added is collected by the water metering device, with the measurement scope being the ratio of the added water mass to the dry weight of the mixed powder. This is subsequently used to determine the material adhesion risk and the direction of gel strength control. The type and amount of emulsifier added are collected by the ingredient recording system. The measurement caliber for emulsifier type is the name of the emulsifier in the preset emulsifier list, and the measurement caliber for emulsifier addition amount is the ratio of emulsifier mass to the dry basis mass of the mixed powder. This is subsequently used in the co-diagnosis of granulation defects and taste quality to determine the granulation risk and taste limitation risk related to gel strength. The preset emulsifier list refers to the set of food processing emulsifier names pre-stored in the benchmark parameter set, which is derived from food formulation constraints and historical test records. The die specifications are read from the equipment parameter table, and the measurement caliber is the die cross-sectional shape and die outlet size, which is subsequently used to determine particle size deviation.

[0038] The dry particle image data refers to the particle appearance images and calibration data used for size conversion obtained by the image acquisition device after the current batch of ultra-small reconstituted porridge rice has been dried, cooled, and spread in a single layer. The measurement scope of the dry particle image data is the particle sample obtained according to the preset sampling quality in the current batch and after sieving to remove powder. Each set of dry particle image data is written with a pixel size conversion coefficient. The pixel size conversion coefficient is determined by the scale calibration plate and is subsequently used to convert the pixel length of the particles in the image into the actual particle length and actual particle width. If several sets of particle image data lack pixel size conversion coefficients, or if the particle overlap area ratio exceeds the upper limit of the overlap area in the benchmark parameter set, then the set of dry particle image data will not enter the granulation defect deviation generation process and will trigger re-sampling and re-acquisition; if the image entry conditions are still not met after re-acquisition, the set of dry particle image data will not enter the granulation defect deviation generation process; if there is no usable dry particle image data in the current batch, then the current batch will not enter the subsequent granulation defect and taste quality co-diagnosis process.

[0039] It should be noted that the ultra-small reconstituted porridge rice described in this application refers to a granular porridge rice product formed by mixing and conditioning broken rice as the main raw material, compounded with quinoa, and then extruding, cooking, die-cutting, granulating, and drying. This ultra-small reconstituted porridge rice differs from naturally broken rice formed directly from rice milling; its particle shape is reconstructed through the extrusion and granulation processes. It also differs from pre-cooked porridge products that are already cooked and ready to eat, requiring further steaming before consumption. The ultra-small reconstituted porridge rice has a target particle length range of 2 mm to 4 mm and a target particle width range of 1 mm to 2.5 mm in its dry granule state, meeting the requirements for dispersibility, and is used to form a porridge body where rice grains and rice water coexist without prior soaking. The target particle length and width ranges serve as benchmarks for subsequent judgment of particle size deviation, and the attribute of not requiring prior soaking and forming a porridge body serves as the product basis for subsequent evaluation of light cooking adaptability, rice grain texture, rice water state, and flavor.

[0040] Before collecting cooking quality data, texture data, rice water state data, and flavor data, the batch benchmark documentation module reads the cooking test condition set. This cooking test condition set refers to the set of parameters used to define the cooking test conditions and post-cooking test conditions for the current batch, including sample mass, water mass, heating method, cooking container specifications, draining time, settling time, test temperature, and rice water sampling volume. The cooking test condition set is derived from production process documents and qualified historical batch test conditions and is written into the benchmark parameter set. The cooking test condition set is subsequently used to ensure that cooking quality data, texture data, rice water state data, and flavor data are generated under the same testing caliber.

[0041] The cooking quality data refers to the set of data reflecting the rehydration and cooking performance of the current batch of ultra-small recombinant porridge rice under the defined conditions of cooking test, including cooking time, rehydration rate, and cooking loss. Cooking time is collected by the cooking test device, and the measurement is the duration required for the sample to reach the cooking endpoint as defined in the benchmark parameter set after being added to boiling water; this is subsequently used to generate the light cooking adaptability evaluation results. Rehydration rate is collected by a weighing device, and the measurement is the increase in sample mass after cooking and draining relative to the sample mass before cooking; this is subsequently used to determine rehydration capacity. Cooking loss is collected by a dry matter analyzer, and the measurement is the ratio of the mass of dissolved dry matter in the cooking liquid to the dry basis mass of the sample before cooking; this is subsequently used to determine the rice soup formation state and particle structure retention state.

[0042] The texture data refers to the mechanical response data exhibited by the cooked, ultra-small recombinant rice grains under texture testing conditions, including grain hardness, grain elasticity, and grain cohesion. Texture data is acquired using a texture analyzer. During testing, the cooked, ultra-small recombinant rice grains are placed on a testing platform, and a preset compression probe is used to perform two consecutive compressions on the grains, recording the force change curves during the compression process. Grain hardness is determined based on the force value corresponding to the maximum force point during the first compression, reflecting the grain's ability to resist compression deformation. Grain elasticity is determined based on the recovery ratio between the grain's recovered height and its original height after the first compression, reflecting the rebound state of the cooked grains. Grain cohesion is determined based on the ratio between the area of ​​the second compression curve and the area of ​​the first compression curve, reflecting the ability of the cooked grain's internal structure to maintain continuity. The texture data is subsequently used to generate rice grain texture evaluation results and to assess granulation risks and taste limitations related to gel strength in the co-diagnosis of granulation defects and taste quality.

[0043] The rice water state data refers to the set of data on the consistency, flowability, turbidity, and stability of the liquid phase of the porridge after cooking. This includes apparent viscosity, flow time, turbidity, and separation ratio. Apparent viscosity is collected by a viscosity testing device, with a measurement range of a preset volume of rice water sample taken after cooking and measured at a specified temperature under the cooking test conditions. The unit is millipascals per second (mPa·s), which is used to determine the consistency of the rice water. Flow time is collected by a flowability testing device, with a measurement range of the time required for a preset volume of rice water sample to pass through a standard test channel at the specified temperature under the cooking test conditions. The unit is seconds (seconds), which is used to determine the flowability of the rice water. Turbidity is collected by an optical detection device, with a measurement range of the turbidity test result obtained under the specified conditions of the cooking test conditions. This value characterizes the degree of release of soluble and suspended substances in the rice water and is used to determine the appearance of the rice water. The separation ratio during settling was acquired by an image acquisition device. The measurement caliber was the ratio between the height of the supernatant layer formed after the rice soup sample was set to a set of settling time under the cooking test conditions and the total height of the sample. This ratio was subsequently used to determine the stability of the rice soup.

[0044] The flavor and taste data refers to a set of data reflecting the aroma, taste, and sensory acceptance of the current batch of ultra-small recombinant rice after steaming and cooking, including aroma response data, taste response data, and sensory evaluation data. Aroma response data is collected by an electronic nose detection device, with a measurement scope of the response value of the cooked rice porridge sample to the aroma sensor array under the defined steaming and cooking detection conditions, used to generate flavor evaluation results. Taste response data is collected by an electronic tongue detection device, with a measurement scope of the response value of the rice water sample to the taste sensor array under the defined steaming and cooking detection conditions, used to generate taste evaluation results. Sensory evaluation data is collected by evaluators according to a preset sensory evaluation table, with a measurement scope of the scoring results for rice grain texture, rice water state, aroma, taste, and overall acceptance under the same steaming and cooking conditions. The preset sensory evaluation table refers to a pre-stored scoring table in the benchmark parameter set, which includes scoring items, score ranges, and scoring scopes for rice grain texture, rice water state, aroma, taste, and overall acceptance. Sensory evaluation data is used to correct the flavor and taste evaluation results.

[0045] The benchmark parameter set refers to the data set used to provide target ranges, judgment thresholds, and control boundaries for subsequent granulation defect deviation generation, taste quality risk assessment, collaborative diagnosis of granulation defects and taste quality, and production parameter control information generation. The benchmark parameter set originates from product design requirements, production process documents, equipment calibration records, and qualified historical batch data, and together with the batch data set to be controlled, serves as the basic input for subsequent processing. The benchmark parameter set includes granulation benchmark parameters required for generating granulation defect deviations, taste benchmark parameters required for taste quality evaluation, and control boundary parameters required for production parameter control.

[0046] The granulation defect deviation generation module extracts effective particle instances based on dry particle image data and granulation benchmark parameters, and generates particle size deviation, cutting tail deviation, particle integrity deviation, and adhesion ratio deviation based on the effective particle instances, forming a set of granulation defect deviations.

[0047] The granulation defect deviation generation module is used to extract effective particle instances of the current batch of ultra-small recombinant rice based on the dry particle image data in the batch data package to be controlled and the granulation benchmark parameters in the benchmark parameter set. It then generates a granulation defect deviation set based on the size recognition results, cutting and tailing recognition results, particle integrity recognition results, and adhesion recognition results of the effective particle instances. The granulation defect deviation set refers to the data set characterizing the deviation of the current batch's granulation result from the granulation target state, including particle size deviation, cutting and tailing deviation, particle integrity deviation, and adhesion ratio deviation.

[0048] The granulation defect deviation generation module reads the dry particle image data from the batch data package to be regulated and reads the granulation benchmark parameters from the benchmark parameter set. The granulation benchmark parameters refer to a set of parameters used to limit the granulation target, image basic verification conditions, particle instance extraction conditions, and defect judgment conditions for ultra-small recombinant rice dry particles. These parameters include the target granulation size range, upper limit for tailing rate, upper limit for tailing severity, lower limit for particle integrity, upper limit for adhesion ratio, upper limit for overlap area, minimum particle projection area threshold, maximum particle projection area threshold, tailing length threshold, tailing width ratio threshold, main body width fluctuation threshold, end defect area threshold, contour interruption length threshold, neck width ratio threshold, image segmentation parameters, and deviation level judgment rules. The target granulation size range includes the target particle length range and the target particle width range; the target particle length range is 2 mm to 4 mm, and the target particle width range is 1 mm to 2.5 mm. The image segmentation parameters include the background grayscale range, background color range, and edge judgment threshold. The deviation level judgment rules are used to determine the levels of particle size deviation, particle integrity deviation, adhesion ratio deviation, and cutting tailing deviation. The above parameters are derived from product design requirements, qualified historical batch data, and manually verified qualified samples, and are stored in the benchmark parameter set.

[0049] The granulation defect deviation generation module first performs basic image verification on the dry granule image data. If several sets of granule image data lack pixel size conversion coefficients, size conversion and defect identification are not performed on that set of dry granule image data, and re-acquisition is triggered. If pixel size conversion coefficients are still missing after re-acquisition, the set of dry granule image data is marked as unusable image data. If no usable image data exists for the current batch, the granulation defect deviation generation module does not generate a granulation defect deviation set, and the current batch is not included in the subsequent collaborative diagnosis process of granulation defects and taste quality.

[0050] After the basic image verification is passed, the granulation defect deviation generation module performs particle instance extraction processing on the dry particle image data. Particle instance extraction refers to identifying and separating the image region corresponding to a single particle from the dry particle image data, and treating each image region as a particle object to be analyzed. Specifically, the granulation defect deviation generation module first performs grayscale conversion and image enhancement processing on the dry particle image data, and calculates the grayscale difference, color difference, and edge gradient change between each pixel region and the background region in the image. Pixel regions whose grayscale values ​​deviate from the background grayscale range, whose color features deviate from the background color range, and whose edge gradients are higher than the edge judgment threshold are marked as candidate particle regions. Subsequently, threshold segmentation processing is performed on the candidate particle regions to generate particle foreground regions. Morphological opening and closing operations are then used to remove isolated noise regions and fill in internal voids within the particles, thereby obtaining continuous particle foreground regions. Finally, connected component analysis is performed on the particle foreground regions to extract the independent foreground regions as candidate particle regions.

[0051] For each candidate particle region, the granulation defect deviation generation module calculates its projected area and classifies it based on the minimum and maximum projected area thresholds. When the projected area of ​​a candidate particle region is less than the minimum projected area threshold, it is classified as a powder or debris region and not included in subsequent particle statistics. When the projected area of ​​a candidate particle region is greater than the maximum projected area threshold, it is classified as a suspected adhesion region and proceeds to the subsequent adhesion identification process. When the projected area of ​​a candidate particle region is between the minimum and maximum projected area thresholds, its outer contour is further checked for continuity and closure. If the outer contour meets the closure condition, the candidate particle region is determined as a valid particle instance. A valid particle instance refers to a particle image region whose projected area is between the minimum and maximum projected area thresholds, whose outer contour is continuous and forms a closed boundary, and which can be used for size conversion and participate in granulation defect statistical analysis. The continuous outer contour and the formation of a closed boundary mean that there are no opening defects at the edge of the particle region that would cause the particle interior to connect with the background region. Subsequent particle size recognition, cutting tail recognition, and particle integrity recognition are all based on valid particle instances.

[0052] The particle overlap area ratio is generated based on candidate particle regions and suspected adhesion regions. The particle overlap area ratio refers to the proportion of the area that cannot be attributed to a single particle in a dry particle image when two or more particles are in contact, overlapping, or partially fused, causing the boundary contours between particles to be indistinguishable according to instance segmentation rules, relative to the total area of ​​the foreground region of all particles. The particle foreground region refers to all regions in the image identified as particles; the overlap area refers to the area that cannot be clearly classified as a single particle due to particle contact, overlapping, or partial fusion. The granulation defect deviation generation module uses the area of ​​the region in the suspected adhesion region that cannot be segmented into a valid particle instance as the overlap area, and generates the particle overlap area ratio based on the ratio of the overlap area to the total area of ​​the particle foreground region. If the particle overlap area ratio exceeds the upper limit of the overlap area, the dry particle image data set is not included in the granulation defect statistics, and re-layouting and re-acquisition are triggered; if the particle overlap area ratio still exceeds the upper limit of the overlap area after re-acquisition, the dry particle image data set is marked as unusable image data.

[0053] For each valid particle instance, the granulation defect deviation generation module generates particle length and particle width based on pixel size conversion factors. The particle length refers to the actual length corresponding to the longer side of the minimum bounding rectangle of the valid particle instance; the particle width refers to the actual length corresponding to the shorter side of the minimum bounding rectangle of the valid particle instance. The granulation defect deviation generation module compares the particle length with the target particle length range and the particle width with the target particle width range. When the particle length exceeds the upper limit of the target particle length range, the corresponding valid particle instance is marked as a particle exceeding the upper limit; when the particle length is below the lower limit of the target particle length range, the corresponding valid particle instance is marked as a particle below the lower limit. When the particle width exceeds the upper limit of the target particle width range, the corresponding valid particle instance is marked as a particle exceeding the upper limit; when the particle width is below the lower limit of the target particle width range, the corresponding valid particle instance is marked as a particle below the lower limit.

[0054] The granulation defect deviation generation module statistically analyzes the proportions of particles exceeding the upper limit of length, particles below the lower limit of length, particles exceeding the upper limit of width, and particles below the lower limit of width, and calculates the average deviation of each particle from the target range boundary.

[0055] The average deviation refers to the average difference between the actual size of the corresponding out-of-range particle and the target range boundary. For particles exceeding the length upper limit, the average deviation is the average difference between the particle length and the upper limit of the target length range; for particles below the length lower limit, the average deviation is the average difference between the lower limit of the target length range and the particle length; for particles exceeding the width upper limit, the average deviation is the average difference between the particle width and the upper limit of the target width range; for particles below the width lower limit, the average deviation is the average difference between the lower limit of the target width range and the particle width. Deviations for exceeding the upper limit in length, below the lower limit in length, exceeding the upper limit in width, and below the lower limit in width are generated based on the corresponding proportions and the corresponding average deviations. The particle size deviation is composed of deviations for exceeding the upper limit in length, below the lower limit in length, exceeding the upper limit in width, and below the lower limit in width. It is used to indicate the direction and degree of deviation of the particle size of the current batch relative to the target granulation size range.

[0056] A cutting and tailing identification process is performed on valid particle instances. The cutting and tailing refers to a granulation defect where, after ultra-small reconstituted rice is extruded from a die and cut by a cutter, a tail-like dragging region forms at the end of the particle. The tail-like dragging region is a slender area located at the end of the particle's main body region, extending outwards along the particle's length direction, with a width less than the average width of the main body region, and continuously connected to the main body region. The particle's main body region refers to the area continuously distributed along the long side of the minimum bounding rectangle in valid particle instances, with width variations within the main body width fluctuation threshold. The granulation defect deviation generation module identifies the end-extending regions at both ends of valid particle instances. These end-extending regions are located at both ends of the particle's main body region, extending beyond the boundary of the main body region and continuing along the particle's length direction. Their starting position is where the particle's local width first falls below the product of the main body average width and the tailing width ratio threshold, and their ending position is the particle's outer contour endpoint. When the extension length of the end-extending region exceeds the tailing length threshold, and the average width of the end-extending region is less than the product of the main body average width and the tailing width ratio threshold, the corresponding valid particle instance is marked as a tailing particle.

[0057] The tailing incidence rate is generated based on the number of tailing particles and the total number of valid particle instances. The tailing severity is generated based on the extension length and area of ​​the tail-like dragging region within the tailing particles. For each tailing particle, the ratio of the tail-like dragging region length to the length of the corresponding particle's main body region is calculated, as is the ratio of the tail-like dragging region area to the total area of ​​the corresponding particle. These two ratios are then weighted and summed according to preset weights to generate the single-particle tailing severity. The average of the single-particle tailing severity for all tailing particles in the current batch is then calculated to generate the tailing severity for the current batch. The preset weights include length weights and area weights, the sum of which is one. The length and area weights are determined based on the contribution of the tail-like dragging length and area in the manually reviewed defect samples to the tailing determination result and are stored in a benchmark parameter set.

[0058] Cutting tailing deviation includes cutting tailing occurrence rate, tailing severity, tailing occurrence rate exceeding limit, tailing severity exceeding limit, and cutting tailing deviation level. The cutting tailing occurrence rate characterizes the proportion of particles with cutting tailing defects in the current batch, while the tailing severity characterizes the degree of end dragging of the tailing particles. When the cutting tailing occurrence rate exceeds the upper limit, the tailing occurrence rate exceeding limit is determined to be exceeding the limit; when the cutting tailing occurrence rate is not higher than the upper limit, the tailing occurrence rate exceeding limit is determined to be not exceeding the limit. When the tailing severity exceeds the upper limit, the tailing severity exceeding limit is determined to be exceeding the limit; when the tailing severity is not higher than the upper limit, the tailing severity exceeding limit is determined to be not exceeding the limit. If both the tail-wagging incidence and severity exceed limits are within limits, the cutting tail-wagging deviation level is determined to be within limits. If only the tail-wagging incidence exceeds limits, the cutting tail-wagging deviation level is determined to be incidence exceeding limits. If only the tail-wagging severity exceeds limits, the cutting tail-wagging deviation level is determined to be severity exceeding limits. If both the tail-wagging incidence and severity exceed limits, the cutting tail-wagging deviation level is determined to be a combined exceedance. The cutting tail-wagging deviation is used to characterize the degree of tail-like dragging defects formed during the cutting and granulation process of the current batch of ultra-small recombinant porridge rice. In the subsequent collaborative diagnosis type generation process, the cutting tail-wagging deviation is jointly judged with particle size deviation, particle integrity deviation, adhesion ratio deviation, production parameter set, and taste quality risk set to determine the collaborative diagnosis type of the current batch and provide granulation-side input for subsequent forward correction control or conflict constraint control.

[0059] Further identification of particle integrity is performed. Particle integrity refers to whether a valid particle instance maintains a short, granular body shape and has a continuous outer contour. The granulation defect deviation generation module judges the continuity of the outer contour, the area of ​​end defects, and the state of body fracture of valid particle instances. If the length of the outer contour interruption of a valid particle instance exceeds the contour interruption length threshold, the area of ​​end defects exceeds the end defect area threshold, or the main body region is divided into two or more discontinuous regions, the corresponding valid particle instance is marked as an incomplete particle; if a valid particle instance does not meet any of the above conditions, the corresponding valid particle instance is marked as a complete particle. The length of the outer contour interruption refers to the length of the gap in the outer contour of a valid particle instance that cannot form a continuous boundary; the area of ​​end defects refers to the area of ​​the region missing from the end of a valid particle instance relative to the short, granular body contour; the state of body fracture refers to the state in which the main body region of a valid particle instance is divided into two or more discontinuous regions by the background region.

[0060] The particle integrity rate is generated based on the number of intact particles and the total number of effective particle instances. The particle integrity rate is the ratio of the number of intact particles to the total number of effective particle instances. If the particle integrity rate is lower than the lower limit of particle integrity rate in the benchmark parameter set, a particle integrity deficiency deviation is generated; if the particle integrity rate is not lower than the lower limit, the particle integrity deviation is determined to be within the limit. The deviation amount of the particle integrity deficiency deviation is determined based on the difference between the lower limit of particle integrity rate and the particle integrity rate of the current batch. The granulation defect deviation generation module further determines the particle integrity deviation level according to the deviation level judgment rules: when the particle integrity rate is not lower than the lower limit, the particle integrity deviation level is within the limit; when the particle integrity rate is lower than the lower limit and the deviation amount does not exceed the particle integrity control threshold, the particle integrity deviation level is slightly exceeded; when the particle integrity rate is lower than the lower limit and the deviation amount exceeds the particle integrity control threshold, the particle integrity deviation level is severely exceeded. The particle integrity deviation includes the particle integrity rate, particle integrity deficiency deviation, and particle integrity deviation level, and serves as the granulation-side input for subsequent collaborative diagnostic type generation and conflict verification.

[0061] For suspected adhesion regions, adhesion identification is performed. An adhesion region refers to a continuous area formed by two or more particle instances in an image where their boundaries meet, are partially fused, or are insufficiently separated. The projected area, the location of the outer contour indentation, and the number of contracted necks of the suspected adhesion region are read. A contracted neck refers to a connection point where the outer contour of the suspected adhesion region is relatively concave, and its local width is lower than the ratio threshold between the width of the adjacent main body and the neck width. When the projected area of ​​a suspected adhesion region is greater than the maximum projected area threshold of the particle, and there are two or more contracted necks in the outer contour, the suspected adhesion region is marked as an adhesion region. For adhesion regions that can be segmented into multiple particle instances according to the contracted necks, segmentation is performed, and the segmented particle instances are re-entered into the valid particle instance set. For adhesion regions where the projected area of ​​the segmented sub-region is less than the minimum projected area threshold of the particle, the outer contour of the sub-region is not closed, or the number of sub-regions does not match the number of contracted necks, instance segmentation is not performed, and the adhesion region is included in the adhesion statistics.

[0062] The adhesion ratio is generated based on the number of adhered particles and the total number of particles in the current batch. The number of adhered particles refers to the total number of particles identified as adhered areas that could not be restored to independent valid particle instances using the segmentation method that separates the adhered area into multiple independent particle instances according to the contraction neck position. For each unsegmented adhered area, the granulation defect deviation generation module calculates the ratio between the projected area of ​​the unsegmented adhered area and the average projected area of ​​the valid particle instances, and rounds this ratio to the nearest integer as the number of particles corresponding to that unsegmented adhered area; the sum of the particle numbers corresponding to multiple unsegmented adhered areas is determined as the number of adhered particles. The total number of particles in the current batch refers to the sum of the number of valid particle instances and the number of adhered particles. The adhesion ratio is the proportion of the number of adhered particles to the total number of particles in the current batch. When the adhesion ratio is higher than the upper limit of the adhesion ratio, an adhesion ratio exceeding the limit deviation is generated; when the adhesion ratio is not higher than the upper limit of the adhesion ratio, no adhesion ratio exceeding the limit deviation is generated, and the deviation amount of the adhesion ratio exceeding the limit deviation is recorded as zero. The deviation amount for exceeding the adhesion ratio limit is determined based on the difference between the adhesion ratio and the upper limit of the adhesion ratio. The granulation defect deviation generation module further determines the adhesion ratio deviation level according to the deviation level judgment rules: when the adhesion ratio is not higher than the upper limit of the adhesion ratio, the adhesion ratio deviation level is not exceeding the limit; when the adhesion ratio is higher than the upper limit of the adhesion ratio and the deviation amount does not exceed the adhesion ratio control threshold, the adhesion ratio deviation level is slightly exceeding the limit; when the adhesion ratio is higher than the upper limit of the adhesion ratio and the deviation amount exceeds the adhesion ratio control threshold, the adhesion ratio deviation level is severely exceeding the limit. The adhesion ratio deviation includes the adhesion ratio, the adhesion ratio exceeding the limit deviation, and the adhesion ratio deviation level, and serves as the granulation-side input for subsequent collaborative diagnostic type generation and conflict verification.

[0063] After completing particle size identification, cutting tailing identification, particle integrity identification, and adhesion identification, the granulation defect deviation generation module generates a granulation defect deviation set. This set includes particle size deviation, cutting tailing deviation, particle integrity deviation, and adhesion ratio deviation. Specifically, particle size deviation records the direction and degree of deviation of particle length and width from the target granulation size range; cutting tailing deviation records the occurrence rate, severity, and level of cutting tailing deviation; particle integrity deviation records the particle integrity rate, insufficient particle integrity deviation, and level of particle integrity deviation; and adhesion ratio deviation records the adhesion ratio, excessive adhesion ratio deviation, and level of adhesion ratio deviation. The granulation defect deviation set is written into the batch data package to be controlled and serves as the granulation-side input for subsequent taste quality risk generation, collaborative diagnosis type generation, conflict verification, and production parameter control.

[0064] The taste quality risk assessment module constructs a taste detection feature vector based on cooking quality data, texture data, rice soup state data, and flavor data. It compares the feature vector with the taste target feature vector to generate a taste deviation vector. Based on the taste deviation vector, it generates a taste quality risk set including a taste quality risk profile, dominant taste risk type, and taste risk level.

[0065] The taste quality risk assessment module generates a taste quality risk set for the current batch of ultra-small recombinant porridge rice based on the cooking quality data, texture data, rice soup state data, and flavor data in the batch data package to be regulated. This taste quality risk set refers to a data set characterizing the taste deviation of the current batch of ultra-small recombinant porridge rice from the acceptable taste target state. It includes a taste quality risk profile, dominant taste risk type, and taste risk level. The taste quality risk profile includes deviations in light cooking and rehydration, rice grain texture, rice soup harmony, and flavor acceptance. The taste quality risk set serves as the taste-side input for subsequent collaborative diagnostic type generation, granulation and taste conflict verification, and production parameter control.

[0066] The taste quality risk assessment module first reads the cooking quality data, texture data, rice soup state data, and flavor data from the batch data package to be controlled, and then reads the taste benchmark parameters from the benchmark parameter set. The taste benchmark parameters are a set of parameters used to define the qualified taste target state and the rules for generating taste risks. These include the taste target profile, field order rules, field direction attributes, allowable deviation boundaries, dimension field attribution rules, dimension weight rules, dominant risk identification rules, and taste risk grading rules. The taste target profile is a set of target data formed based on qualified historical batch data, product design requirements, and manual review results, used as a comparison benchmark for generating the taste quality risk profile of the current batch.

[0067] The field order rule refers to the arrangement order of the representative values ​​in the taste detection feature vector. The taste detection feature vector is generated in the following order: cooking time, rehydration rate, cooking loss, rice grain hardness, rice grain elasticity, rice grain cohesion, apparent viscosity of rice soup, rice soup flow time, rice soup turbidity, standing stratification ratio, aroma response data, taste response data, and sensory evaluation data. For aroma response data, it is written into the taste detection feature vector in ascending order of electronic nose sensor channel number; for taste response data, it is written into the taste detection feature vector in ascending order of electronic tongue sensor channel number; for sensory evaluation data, it is written into the taste detection feature vector in the following order: rice grain texture score, rice soup state score, aroma score, taste score, and overall acceptability score.

[0068] The field orientation attribute refers to the attribute used to determine the direction of deviation when the detected representative value deviates from the target boundary. Field orientation attributes include smaller is better, larger is better, and range-appropriate. Cooking time, cooking loss, and settling stratification ratio are set as smaller is better fields; rehydration rate, rice grain hardness, rice soup apparent viscosity, rice soup flow time, rice soup turbidity, aroma response data, and taste response data are set as range-appropriate fields; rice grain elasticity, rice grain cohesion, and sensory evaluation data are set as larger is better fields. For sensor channels in aroma response data and taste response data, if the field orientation attribute is stored separately for each sensor channel in the benchmark parameter set, the field orientation attribute corresponding to that single sensor channel is preferentially used; if the field orientation attribute is not stored separately, it is processed as a range-appropriate field.

[0069] The allowable deviation boundary refers to the benchmark boundary used to convert the deviation range of representative test values ​​with different dimensions into a unified risk measurement caliber. Each representative test value has corresponding target boundary information and allowable deviation boundary stored in the taste target profile. For fields where smaller is better, the target boundary information is the upper limit; for fields where larger is better, the target boundary information is the lower limit; for fields of appropriate range, the target boundary information is the target range. The allowable deviation boundary is determined based on the fluctuation range of the corresponding representative test value in qualified historical batch data, the product design allowable error, and the results of manual review, and is stored in the benchmark parameter set.

[0070] The taste quality risk assessment module constructs a taste detection feature vector according to the field order rules. The taste detection feature vector refers to a data set formed by arranging representative values ​​of the current batch's taste quality-related tests in a unified order. By constructing the taste detection feature vector, cooking quality data, texture data, rice soup state data, and flavor data are unified into taste-side inputs under the same field order.

[0071] A taste target feature vector is constructed based on the taste target profile. The taste target feature vector refers to a set of target data with the same field order as the taste detection feature vector, including target boundary information and allowable deviation boundaries corresponding to each representative detection value. The taste target feature vector corresponds one-to-one with the taste detection feature vector and is used to generate a taste deviation vector field by field.

[0072] A taste deviation vector is generated based on the taste detection feature vector and the taste target feature vector. The taste deviation vector is a data set recording the normalized deviation values ​​of each representative value in the taste detection feature vector relative to the corresponding target boundary. The taste deviation vector has the same field order as the taste detection feature vector. Each field deviation value in the taste deviation vector corresponds to a representative detection value. These field deviation values ​​include deviations for cooking time, rehydration rate, cooking loss, rice grain hardness, rice grain elasticity, rice grain cohesion, apparent viscosity of rice soup, flow time of rice soup, turbidity of rice soup, separation ratio during settling, aroma response, taste response, and sensory evaluation.

[0073] For fields where smaller is better, when the detected representative value is not higher than the corresponding target upper limit, the field deviation value of the field is recorded as zero; when the detected representative value is higher than the corresponding target upper limit, the difference between the detected representative value and the target upper limit is used as the original deviation difference value of the field, and the allowable deviation boundary corresponding to the field is used as the normalization benchmark. The ratio between the original deviation difference value and the allowable deviation boundary is used as the field deviation value of the field; if the resulting field deviation value is greater than one, the field deviation value of the field is recorded as one. For fields where larger is better, when the detected representative value is not lower than the corresponding target lower limit, the field deviation value of the field is recorded as zero; when the detected representative value is lower than the corresponding target lower limit, the difference between the target lower limit and the detected representative value is used as the original deviation difference value of the field, and the allowable deviation boundary corresponding to the field is used as the normalization benchmark. The ratio between the original deviation difference value and the allowable deviation boundary is used as the field deviation value of the field; if the resulting field deviation value is greater than one, the field deviation value of the field is recorded as one.

[0074] For interval-appropriate fields, when the detected representative value is within the corresponding target range, the field deviation value of the field is recorded as zero; when the detected representative value is higher than the upper limit of the target range, the difference between the detected representative value and the upper limit of the target range is taken as the original deviation difference value of the field, and the allowable deviation boundary corresponding to the field is used as the normalization benchmark, and the ratio between the original deviation difference value and the allowable deviation boundary is taken as the field deviation value of the field; when the detected representative value is lower than the lower limit of the target range, the difference between the lower limit of the target range and the detected representative value is taken as the original deviation difference value of the field, and the allowable deviation boundary corresponding to the field is used as the normalization benchmark, and the ratio between the original deviation difference value and the allowable deviation boundary is taken as the field deviation value of the field; if the obtained field deviation value is greater than one, the field deviation value of the field is recorded as one.

[0075] The allowable deviation boundary and the corresponding representative detection value have the same unit of measurement, and the allowable deviation boundary is a value greater than zero. By comparing the original deviation difference with the allowable deviation boundary, the representative detection values ​​of different dimensions are converted into dimensionless field deviation values, so that the cooking quality data, texture data, rice soup state data, and flavor data can participate in the subsequent generation of food quality risk profiles according to the same risk measurement caliber.

[0076] The dimensional field attribution rules refer to the rules for assigning the deviation values ​​of each field in the taste deviation vector to different taste dimensions. Specifically, the deviation values ​​of the cooking time, rehydration rate, and cooking loss fields are assigned to the light cooking and rehydration dimension; the deviation values ​​of the rice grain hardness, rice grain elasticity, and rice grain cohesion fields are assigned to the rice grain texture dimension; the deviation values ​​of the rice soup apparent viscosity, rice soup flow time, rice soup turbidity, and standing layering ratio fields are assigned to the rice soup harmony dimension; and the comprehensive deviation values ​​of aroma response, taste response, and sensory evaluation are assigned to the flavor acceptance dimension.

[0077] For aroma response data including multiple electronic nose sensor channels, the taste quality risk assessment module first reads the deviation values ​​of each aroma response channel field in the order of the electronic nose sensor channels, and then performs a weighted sum of the deviation values ​​of each aroma response channel field according to the aroma channel weight to generate a comprehensive aroma response deviation value. When no aroma channel weight is configured in the benchmark parameter set, the deviation values ​​of each aroma response channel field are summed after being weighted equally according to the number of channels. For taste response data including multiple electronic tongue sensor channels, the taste quality risk assessment module first reads the deviation values ​​of each taste response channel field in the order of the electronic tongue sensor channels, and then performs a weighted sum of the deviation values ​​of each taste response channel field according to the taste channel weight to generate a comprehensive taste response deviation value. When no taste channel weight is configured in the benchmark parameter set, the deviation values ​​of each taste response channel field are summed after being weighted equally according to the number of channels. When sensory evaluation data includes multiple scoring items, the taste quality risk assessment module first reads the deviation values ​​of the rice grain texture scoring field, the rice soup state scoring field, the aroma scoring field, the flavor scoring field, and the overall acceptability scoring field. Then, it performs a weighted sum according to the weight of each sensory scoring item to generate a comprehensive sensory evaluation deviation value. When no sensory scoring item weights are configured in the benchmark parameter set, the deviation values ​​of each sensory scoring item field are summed after being weighted equally according to the number of scoring items.

[0078] The dimensional weighting rules refer to the weighting allocation rules for the weighted sum of deviation values ​​or comprehensive deviation values ​​of different fields within the same taste dimension. Dimensional weighting rules include weighting rules for the light cooking and rehydration dimension, the rice grain texture dimension, the rice soup harmony dimension, and the flavor acceptance dimension. The light cooking and rehydration dimension weighting rules include weights for cooking time, rehydration rate, and cooking loss, with the sum of these three being equal to one. The rice grain texture dimension weighting rules include weights for rice grain hardness, rice grain elasticity, and rice grain cohesion, with the sum of these three being equal to one. The rice soup harmony dimension weighting rules include weights for apparent viscosity, flow time, turbidity, and stratification ratio, with the sum of these four being equal to one. The flavor acceptance dimension weighting rules include weights for aroma response, taste response, and sensory evaluation, with the sum of these three being equal to one.

[0079] In one optional implementation, in the weighting rules for the light cooking and rehydration dimension, the weight of cooking time is 0.4, the weight of rehydration rate is 0.3, and the weight of cooking loss is 0.3; in the weighting rules for the rice grain texture dimension, the weight of rice grain hardness is 0.4, the weight of rice grain elasticity is 0.3, and the weight of rice grain cohesion is 0.3; in the weighting rules for the rice soup harmony dimension, the weight of apparent viscosity of rice soup is 0.3, the weight of rice soup flow time is 0.2, the weight of rice soup turbidity value is 0.2, and the weight of standing stratification ratio is 0.3; in the weighting rules for the flavor acceptance dimension, the weight of aroma response is 0.25, the weight of taste response is 0.25, and the weight of sensory evaluation is 0.5. The aforementioned preset weights are derived from product design requirements and manual review experience. Among them, in the light cooking and rehydration dimension, the cooking time directly corresponds to the light cooking target, so the weight of cooking time is higher than the weight of rehydration rate and cooking loss. In the rice grain texture dimension, rice grain hardness directly affects the chewing texture, so the weight of rice grain hardness is higher than the weight of rice grain elasticity and rice grain cohesion. In the rice soup harmony dimension, the apparent viscosity and the proportion of stratification after standing reflect the consistency and stability of rice soup, respectively, so their weights are higher than the weights of rice soup flow time and rice soup turbidity. In the flavor acceptance dimension, sensory evaluation data directly reflects the overall acceptance status, so the weight of sensory evaluation is higher than the weights of aroma response and taste response.

[0080] Based on the taste deviation vector, aroma response comprehensive deviation value, taste response comprehensive deviation value, sensory evaluation comprehensive deviation value, and dimensional weighting rules, the following deviation values ​​are generated: light cooking and rehydration deviation value, rice grain texture deviation value, rice soup harmony deviation value, and flavor acceptance deviation value.

[0081] Specifically, the taste quality risk assessment module reads the deviation values ​​of the cooking time field, the rehydration rate field, and the cooking loss field. It multiplies the cooking time field deviation value by the cooking time weight to obtain the cooking time weighted deviation value; it multiplies the rehydration rate field deviation value by the rehydration rate weight to obtain the rehydration rate weighted deviation value; it multiplies the cooking loss field deviation value by the cooking loss weight to obtain the cooking loss weighted deviation value; and then it adds the cooking time weighted deviation value, the rehydration rate weighted deviation value, and the cooking loss weighted deviation value to generate the light cooking rehydration deviation amount.

[0082] Read the deviation values ​​of the rice grain hardness field, rice grain elasticity field, and rice grain cohesion field. Multiply the deviation value of the rice grain hardness field by the rice grain hardness weight to obtain the rice grain hardness weighted deviation value; multiply the deviation value of the rice grain elasticity field by the rice grain elasticity weight to obtain the rice grain elasticity weighted deviation value; multiply the deviation value of the rice grain cohesion field by the rice grain cohesion weight to obtain the rice grain cohesion weighted deviation value; then add the rice grain hardness weighted deviation value, rice grain elasticity weighted deviation value, and rice grain cohesion weighted deviation value together to generate the rice grain texture deviation amount.

[0083] Read the deviation values ​​of the apparent viscosity, flow time, turbidity, and sedimentation ratio fields of rice soup. Multiply the apparent viscosity deviation value by the apparent viscosity weight to obtain the apparent viscosity weighted deviation value; multiply the flow time deviation value by the flow time weight to obtain the flow time weighted deviation value; multiply the turbidity deviation value by the turbidity weight to obtain the turbidity weighted deviation value; multiply the sedimentation ratio deviation value by the sedimentation ratio weight to obtain the sedimentation weighted deviation value; and finally, add the apparent viscosity weighted deviation value, flow time weighted deviation value, turbidity weighted deviation value, and sedimentation weighted deviation value together to generate the rice soup coordination deviation.

[0084] Read the comprehensive deviation values ​​of aroma response, taste response, and sensory evaluation. Multiply the comprehensive deviation value of aroma response by the aroma response weight to obtain the weighted deviation value of aroma response; multiply the comprehensive deviation value of taste response by the taste response weight to obtain the weighted deviation value of taste response; multiply the comprehensive deviation value of sensory evaluation by the sensory evaluation weight to obtain the weighted deviation value of sensory evaluation; and finally add the weighted deviation values ​​of aroma response, taste response, and sensory evaluation to generate the flavor acceptance deviation.

[0085] A risk profile of food quality is generated based on deviations in light cooking and rehydration, rice grain texture, rice soup harmony, and flavor acceptance. The light cooking and rehydration deviation characterizes the degree of deviation in short-term cooking, rehydration, and cooking retention of the current batch; the rice grain texture deviation characterizes the degree of deviation in hardness, elasticity, and cohesion of rice grains after cooking; the rice soup harmony deviation characterizes the degree of deviation in consistency, fluidity, turbidity, and stability of rice soup after standing; and the flavor acceptance deviation characterizes the degree of deviation in aroma, taste, and sensory acceptance of the current batch.

[0086] The dominant taste risk type is determined according to the dominant risk identification rules. If the deviations in light cooking rehydration, rice grain texture, rice soup harmony, and flavor acceptance are all zero, the dominant taste risk type is determined to be the "no taste dominant risk" type. If the deviation in light cooking rehydration is the maximum value among the four types of deviations, the dominant taste risk type is determined to be the "light cooking rehydration deviation" type; if the deviation in rice grain texture is the maximum value among the four types of deviations, the dominant taste risk type is determined to be the "rice grain texture deviation" type; if the deviation in rice soup harmony is the maximum value among the four types of deviations, the dominant taste risk type is determined to be the "rice soup harmony deviation" type; if the deviation in flavor acceptance is the maximum value among the four types of deviations, the dominant taste risk type is determined to be the "flavor acceptance deviation" type. If two or more deviations simultaneously reach their maximum values, the dominant taste risk type is determined according to the priority stored in the dominant risk identification rules. The priority is set in the order of light cooking rehydration deviation, rice grain texture deviation, rice soup harmony deviation, and flavor acceptance deviation, or it can be preset in the benchmark parameter set according to product design requirements.

[0087] The risk level of the food taste is determined according to the risk grading rules. Specifically, the food taste quality risk assessment module reads the maximum deviation from the following: light cooking and rehydration deviation, rice grain texture deviation, rice soup harmony deviation, and flavor acceptance deviation. This maximum deviation is then used as the overall food taste deviation for the current batch. The overall food taste deviation is used to characterize the most prominent food taste weakness of the current batch.

[0088] When the overall deviation in taste is not higher than the taste qualification threshold, the taste risk level is determined to be "not exceeding the limit"; when the overall deviation in taste is higher than the taste qualification threshold but not higher than the taste control threshold, the taste risk level is determined to be "slightly exceeding the limit"; when the overall deviation in taste is higher than the taste control threshold, the taste risk level is determined to be "severely exceeding the limit". The taste qualification threshold refers to the deviation boundary used to determine whether the taste quality of the current batch is still within the acceptable target state. When the overall deviation in taste is not higher than the taste qualification threshold, it indicates that although there may be fluctuations in individual tests in the current batch, the overall taste deviation is still within the acceptable range. The taste control threshold refers to the deviation boundary used to determine whether the taste deviation of the current batch has reached the deviation boundary that requires participation in production parameter control. When the overall deviation in taste is higher than the taste control threshold, it indicates that the current batch has a significant taste quality risk and needs to participate in the control as a taste constraint in the subsequent production parameter control process.

[0089] The taste qualification threshold and taste control threshold are set based on qualified historical batch data, product design requirements, and manual review results, and stored in a benchmark parameter set. Specifically, the system reads the comprehensive taste deviation of multiple qualified historical batches, determines the upper limit of the distribution of the comprehensive taste deviation of qualified historical batches, and sets the taste qualification threshold in conjunction with the allowable taste fluctuation range in the product design requirements; it reads historical batches that are determined to require adjustment of formula parameters, extrusion parameters, or cooking quality control parameters in the manual review results, uses the comprehensive taste deviation corresponding to the above historical batches as a control reference sample, and sets the taste control threshold based on the deviation distribution of the control reference sample. The taste control threshold is greater than the taste qualification threshold, used to distinguish between slight taste fluctuations and taste risks that require control and correction.

[0090] After completing the above processing, a taste quality risk set is generated. This set includes a taste quality risk profile, dominant taste risk types, and taste risk levels. The taste quality risk profile includes deviations in light cooking and rehydration, rice grain texture deviations, rice soup harmony deviations, and flavor acceptance deviations. The taste quality risk set is written into the batch data package to be regulated and serves as the taste-side input for subsequent collaborative diagnostic type generation, granulation and taste conflict verification, and production parameter control.

[0091] The collaborative diagnosis module determines the dominant deviation on the granulation side, the dominant risk on the taste side, the collaborative diagnosis type, and the main control target based on the set of granulation defect deviations and the set of taste quality risks. It generates a set of main control candidate correction directions and conflict identifiers based on the set of parameter response coefficients, thus forming a set of collaborative diagnosis results.

[0092] It should be noted that the flowchart for forming the collaborative diagnostic result set is as follows: Figure 3 As shown.

[0093] The collaborative diagnosis module generates a collaborative diagnosis result set for the current batch of ultra-small recombinant porridge rice based on the granulation defect deviation set, the taste quality risk set, the production parameter set, and the benchmark parameter set. This collaborative diagnosis result set refers to the data set characterizing the granulation-side deviation, taste-side risk, collaborative diagnosis type, and control direction conflict status of the current batch. It includes the dominant granulation-side deviation, dominant taste-side risk, collaborative diagnosis type, main control target, main control deviation items, secondary control deviation items, main control candidate correction direction set, and conflict identifier. The collaborative diagnosis result set serves as the diagnostic input for the subsequent conflict constraint correction module to generate production parameter correction information.

[0094] The collaborative diagnosis module reads the granulation defect deviation set, the taste quality risk set, and the production parameter set, and also reads the collaborative diagnosis benchmark parameters from the benchmark parameter set. The collaborative diagnosis benchmark parameters refer to the set of parameters used to generate diagnosis types, main control targets, and conflict indicators based on granulation-side deviations and taste-side risks. These parameters include granulation-side dominant deviation judgment rules, taste-side dominant risk judgment rules, collaborative diagnosis type judgment rules, parameter response coefficient set, basic evaluation step size, production parameter control boundaries, secondary control constraint thresholds, confidence lower limit, and response judgment thresholds.

[0095] The parameter response coefficient set refers to a data set representing the unit response of different production parameters to granulation-side deviations and taste-side risks under different correction directions. Each response record in the parameter response coefficient set includes the production parameter name, correction direction, deviation item, unit correction magnitude, unit response coefficient, predicted response amount, and response confidence. The production parameter name includes the ratio of broken rice to quinoa, barrel temperature, material moisture content, screw speed, feeding speed, emulsifier type, emulsifier content, die size, and cutting speed. The correction direction refers to the adjustment direction of the corresponding production parameter relative to the current batch production parameter set; for barrel temperature, material moisture content, screw speed, feeding speed, emulsifier content, and cutting speed, the correction direction includes increasing, decreasing, and maintaining; for the ratio of broken rice to quinoa, the correction direction includes increasing the quinoa proportion, decreasing the quinoa proportion, and maintaining; for the emulsifier type and die size, the correction direction includes replacing and maintaining. The deviation items include particle size deviation, cutting tail deviation, particle integrity deviation, adhesion ratio deviation, light cooking and rehydration deviation, rice grain texture deviation, rice soup harmony deviation, and flavor acceptance deviation.

[0096] The unit response coefficient refers to the change in deviation of the corresponding deviation item after adjusting the corresponding production parameter by one unit correction magnitude according to the corresponding correction direction. A negative unit response coefficient indicates that the correction direction can reduce the corresponding deviation item; a positive unit response coefficient indicates that the correction direction will increase the corresponding deviation item; a zero unit response coefficient indicates that the change in the corresponding deviation item does not exceed the response judgment threshold. The predicted response amount refers to the predicted change in the deviation item after the object replacement production parameter is executed according to the corresponding replacement direction, applicable to emulsifier type and die specification. The response reliability refers to the degree of reliability of the corresponding response record for the current batch conflict prediction; response records with a reliability not lower than the lower reliability limit participate in subsequent candidate direction screening and conflict judgment.

[0097] The parameter response coefficient set is generated based on historical batch data and process test data. Specifically, the system reads historical batch samples with production parameter adjustment records, and obtains the production parameter set, granulation defect deviation set, and taste quality risk set for each historical batch before and after parameter adjustment; then, it determines the parameter difference before and after the production parameter adjustment, and determines the deviation difference of granulation-side deviation and taste-side risk before and after adjustment. For numerical production parameters, the system uses the ratio between the deviation difference of a certain action deviation item and the corresponding production parameter difference as the single-sample response coefficient; then, it performs statistical processing on multiple single-sample response coefficients under the same production parameter, the same correction direction, and the same action deviation item to obtain the unit response coefficient. The statistical processing includes removing abnormal production samples and abnormal detection samples, and using the median or the average value after removing extreme values ​​as the unit response coefficient. For object replacement-type production parameters such as emulsifier type and die specification, the system reads the deviation difference of multiple historical batches under the same replacement direction, and uses the median or the average value after removing extreme values ​​as the predicted response quantity corresponding to that replacement direction. The credibility of a response is determined based on the number of historical samples corresponding to the same response record, the fluctuation range of the single sample response coefficient, and the results of manual review. When the credibility of a response is lower than the lower limit of credibility, the corresponding response record will not participate in the conflict judgment.

[0098] The collaborative diagnostic module generates dominant granulation deviations based on the set of granulation defect deviations. These dominant granulation deviations refer to the main control deviations on the granulation side determined by their deviation level number and magnitude within the current batch's set of granulation defect deviations. The collaborative diagnostic module reads particle size deviation, cutting tailing deviation, particle integrity deviation, and adhesion ratio deviation, and reads the corresponding deviation level and magnitude for each. For particle size deviation, particle integrity deviation, and adhesion ratio deviation, the deviation levels are assigned levels 0, 1, and 2, respectively, for no exceedance, slight exceedance, and severe exceedance. For cutting tailing deviation, no exceedance corresponds to level 0, exceedance in occurrence rate and severity corresponds to level 1, and combined exceedance corresponds to level 2. The magnitude of the cutting tailing deviation is determined by the larger of the tailing occurrence rate exceedance deviation and the tailing severity exceedance deviation.

[0099] When the grade numbers of particle size deviation, cutting tailing deviation, particle integrity deviation, and adhesion ratio deviation are all zero, the dominant deviation on the granulation side is determined to be no dominant granulation deviation. When there is a granulation deviation with a grade number greater than zero, the collaborative diagnosis module selects the granulation deviation with the highest grade number as the dominant deviation on the granulation side; when two or more granulation deviations have the same highest grade number, the granulation deviation with the largest deviation amount is selected as the dominant deviation on the granulation side; when two or more granulation deviations have the same grade number and deviation amount, the dominant deviation on the granulation side is determined in the order of cutting tailing deviation, particle integrity deviation, adhesion ratio deviation, and particle size deviation.

[0100] The collaborative diagnosis module generates dominant risk on the taste side based on the taste quality risk set. The dominant risk on the taste side refers to the main control risk item on the taste side determined according to the dominant taste risk type and taste risk level in the current batch's taste quality risk set. The collaborative diagnosis module reads the taste quality risk profile, dominant taste risk type, and taste risk level; where the taste risk level corresponds to level numbers zero, one, and two for no exceedance, slight exceedance, and severe exceedance, respectively. When the dominant taste risk type is "no dominant risk" and the taste risk level is "no exceedance," the dominant risk on the taste side is determined to be "no dominant risk." When the dominant taste risk type is not "no dominant risk," the dominant taste risk type is determined as the dominant risk on the taste side, and the level number corresponding to the taste risk level is used as the taste side deviation level number.

[0101] The collaborative diagnosis module generates collaborative diagnosis types based on the dominant deviation on the granulation side and the dominant risk on the taste side. These collaborative diagnosis types refer to the diagnostic results obtained by classifying the deviation status of the current batch according to the dominant deviation on the granulation side and the dominant risk on the taste side, including two-sided non-exceeding type, granulation-side dominant type, taste-side dominant type, and granulation-taste two-sided deviation type.

[0102] When the dominant deviation on the granulation side is "no granulation dominant deviation" and the dominant risk on the taste side is "no taste dominant risk," the collaborative diagnosis type is determined to be "two-sided non-exceeding limit type." This "two-sided non-exceeding limit type" means that the current batch has not formed any dominant deviations requiring entry into the production parameter control process on either the granulation or taste side, and the subsequent conflict constraint correction module does not generate mandatory correction information. When the dominant deviation on the granulation side is not "no granulation dominant deviation" and the dominant risk on the taste side is "no taste dominant risk," the collaborative diagnosis type is determined to be "granulation side dominant type." When the dominant deviation on the granulation side is not "no granulation dominant deviation" and the dominant risk on the taste side is not "no taste dominant risk," the collaborative diagnosis type is determined to be "granulation and taste two-sided deviation type."

[0103] The primary control objective and primary control deviation items are determined based on the collaborative diagnosis type. The primary control objective refers to the control object whose deviation is prioritized for reduction by the subsequent conflict constraint correction module, including maintaining current parameters, granulation-side primary control, and taste-side primary control. Primary control deviation items refer to deviation items corresponding to the primary control objective and requiring priority reduction. When the collaborative diagnosis type is bilateral and not exceeding limits, the primary control objective is set to maintain current parameters, and no primary control deviation items are set. When the collaborative diagnosis type is granulation-side dominant, the primary control objective is set to granulation-side primary control, and the primary control deviation item is set to granulation-side dominant deviation. When the collaborative diagnosis type is taste-side dominant, the primary control objective is set to taste-side primary control, and the primary control deviation item is set to taste-side dominant risk. When the collaborative diagnosis type is granulation-based bilateral deviation in taste, the grade number of the dominant deviation on the granulation side is compared with the grade number corresponding to the taste risk level. If the grade number of the dominant deviation on the granulation side is higher than the grade number corresponding to the taste risk level, the primary control objective is determined to be the granulation-based primary control, the primary control deviation item is determined to be the granulation-based dominant deviation, and the secondary control deviation item is determined to be the taste-based dominant risk. If the grade number corresponding to the taste risk level is higher than the grade number of the dominant deviation on the granulation side, the primary control objective is determined to be the taste-based primary control, the primary control deviation item is determined to be the taste-based dominant risk, and the secondary control deviation item is determined to be the granulation-based dominant deviation. If the grade numbers are the same, the primary control objective is determined to be the granulation-based primary control, the primary control deviation item is determined to be the granulation-based dominant deviation, and the secondary control deviation item is determined to be the taste-based dominant risk. For bilateral non-exceeding-limit, granulation-based dominant, and taste-based dominant types, no secondary control deviation items are set.

[0104] When generating the set of candidate correction directions for the main control, the collaborative diagnostic module determines the proposed evaluation correction magnitude for each candidate correction direction. The proposed evaluation correction magnitude refers to the adjustment range of the candidate parameters used for conflict prediction. For numerical production parameters, the proposed evaluation correction magnitude is determined based on the basic evaluation step size and the production parameter control boundary. The basic evaluation step size is pre-set by the benchmark parameter set according to the level sequence of the main control deviation item, and does not exceed the single allowable correction limit of the corresponding production parameter. When the correction direction is increasing, the proposed evaluation correction magnitude is the smaller of the remaining increase range between the current production parameter value and the parameter upper limit and the basic evaluation step size; when the correction direction is decreasing, the proposed evaluation correction magnitude is the smaller of the remaining decrease range between the current production parameter value and the parameter lower limit and the basic evaluation step size. For the ratio of broken rice to quinoa, the proposed evaluation correction magnitude is the smaller of the remaining adjustable range between the current quinoa percentage and the corresponding ratio boundary and the basic ratio evaluation step size. For object-substitutable production parameters such as emulsifier type and die specifications, the proposed evaluation correction magnitude does not use a numerical step size, but directly calls the predicted response amount of the corresponding substitution direction.

[0105] The predicted change in the master control parameter refers to the expected change in the master control deviation item after the candidate correction direction is executed according to the proposed evaluation correction magnitude. The predicted change in the master control parameter is generated by the collaborative diagnostic module based on the parameter response coefficient set, candidate correction directions, and the proposed evaluation correction magnitude. For numerical production parameters, the predicted change in the master control parameter is determined based on the corresponding unit response coefficient and the proposed evaluation correction magnitude. For object-substitutable production parameters such as emulsifier type and die specification, the predicted change in the master control parameter is determined based on the predicted response of the corresponding substitute object. A negative predicted change in the master control parameter indicates that the candidate correction direction is expected to reduce the master control deviation item; a positive predicted change in the master control parameter indicates that the candidate correction direction is expected to increase the master control deviation item; a zero predicted change in the master control parameter indicates that the impact of the candidate correction direction on the master control deviation item does not exceed the response judgment threshold. The predicted change in the master control parameter is used to screen candidate correction directions that can reduce the master control deviation item and to determine the production parameters to be controlled.

[0106] The predicted change in sub-control refers to the expected change in the sub-control deviation item after the candidate correction direction is executed according to the proposed evaluation correction magnitude. The predicted change in sub-control is generated by the collaborative diagnosis module when the collaborative diagnosis type is granulation-based taste two-sided deviation type. It is used to determine whether reducing the main control deviation item will cause the sub-control deviation item to continue to increase. A negative predicted change in sub-control indicates that the candidate correction direction is expected to reduce the sub-control deviation item; a positive predicted change indicates that the candidate correction direction is expected to increase the sub-control deviation item; a zero predicted change indicates that the impact of the candidate correction direction on the sub-control deviation item does not exceed the response judgment threshold. The predicted change in sub-control is used to generate the candidate direction state and participates in the generation of constraint control step size in conflict constraint control.

[0107] The remaining allowable deviation of the sub-control refers to the amount of deviation that a sub-control deviation item is still allowed to increase in, provided it does not exceed the sub-control constraint threshold. The remaining allowable deviation is determined based on the difference between the sub-control constraint threshold and the current deviation of the sub-control deviation item. The current deviation of the sub-control deviation item is read from the batch data packet to be regulated; when the sub-control deviation item is a dominant risk on the taste side, the current deviation is the corresponding taste dimension deviation; when the sub-control deviation item is a dominant deviation on the granulation side, the current deviation is the corresponding granulation deviation. When the remaining allowable deviation is greater than zero, it indicates that the sub-control deviation item still has room for allowable increase; when the remaining allowable deviation is less than or equal to zero, it indicates that the sub-control deviation item has reached or exceeded the sub-control constraint threshold. The remaining allowable deviation is used in conflict constraint control to limit the control step size of candidate correction directions, ensuring that the increase in sub-control deviation corresponding to the control step size does not exceed the remaining allowable deviation.

[0108] A set of candidate correction directions for main control is generated based on the main control objective, main control deviation items, and parameter response coefficient set. The set of candidate correction directions refers to the data set predicting production parameters that can reduce the main control deviation items and their correction directions. Each candidate correction direction in the set includes a candidate production parameter name, candidate correction direction, proposed correction magnitude, main control predicted change, secondary control predicted change, and candidate direction status. When the main control objective is to maintain the current parameters, the set of candidate correction directions is empty. When the main control objective is granulation-side main control or taste-side main control, response records in the parameter response coefficient set that have deviation items consistent with the main control deviation items and whose response confidence is not lower than the lower confidence limit are read. Production parameters with negative predicted changes for the main control deviation items and their correction directions are then selected to generate the set of candidate correction directions for main control.

[0109] The collaborative diagnostic module generates predicted changes based on the unit response coefficient and the proposed evaluation correction magnitude. For numerical production parameters, the module reads the unit response coefficient corresponding to the candidate correction direction and multiplies it by the proposed evaluation correction magnitude to obtain the predicted change for the corresponding deviation item under that candidate correction direction. A negative predicted change indicates that the deviation item is expected to decrease after implementing the candidate correction direction; a positive predicted change indicates that the deviation item is expected to increase; and a zero predicted change indicates that the deviation item is not expected to change beyond the response judgment threshold. For object replacement-type production parameters, the collaborative diagnostic module uses the predicted response magnitude of the corresponding replacement direction as the predicted change.

[0110] The collaborative diagnostic module generates conflict identifiers based on the predicted changes. These conflict identifiers indicate whether there exists a candidate correction direction in the set of primary control candidate correction directions that can reduce the primary control deviation while keeping the secondary control deviation within the constraint boundary. This includes both conflict-free and conflict-prone options. When no secondary control deviation is set, the conflict identifier is determined to be conflict-free. When a secondary control deviation is set, the collaborative diagnostic module reads each candidate correction direction in the set of primary control candidate correction directions, generating the primary control predicted change for the primary control deviation and the secondary control predicted change for the secondary control deviation. If the primary control predicted change is not less than zero, it indicates that the candidate correction direction cannot reduce the primary control deviation, and the collaborative diagnostic module does not consider this candidate correction direction as a valid primary control candidate direction. If the change in the main control prediction is less than zero, the change in the sub-control prediction is further determined. When the change in the sub-control prediction is not greater than zero, the candidate correction direction is marked as a conflict-free candidate direction. When the change in the sub-control prediction is greater than zero, the collaborative diagnosis module adds the current deviation of the sub-control deviation item to the sub-control prediction change to obtain the predicted sub-control deviation. If the predicted sub-control deviation is not higher than the sub-control constraint threshold, the candidate correction direction is marked as a conflict-free candidate direction. If the predicted sub-control deviation is higher than the sub-control constraint threshold, the candidate correction direction is marked as a conflict candidate direction.

[0111] The sub-control constraint threshold refers to the upper limit of deviation that a sub-control deviation item is allowed to reach during the correction process of the main control target, and it is stored in the benchmark parameter set. For sub-control deviation items on the taste side, the sub-control constraint threshold adopts the taste control threshold; for sub-control deviation items on the granulation side, the sub-control constraint threshold adopts the corresponding granulation control threshold. The current deviation amount of the sub-control deviation item refers to the deviation amount or amount of deviation that has already formed in the current batch; when the sub-control deviation item is a dominant risk on the taste side, the current deviation amount of the sub-control deviation item is the deviation amount of the corresponding taste dimension; when the sub-control deviation item is a dominant deviation on the granulation side, the current deviation amount of the sub-control deviation item is the deviation amount of the corresponding granulation deviation.

[0112] After the collaborative diagnosis module completes the judgment of all candidate correction directions, it generates a conflict identifier based on the existence status of conflict-free and conflicting candidate directions. If there is a conflict-free candidate direction in the master control candidate correction direction set, the conflict identifier is determined to be conflict-free, and the conflict-free candidate direction is output to the subsequent conflict constraint correction module first. If all candidate correction directions in the master control candidate correction direction set are conflicting candidate directions, the conflict identifier is determined to be conflicting. A conflict identifier of conflicting indicates that there is currently no candidate correction direction that can reduce the master control deviation item while keeping the sub-control deviation item within the sub-control constraint threshold. The subsequent conflict constraint correction module needs to perform correction step size compression on the candidate correction directions or generate the next batch of adjustment information.

[0113] After completing the above processing, the collaborative diagnosis module generates a collaborative diagnosis result set. This set includes the dominant deviation on the granulation side, the dominant risk on the taste side, the collaborative diagnosis type, the main control objective, the main control deviation items, the secondary control deviation items, the set of candidate correction directions for the main control, and conflict identifiers. The collaborative diagnosis result set is written into the batch data package to be regulated and output to the subsequent conflict constraint correction module.

[0114] The conflict constraint correction module determines the control type based on the collaborative diagnosis result set, the production parameter set, and the baseline parameter set. It then determines the production parameters to be controlled and the control step size from the master control candidate correction direction set according to the control type, generates production parameter control information, and executes production parameter control.

[0115] It should be noted that the flowchart for generating production parameter control information is as follows: Figure 4 As shown.

[0116] The conflict constraint correction module generates production parameter control information for the current batch of ultra-small reconstituted porridge rice based on the collaborative diagnostic result set, production parameter set, and baseline parameter set. This production parameter control information refers to a data set used to instruct the production control system to perform maintenance control, forward correction control, or conflict constraint control on the production parameters of subsequent production stages or the next production batch. It includes control type, primary control target, primary control deviation item, secondary control deviation item, production parameter to be controlled, current parameter value, current user object, control direction, control step size, control target value, replacement object, and conflict constraint identifier. The current parameter value, control step size, and control target value are used for numerical production parameters, while the current user object and replacement object are used for emulsifier type and die specification. The production parameter control information serves as the control input for the production control system to perform quality regulation of ultra-small reconstituted porridge rice granulation.

[0117] The conflict constraint correction module first reads the collaborative diagnosis type, main control target, main control deviation item, secondary control deviation item, main control candidate correction direction set and conflict identifier from the collaborative diagnosis result set, and reads the current production parameter value and current user object from the production parameter set; the current production parameter value corresponds to the barrel temperature, material moisture addition amount, screw speed, feeding speed, emulsifier addition amount, cutting speed and broken rice and quinoa compound ratio, and the current user object corresponds to the emulsifier type and die specification.

[0118] The conflict constraint correction module also reads conflict constraint correction benchmark parameters from the benchmark parameter set. These benchmark parameters are a set of parameters used to define the control range of production parameters and the scope of conflict constraint control, including production parameter control boundaries, secondary control constraint thresholds, control step size generation rules, and production parameter control priorities. Production parameter control boundaries are determined based on the equipment's rated operating range, the permissible range of the food formula, production process documents, and the parameter distribution of qualified historical batches. Production parameter control priorities are stored in the benchmark parameter set in the form of a priority sequence list.

[0119] The production parameter control boundary refers to the allowable range of values ​​or selectable objects for adjusting production parameters in the current batch or the next production batch. It is used to limit the adjustable range of barrel temperature, material moisture addition, screw speed, feeding speed, emulsifier addition, cutting speed, and the ratio of broken rice to quinoa, and to limit the selectable range of emulsifier types and die specifications. The secondary control constraint threshold refers to the upper limit of deviation that a secondary control deviation item can reach during the correction of the primary control target. When the secondary control deviation item belongs to the taste-side risk, the secondary control constraint threshold adopts the taste control threshold; when the secondary control deviation item belongs to the granulation-side deviation, the secondary control constraint threshold adopts the granulation control threshold corresponding to that secondary control deviation item. Specifically, particle size deviation corresponds to the particle size control threshold, cutting tailing deviation corresponds to the cutting tailing control threshold, particle integrity deviation corresponds to the particle integrity control threshold, and adhesion ratio deviation corresponds to the adhesion ratio control threshold. The control step size generation rule refers to the rule for generating the actual control step size based on the proposed evaluation correction magnitude of the candidate correction direction, the predicted change in secondary control, and the remaining allowable deviation of secondary control. The production parameter control priority refers to the sorting rule used to determine the production parameters to be controlled when the main control improvement degree of two or more candidate correction directions is the same and the influence state of the secondary control cannot be distinguished. It is preset based on the degree of influence of the production parameters on granulation stability, the convenience of the production control system, and historical control records.

[0120] The master control objective is used to define the effect target of this production parameter control, and is not the actual production parameter to be adjusted. The conflict constraint correction module determines the master control deviation items based on the master control objective, and uses the predicted decrease of the master control deviation items as the basis for screening the production parameters to be controlled. The production parameters to be controlled refer to the specific production parameters screened from the master control candidate correction direction set, including the ratio of broken rice to quinoa, barrel temperature, material moisture content, screw speed, feeding speed, emulsifier type, emulsifier content, die specifications, and cutting speed.

[0121] When the primary control objective is granulation-side control, the primary control deviation item is one of the following: particle size deviation, cutting tailing deviation, particle integrity deviation, and adhesion ratio deviation. The conflict constraint correction module selects candidate correction directions from the primary control candidate correction direction set that can make the primary control predicted change of the primary control deviation item negative, and uses the corresponding candidate production parameter name as a candidate source for determining the production parameter to be controlled. When the primary control objective is taste-side control, the primary control deviation item is one of the following: light cooking and rehydration deviation, rice grain texture deviation, rice soup harmony deviation, and flavor acceptance deviation. The conflict constraint correction module selects candidate correction directions from the primary control candidate correction direction set that can make the primary control predicted change of the primary control deviation item negative, and uses the corresponding candidate production parameter name as a candidate source for determining the production parameter to be controlled.

[0122] The control type is determined based on the collaborative diagnosis type and conflict identifier, and the candidate direction invocation range is determined based on the control type. If the collaborative diagnosis type is bilateral non-exceeding type, the control type is determined to be maintenance control, and the production parameter control information is recorded as maintaining the current parameters. If the collaborative diagnosis type is granulation-side dominant type or taste-side dominant type, the control type is determined to be forward correction control, and the conflict constraint correction module writes the candidate correction directions with negative predicted changes from the master control candidate correction direction set into the forward candidate direction set. If the collaborative diagnosis type is granulation-taste bilateral deviation type, and the conflict identifier is non-conflict, the control type is determined to be forward correction control, and the conflict constraint correction module writes the candidate correction directions with non-conflict status from the master control candidate correction direction set into the forward candidate direction set. If the collaborative diagnosis type is granulation-taste bilateral deviation type, and the conflict identifier is conflicted, the control type is determined to be conflict constraint control, and the conflict constraint control process is entered.

[0123] The forward candidate direction set refers to the set of candidate correction directions that can reduce the main control deviation and allow direct generation of production parameter control information without requiring step size compression. Each candidate correction direction in the forward candidate direction set includes a candidate production parameter name, candidate correction direction, proposed evaluation correction magnitude, main control predicted change, secondary control predicted change, and candidate direction status. When no secondary control deviation is set, the secondary control predicted change in the forward candidate direction set is recorded as zero, and the candidate direction status is recorded as a conflict-free candidate direction.

[0124] When the control type is forward correction control, the conflict constraint correction module generates production parameter control information based on the forward candidate direction set. If the forward candidate direction set is empty, the conflict constraint correction module does not generate the production parameters to be controlled, records the control type in the production parameter control information as forward correction control, writes the conflict constraint identifier as "no executable forward control direction," and records that the current production parameter set remains unchanged. If the forward candidate direction set is not empty, the conflict constraint correction module selects the candidate correction direction with the negative predicted change and the largest negative improvement magnitude as the direction to be controlled; the negative improvement magnitude refers to the magnitude of the absolute value of the predicted change. When the negative improvement magnitudes of two or more candidate correction directions are the same, if a secondary control deviation item has been set, the candidate correction direction with the lower predicted secondary control deviation is selected; the predicted secondary control deviation is the deviation obtained by adding the current deviation of the secondary control deviation item to the predicted change. If no secondary control deviation item has been set, the candidate correction direction with the smaller expected correction magnitude is selected; when there are still two or more candidate correction directions, the direction to be controlled is determined according to the production parameter control priority.

[0125] When the control direction corresponds to a numerical production parameter, the conflict constraint correction module determines the candidate production parameter name as the production parameter to be controlled, the candidate correction direction as the control direction, and the proposed correction magnitude as the control step size. It then generates a control target value based on the current parameter value, control direction, and control step size. If the control direction is increasing, the current parameter value is added to the control step size to obtain the control target value; if the control direction is decreasing, the current parameter value is subtracted from the control step size to obtain the control target value. When the production parameter to be controlled is the ratio of broken rice to quinoa, if the control direction is to increase the quinoa proportion, the current quinoa proportion is added to the control step size to obtain the control target value; if the control direction is to decrease the quinoa proportion, the current quinoa proportion is subtracted from the control step size to obtain the control target value. If the control target value exceeds the production parameter control boundary, the boundary value corresponding to the production parameter control boundary is used as the control target value. When the direction to be controlled corresponds to the type of emulsifier or the die specification, the conflict constraint correction module determines the candidate production parameter name as the production parameter to be controlled, determines the candidate correction direction as the control direction, and writes the replacement object corresponding to the candidate correction direction into the production parameter control information.

[0126] When the control type is conflict constraint control, it means that there are no conflict-free candidate directions in the main control candidate correction direction set. Candidate correction directions whose status is conflict candidate direction are read from the main control candidate correction direction set. The conflict candidate direction refers to a candidate correction direction that can reduce the main control deviation item, but, after execution according to the proposed evaluation correction magnitude, will cause the sub-control deviation item to exceed the sub-control constraint threshold. The conflict constraint correction module does not directly output conflict candidate directions according to the proposed evaluation correction magnitude, but instead performs step-size compression processing on the conflict candidate directions based on the remaining allowable deviation of the sub-control.

[0127] The step size compression process refers to compressing the original proposed evaluation correction magnitude into a constraint control step size that will not cause the sub-control deviation item to exceed the sub-control constraint threshold when the candidate correction direction will increase the sub-control deviation item. The conflict constraint correction module reads the current deviation amount of the sub-control deviation item and the sub-control constraint threshold, and determines the difference between the sub-control constraint threshold and the current deviation amount of the sub-control deviation item as the remaining allowable deviation amount of the sub-control. The current deviation amount of the sub-control deviation item is read from the batch data packet to be regulated; when the sub-control deviation item is a taste-side dominant risk, the current deviation amount is the corresponding taste dimension deviation amount; when the sub-control deviation item is a granulation-side dominant deviation, the current deviation amount is the corresponding granulation deviation amount.

[0128] When the remaining allowable deviation of the sub-control is less than or equal to zero, it indicates that the sub-control deviation item has reached or exceeded the sub-control constraint threshold. The conflict constraint correction module will not output candidate correction directions that would increase the sub-control deviation item, and will write the conflict constraint identifier in the production parameter control information as "sub-control constraint not satisfied". When the remaining allowable deviation of the sub-control is greater than zero, the conflict constraint correction module performs the following processing according to the control step size generation rules: read the predicted change of the sub-control for the conflict candidate direction and the proposed evaluation correction magnitude. The predicted change of the sub-control refers to the expected increase in deviation of the sub-control deviation item after executing the candidate correction direction according to the proposed evaluation correction magnitude; the predicted change of the sub-control corresponding to the conflict candidate direction is a positive value. The conflict constraint correction module determines the ratio between the remaining allowable deviation of the sub-control and the predicted change of the sub-control as the step size compression coefficient; when the step size compression coefficient is greater than one, the step size compression coefficient is recorded as one; when the step size compression coefficient is not greater than one, the step size compression coefficient is retained. Subsequently, the conflict constraint correction module multiplies the proposed evaluation correction magnitude by the step size compression coefficient to generate the constraint control step size.

[0129] The conflict constraint correction module generates the predicted change in the master control constraint based on the constraint control step size. Specifically, the conflict constraint correction module multiplies the predicted change in the master control constraint by the step size compression coefficient to obtain the predicted change in the master control constraint. The predicted change in the master control constraint is used to represent the expected improvement of the master control deviation item after the conflict candidate direction is compressed by the step size. When the predicted change in the master control constraint is still negative, it means that the compressed control direction can still reduce the master control deviation item, and the conflict constraint correction module determines the conflict candidate direction as a constrainable control direction; when the predicted change in the master control constraint is not less than zero, it means that the compressed control direction cannot reduce the master control deviation item, and the conflict candidate direction is not output.

[0130] If no conflict candidate direction exists, or if none of the conflict candidate directions are determined to be constrainable control directions, the conflict constraint correction module does not generate production parameters to be controlled. Instead, it records the control type in the production parameter control information as conflict constraint control, writes the conflict constraint identifier as a control direction without executable constraints, and records that the current production parameter set remains unchanged. If a constrainable control direction exists, the conflict constraint correction module determines that constrainable control direction as the direction to be controlled. If multiple constrainable control directions exist, the conflict constraint correction module selects the constrainable control direction with the largest negative improvement magnitude and the predicted change of the main control constraint as the direction to be controlled. When the negative improvement magnitudes of two or more constrainable control directions are the same, the constrainable control direction with the larger constraint control step size is selected.

[0131] When the control direction corresponds to a numerical production parameter, the conflict constraint correction module generates a control target value based on the current parameter value, control direction, and constraint control step size. If the control direction is increasing, the current parameter value is added to the constraint control step size to obtain the control target value; if the control direction is decreasing, the current parameter value is subtracted from the constraint control step size to obtain the control target value. When the production parameter to be controlled is the ratio of broken rice to quinoa, if the control direction is to increase the quinoa proportion, the current quinoa proportion is added to the constraint control step size to obtain the control target value; if the control direction is to decrease the quinoa proportion, the current quinoa proportion is subtracted from the constraint control step size to obtain the control target value. If the control target value exceeds the production parameter control boundary, the boundary value corresponding to the production parameter control boundary is used as the control target value. The conflict constraint correction module writes the conflict constraint identifier as an executed step size compression.

[0132] For object-substitutable production parameters such as emulsifier type and die specifications, the conflict constraint correction module does not perform step size compression. When the object-substitutable candidate correction direction is a non-conflict candidate direction, the corresponding replacement object is written into the production parameter control information by the forward correction control flow. When the object-substitutable candidate correction direction is a conflict candidate direction, the corresponding replacement object is not used as the direct execution object, and the object-substitutable candidate correction direction is not determined as the direction to be controlled; when there are no other constrainable control directions, the conflict constraint correction module writes the conflict constraint identifier in the production parameter control information as a restricted replacement object. Through the above processing, the conflict constraint correction module avoids directly adopting object replacement schemes that would increase the deviation of secondary control items when both the granulation side and the taste side have already deviated.

[0133] Production parameter control information is generated based on the control type. For maintenance control, the production parameter control information records the control type as maintenance control, the primary control objective as maintaining the current parameter, and the production parameter to be controlled as empty, and records that the current production parameter set remains unchanged. For forward correction control, the production parameter control information records the control type as forward correction control, the primary control objective, the primary control deviation item, the production parameter to be controlled, the current parameter value, the current user object, the control direction, the control step size, and the control objective value; if the production parameter to be controlled is an object replacement type production parameter, then the current user object and the replacement object are recorded. For conflict constraint control, the production parameter control information records the control type as conflict constraint control, the primary control objective, the primary control deviation item, the secondary control deviation item, the production parameter to be controlled, the current parameter value, the current user object, the control direction, the constraint control step size, the control objective value, and the conflict constraint identifier.

[0134] The conflict constraint correction module generates control execution information based on the type of the production parameter to be controlled. If the production parameter to be controlled is barrel temperature, material moisture content, screw speed, feeding speed, or cutting speed, the production control system writes the control target value into the set value of the corresponding equipment controller and uses it for extrusion granulation control in subsequent production stages of the current batch. If the production parameter to be controlled is the ratio of broken rice to quinoa, emulsifier type, emulsifier content, or die specifications, the production control system writes the control target value or replacement object into the formula parameters and process parameter settings of the next production batch. If the conflict constraint is identified as a sub-control constraint not being met, no executable constraint control direction, or a restricted replacement object, the production control system will not execute it, which will increase the deviation of the sub-control from the project's control direction.

[0135] After completing the above processing, the conflict constraint correction module outputs production parameter control information. This production parameter control information is written into the batch data packet to be adjusted and sent to the production control system. The production control system adjusts the continuous control parameters of the equipment in subsequent production stages of the current batch based on the production parameter control information, or uses the production parameter control information as the basis for setting the formula parameters, emulsifier parameters, die parameters, and equipment control parameters for the next production batch.

[0136] Example 2:

[0137] like Figure 2 As shown, this embodiment provides a method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality, including:

[0138] Establish a batch data package and a set of baseline parameters to be adjusted. The batch data package includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data.

[0139] Effective particle instances are extracted based on dry particle image data and granulation benchmark parameters. Based on the effective particle instances, particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation are generated to form a set of granulation defect deviations.

[0140] Based on the cooking quality data, texture data, rice soup state data, and flavor data, a taste detection feature vector is constructed. This vector is then compared with the taste target feature vector to generate a taste deviation vector. Finally, a taste quality risk set is generated based on the taste deviation vector.

[0141] Based on the set of granulation defect deviations and the set of taste quality risks, the dominant deviations on the granulation side, the dominant risks on the taste side, the collaborative diagnosis types and the main control targets are determined. Based on the set of parameter response coefficients, a set of main control candidate correction directions and conflict identifiers are generated to form a set of collaborative diagnosis results.

[0142] The control type is determined based on the set of collaborative diagnostic results, the set of production parameters, and the set of baseline parameters. The production parameters to be controlled and the control step size are determined from the set of main control candidate correction directions according to the control type. Production parameter control information is generated and production parameter control is executed.

[0143] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality, characterized in that, include: Establish a batch data package and a set of baseline parameters to be adjusted. The batch data package includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data. Effective particle instances are extracted based on dry particle image data and granulation benchmark parameters. Based on the effective particle instances, particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation are generated to form a set of granulation defect deviations. Based on the cooking quality data, texture data, rice soup state data, and flavor data, a taste detection feature vector is constructed. This vector is then compared with the taste target feature vector to generate a taste deviation vector. Finally, a taste quality risk set is generated based on the taste deviation vector. Based on the set of granulation defect deviations and the set of taste quality risks, the dominant deviations on the granulation side, the dominant risks on the taste side, the collaborative diagnosis types and the main control targets are determined. Based on the set of parameter response coefficients, a set of main control candidate correction directions and conflict identifiers are generated to form a set of collaborative diagnosis results. The control type is determined based on the set of collaborative diagnostic results, the set of production parameters, and the set of baseline parameters. The production parameters to be controlled and the control step size are determined from the set of main control candidate correction directions according to the control type. Production parameter control information is generated and production parameter control is executed.

2. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 1, characterized in that, The set of granulation defect deviations includes: Read the dry particle image data and granulation reference parameters, and verify the pixel size conversion coefficient of the dry particle image data; after the verification is passed, extract the particle foreground and analyze the connected region of the dry particle image data to obtain the candidate particle region. Candidate particle regions are screened based on the minimum and maximum projected area thresholds to determine valid particle instances. The particle length and width of the valid particle instances are generated based on the pixel size conversion factor and compared with the target range of granulation size to generate particle size deviation.

3. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 2, characterized in that, The generation of the granulation defect deviation set also includes: Tail-waving particles are identified based on the main body region and end extension region of the effective particle instance, and tail-waving deviation is generated based on the number of tail-waving particles, the length of the tail-like drag region, and the area of ​​the tail-like drag region; particle integrity deviation is generated based on the outer contour interruption length, end defect area, and main body fracture state of the effective particle instance. The adhesion ratio deviation is generated based on the projected area of ​​the suspected adhesion area, the number of shrinkage necks, and the number of adhesion particles; the particle size deviation, cutting tail deviation, particle integrity deviation, and adhesion ratio deviation are written into the granulation defect deviation set.

4. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 1, characterized in that, The generation of taste deviation vectors includes: According to the field order rules, the representative values ​​of the detection in the steaming and cooking quality data, texture data, rice soup state data and flavor data are arranged into a taste detection feature vector; Based on the taste target profile, construct a taste target feature vector with the same field order as the taste detection feature vector; read the detection representative value in the taste detection feature vector and the target boundary information in the taste target feature vector field by field, determine the original deviation difference of the detection representative value according to the field direction attribute and the target boundary information, and determine the field deviation value by the ratio between the original deviation difference and the corresponding allowable deviation boundary; arrange the field deviation values ​​according to the field order of the taste detection feature vector to generate the taste deviation vector.

5. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 1, characterized in that, The generated food quality risk set includes: According to the dimensional field attribution rules, the field deviation values ​​corresponding to the light cooking and rehydration dimension, rice grain texture dimension, and rice soup coordination dimension are read from the taste deviation vector, and the aroma response comprehensive deviation value, taste response comprehensive deviation value, and sensory evaluation comprehensive deviation value corresponding to the flavor acceptance dimension are read. According to the dimensional weighting rules, the deviation values ​​within each flavor dimension are weighted and summed to generate deviations in light cooking and rehydration, rice grain texture, rice soup harmony, and flavor acceptance, which are then written into the flavor quality risk profile. The dominant flavor risk type is determined based on the largest deviation among the four types, and the largest deviation is used as the comprehensive flavor deviation. The flavor risk level is determined based on the comparison between the comprehensive flavor deviation and the flavor pass threshold and flavor control threshold, generating a flavor quality risk set that includes the flavor quality risk profile, the dominant flavor risk type, and the flavor risk level.

6. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 1, characterized in that, The collaborative diagnostic results set includes: Read the particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation from the granulation defect deviation set, and determine the dominant deviation on the granulation side according to the level number and deviation amount of each granulation deviation; read the dominant taste risk type and taste risk level from the taste quality risk set, and determine the dominant risk on the taste side. The collaborative diagnosis type is determined based on the dominant deviation on the granulation side and the dominant risk on the taste side. The collaborative diagnosis type includes two-sided non-exceeding type, granulation side dominant type, taste side dominant type, and granulation and taste two-sided deviation type. The main control target is determined based on the collaborative diagnosis type, and the main control deviation item corresponding to the main control target is determined.

7. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 6, characterized in that, The formation of a collaborative diagnostic results set also includes: When the collaborative diagnosis type is granulation taste two-sided deviation type, determine the secondary control deviation items according to the main control target; read the parameter response coefficient set, and determine the proposed evaluation correction magnitude for candidate correction directions; Based on the unit response coefficient and the proposed evaluation correction magnitude, candidate correction directions are generated for the main control prediction change of the main control deviation items and the sub-control prediction change of the sub-control deviation items; Candidate correction directions for the master control prediction change are selected if the response confidence is not lower than the confidence lower limit, and a set of candidate correction directions for the master control is generated. The candidate direction status of each candidate correction direction is determined based on the change in the secondary control prediction and the secondary control constraint threshold, and a conflict identifier is generated based on the candidate direction status. The set of collaborative diagnostic results is constructed by considering the dominant deviations on the granulation side, the dominant risks on the taste side, the collaborative diagnostic types, the main control objectives, the main control deviation items, the secondary control deviation items, the set of main control candidate correction directions, and the conflict identifiers.

8. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 1, characterized in that, The generated production parameter control information includes: Read the collaborative diagnosis type, main control target, main control deviation item, main control candidate correction direction set and conflict identifier from the collaborative diagnosis result set, and read the current production parameter value and current user from the production parameter set; determine the control type based on the collaborative diagnosis type and conflict identifier; When the control type is maintenance control, the current set of production parameters will remain unchanged and will be written into the production parameter control information. When the control type is forward correction control, candidate correction directions with negative predicted changes are selected from the master control candidate correction direction set to form a forward candidate direction set; then, the candidate correction direction with the largest absolute value of the master control predicted change is selected from the forward candidate direction set as the direction to be controlled, and the production parameters to be controlled, control direction, control step size and control target value are determined according to the direction to be controlled, and the production parameter control information corresponding to the forward correction control is generated.

9. The method for controlling the granulation quality of ultra-small recombinant porridge rice by synergistic evaluation of granulation defects and eating quality according to claim 8, characterized in that, The generation of production parameter control information also includes: When the control type is conflict constraint control, the conflict candidate directions in the set of primary control candidate correction directions are read, and the remaining allowable deviation of the secondary control is generated according to the secondary control constraint threshold and the current deviation of the secondary control deviation item. When the remaining allowable deviation of the secondary control is greater than zero, the step size compression coefficient is generated according to the remaining allowable deviation of the secondary control and the predicted change of the secondary control, and the proposed evaluation correction magnitude is compressed according to the step size compression coefficient to generate the constraint control step size. The predicted change of the primary control constraint is generated according to the constraint control step size, and when the predicted change of the primary control constraint is negative, the constrainable control direction is determined. Based on the constrainable control direction, the production parameter control information corresponding to the conflict constraint control is generated.

10. A granulation quality control system for ultra-small recombinant porridge rice with synergistic evaluation of granulation defects and eating quality, used to implement the granulation quality control method for ultra-small recombinant porridge rice with synergistic evaluation of granulation defects and eating quality as described in any one of claims 1 to 9, characterized in that, include: The batch baseline filing module establishes a batch data package and a set of baseline parameters to be adjusted. The batch data package to be adjusted includes the current batch production parameter set, dry particle image data, cooking quality data, texture data, rice soup state data, and flavor data. The granulation defect deviation generation module extracts effective particle instances based on dry particle image data and granulation benchmark parameters, and generates particle size deviation, cutting tail deviation, particle integrity deviation and adhesion ratio deviation based on the effective particle instances, forming a set of granulation defect deviations. The food quality risk assessment module constructs a food quality detection feature vector based on cooking quality data, texture data, rice soup state data, and flavor data. It compares the feature vector with the target food quality vector to generate a food quality deviation vector. Based on the food quality deviation vector, it generates a food quality risk set including a food quality risk profile, dominant food quality risk type, and food quality risk level. The collaborative diagnosis module determines the dominant deviation on the granulation side, the dominant risk on the taste side, the collaborative diagnosis type, and the main control target based on the set of granulation defect deviations and the set of taste quality risks. It generates a set of main control candidate correction directions and conflict identifiers based on the set of parameter response coefficients, thus forming a set of collaborative diagnosis results. The conflict constraint correction module determines the control type based on the collaborative diagnosis result set, the production parameter set, and the baseline parameter set. It then determines the production parameters to be controlled and the control step size from the master control candidate correction direction set according to the control type, generates production parameter control information, and executes production parameter control.

Images