Mung bean flour processing and baking machine

Abstract

The invention discloses a mung bean powder processing and baking machine which comprises a central control unit, and a raw material pretreatment unit, a micro-crushing and airflow grading unit, a baking powder film forming unit, a drying unit and an integrated packaging unit which are respectively in communication connection with the central control unit. The whole process control from raw materials to finished products is realized; the central control unit is introduced, and the whole-process processing device covering raw material processing to finished product packaging is constructed, so that data intercommunication and collaborative operation among all process units are realized, the current situation of isolated operation of traditional equipment is thoroughly changed, and the overall automation degree and operation efficiency of the production system are improved.

Description

Technical Field

[0001] This invention relates to the field of mung bean powder processing technology, and in particular to a mung bean powder processing and baking machine. Background Technology

[0002] The processing of mung bean flour into rice noodles is a traditional food processing industry in my country. Its production process mainly includes raw material processing, grinding into a paste, baking into shape, and drying. Currently, most of this industry still uses semi-mechanized or partially automated production equipment, and the coordination and cooperation between each process mainly rely on manual experience for control.

[0003] Existing mung bean powder processing equipment typically consists of independent single-machine combinations, lacking unified system control. For example, the soaking process often relies on a fixed duration, making it difficult to adapt to the differences in initial moisture content of different batches of raw mung beans. The particle size control precision of the grinding process is insufficient and cannot be dynamically adjusted according to real-time operating conditions. The process parameters (such as temperature, flow rate, and humidity) of the baking and drying processes remain unchanged after being set, failing to respond to environmental fluctuations and changes in raw material characteristics. This results in poor stability of the produced baking powder products in terms of thickness, moisture content, taste, and integrity, high energy consumption, and high dependence on the technical experience of operators, making it difficult to achieve large-scale, standardized, and high-quality production.

[0004] Therefore, in response to the problems mentioned above, this invention proposes a mung bean flour processing and baking machine. Summary of the Invention

[0005] To overcome the problems of low automation, isolated operation of each process step, and rigid production parameters in existing equipment, which lead to unstable product quality and low efficiency, this invention proposes a mung bean powder processing and baking machine. By introducing sensing detection and feedback mechanisms in each key process step and dynamically optimizing the execution strategy based on real-time process data, the whole process control from raw materials to finished products can be achieved.

[0006] The technical solution of this invention is: a mung bean flour processing and baking machine, comprising: The central control unit is used to receive and process data from the raw material pretreatment unit, the micronization and air classification unit, the powder film forming unit, the drying unit, and the integrated packaging unit. Based on the built-in process parameter database and optimization algorithm, which refers to the database stored in the central control unit's memory, containing a standard parameter set and a success case library formed by historical production data accumulation, and a software program that is data-driven and can call pre-stored process mathematical models to process, analyze, and make decisions on real-time multi-source data, the central control unit issues control commands to the execution unit. The raw material pretreatment unit is connected in communication with the central control unit and is used to receive mung bean raw materials; The micro-pulverizing and air classifying unit is connected to the central control unit. It is used to receive pre-treated mung beans and adjust the pulverizing intensity in real time through feedback from the motor speed control module and the online particle size monitoring module to obtain mung bean micro-powder with the target particle size distribution. The powder forming unit is connected to the central control unit and is used to receive mung bean slurry and form it into a film. The drying unit is connected to the central control unit and is used to receive the shaped wet powder sheets. The drying unit integrates a drying medium temperature and humidity sensor, an online powder sheet moisture detector and a dehumidification regulating valve. The central control unit adjusts the drying parameters in real time according to the drying kinetic model until the powder sheets reach the preset final moisture content. The integrated packaging unit communicates with the central control unit and is used to measure and package the dried baking powder.

[0007] Preferably, the raw material pretreatment unit includes at least a moisture detection module, a visual recognition and sorting module, and an intelligent soaking module for dynamically calculating the soaking time based on the difference between the initial moisture content and the target moisture content of the mung beans.

[0008] Preferably, the powder forming unit includes at least a slurry flow control valve, a film forming temperature control module, and a film forming thickness monitoring sensor. The central control unit controls the slurry supply per unit time and the heating temperature of the film forming surface based on the ambient temperature and humidity and the characteristics of the slurry, in order to form a uniform powder sheet.

[0009] Preferably, the integrated packaging unit includes a finished product visual quality inspection module. The inspection results of this module are fed back to the central control unit for traceability calibration of the process parameters of the aforementioned processing unit.

[0010] Preferably, the intelligent soaking module in the raw material pretreatment unit includes the following steps for "dynamically calculating soaking time": The central control unit receives the initial moisture value of the mung beans measured by the moisture detection module, calculates the amount of water to be absorbed based on the target moisture value, and then combines the current water temperature with the mung bean type and integrity information obtained by the visual recognition and sorting module. Through a pre-stored moisture absorption kinetic model, the optimal soaking time is calculated and the actuator is controlled. The moisture absorption kinetic model is a multivariate mathematical relationship used to describe and quantify the moisture absorption law and rate of mung bean raw materials during the soaking process. The model uses the real-time detected initial moisture value of the mung beans, the preset target moisture value, the current soaking water temperature, and the mung bean type and integrity information provided by the visual recognition and sorting module as input parameters. Through built-in calculation logic, the optimal time parameter required for the mung beans to reach the best soaking state under the current specific working conditions is dynamically solved, and this calculation result is used as the core basis for controlling the actuator of the intelligent soaking module.

[0011] Preferably, the moisture detection module uses a near-infrared spectrometer for online detection, and the visual recognition and sorting module uses a high-resolution CCD camera combined with machine learning algorithms to distinguish the variety and size of mung beans and remove unqualified raw materials such as moldy or insect-infested beans.

[0012] Preferably, in the micro-pulverization and air classification unit, the online particle size monitoring module adopts the principle of laser diffraction to monitor the particle size distribution of the pulverized mung bean powder in real time and feeds the data back to the central control unit. After receiving the real-time particle size distribution data from the online particle size monitoring module, the central control unit compares and analyzes it with the target particle size distribution range pre-stored in the process parameter database, and adjusts the operating speed of the pulverizer or classifier motor by changing the output frequency of the motor speed control module according to the analysis results.

[0013] Preferably, the target particle size distribution is a pre-set range of key indicators such as D50 and D90 based on the final product's taste requirements (such as smoothness and toughness). The motor speed control module is a frequency converter, which adjusts the crushing force and grading accuracy by changing the speed of the main crushing motor and the classifier motor.

[0014] Preferably, in the powder forming unit, the control of flow rate and temperature by the central control unit includes: the central control unit receives signals from the ambient temperature and humidity sensor and the estimated viscosity of the slurry from the micro-pulverizing unit, and calculates the slurry flow rate and the surface temperature of the forming roller or steel belt required to form a uniform powder sheet of a preset thickness under the current conditions through a pre-stored film forming process model. The slurry flow rate refers to the volume of mung bean slurry delivered to the forming roller or steel belt per unit time by controlling the opening of the slurry flow control valve, and the surface temperature of the forming roller or steel belt refers to the specific heating temperature maintained on the surface of the film forming carrier in contact with the slurry by adjusting the power output of the film forming temperature control module.

[0015] Preferably, the slurry viscosity estimate is calculated based on online particle size data and concentration data using an empirical formula, or is directly measured by an online viscometer. The film-forming temperature control module uses a multi-segment controlled thermal oil furnace or infrared heater to ensure the uniformity of the film-forming surface temperature.

[0016] Preferably, in the drying unit, the drying kinetic model operated by the central control unit can fit the real-time moisture decrease curve fed back by the online moisture detector of the rice noodle sheet with the model pre-curve. If the deviation exceeds the threshold, the temperature, humidity or flow rate of the drying medium is adjusted to bring the actual drying curve closer to the optimal process curve. The real-time moisture decrease curve is the actual data trajectory that is continuously monitored and fed back by the online moisture detector of the rice noodle sheet, representing the change of the moisture content of the rice noodle sheet with time during the drying process. The model pre-curve is the ideal theoretical trajectory that is pre-calculated by the central control unit based on the initial conditions of the rice noodle sheet to be dried and its pre-stored drying kinetic model, representing the change of the moisture content of the rice noodle sheet with time under the optimal drying parameters.

[0017] Preferably, the finished product visual quality inspection module in the integrated packaging unit is used to identify surface defects, uneven color, and damage of the finished baking powder. When the number or area of ​​a specific type of defect exceeds a set threshold, the central control unit can trace the processing parameter records of that batch of products in the baking powder film-forming unit and drying unit, and use machine learning algorithms to analyze the causes of the defects, thereby proposing or adjusting the relevant process parameters. The specific algorithm flow is as follows: The central control unit acquires classified defect data from the finished product visual quality inspection module of the integrated packaging unit, and simultaneously extracts process parameter data for the corresponding time period from the historical operation databases of the powder forming and drying units, forming a timestamped multivariate dataset. The data is then preprocessed, including numerical normalization, discretization, and feature labeling, to form a training sample set. A decision tree algorithm is then used to construct a classification model. The feature splitting order is determined by calculating the information gain or Gini impurity of each process parameter feature. The information gain calculation formula is as follows: ; The total number of data samples representing the parent node; This represents the number of data samples contained in the j-th child node; m represents the total number of child nodes generated after splitting based on a certain feature f; For the parent node dataset; The dataset of child nodes after splitting; This is a function of Gini impurity or information entropy. f represents the feature to be split.

[0018] Preferably, the frying machine also includes a remote receiving platform, which has a case library of processing parameters based on mung bean raw materials from different origins and batches. When the visual recognition and sorting module identifies the category characteristics of the newly added raw materials, the central control unit can match or adaptively generate a set of recommended initial processing parameters from the case library.

[0019] The beneficial effects of this invention are: 1. This invention, by introducing a central control unit and constructing a processing device covering the entire process from raw material handling to finished product packaging, realizes data communication and collaborative operation between various process units, changing the traditional situation of isolated equipment operation, thereby improving the overall automation level and operating efficiency of the production system; 2. This invention achieves refined processing of mung bean raw materials through a moisture detection and visual recognition sorting module in the raw material pretreatment unit, combined with an intelligent soaking algorithm. It can automatically adapt to the differences in the initial state of different batches or types of raw materials, ensuring the stability of subsequent processes from the source, reducing product quality risks caused by raw material fluctuations and reliance on skilled workers; 3. Based on an online particle size monitoring process in the micro-pulverization and air classification unit, this invention can sense and adjust the pulverization intensity in real time, thereby ensuring that the particle size distribution of mung bean powder remains stable within the target range, thus guaranteeing the uniformity of slurry characteristics. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the system framework of the present invention; Figure 2 This is a schematic diagram of the workflow of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 and Figure 2 This invention provides an embodiment of a mung bean flour processing and baking machine: This invention relates to an integrated rice flour milling machine that connects the entire process of raw material pretreatment, micro-pulverization and air classification, rice flour film formation, drying and integrated packaging through a central control unit. After the device is started, the mung bean raw material first enters the raw material pretreatment unit to complete sorting and precise soaking. Then, the processed mung beans enter the micro-pulverization and air classification unit, where they are processed into uniformly sized micro-powder and mixed into a slurry. The mung bean slurry is then transported to the rice flour film formation unit, where a uniform wet rice flour sheet is formed under dynamic collaborative control. The formed wet rice flour sheet immediately enters the drying unit and is dried to a safe moisture content according to the optimal drying curve. Finally, the dried rice flour is packaged in the integrated packaging unit.

[0023] In this embodiment, the central control unit continuously receives data from various unit sensors. Its built-in optimization algorithm first fuses and aligns the real-time data (such as raw material moisture, ambient temperature and humidity, real-time particle size, drying moisture, etc.) with timestamps to comprehensively evaluate the current operating status of the entire system. Then, it calls the corresponding process models (such as drying kinetics model, moisture absorption kinetics model, film formation process model) to calculate and generate process parameter adjustments required to achieve the optimal target. Finally, these decisions are converted into control commands and distributed to various actuators (such as valves, frequency converters, heaters), and the system feedback after the commands are executed is continuously monitored. For example, the central control unit first obtains the real-time moisture sequence data fed back by the online moisture detector for rice noodle sheets. The data from the temperature and humidity sensor of the drying medium is compared with a pre-stored drying kinetic model, namely the Page model. (Where MR is the moisture ratio, k is the drying rate coefficient, and n is the time exponent) The generated theoretical drying curve Real-time fitting is performed, and the integral of the deviation between the actual curve and the theoretical curve is calculated using the least squares method. When the SSE exceeds a preset threshold, an optimization algorithm based on gradient descent is initiated, with minimizing the SSE as the objective function, through iterative calculation. and (in The drying temperature, The partial derivative of the dehumidification valve opening is used to dynamically solve for the parameter adjustment amount that allows the actual drying trajectory to return to the optimal path. and Ultimately, the adjustment amount is mapped into specific control commands for the heater power controller and the dehumidification regulating valve actuator.

[0024] The core function of the central control unit lies in achieving a fundamental leap from single-point control to system-level collaborative optimization. By processing global information from multiple variables and cross-processes and making comprehensive intelligent decisions, it ensures that all units, such as raw material pretreatment, micronization and air classification, powder film formation, closed-loop drying and integrated packaging and quality feedback, can work together, so that the entire production system always operates within the optimal process window, thereby significantly improving production efficiency, product quality stability and energy utilization efficiency.

[0025] In this embodiment, the raw material pretreatment unit includes, in the following working order, a feeding port, a visual recognition and sorting module, a moisture detection module, a soaking tank, and an intelligent soaking module. The visual recognition and sorting module includes a high-resolution CCD camera, a specific wavelength LED light source, and an image processing computer. The moisture detection module preferably uses a near-infrared online spectrometer, which can detect moisture non-destructively and quickly during material transportation.

[0026] During operation, mung beans are evenly distributed through the feeding port onto a conveyor belt. A CCD camera captures their images, and an image processor runs a pre-trained machine learning image classification algorithm to identify moldy, insect-damaged, discolored beans and impurities. The processor then controls a spray valve or robotic arm to remove these defects. Simultaneously, qualified mung beans flow through a near-infrared online spectrometer detection area. The spectrometer acquires spectral data in real time and directly calculates the initial moisture content using a calibrated model. The central control unit receives the sorting results (type, integrity) and moisture data. The intelligent soaking module then begins calculations based on a moisture absorption kinetic model, which can be simplified to: T = f(ΔM, Tw, The formula is Kp), where T is the required soaking time, ΔM is the difference between the target moisture content and the initial moisture content, Tw is the current water temperature, and Kp is the absorption coefficient determined based on the type and integrity of the mung beans (this coefficient is pre-stored in the database). The central control unit uses this formula to calculate the optimal soaking time. Finally, the central control unit controls the feed valve and water inlet valve to add a fixed amount of mung beans and water to the soaking tank and starts timing. After the optimal soaking time is reached, a command is issued to drain the soaking water.

[0027] The machine learning image classification algorithm in the above embodiments includes, specifically: acquiring RGB three-channel digital images of mung bean raw materials using a high-resolution CCD camera, and preprocessing the original images, including noise reduction using Gaussian filtering and image size standardization to a uniform pixel size; then using a pre-trained convolutional neural network model for feature extraction and classification. This network uses multiple convolutional layers (its convolution operation formula is: ; Where I is the input image matrix; K is the convolution kernel weight matrix; To output the spatial location on the feature map; It is the index used when traversing the convolution kernel K; The output feature map is a two-dimensional matrix.

[0028] Hierarchical features from edge to texture are extracted layer by layer, then non-linearly transformed using the ReLU activation function f(x) = max(0,x), and the feature maps are then fed into a fully connected layer, finally passing through the Softmax function. ; This represents the output probability of the input vector z belonging to the i-th category after being calculated by the Softmax function. Let z be the numerical value in the input vector corresponding to the i-th category; K is the total number of categories; For each Perform exponentiation to convert any real number into a positive number; This operation sums the exponents for all categories, normalizing all output probabilities to ensure that the sum is 1.

[0029] The system calculates the probability distribution of each category (qualified mung beans, moldy beans, insect-damaged beans, discolored beans, and impurities); the system takes the category corresponding to the maximum probability as the identification result, and immediately sends a rejection instruction to the sorting execution mechanism when the category is identified as a defect, thus completing the intelligent sorting process based on deep learning images.

[0030] Through the above steps, the raw material pretreatment unit avoids the problems of insufficient soaking (poor subsequent grinding effect) or excessive soaking (nutrient loss and fermentation) caused by the traditional fixed-time method, thereby improving the stability of the process and the consistency of product quality from the source.

[0031] In this embodiment, the micro-pulverization and air classification unit mainly includes a feeder, a main pulverizer, a classifier, an induced draft fan, an online particle size monitoring module, and a motor speed control module. The online particle size monitoring module adopts a particle size analyzer based on the principle of laser diffraction, and its sampling head is installed in the main air duct after pulverization.

[0032] During operation, the pulverized mung bean powder is carried by airflow through the sampling area. A laser particle size analyzer continuously measures the particle size distribution of the powder and transmits key indicators (such as D50 median diameter and D90 diameter) to the central control unit in real time. The central control unit compares the received real-time D50 and D90 values ​​with the target range (such as D50 ± 2 μm) set in the process parameter database. If the real-time value continuously deviates from the target range, the control unit will activate an adjustment algorithm. For example, if D50 is consistently too high, it is determined that the pulverizing force is insufficient, and the algorithm will calculate a new frequency setting value for the main pulverizer inverter. Where k is the proportionality coefficient, Original frequency, new frequency command The frequency converter sent to the main crusher increases the motor speed, enhances the crushing intensity, and reduces the output particle size, and vice versa.

[0033] Through the above steps, the micro-pulverization and air classification unit can automatically compensate for fluctuations in the pulverization effect caused by wear of grinding parts, changes in the hardness of raw materials, etc., thereby ensuring that the obtained mung bean micro powder has a highly uniform particle size.

[0034] In this embodiment, the slurry film forming unit includes a slurry buffer tank, a slurry flow control valve, a film forming device, a film forming temperature control module, a film forming thickness monitoring sensor, and an ambient temperature and humidity sensor.

[0035] During operation, the central control unit reads the ambient temperature and humidity (Ta, RH) in real time and obtains the estimated viscosity μ of the current slurry from the data of the previous unit (this estimate can be calculated based on particle size data and solid content using the empirical formula μ=f(D50,C)). Then, the central control unit calls the film formation process model, which describes the coupling relationship between flow rate (Q), temperature (T), ambient conditions (Ta, RH), slurry viscosity (μ), and film thickness (H), which can be expressed as: H = g(Q, T, Ta, RH, μ). The goal of the control unit is to maintain H constant at the set value. When Ta increases or μ decreases, the model calculates that the flow rate Q and film-forming temperature T need to be reduced simultaneously to prevent the powder film from becoming too thin or drying too quickly, which could lead to cracking. The specific adjustment amount is obtained through model back-calculation, and the central control unit will use the calculated new flow rate setpoint. and new temperature setting The new parameters are simultaneously sent to the metering pump and temperature control system, enabling them to operate according to the new parameters.

[0036] The above steps solve the problem of isolated flow and temperature control in traditional equipment, effectively address the fluctuations in the environment and raw materials, and ensure the uniformity of the powder sheet thickness and mechanical integrity.

[0037] In this embodiment, the drying unit is a tunnel-type or box-type drying equipment, which is equipped with a conveyor belt and integrates a drying medium temperature and humidity sensor located in the air supply and return ducts, an online moisture detector for rice noodle skin located at the end of the drying section (using a non-contact infrared moisture meter), a heater, a humidifier, a dehumidification regulating valve, and a circulating fan.

[0038] During operation, the infrared moisture meter continuously monitors the moisture content of the powder skin as it is about to leave the drying zone. (t), and plot the real-time moisture decrease curve. The central control unit runs a drying kinetic model, which is a simplified thin-layer drying equation, such as the Page model: Where MR is the moisture ratio, and k and n are model parameters. The control unit fits the real-time moisture data with the ideal drying curve predicted by the model. If the real-time moisture curve is consistently higher than the ideal curve (i.e., drying is too slow), the control unit determines that the current drying conditions are insufficient. The algorithm calculates the adjustment amount required to increase the temperature of the drying medium or decrease the humidity of the medium (by increasing the opening of the dehumidification regulating valve) based on the magnitude of the deviation. The adjustment command is sent to the heater power controller and the dehumidification regulating valve actuator to change the environment inside the drying chamber. The system continuously monitors the effect after adjustment until the moisture content of the powder reaches the preset final moisture content.

[0039] The algorithm described above calculates the adjustment amount needed to increase the temperature of the drying medium or decrease its humidity based on the magnitude of the deviation. Specifically, it involves calculating the integral of the deviation between the actual curve and the theoretical curve using the least squares method. When the SSE exceeds a preset threshold, an optimization algorithm based on gradient descent is initiated, with minimizing the SSE as the objective function, through iterative calculation. and (in The drying temperature, The partial derivative of the dehumidification valve opening is used to dynamically solve for the parameter adjustment amount that allows the actual drying trajectory to return to the optimal path. and The sensitivity of the deviation to each control variable is precisely quantified, and finally, based on the sign and magnitude of the partial derivatives, the specific adjustment direction and amount required to reduce SSE are determined, i.e., when... When the value is negative, calculate the temperature rise. ,when When the value is positive, calculate the increase in the opening of the exhaust valve. This enables precise parameter correction based on deviation quantification analysis.

[0040] By following the steps above, we can avoid the energy waste or product damage caused by setting fixed times based on experience in the traditional method. This allows us to minimize the drying time while ensuring product quality (no cracking, no mold).

[0041] In this embodiment, the integrated packaging unit includes a conveyor belt, an automatic weighing machine, a packaging machine, and a finished product visual quality inspection module, wherein the visual quality inspection module includes a high-speed line scan camera, a high-uniformity LED light source, and an image processing computer.

[0042] During operation, the finished baking powder passes through a vision inspection area on a conveyor belt. A line scan camera continuously captures images of its surface, and an image processing computer runs machine learning algorithms. The central control unit acquires classified defect data from the finished product visual quality inspection module of the integrated packaging unit, and simultaneously extracts process parameter data for the corresponding time period from the historical operation databases of the powder forming and drying units, forming a timestamped multivariate dataset. The data is then preprocessed, including numerical normalization, discretization, and feature labeling, to form a training sample set. A decision tree algorithm is then used to construct a classification model. The feature splitting order is determined by calculating the information gain or Gini impurity of each process parameter feature. The information gain calculation formula is as follows: ; The total number of data samples representing the parent node; This represents the number of data samples contained in the j-th child node; m represents the total number of child nodes generated after splitting based on a certain feature f; For the parent node dataset; The dataset of child nodes after splitting; This is a function of Gini impurity or information entropy. f represents the feature to be split.

[0043] The system automatically identifies and classifies defects such as bubbles, black spots, discoloration, or breakage. When the incidence of a certain type of defect (such as "large-area bubbles") exceeds a set threshold, the central control unit initiates a traceability program. This program retrieves all historical process parameter records for that batch of products during the production process, including the soldering unit (potential causes of bubbles: excessively high temperature, unstable slurry flow) and the drying unit (potential causes of bubbles: a sudden temperature rise in the initial drying stage). Then, it uses machine learning diagnostic algorithms to analyze the correlation between these parameters and the occurrence of defects. For example, it might find that "when the film-forming temperature exceeds..." When the bubble defect rate increases significantly, the central control unit will issue an alarm and parameter adjustment suggestions to the operator based on the diagnostic results (such as "it is recommended to reduce the film formation temperature by 3°C"), or automatically adjust the process parameter settings of the relevant units after obtaining authorization.

[0044] The aforementioned machine learning diagnostic algorithm includes: the central control unit first constructs a database containing historical process parameters (including film formation temperature). Slurry flow rate Q, drying temperature A dataset containing environmental humidity (RH, etc.) and corresponding defect types (such as bubbles, black spots, uneven color) labels was generated. The Apriori algorithm from association rule mining was then used for analysis. This algorithm quantifies the association strength between parameter combinations and defects by calculating two key indicators: support and confidence. The calculation formulas are as follows: ; ; Where A represents a specific combination of process parameters (such as "..."). B represents a specific defect type. (X) represents the number of transactions in the dataset that contain itemset X, and N represents the total number of transactions. The algorithm filters out strongly correlated rules from all possible parameter combinations by setting minimum support thresholds and minimum confidence thresholds (e.g., "when film formation temperature > And slurry flow rate < At that time, the confidence level of bubble defects was 85% and the support level was 12%. Finally, the central control unit established a diagnostic knowledge base for the causes of defects based on these strong correlation rules, which was used for parameter anomaly location and optimization decision-making in real-time production.

[0045] In this embodiment, the remote receiving platform integrates a processing parameter case library. When the visual recognition and sorting module of the field system identifies the newly added raw material as "Northeast mung beans", the central control unit can initiate a query to the platform. The platform uses a matching algorithm to find the most similar past successful production cases in the case library and recommends its process parameter set to the field system for use.

[0046] This invention provides embodiments: This example illustrates a factory that originally used "North China mung beans" for production but now needs to switch to "Northeast China mung beans": The newly added "Northeast mung bean" raw material first enters the raw material pretreatment unit. The visual recognition and sorting module in this unit captures the image features of the mung beans and runs the built-in machine learning image classification algorithm to automatically identify its category as "Northeast mung bean". Then, it immediately reports this key raw material category information to the central control unit.

[0047] After receiving information about new raw material categories, the central control unit immediately sends a query request to the remote receiving platform via the network. The remote receiving platform uses its stored processing parameter case library, which is stored in advance in the device. For example, the processing parameters for Mung Bean No. 1 are set in this library. When processing Mung Bean No. 1, the system directly applies these parameters.

[0048] The system then runs a K-nearest neighbor matching algorithm to find historical successful production cases that are most similar to the current "Northeast mung beans" in terms of characteristics. Finally, it recommends a complete set of corresponding and verified initial process parameters (including soaking coefficient, target particle size, film-forming temperature and flow rate ratio, drying curve, etc.) to the central control unit.

[0049] The central control unit receives the recommended parameter set from the remote operation and maintenance platform and loads it as the initial process setting value for this production operation. Then, it sends various specific parameters, such as target particle size distribution, film formation temperature and flow rate coordination relationship, to the corresponding micro-pulverization and airflow classification units and powder film formation units. The system immediately starts production based on this new set of parameters.

[0050] After production stabilizes, the finished product visual quality inspection module in the integrated packaging unit starts working. It continuously collects surface images of the finished product powder using a line scan camera and analyzes them using a machine learning defect recognition model. It finds that the proportion of defects of the "slightly yellow color" type in this batch of products exceeds the preset quality pass threshold and immediately reports this quality abnormality to the central control unit.

[0051] Upon receiving a quality anomaly alarm, the central control unit immediately initiates a traceability and diagnostic procedure. It retrieves historical process parameter records of the batch of products as they flow through the soldering and drying units during production, and uses machine learning diagnostic algorithms to perform correlation analysis between these data and defect phenomena. Ultimately, it diagnoses that "the film-forming temperature setpoint is at the upper limit of the recommended range" as the main potential cause of the slightly yellow color.

[0052] Based on the diagnostic analysis, the central control unit can automatically perform the correction operation without waiting for manual intervention. It precisely lowers the film-forming temperature setting of the powder forming unit by 2°C, thereby eliminating the slight yellowing defect.

[0053] After parameter adjustments, the system continued production and confirmed that the color of the finished product returned to normal and the quality defect rate dropped below the threshold. At this point, the central control unit saved the successful fine-tuning of parameters (the parameter set after a 2°C reduction) for "Northeast mung beans" as a new production case in the local database and the case library of the remote receiving platform, and named it "Northeast Mung Beans - Optimized Version". This provides a more accurate initial parameter setting for the production of the same raw material in the future.

[0054] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A mung bean flour processing and baking machine, characterized in that, Including: The central control unit is used to receive and process data from the raw material pretreatment unit, the micronization and air classification unit, the powder film forming unit, the drying unit, and the integrated packaging unit. Based on the built-in process parameter database and optimization algorithm, which refers to the database stored in the central control unit's memory, containing a standard parameter set and a success case library formed by historical production data accumulation, and a software program that is data-driven and can call pre-stored process mathematical models to process, analyze, and make decisions on real-time multi-source data, the central control unit issues control commands to the execution unit. The raw material pretreatment unit is connected in communication with the central control unit and is used to receive mung bean raw materials; The micro-pulverizing and air classifying unit is connected to the central control unit. It is used to receive pre-treated mung beans and adjust the pulverizing intensity in real time through feedback from the motor speed control module and the online particle size monitoring module to obtain mung bean micro-powder with the target particle size distribution. The powder forming unit is connected to the central control unit and is used to receive mung bean slurry and form it into a film. The drying unit is connected to the central control unit and is used to receive the shaped wet powder sheets. The drying unit integrates a drying medium temperature and humidity sensor, an online powder sheet moisture detector and a dehumidification regulating valve. The central control unit adjusts the drying parameters in real time according to the drying kinetic model until the powder sheets reach the preset final moisture content. The integrated packaging unit communicates with the central control unit and is used to measure and package the dried baking powder.

2. The mung bean flour processing and baking machine according to claim 1, characterized in that: The raw material pretreatment unit includes at least a moisture detection module, a visual recognition and sorting module, and an intelligent soaking module for dynamically calculating the soaking time based on the difference between the initial moisture content and the target moisture content of the mung beans.

3. The mung bean flour processing and baking machine according to claim 1, characterized in that: The powder forming unit includes at least a slurry flow control valve, a film forming temperature control module, and a film forming thickness monitoring sensor. The central control unit controls the slurry supply per unit time and the heating temperature of the film forming surface based on the ambient temperature and humidity and the characteristics of the slurry, in order to form a uniform powder sheet.

4. The mung bean flour processing and baking machine according to claim 1, characterized in that: The integrated packaging unit includes a finished product visual quality inspection module. The inspection results of this module are fed back to the central control unit for traceability calibration of the process parameters of the aforementioned processing unit.

5. A mung bean flour processing and baking machine according to claim 2, characterized in that, The intelligent soaking module in the raw material pretreatment unit has a "dynamic calculation of soaking time" step that specifically includes: the central control unit receives the initial moisture value of the mung beans measured by the moisture detection module, calculates the amount of water to be absorbed based on the target moisture value, and then combines the current water temperature with the mung bean type and integrity information obtained by the visual recognition and sorting module. Through a pre-stored moisture absorption kinetic model, the optimal soaking time is calculated and the actuator is controlled. The moisture absorption kinetic model is a multivariate mathematical relationship used to describe and quantify the moisture absorption law and rate of mung bean raw materials during the soaking process. The model uses the real-time detected initial moisture value of the mung beans, the preset target moisture value, the current soaking water temperature, and the mung bean type and integrity information provided by the visual recognition and sorting module as input parameters. Through built-in calculation logic, the optimal time parameter required for the mung beans to reach the best soaking state under the current specific working conditions is dynamically solved, and this calculation result is used as the core basis for controlling the actuator of the intelligent soaking module.

6. The mung bean flour processing and baking machine according to claim 1, characterized in that: In the micro-pulverization and air classification unit, the online particle size monitoring module uses the principle of laser diffraction to monitor the particle size distribution of the pulverized mung bean powder in real time and feeds the data back to the central control unit. After receiving the real-time particle size distribution data from the online particle size monitoring module, the central control unit compares and analyzes it with the target particle size distribution range pre-stored in the process parameter database, and adjusts the operating speed of the pulverizer or classifier motor by changing the output frequency of the motor speed control module according to the analysis results.

7. A mung bean flour processing and baking machine according to claim 3, characterized in that, In the powder forming unit, the control of flow rate and temperature by the central control unit includes: the central control unit receives signals from the ambient temperature and humidity sensor and the estimated viscosity of the slurry from the micro-pulverizing unit, and calculates the slurry flow rate and surface temperature of the forming roller or steel belt required to form a uniform powder sheet of a preset thickness under the current conditions through a pre-stored film forming process model. The slurry flow rate refers to the volume of mung bean slurry delivered to the forming roller or steel belt per unit time by controlling the opening of the slurry flow control valve, while the surface temperature of the forming roller or steel belt refers to maintaining a specific heating temperature on the surface of the film forming carrier in contact with the slurry by adjusting the power output of the film forming temperature control module.

8. A mung bean flour processing and baking machine according to claim 1, characterized in that: In the drying unit, the drying kinetic model running by the central control unit can fit the real-time moisture decrease curve fed back by the online moisture detector of the rice noodle sheet with the model pre-curve. If the deviation exceeds the threshold, the temperature, humidity or flow rate of the drying medium is adjusted to bring the actual drying curve closer to the optimal process curve. The real-time moisture decrease curve is the actual data trajectory that characterizes the change of the moisture content of the rice noodle sheet over time during the drying process, continuously monitored and fed back by the online moisture detector of the rice noodle sheet. The model pre-curve is the ideal theoretical trajectory that characterizes the change of the moisture content of the rice noodle sheet over time under the optimal drying parameters, which is pre-calculated by the central control unit based on the initial conditions of the rice noodle sheet to be dried and its pre-stored drying kinetic model.

9. A mung bean flour processing and baking machine according to claim 4, characterized in that: The finished product visual quality inspection module in the integrated packaging unit is used to identify surface defects, uneven color, and damage of the finished baking powder. When the number or area of ​​a specific type of defect exceeds a set threshold, the central control unit can trace the processing parameter records of that batch of products in the baking powder film-forming unit and drying unit, and use machine learning algorithms to analyze the causes of the defects, thereby proposing or adjusting the relevant process parameters. The specific algorithm flow is as follows: The central control unit acquires classified defect data from the finished product visual quality inspection module of the integrated packaging unit, and simultaneously extracts process parameter data for the corresponding time period from the historical operation databases of the powder forming and drying units, forming a timestamped multivariate dataset. The data is then preprocessed, including numerical normalization, discretization, and feature labeling, to form a training sample set. A decision tree algorithm is then used to construct a classification model. The feature splitting order is determined by calculating the information gain or Gini impurity of each process parameter feature. The information gain calculation formula is as follows: ； The total number of data samples representing the parent node; This represents the number of data samples contained in the j-th child node; m represents the total number of child nodes generated after splitting based on a certain feature f; For the parent node dataset; The dataset of child nodes after splitting; This is a function of Gini impurity or information entropy. f represents the feature to be split.

10. A mung bean flour processing and baking machine according to claim 5, characterized in that: The frying machine also includes a remote receiving platform, which has a case library of processing parameters based on mung bean raw materials from different origins and batches. When the visual recognition and sorting module identifies the category characteristics of the newly added raw materials, the central control unit can match or adaptively generate a set of recommended initial processing parameters from the case library.

Images