Traditional Chinese medicine grinding control system

Abstract

The invention discloses a traditional Chinese medicine grinding control system, which relates to the technical field of traditional Chinese medicine grinding control and comprises a medicinal material characteristic sensing module, an intelligent calculation analysis module, a dynamic execution adjustment module, a multi-mode optimization control module and a system coordination updating module. The medicinal material characteristic sensing module collects physical and texture characteristic data of medicinal materials; the intelligent calculation and analysis module fuses the medicinal material processing knowledge base and historical grinding process data to generate initial grinding control parameters; the dynamic execution adjustment module adjusts the operation parameters of the grinding mechanism in combination with the real-time grinding state data; the multi-modal optimization control module monitors the deviation between actual quality parameters and target quality parameters of powder, and a self-adaptive decision algorithm is used for optimizing regulation and control parameters of a grinding area; the system coordination updating module collects whole-process data and performance index evaluation and generates a system optimization instruction, and the instruction corresponding module adjusts the rotating speed and pressure parameters of a grinding area of the vertical grinding machine. According to the system, intelligent and precise control over medicinal material grinding is achieved, and the stability and consistency of the grinding quality are improved.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine grinding control technology, specifically a traditional Chinese medicine grinding control system. Background Technology

[0002] In the clinical diagnosis and treatment of traditional Chinese medicine and the production of Chinese medicine preparations, the grinding process of medicinal materials has a direct impact on the final efficacy. The quality parameters such as the particle size uniformity and fineness of the ground medicinal powder are closely related to the dissolution efficiency of the effective components and the absorption effect by the human body. Generally, Chinese medicinal materials are ground using grinding equipment, such as vertical grinding mills. These vertical grinding mills are generally composed of a mounting frame, grinding mechanism, feed hopper, and grinding disc drive motor, and are relatively mature products. Although the vertical grinding mill is structurally convenient for the material to flow from top to bottom, it is limited by the single vertical cavity design and mostly relies on the experience of the operator, lacking precise consideration of the characteristics of the medicinal materials themselves. Different Chinese medicinal materials have significant differences in physical properties, such as density, hardness, brittleness, and initial texture characteristics such as fibrous, granular, and powdery forms. These characteristics directly determine the appropriate grinding method and grinding parameters. However, in the traditional grinding process, operators often choose fixed grinding speed, grinding time, and grinding pressure parameters based solely on experience, and cannot make targeted adjustments according to the specific characteristics of the medicinal materials.

[0003] Traditional grinding processes lack the ability to monitor and dynamically control the grinding state in real time. During grinding, medicinal materials gradually transform from lumps or granules into powder, and their particle size, uniformity, and other properties continuously change. Traditional grinders cannot collect this grinding state data in real time, resulting in grinding parameters remaining fixed and making it difficult to optimize and adjust them according to changes in the grinding process. When the characteristics of the medicinal materials change or the grinder experiences slight wear, over-grinding or under-grinding is very likely to occur. Over-grinding will destroy the effective components of the medicinal materials, while under-grinding will result in uneven particle size of the medicinal material powder, which cannot meet the requirements of subsequent processing or preparation.

[0004] While some existing vertical grinding machines possess a degree of automation and attempt to incorporate simple parameter control functions, most rely solely on single preset parameters. They lack a comprehensive knowledge base for medicinal material processing and fail to integrate historical grinding process data for analysis, resulting in unscientific and irrational parameter settings. Furthermore, these devices cannot accurately detect and optimize the quality of the ground medicinal powder, hindering the formation of a complete closed loop of "medicinal material characteristic perception - grinding parameter calculation - grinding process control - powder quality optimization." This prevents intelligent and precise control of the grinding process, making it difficult to meet the high demands of modern Chinese medicine production for consistent and stable grinding quality, thus hindering the standardization and modernization of Chinese medicine production. Summary of the Invention

[0005] The purpose of this invention is to provide a traditional Chinese medicine grinding control system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a traditional Chinese medicine grinding control system for controlling a vertical grinding machine to perform grinding operations. The vertical grinding machine includes a mounting frame, a grinding mechanism, a feeding hopper, and a grinding disc drive motor. The system includes: The medicinal material characteristic sensing module is used to collect physical and texture characteristic data of the medicinal materials to be processed. The intelligent calculation and analysis module is used to receive the physical property data and the texture property data, integrate the preset medicinal material processing knowledge base and historical grinding process data, and generate initial grinding control parameters. The dynamic execution adjustment module is used to receive the initial grinding control parameters and, in conjunction with the real-time collected grinding status data, dynamically adjust multiple operating parameters of the grinding mechanism. The multimodal optimization control module is used to receive the adjusted and ground medicinal powder sample and its target quality parameters, obtain the actual quality characteristics of the powder through a real-time monitoring system, use an adaptive decision algorithm to determine the deviation between the actual quality characteristics and the target quality parameters, and dynamically optimize the control parameters of the grinding area. The system coordination and update module is used to collect and evaluate the full-process operation data and performance indicators of the medicinal material characteristic sensing module, the intelligent calculation and analysis module, the dynamic execution adjustment module, and the multimodal optimization control module, and generate system optimization instructions. The corresponding modules of the instructions control the grinding disc drive motor, grinding pressure regulating hydraulic cylinder, and feeding speed control motor of the vertical grinder, and adjust the rotation speed and pressure parameters of the grinding area of ​​the vertical grinder.

[0007] Preferably, the intelligent computing parsing module specifically includes: The parameter generation unit is used to process the physical property data, the texture property data, the medicinal material processing knowledge base and the historical grinding process data through an integrated learning model, and output preliminary grinding control parameters. The parameter optimization unit is used to judge the preliminary grinding control parameters based on the grinding effect optimization target and equipment operation constraints. If the requirements are met, the final grinding control parameters are output to the dynamic execution adjustment module. Otherwise, the final grinding control parameters are output after iterative optimization using an adaptive disturbance exploration strategy.

[0008] Preferably, the adaptive perturbation exploration optimization strategy specifically includes: The grinding control parameters to be adjusted and the adjustment direction are intelligently selected based on a preset rule set; The adjustment range of the grinding control parameters is dynamically and adaptively set, and candidate parameter combinations are generated iteratively. The candidate parameter combinations are verified and evaluated, and the optimal parameter set is updated until the termination condition is met.

[0009] Preferably, the dynamic execution adjustment module specifically includes: The data acquisition unit is used to continuously acquire and parse feedback data reflecting the current state of the grinding material from the sensor group; The state comparison unit is used to compare the feedback data with the expected grinding target state received from the intelligent computing and parsing module in real time, and determine the deviation between the actual grinding state and the expected grinding target state. The parameter adjustment unit is used to calculate the required adjustment amount for grinding speed, grinding pressure and grinding time in real time based on the deviation, and apply this adjustment amount to the corresponding grinding actuator.

[0010] Preferably, the parameter adjustment unit specifically includes: The characteristic estimation unit is used to analyze the historical adjustment data of the grinding speed, the grinding pressure and the grinding time, as well as the corresponding real-time grinding status feedback data provided by the sensor group, and to continuously estimate and update the key dynamic characteristic parameters of the grinding process online using a recursive estimation algorithm. The parameter setting unit automatically and periodically calculates and sets various gain parameters of the proportional-integral-derivative (PID) control algorithm based on the key dynamic characteristic parameters and in accordance with preset proportional-integral-derivative (PID) parameter tuning rules.

[0011] Preferably, the process of acquiring the actual quality characteristics of the powder through the real-time monitoring system in the multimodal optimization control module includes: During the grinding process, image data of medicinal powder in each grinding area is captured and processed in real time; The key observation area is located, and the powder particle size distribution parameters and uniformity parameters are calculated from it. These parameters are then fed back to the adaptive decision algorithm as the actual quality characteristics.

[0012] Preferably, the specific processing flow of the multimodal optimization control module for dynamically optimizing the control parameters of the grinding area using an adaptive decision algorithm is as follows: For each grinding zone, the adaptive decision algorithm calculates and generates a set of correction and control parameters aimed at minimizing the deviation quantification result based on the deviation quantification result between the current powder quality characteristics monitored in real time by machine vision and the preset target powder quality characteristics, combined with the grinding process response model. The modified control parameters are applied to the control actuator of the corresponding grinding area.

[0013] Preferably, the system coordination and update module specifically includes: The data processing unit is used to process the full-process operation data and performance indicators using a multi-source data fusion model, and output system risk assessment indicators and optimization gain indicators. The instruction generation unit is used to process the system risk assessment index and the optimization gain index by combining the intelligent decision-making model with historical optimization cases, and generate the system optimization instruction.

[0014] Preferably, the physical property data collected in the medicinal material property sensing module includes at least the medicinal material hardness value and moisture content value, and the texture property data includes at least the fiber toughness value and density value.

[0015] Preferably, the multiple operating parameters adjusted in the dynamic execution adjustment module include at least the grinding disc rotation speed parameter, grinding pressure parameter, feeding speed parameter, and grinding time parameter.

[0016] Compared with the prior art, the beneficial effects of the present invention are: The intelligent calculation and analysis module integrates a pre-set medicinal herb processing knowledge base with historical grinding process data to generate initial grinding control parameters, overcoming the shortcomings of existing automated vertical grinding machines that rely solely on a single pre-set parameter for control. The medicinal herb processing knowledge base gathers suitable grinding process rules for various medicinal materials, while the historical grinding process data contains numerous practically validated examples of effective grinding parameters. The integration of these two elements enables intelligent and precise calculation of initial grinding parameters. Through this module, initial grinding control parameters are no longer abstract values ​​detached from practical application experience, but rather scientific parameters combining industry knowledge and practical experience. This significantly improves the rationality and applicability of initial parameter settings, reducing grinding problems caused by improper parameter settings.

[0017] The dynamic adjustment module combines real-time collected grinding status data to dynamically adjust multiple operating parameters of the grinding mechanism, solving the problem that traditional grinding processes cannot adjust parameters according to changes in the grinding process. During grinding, the state of the medicinal materials continuously changes. By acquiring grinding status data in real time, this module can promptly detect deviations in the grinding process, such as excessively rapid or slow changes in particle size, and quickly and dynamically adjust multiple operating parameters such as grinding speed, grinding pressure, and grinding gap to ensure that the grinding process is always in optimal condition, avoiding over-grinding or under-grinding, and ensuring the stability and continuity of the grinding process. The multimodal optimization control module monitors the actual quality characteristics of the ground medicinal powder in real time and compares them with the target quality parameters. It then uses an adaptive decision-making algorithm to dynamically optimize the control parameters of the grinding area, constructing a closed-loop optimization mechanism for grinding quality. Compared to existing equipment that cannot accurately detect and optimize powder quality, this module can precisely capture the deviation between the actual and target powder quality. Through the adaptive decision-making algorithm, it quickly formulates optimization plans and makes targeted adjustments to the control parameters of the grinding area. This ensures that the quality of the ground medicinal powder continuously approaches the target quality, significantly improving the consistency and pass rate of the medicinal powder quality and meeting the high requirements for grinding quality in modern Chinese medicine production.

[0018] The system coordination and update module collects and evaluates the entire process's operational data and performance indicators, generating system optimization instructions to achieve self-improvement and continuous upgrading of the entire control system. This module comprehensively monitors the operational status of each stage, including medicinal material characteristic sensing, parameter calculation, process control, and quality optimization. Through comprehensive evaluation of operational data and performance indicators, it promptly identifies deficiencies in various system modules, such as insufficient data acquisition accuracy in the sensing module and low parameter generation efficiency in the calculation module. It then generates targeted system optimization instructions, which correspond to the control modules controlling the grinding disc drive motor, grinding pressure regulating hydraulic cylinder, and feeding speed control motor of the vertical grinder. By improving and refining each module, the performance of the entire control system is continuously enhanced, maintaining a highly efficient and precise operating state over the long term. This adapts to the ever-changing needs of traditional Chinese medicine production, promoting the intelligent and standardized development of the traditional Chinese medicine grinding process and contributing to the standardization and modernization of modern traditional Chinese medicine production. Attached Figure Description

[0019] Figure 1 This is a timing diagram of the traditional Chinese medicine grinding control system described in this invention; Figure 2 A flowchart of the data acquisition process for the medicinal material characteristic sensing module; Figure 3 A flowchart illustrating the working principle of the dynamic execution adjustment module; Figure 4 This is an overall diagram of an existing vertical grinding machine.

[0020] In the diagram: 1. Mounting frame; 2. Grinding mechanism; 3. Feed hopper; 4. Grinding disc drive motor. Detailed Implementation

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

[0022] Please see Figure 1 This invention provides a TCM grinding control system for controlling a vertical grinder to perform grinding operations. The vertical grinder includes a mounting frame 1, a grinding mechanism 2, a feeding hopper 3, and a grinding disc drive motor 4. The system includes: a medicinal material characteristic sensing module, an intelligent calculation and analysis module, a dynamic execution adjustment module, a multimodal optimization control module, and a system coordination and update module.

[0023] The medicinal herb characteristic sensing module collects physical and texture data of the medicinal herbs to be processed using dedicated sensing devices. Physical data includes hardness and moisture content, while texture data includes fiber toughness and density. This data is transmitted to the intelligent computing and analysis module via a data interface. Upon receiving this data, the intelligent computing and analysis module integrates and analyzes the data using its built-in medicinal herb processing knowledge base and historical grinding process data. It then generates initial grinding control parameters through an integrated learning model and outputs them to the dynamic execution and adjustment module. The dynamic execution and adjustment module initially sets the grinding mechanism based on the initial grinding control parameters. Simultaneously, it collects grinding status data in real time during operation and dynamically adjusts multiple operational parameters, including grinding disc rotation speed, grinding pressure, feeding speed, and grinding time, to maintain a stable grinding process. The multimodal optimization control module receives the adjusted and ground medicinal herb powder sample and its target quality parameters. It obtains the actual quality characteristics of the powder through a real-time monitoring system, uses an adaptive decision-making algorithm to determine the deviation between the actual quality characteristics and the target quality parameters, and dynamically optimizes the control parameters of the grinding area accordingly. The system coordination and update module is responsible for collecting the full-process operation data and performance indicators of each module, and generating system optimization instructions through evaluation and analysis, so as to achieve continuous improvement of the overall system performance.

[0024] Example 1: See Figure 2The intelligent computing and analysis module undertakes the core data processing and decision-making functions, acting as a bridge between upstream and downstream processes. This module, through its closely collaborative parameter generation and optimization units, performs in-depth analysis of the upstream data and generates reliable grinding control commands. The medicinal herb characteristic perception module first acquires multi-dimensional characteristic data of the medicinal herbs to be processed through a high-precision sensor array. This physical characteristic data precisely includes the hardness and moisture content values ​​of the herbs, while the texture characteristic data details fiber toughness and density values. All collected data, after standardized preprocessing, is transmitted in real-time to the input interface of the intelligent computing and analysis module via a high-speed data bus. The parameter generation unit, as the core of the module's data processing, immediately activates its integrated learning model. This model is a multi-algorithm fusion framework trained on a large amount of medicinal herb processing data. It simultaneously receives the incoming physical characteristic data and texture characteristic data and calls upon two large pre-stored databases within the system: the medicinal herb processing knowledge base and the historical grinding process database. The medicinal herb processing knowledge base stores expert knowledge on classic processing theories, suitable grinding particle size ranges, and physicochemical property changes of various Chinese medicinal herbs. The historical grinding process database has accumulated massive amounts of data, including detailed parameter records, equipment operating status, and final powder quality results from past successful grinding cases. The integrated learning model performs feature extraction, correlation analysis, and weighted calculation on all these heterogeneous data. Its internal sub-models, such as decision trees and neural networks, learn from the data from different perspectives and provide preliminary grinding control parameter suggestions. Finally, the model fusion algorithm outputs a set of coordinated preliminary grinding control parameters, which cover the main adjustable variables of the grinder.

[0025] The initial grinding control parameters are then sent to the parameter optimization unit for feasibility and optimization verification. This unit has a strict set of judgment logic that compares the received parameter set with the preset grinding effect optimization targets item by item. The optimization targets include multiple indicators such as the desired powder particle size distribution, energy consumption limit, and processing efficiency. At the same time, these parameters must also meet the equipment operating constraints. The equipment operating constraints define the safe working range, maximum load capacity, and linkage limitations between the various actuators of the grinder. The parameter optimization unit determines whether this set of initial parameters can achieve the optimization targets without violating any constraints through simulation calculations and rule reasoning. If the calculation results fully meet the requirements, the set of parameters is immediately confirmed as the final grinding control parameters and sent to the dynamic execution adjustment module through the communication interface. When the initial parameters fail validation, an adaptive perturbation exploration optimization strategy is automatically activated for iterative parameter optimization. This strategy first intelligently identifies the key parameters most likely to affect the achievement of the target in the current parameter combination based on a pre-defined rule set. This rule set, constructed from domain expert knowledge, defines the sensitivity relationship between different medicinal material characteristics and grinding parameters. For example, for medicinal materials with high hardness and low fiber toughness, grinding pressure is usually the preferred adjustment parameter. Simultaneously, the strategy intelligently determines the adjustment direction of the parameters—whether to increase or decrease the value—based on the magnitude and direction of the current deviation. Subsequently, the strategy enters a dynamic adaptive adjustment amplitude setting stage, calculating a non-fixed amplitude in real time based on historical optimization experience and the current state. This amplitude may be larger in the early stages of iteration to quickly approach the optimal region, while decreasing in later stages to achieve a finer search.

[0026] Based on the determined adjustment direction and magnitude, the strategy generates the first set of candidate parameter combinations and inputs them into a grinding process simulator for virtual verification. The simulator quickly predicts the grinding results and equipment operating status under this set of parameters based on the equipment physical model and material state model. The evaluation algorithm then calculates a comprehensive performance score based on the prediction results. The scoring function comprehensively considers multiple objective factors such as powder quality, energy consumption, and time cost. If the score shows performance improvement and does not violate constraints, the set of parameters is retained as the current optimal candidate, and the strategy then generates new candidate combinations in its neighborhood for exploration. If the score does not improve, the strategy adjusts the search direction or magnitude and regenerates parameter combinations. This generation-verification-evaluation cycle continues, with each iteration using the results from the previous iteration to guide the new search, forming an adaptive process that gradually approaches the optimal solution. The iteration termination condition is set to the comprehensive score no longer significantly improving after a certain number of consecutive iterations or reaching the maximum number of iterations. Finally, the strategy outputs the parameter combination with the best performance found during the iteration process that satisfies all constraints as the final grinding control parameters.

[0027] Example 2: See Figure 3 and Figure 4The dynamic adjustment module receives the final grinding control parameters from the intelligent calculation and analysis module and dynamically adjusts them based on the real-time grinding status to accurately achieve the grinding target. This module achieves closed-loop control of the grinding process through three internally cooperating units: a data acquisition unit, a state comparison unit, and a parameter adjustment unit. Its core task is to ensure that multiple operational parameters, such as grinding disc rotation speed, grinding pressure, feeding speed, and grinding time, are always optimally configured. The data acquisition unit continuously acquires raw signals from a multimodal sensor array deployed within the grinding chamber via a high-speed data acquisition card. These sensors include high-precision torque sensors, acoustic emission sensors, vibration sensors, and infrared temperature sensors, which capture physical phenomena during the grinding process from different perspectives. After amplification and filtering by the signal conditioning circuit, the acquired raw signals are analyzed and feature extracted by the embedded processor, converting them into quantitative feedback data reflecting the current state of the grinding material, such as material filling degree, interparticle friction coefficient, instantaneous energy consumption, and heat accumulation status—key state variables.

[0028] The state comparison unit receives real-time feedback data streams from the data acquisition unit and simultaneously obtains the desired grinding target state data from the intelligent calculation and parsing module. The desired target state is a multi-dimensional vector that defines the ideal powder quality indicators such as the median particle size, standard deviation of particle size distribution, particle shape factor, and maximum allowable energy consumption threshold that should be achieved in the current grinding stage. The multivariate comparison algorithm built into the state comparison unit performs real-time matching calculations between the real-time feedback data and the desired target state dimension by dimension. The algorithm first performs time series alignment and scale normalization on the feedback data to eliminate comparison errors caused by sampling delay and different dimensions. Then, it determines the multi-dimensional deviation vector between the actual grinding state and the desired grinding target state by calculating metrics such as Euclidean distance or cosine similarity. This deviation vector quantifies the magnitude and direction of the difference between the current state and the target state in each quality dimension.

[0029] The parameter adjustment unit calculates the required adjustments to the grinding speed, grinding pressure, and grinding time in real time based on the multidimensional deviation vector output by the state comparison unit. This unit uses a model predictive control (MPC)-based strategy to generate adjustment commands. Internally, a simplified dynamic mathematical model of the grinding process is constructed, describing the dynamic relationship between changes in operating parameters (such as speed and pressure) and state variables (such as particle size and energy consumption). When the deviation vector is input, the adjustment algorithm linearizes the model at the current operating point and continuously optimizes the sequence of operating parameters within a future time window to minimize the deviation between the predicted and target states. The optimization process strictly considers equipment physical constraints, such as the maximum motor speed, hydraulic system pressure limits, and mechanical structure strength limitations, ensuring that the generated adjustments are both effective and safe. The calculated adjustment amount is output as a standard industrial control signal through a digital-to-analog converter (DAC), such as a 4-20mA current signal or a 0-10V voltage signal. These signals are applied to the corresponding grinding actuators: the frequency converter adjusts the speed of the grinding disc motor, the electro-hydraulic servo valve adjusts the pressure of the grinding hydraulic cylinder, the precision feeder adjusts the feeding speed, and the grinding time is corrected by the timing function of the programmable logic controller (PLC).

[0030] During parameter adjustment, if the system detects a persistent deviation or unsatisfactory optimization results through state comparison, it will intelligently trigger an embedded adaptive disturbance exploration optimization strategy to further improve adjustment accuracy. This strategy is not continuously running but intervenes as a supplementary optimization mechanism when the performance of the conventional control algorithm is insufficient. The adaptive disturbance exploration optimization strategy first analyzes a preset set of expert rules, which contains a large amount of prior knowledge, such as "when the particle size is too large and the energy consumption is too high, prioritize fine-tuning the pressure parameter rather than the speed parameter" or "when the particle uniformity is poor, the feeding speed and grinding time need to be adjusted in tandem." The strategy intelligently selects the type of operating parameter to be adjusted and its adjustment direction (increase or decrease) based on the specific pattern of the current deviation. Subsequently, the strategy dynamically and adaptively sets the adjustment amplitude. The amplitude setter calculates a non-fixed amplitude in real time based on historical adjustment effect data, the current deviation size, and the current load status of the equipment. This amplitude is larger in the early stage of the search to quickly jump out of possible undesirable local areas, and automatically decreases when approaching the optimal point to achieve fine-tuning.

[0031] Based on the selected parameters, direction, and amplitude, the strategy generates a set of candidate parameter combinations and applies them to the grinding process (usually within a very short test cycle). The data acquisition unit simultaneously collects new state feedback data under these candidate combinations. The state comparison unit immediately calculates the deviation between the new state and the target and compares it with the deviation of the old state to evaluate the effectiveness of the candidate combination. The evaluation results are fed back to the core decision logic of the strategy, which determines the next action based on the evaluation results: if the deviation decreases, it continues exploring along the current direction; if the deviation increases, it reverses the direction or switches to another parameter to be adjusted. This generation-testing-evaluation cycle iterates at a high frequency, continuously updating the current optimal parameter set. The termination condition for the iteration process is set to the deviation falling below a preset threshold or the gain no longer being significant after several consecutive iterations. Finally, the strategy solidifies the searched optimal parameter combination and outputs it to the execution mechanism, enabling the dynamic execution adjustment module to maintain high quality and efficiency in the grinding process under complex and changing working conditions.

[0032] Taking the grinding of Astragalus membranaceus (Huang Qi) as an example, Astragalus membranaceus, as a rhizome-based medicinal herb, is characterized by medium hardness, rich fiber structure, and a certain degree of toughness. The intelligent calculation and analysis module generates an initial set of grinding control parameters based on the characteristic data of Astragalus membranaceus (hardness value, moisture content, fiber toughness value, density value) and historical processing knowledge: the grinding disc speed is set to 1000 RPM, the grinding pressure is set to 0.6 MPa, the feeding speed is set to 50 g / min, and the preset grinding time is 25 minutes. The dynamic execution adjustment module receives these parameters and starts the grinding program. Five minutes after the grinding begins, the data acquisition unit continuously collects real-time feedback data through a sensor array. The torque sensor shows a slow upward trend in the motor load current, the acoustic emission sensor detects that the intensity of a specific high-frequency signal generated by inter-particle friction is lower than expected, and the vibration sensor's spectrum analysis indicates a slight non-uniformity in the flow state of the material within the grinding chamber. The state comparison unit compares this real-time feedback data with the desired grinding target state, which includes the powder particle size D. 50 The target particle size distribution range is 90 micrometers, energy consumption is maintained below 85% of the rated value, and particle size distribution range is less than 1.7. Comparative analysis revealed a quantifiable deviation between the actual grinding state and the target: the currently estimated powder particle size D50 is approximately 125 micrometers (deviation +35 micrometers), the real-time energy consumption is 92% of the rated value (deviation +7%), and the particle size distribution range is estimated to be 1.9 (deviation +0.2).

[0033] The parameter adjustment unit responds immediately to these deviations, and its built-in model predictive control algorithm calculates based on a dynamic model of the grinding process. Starting with the current operating parameters (1000 RPM, 0.6 MPa), the algorithm continuously optimizes the operating sequence for the next 30 seconds, with the objective function being to minimize the weighted sum of particle size deviation, energy consumption deviation, and span deviation. After multiple rounds of iterative calculations, the algorithm outputs a set of adjustments: increasing the grinding disc speed to 1080 RPM, raising the grinding pressure to 0.68 MPa, and slightly reducing the feeding speed to 47 g / min to prolong the material's residence time in the grinding zone. These adjustments are converted into 4-20 mA control signals via an analog signal output module and applied to the grinding motor inverter, hydraulic servo valve, and precision feeder, respectively.

[0034] During the next 8 minutes of operation, the state comparison unit continuously monitored and found that the particle size deviation had decreased to +15 micrometers, but the energy consumption deviation still existed (+5%), and the span deviation did not improve significantly (+0.15). Faced with this complex situation of multiple conflicting objectives, the parameter adjustment unit intelligently triggered the embedded adaptive perturbation exploration optimization strategy. The strategy first called the preset rule set for analysis. One relevant rule in the rule set pointed out: "When particle size and energy consumption objectives conflict and uniformity is poor, priority should be given to exploring the coordinated adjustment of feeding speed and grinding pressure." Based on this, the strategy determined that the parameters to be adjusted were feeding speed and grinding pressure, and the adjustment direction was to appropriately reduce the feeding speed and fine-tune the pressure. The strategy dynamically set the adjustment range. The range calculator referred to the effective adjustment range under similar working conditions in historical data, and initially set the feeding speed adjustment range to -4 g / min and the pressure adjustment range to ±0.05 MPa. The strategy generated the first set of candidate parameter combinations: feeding speed 43 g / min (original value 47 g / min), pressure 0.73 MPa (increased by 0.05 MPa). The combination was implemented within a short test cycle (90 seconds), with the data acquisition unit simultaneously collecting new state data. The state comparison unit's evaluation found that the particle size deviation was further reduced to +8 micrometers and the energy consumption deviation was reduced to +3%, but the span deviation slightly increased to +0.18.

[0035] The evaluation results were fed back to the strategy decision-making logic. The logic determined that while particle size and energy consumption targets had improved, uniformity had deteriorated. Therefore, it was decided to maintain the feeding speed adjustment direction in the next round of exploration, but reverse the pressure adjustment direction. The strategy generated a second set of candidate parameters: feeding speed 43 g / min, pressure 0.63 MPa (reduced by 0.05 MPa). After a test cycle evaluation, the particle size deviation was +10 micrometers, the energy consumption deviation was +2%, and the span deviation improved by +0.12. The strategy continued to iterate, generating a third set of candidate parameters: feeding speed maintained at 43 g / min, pressure finely adjusted to 0.65 MPa. Under this combination, the indicators reached a relative balance: particle size deviation +5 micrometers, energy consumption deviation +1.5%, and span deviation +0.09. After multiple rounds of iterative exploration, the strategy determined that the current parameter combination (rotation speed 1080 RPM, pressure 0.65 MPa, feeding speed 43 g / min) had brought the overall deviation below the preset threshold, so the optimization cycle was terminated and this parameter combination was fixed and output. Simultaneously, the system coordination and update module evaluates the data from the entire process and instructs the corresponding modules to maintain or control the grinding disc drive motor 4, the grinding pressure regulating hydraulic cylinder, and the feeding speed control motor of the vertical grinder. It fine-tunes the rotational speed and pressure parameters of the grinding area of ​​the vertical grinder, and the dynamic execution adjustment module continues to operate with these parameters until the grinding process is successfully completed. During this period, the status comparison unit continuously monitors and performs necessary fine-tuning to ensure that the final powder quality closely approximates the preset target. The entire adjustment process demonstrates the module's adaptability in handling multi-variable, multi-objective optimization problems.

[0036] While the dynamic execution adjustment module solidifies and maintains the parameter combination (1080 RPM rotation speed, 0.65 MPa pressure, and 43 g / min feeding speed) during the grinding process, the system coordination and update module synchronously collects the original characteristic data of Astragalus membranaceus, such as hardness, moisture content, fiber toughness, and density, collected by the medicinal material characteristic sensing module. The intelligent calculation and analysis module generates initial parameters and parameter records during the iterative optimization process. The dynamic execution adjustment module provides real-time grinding status feedback data (such as torque, acoustic emission, and vibration signal data) and parameter adjustment logs. The multi-modal optimization control module monitors quality data such as powder particle size and uniformity, as well as deviation analysis results. These full-process operation data and system performance indicators (such as parameter adjustment response speed and powder quality compliance) are comprehensively evaluated. The evaluation found that there is still room for optimization in the residence time of the material in the grinding zone at the current feeding speed, and the motor load is slightly high when the grinding disc rotation speed is maintained at 1080 RPM. Based on this, a system optimization command is generated. The instructions are sent to the corresponding modules: the dynamic execution adjustment module controls the feeding speed control motor, fine-tuning the feeding speed from 43g / min to 41g / min to better match the current grinding rhythm; simultaneously, it controls the grinding disc drive motor 4, fine-tuning its speed from 1080RPM to 1060RPM to reduce the motor's operating load; furthermore, the dynamic execution adjustment module controls the grinding pressure regulating hydraulic cylinder to maintain the current grinding pressure of 0.65MPa. After receiving the instructions, each module performs corresponding operations on the grinding disc drive motor, grinding pressure regulating hydraulic cylinder, and feeding speed control motor, ultimately achieving precise adjustment of the grinding speed, pressure, and feeding speed parameters in the grinding area of ​​the vertical grinder. This ensures that the grinding process maintains powder quality while better aligning with the equipment's operating characteristics.

[0037] Example 3: The parameter adjustment unit achieves advanced adaptive control of the grinding process through two precisely cooperating sub-units: the characteristic estimation unit and the parameter setting unit. The core function of this unit is to identify process characteristics online based on real-time feedback data and automatically adjust the controller parameters. The characteristic estimation unit continuously receives two input data streams. One stream comes from the internal memory, which contains a sequence of historical adjustment data for grinding speed, grinding pressure, and grinding time. This data is stored by timestamp and includes the magnitude, direction, and corresponding instantaneous effect of each parameter adjustment. The other data stream comes in real time from a sensor group installed on the grinder. This sensor group provides rich real-time grinding status feedback data, including multi-dimensional information such as motor current, vibration spectrum, cavity temperature, and acoustic emission signals. The characteristic estimation unit uses an improved recursive least squares algorithm to process the input data online. This algorithm can gradually correct its internal state estimates without storing all historical data, thereby achieving continuous online estimation and updating of key dynamic characteristic parameters of the grinding process. These key dynamic characteristic parameters are not directly measurable physical quantities, but rather abstract features that profoundly reflect the instantaneous behavior of the grinding system, such as the grindability coefficient of the material, the system energy conversion efficiency, and the wear state factor. Together, they constitute an internal model describing the dynamic response of the system.

[0038] The mathematical expression of this internal model aims to capture the dynamic relationship between operating parameters (such as rotational speed and pressure) and state variables (such as particle size and energy consumption). A simplified relationship describing the instantaneous change in grinding efficiency can be expressed as: ; in: Represents time slice The grinding efficiency coefficient within is a dimensionless indicator value; It is the first The weighting factors for each material property (such as hardness and fiber toughness) are obtained by fitting historical data; and These represent time slices. Normalized variation of internal grinding pressure and rotation speed relative to their reference values; It is the wear state compensation factor of the current system, which is a parameter that changes slowly over time; This is the length of the observation time window. This efficiency coefficient, along with other estimated characteristic parameters, provides a quantitative basis for controller parameter tuning. To further quantify the dynamic interaction between operating parameters and state variables, the internal model introduces a differential equation based on mass balance. The particle size change rate is related to the current particle size, rotational speed, and pressure, and its expression is: , in, The median particle size of the powder at time t (unit: μm). The grinding pressure at time t (unit: MPa) The rotational speed of the grinding disc at time t (unit: RPM). The energy consumption at time t (unit: W). Pressure-speed coupling coefficient (unit: μm) -1 •MPa -1 •RPM -1 ・s -1 ), Energy consumption impact coefficient (unit: μm·W) -1 ・s -1 ).

[0039] Meanwhile, the dynamic energy consumption can be described in a combined manner as follows: , in, Equipment characteristic constants (unit: W·MPa) -1 •RPM -1 This was determined through calibration experiments. Equipment characteristic constants (unit: W·μm) -1 This was determined through calibration experiments. This supplementary model provides a quantitative linkage between rotational speed, pressure, particle size, and energy consumption, laying the physical foundation for the calculation of the parameter adjustment unit.

[0040] The internal model constructs a dynamic linkage mechanism by quantifying the causal relationship between operating parameters and state variables. Specifically, the dynamic linkage between rotation speed and particle size is as follows: When parameters such as grinding pressure and feeding speed are fixed, increasing the rotation speed will simultaneously increase the number of impacts and shearing speed experienced by the material in the grinding chamber. For brittle medicinal materials, for every reasonable range of rotation speed increase, the particle crushing rate will increase accordingly, and the particle size refinement efficiency will improve. However, when the rotation speed exceeds the critical value, the material will adhere to the grinding chamber wall due to excessive centrifugal force, the effective grinding contact area will decrease, the particle size refinement rate will tend to level off or even decrease, and at the same time, energy consumption will increase sharply due to the increased motor load.

[0041] The dynamic linkage between pressure and particle size: Under the premise of appropriate rotation speed, a reasonable increase in grinding pressure can enhance the squeezing and grinding effect of the grinding media on the material, especially suitable for medicinal materials with high hardness and dense fibrous structure. Increasing the pressure will make the material particles easier to crush and shear, and the median particle size will show a regular decrease; however, when the pressure exceeds the material's tolerance limit, it will cause the material to agglomerate and the grinding chamber to be blocked. Not only will it fail to reduce the particle size, but it will also cause a surge in energy consumption and aggravate equipment wear.

[0042] The dynamic linkage between rotational speed, pressure, and energy consumption directly responds to changes in effective grinding work, which is determined by both rotational speed and pressure. When material characteristics are stable, energy consumption is positively correlated with rotational speed, as increasing rotational speed leads to increased motor output power. It is also positively correlated with pressure, as increasing pressure leads to increased grinding resistance and increased motor load. Furthermore, there is a synergistic effect between the two. When rotational speed and pressure are within a matching range, such as using low pressure and medium-high rotational speed for soft medicinal materials and high pressure and medium-low rotational speed for hard medicinal materials, the proportion of effective grinding work is high, resulting in high energy efficiency and energy consumption required for particle size refinement. If the parameters are mismatched, such as using high pressure and high rotational speed for soft medicinal materials, ineffective energy consumption, such as friction loss and material idling consumption, increases, leading to decreased energy efficiency.

[0043] When operating parameters are dynamically linked to particle size uniformity, excessively high rotational speed or pressure can lead to a decrease in particle size uniformity. At excessively high rotational speed, large particles are difficult to grind sufficiently due to their high inertia, while small particles may be over-ground or even agglomerate. At excessively high pressure, some materials are excessively compressed and clump together, while others are not effectively ground due to insufficient contact, ultimately resulting in an increased particle size distribution. However, by dynamically adjusting the matching relationship between rotational speed and pressure, the material can achieve a uniform grinding intensity within the grinding chamber, thereby improving particle size uniformity.

[0044] The characteristic estimation unit establishes a mapping relationship between key dynamic characteristic parameters and operating parameters and state variables through the fusion analysis of historical adjustment data and real-time feedback data. For example, when the sensor detects that the particle size refinement rate has slowed down and the motor current has not decreased, the system will determine that the grindability coefficient has increased or the wear state factor has increased. Then, the parameter adjustment unit will optimize the rotation speed or pressure to compensate for the decrease in grinding efficiency caused by the change in characteristic parameters.

[0045] The parameter setting unit periodically receives the latest estimated array of key dynamic characteristic parameters from the characteristic estimation unit. These parameter arrays clearly depict the dynamic behavior of the grinding system at the current moment. The parameter setting unit embeds a set of preset proportional-integral-derivative (PID) parameter tuning rules. These rules are based on an expert knowledge base summarized by control engineers from extensive simulation and field debugging experience. They map different dynamic characteristic parameter patterns to recommended combinations of PID gain parameters. For example, when the estimated system response delay is large and the energy conversion efficiency is low, the tuning rules recommend reducing the proportional gain and appropriately increasing the integral time to prevent overshoot oscillations. Conversely, if the system response is agile and the damping is good, the rules suggest increasing the proportional gain to improve the response speed. The tuning rules are not a simple table lookup, but an intelligent mapping process involving fuzzy inference and interpolation calculation. It can handle continuously changing characteristic parameters and achieve smooth switching of PID parameters.

[0046] The automatic tuning process is executed at a fixed cycle (e.g., once per second). The parameter setting unit calculates a new set of proportional gains based on the latest received characteristic parameters and according to the tuning rules. Integral Time and differential time The new parameters are then downloaded to the underlying PID control algorithm, which independently controls multiple closed-loop circuits such as grinding speed, grinding pressure, and feeding speed. Each circuit has its own PID instance, but its parameters are uniformly and coordinated by the parameter setting unit to ensure the synergy of the multivariable system control. Simultaneously, in the multimodal optimization control module, the acquisition of the actual quality characteristics of the powder is carried out concurrently. This process is accomplished through a real-time monitoring system integrating a high-resolution industrial camera and a high-speed image processor. The camera continuously captures image data of splashed or settled medicinal powder in each grinding area at an extremely high frame rate. The captured raw images are immediately sent to the image processor for pipelined processing. The processing algorithm first performs image enhancement and noise filtering to improve the signal-to-noise ratio, then uses an adaptive threshold segmentation algorithm to separate powder particles from the complex background, and finally identifies the contours of tens of thousands of individual particles through a connected component labeling algorithm. Based on the identified particle contours, the algorithm calculates the equivalent circle diameter, aspect ratio, and other morphological parameters for each particle, and statistically analyzes these parameters across all particles to form a global histogram. From these histograms, the algorithm further extracts two key parameters for quantitatively describing powder quality: powder particle size distribution parameters (usually expressed as D...). 50 D 90 The calculated actual quality parameters are represented by percentiles and uniformity parameters (usually expressed as the standard deviation of the granularity distribution or the span coefficient). These calculated actual quality parameters are packaged in real time and transmitted as feedback signals to the adaptive decision-making algorithm, where they are compared with the preset target quality parameters, thus initiating a new round of optimization.

[0047] Example 4: Within the operational framework of the multimodal optimization control module, the adaptive decision algorithm undertakes the core task of dynamically optimizing the grinding zone control parameters based on real-time quality deviations. The execution of this algorithm relies on a continuously running closed-loop feedback and optimization architecture. Taking the processing of a typical Chinese medicinal herb, Angelica sinensis, as an example, the system precisely controls its powder quality during the grinding process. As a rhizome-type medicinal herb, Angelica sinensis has material properties between fibrous and brittle. The target powder quality is typically set as a particle size distribution D. 50The particle size distribution should be between 80 and 120 micrometers, with a range of less than 1.8. To address the imaging challenges caused by high dust levels and high-speed material movement within the vertical grinding mill cavity, a targeted real-time monitoring system is designed. The camera lens employs a double-layer protective structure: an outer layer of wear-resistant sapphire glass and an inner layer with a high-pressure gas purging channel. A micro-pump continuously outputs dry, clean nitrogen gas (pressure 0.1~0.15MPa) along the tangential direction of the lens edge, forming an air curtain to isolate dust adhesion. Simultaneously, the lens surface is coated with an anti-fouling and hydrophobic coating to reduce dust adhesion, achieving double protection in conjunction with the air curtain. A ring-shaped coaxial LED light source is used, with a wavelength selected in the 550~600nm visible light band. This band exhibits significant differences in reflectivity for medicinal herb powders and has strong resistance to dust scattering interference. The light source brightness is controlled using closed-loop feedback, with a light intensity sensor monitoring the reflected light intensity inside the cavity in real time and dynamically adjusting the LED drive current (adjustment range...). The illumination intensity is kept stable at 800-1200 lux (50-200mA) to avoid image noise caused by light fluctuations due to material movement. A global shutter industrial camera (frame rate ≥200fps) is selected, paired with an f=16mm fixed-focus lens. The focal length is adjusted to the core motion trajectory of the grinding area (15-20cm from the lens) to ensure that the splashed powder is within the clear imaging range. The camera exposure time is set to 10-20μs to shorten the exposure time to freeze the high-speed moving powder particles (maximum movement speed ≤5m / s) and avoid motion blur. At the same time, an image enhancement algorithm is used to perform multi-frame superposition noise reduction on the acquired original image (3-5 frames superimposed) to improve the image signal-to-noise ratio and ensure that the particle outline is clearly distinguishable.

[0048] The real-time monitoring system continuously captures high-resolution images of splashed powder using a linear array camera mounted above each grinding zone. The image data is transmitted to an embedded image processing unit at a rate of 200 frames per second. The processing unit uses morphological operations and a watershed algorithm to segment adhered particles and calculates the equivalent diameter of each particle based on the calibrated pixel dimensions. Finally, it statistically generates the powder particle size distribution parameters (D) for the current region. 10 D 50 D 90 ) and uniformity parameter (span coefficient = (D 90 -D 10 ) / D 50 ).

[0049] The adaptive decision-making algorithm receives the calculated current powder quality characteristic data from the real-time monitoring system and compares it with the preset target quality characteristics; the comparison process generates a multi-dimensional deviation vector, such as D measured in the current area. 50 If the value is 150 micrometers and the target is 100 micrometers, then D 50With a deviation of +50 micrometers, a current span coefficient of 2.1, and a target of 1.8, the uniformity deviation is +0.3. The algorithm integrates a grinding process response model, a simplified mathematical model trained on extensive experimental data. This model describes the effect of changes in grinding operation parameters (such as zone grinding disc speed and pressure) on powder quality indicators (such as D). 50 The influence relationship between the span coefficient and the distance coefficient is essentially a set of approximate transfer functions. For materials like Angelica sinensis, the model might indicate that increasing the grinding pressure in this region can usually effectively reduce D. 50 A higher value may worsen uniformity, while moderately increasing the rotational speed may improve uniformity, but this has less impact on D. 50 The impact is relatively small. Based on the quantized deviation vector and response model, the algorithm initiates its optimization calculation core; this core is essentially a real-time optimizer whose objective function is set to minimize the magnitude of the current quality deviation vector, and the decision variables are the grinding area operation parameters to be adjusted (such as rotation speed and pressure). The optimizer linearizes the response model at the current operating point and uses numerical methods such as gradient descent or the simplex method to calculate a set of operation parameter corrections that can most effectively reduce the deviation. For example, for the aforementioned D... 50 In cases where the value is too high and the uniformity is off, model calculations may indicate that simply increasing the pressure can reduce D. 50 However, this will further worsen the uniformity, therefore a multi-parameter coordinated adjustment scheme must be generated: moderately increasing the pressure to reduce D. 50 At the same time, the rotation speed is slightly increased to compensate for uniformity.

[0050] To quantify the impact of operating parameters on powder quality, a multivariate linear regression model was used for the response. For any grinding region, D 50 The relationship between the pressure adjustment amount and the pressure adjustment amount is as follows: , The relationship between the change in uniformity (represented by the span coefficient S) and the speed adjustment is as follows: , in, This represents the change in D50 value (unit: μm), with negative values ​​indicating a decrease in particle size. This represents the change in the span coefficient (dimensionless); a negative value indicates an improvement in uniformity. This refers to the grinding pressure adjustment amount (unit: MPa). This refers to the adjustment amount of the grinding disc speed (unit: RPM). Response coefficient (unit: μm·MPa) -1 ), Response coefficient (unit: μm·RPM) -1 ), Response coefficient (unit: MPa)-1 ), The response coefficient (RPM) -1 The response coefficients are obtained by fitting experimental data, and the coefficients vary for different types of medicinal materials.

[0051] The calculated and generated corrected control parameters are immediately encapsulated into control commands and sent to the control actuators in the corresponding grinding area via an industrial network bus. These actuators typically include servo drives (controlling the grinding disc speed), electro-hydraulic proportional valves (controlling the grinding pressure), and frequency converters (controlling the feeding speed).

[0052] The actuator precisely adjusts its output according to instructions, thereby changing the grinding mechanical environment of the area. To clearly illustrate a typical optimization decision-making process, refer to Table 1, which records the parameter changes and quality response of a virtual grinding area over several consecutive optimization cycles.

[0053] Table 1: Optimization process of grinding zone control parameters

[0054] The entire optimization process iterates at a high frequency (e.g., several times per second). Within each control cycle, the algorithm re-acquires the latest powder quality data, calculates the new deviations, and reruns the optimization calculation to generate the latest correction parameters. This high-frequency iterative mechanism enables the system to quickly respond to disturbances caused by batch differences in materials, slight wear, or environmental fluctuations during the grinding process, consistently maintaining powder quality that steadily approaches the target value. For multiple parallel grinding zones, each zone runs an independent instance of the adaptive decision-making algorithm, but they share the same global optimization objective, thus achieving fine-grained local control while ensuring the consistency of overall powder output.

[0055] Example 5: Systematically analyze the operational status and interaction relationships of each submodule to maintain and improve overall performance. Taking a complete production batch of Angelica sinensis grinding as an example, this module begins its workflow; the herb characteristic sensing module records the initial physical and texture characteristics of this batch of Angelica sinensis (such as hardness value, moisture content, and fiber toughness value); the intelligent calculation and analysis module generates the corresponding initial grinding control parameter set; the dynamic execution adjustment module records all adjustment logs of parameters such as rotation speed, pressure, and time during the entire grinding process and their corresponding real-time sensor feedback; and the multimodal optimization control module provides continuous monitoring data of powder quality and historical deviations from target values. All these heterogeneous, timestamped, end-to-end operational data and performance indicators are continuously transmitted to the data processing unit of the system coordination and update module via an internal high-speed data bus.

[0056] The data processing unit does not simply store this massive amount of data; instead, it immediately activates its multi-source data fusion model for processing. This model is a complex, hierarchical data processing pipeline that first cleans and aligns the input data, eliminating time asynchrony issues between sensor data from different sampling rates, and mapping all data onto the same time axis. Subsequently, the model extracts features from multiple dimensions: calculating equipment health indicators such as average load rate, peak power, and abnormal vibration frequency from equipment operation logs; calculating control stability indicators such as adjustment frequency and amplitude variance from control parameter adjustment sequences; and calculating quality consistency indicators such as average deviation and duration of exceeding limits from powder quality deviation history. These extracted feature vectors are fed into an analysis engine based on multivariate statistical process control. This engine outputs a set of highly condensed system-level evaluation indicators—system risk assessment indicators and optimization gain indicators—by calculating the covariance matrix and principal components of the feature vectors and comparing them with control limits. The system risk assessment index quantifies the overall degree to which the current operating state deviates from the normal baseline. The higher the value, the greater the risk of system failure or performance degradation. The optimization gain index, on the other hand, estimates the potential performance improvement that may be achieved through parameter retuning or strategy adjustment.

[0057] The instruction generation unit receives system risk assessment indicators and optimization gain indicators from the data processing unit. While these indicators quantify the current state, they do not specify concrete optimization directions or measures. The core task of the instruction generation unit is to transform this state assessment into executable operational instructions. This unit utilizes its built-in intelligent decision-making model for processing. This model is a hybrid system integrating case-based reasoning (CBR) and a rule engine. The case-based reasoning library stores a large number of historical optimization cases. Each case records similar system state indicators (i.e., risk assessment and gain indicators) at a certain point in the past, the optimization measures taken at that time (such as adjusting knowledge base weights, modifying control algorithm parameters, and suggesting maintenance operations), and the final effect evaluation after the measures were implemented. The intelligent decision-making model first performs a similarity search between the current indicators and the historical case library, finding the K most similar historical state scenarios and their corresponding strategies.

[0058] The rules engine then takes over, containing a set of meta-rules defined by domain experts to validate and adjust the applicability of retrieved historical strategies. For example, a meta-rule might state: "If the current equipment wear index exceeds the threshold T, any aggressive parameter tuning strategies suggested in historical cases should be rejected, and preventative maintenance instructions should be prioritized." The model modifies and weights the retrieved strategies, ultimately generating a set of customized system optimization instructions for the current system state. These instructions might include: the dynamic execution adjustment module adjusting the rotational speed of a specific grinding zone in a vertical grinder, or the multimodal optimization control module correcting the pressure in a specific area. These instructions are macro-level and strategic, and might include: suggesting that the dynamic execution adjustment module relax the integral gain of its PID controller to reduce oscillations under specific material conditions; prompting the intelligent calculation and analysis module to update the weights of fibrous medicinal materials in its integrated learning model online; or even generating a warning message suggesting that the operator calibrate the grinding disc gap before the next batch of production.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A traditional Chinese medicine grinding control system, characterized in that, The application relates to a vertical grinder control system for controlling a vertical grinder to perform a grinding operation, wherein the vertical grinder comprises a mounting frame (1), a grinding mechanism (2), a feeding hopper (3) and a grinding disc driving motor (4); The traditional Chinese medicine grinding control system comprises: A medicinal material characteristic sensing module is used for collecting physical characteristic data and texture characteristic data of the to-be-processed traditional Chinese medicinal materials; An intelligent computing and analyzing module is used for receiving the physical characteristic data and the texture characteristic data, fusing a preset medicinal material processing knowledge base and historical grinding process data, and generating initial grinding control parameters; A dynamic execution adjusting module is used for receiving the initial grinding control parameters, combining real-time collected grinding state data, and dynamically adjusting multiple operation parameters of the grinding mechanism; A multi-modal optimization control module is used for receiving a medicinal material powder sample after adjusted grinding and target quality parameters thereof, acquiring actual quality characteristics of the powder through a real-time monitoring system, judging a deviation between the actual quality characteristics and the target quality parameters by using a self-adaptive decision algorithm, and dynamically optimizing regulation and control parameters of a grinding area; A system coordination updating module is used for collecting whole-process operation data and performance indexes of the medicinal material characteristic sensing module, the intelligent computing and analyzing module, the dynamic execution adjusting module and the multi-modal optimization control module, evaluating the whole-process operation data and the performance indexes, generating system optimization instructions, and adjusting rotating speed, pressure and feeding speed parameters of a grinding area of the vertical grinder by controlling the grinding disc driving motor (4), a grinding pressure adjusting hydraulic cylinder and a feeding speed control motor of the vertical grinder.

2. The TCM grinding control system of claim 1, wherein, The intelligent computing and analyzing module specifically comprises: A parameter generating unit is used for processing the physical characteristic data, the texture characteristic data, the medicinal material processing knowledge base and the historical grinding process data by using an ensemble learning model, and outputting preliminary grinding control parameters; A parameter optimization unit is used for judging the preliminary grinding control parameters according to a grinding effect optimization target and equipment operation constraints, outputting final grinding control parameters to the dynamic execution adjusting module if the preliminary grinding control parameters meet the requirements, or iteratively outputting the final grinding control parameters after adopting a self-adaptive disturbance exploration optimization strategy.

3. The TCM mill control system of claim 2, wherein, The self-adaptive disturbance exploration optimization strategy specifically comprises: Intelligently selecting grinding control parameters to be adjusted and adjustment directions based on a preset rule set; Dynamically and adaptively setting adjustment amplitudes of the grinding control parameters, and iteratively generating candidate parameter combinations; Verifying and evaluating the candidate parameter combinations, and updating an optimal parameter set until a termination condition is met.

4. The TCM mill control system of claim 1, wherein, The dynamic execution adjusting module specifically comprises: A data acquisition unit is used for continuously acquiring and analyzing feedback data reflecting current grinding material states from a sensor group; A state comparison unit is used for comparing the feedback data with expected grinding target states received from the intelligent computing and analyzing module in real time, and determining a deviation between an actual grinding state and the expected grinding target state; A parameter adjusting unit is used for calculating adjustment amounts required for grinding rotating speed, grinding pressure and grinding time length in real time according to the deviation, and applying the adjustment amounts to corresponding grinding execution mechanisms.

5. The TCM mill control system of claim 4, wherein, The parameter adjusting unit specifically comprises: a characteristic estimation unit configured to analyze historical adjustment data of the grinding rotation speed, the grinding pressure and the grinding time length and corresponding real-time grinding state feedback data provided by the sensor group, and continuously estimate and update key dynamic characteristic parameters of the grinding process on-line using a recursive estimation algorithm; a parameter setting unit configured to automatically and periodically calculate and set gain parameters of a proportional-integral-derivative control algorithm based on the key dynamic characteristic parameters and according to a preset proportional-integral-derivative parameter setting rule.

6. The TCM mill control system of claim 1, wherein, The process of acquiring the actual quality characteristics of the powder by the real-time monitoring system in the multi-modal optimization control module includes: During the grinding process, image data of the medicinal material powder in each grinding area is captured in real time and processed; A key observation area is located and the powder particle size distribution parameters and uniformity parameters are calculated therefrom as the actual quality characteristic feedback to the adaptive decision algorithm.

7. The TCM mill control system of claim 6, wherein, The specific processing flow of dynamically optimizing the grinding area control parameters by the adaptive decision algorithm in the multi-modal optimization control module is as follows: For each grinding area, the adaptive decision algorithm calculates and generates a set of modified control parameters aiming to minimize the deviation quantification result between the current powder quality characteristics of the area monitored by the machine vision in real time and the preset target powder quality characteristics, in combination with a grinding process response model; The modified control parameters are applied to the control actuators of the corresponding grinding area.

8. The TCM mill control system of claim 1, wherein, The system coordination updating module specifically includes: a data processing unit configured to process the whole-process operation data and the performance indicators using a multi-source data fusion model, and output system risk assessment indicators and optimization gain indicators; an instruction generation unit configured to process the system risk assessment indicators and the optimization gain indicators using an intelligent decision model in combination with historical optimization cases, and generate the system optimization instructions.

9. The TCM mill control system of claim 1, wherein, The physical characteristic data collected in the medicinal material characteristic sensing module at least includes medicinal material hardness values and moisture content values, and the texture characteristic data at least includes fiber toughness values and density values.

10. The TCM mill control system of claim 1, wherein, The multiple operation parameters adjusted in the dynamic execution adjustment module at least include grinding disc rotation speed parameters, grinding pressure parameters, feeding speed parameters and grinding time length parameters. a characteristic estimation unit configured to analyze historical adjustment data of the grinding rotation speed, the grinding pressure and the grinding time length and corresponding real-time grinding state feedback data provided by the sensor group, and continuously estimate and update key dynamic characteristic parameters of the grinding process on-line using a recursive estimation algorithm; a parameter setting unit configured to automatically and periodically calculate and set gain parameters of a proportional-integral-derivative control algorithm based on the key dynamic characteristic parameters and according to a preset proportional-integral-derivative parameter setting rule.

Images