High-efficiency screening method for medicinal liquor raw materials

Abstract

The present application relates to the technical field of medicinal liquor raw material screening, and discloses a high-efficiency medicinal liquor raw material screening method. The method comprises collecting raw material basic parameters of a plurality of candidate medicinal materials, extracting multi-modal feature data to identify effective component distribution states, and generating an effective feature spectrum; calling an active component identification area in the efficacy feature spectrum, obtaining an active component concentration interval and a medicinal property action time length threshold corresponding to a target efficacy direction, detecting matching differences, and generating an efficacy deviation feature set; according to medicinal material numbers associated with the efficacy deviation feature set, extracting component release curves and efficacy peak time points in continuous batches, calculating stability gaps with a preset baseline, screening substandard medicinal materials, and generating a medicinal material proportion dynamic adjustment table; calling baseline proportion values in the medicinal material proportion dynamic adjustment table, monitoring the fusion state of effective components in real-time preparation, marking the starting time nodes of the synergistic reactions of adjacent three components, recording the synergistic activation state, and generating a raw material combination optimization instruction.

Description

Technical Field

[0001] This invention relates to the field of medicinal wine raw material screening technology, specifically to a method for efficient screening of medicinal wine raw materials. Background Technology

[0002] In the traditional field of medicinal wine preparation, raw material selection has always been a core step determining the efficacy and quality of the product. However, the process has long faced technical bottlenecks such as low efficiency and insufficient precision. Currently, the industry generally relies on empirical judgment for the selection of medicinal materials, that is, inferring the content of effective components and applicability by manually observing the appearance, odor characteristics, or simple physicochemical indicators of the medicinal materials. This method is not only time-consuming and labor-intensive, but also makes it difficult to quantify the internal component distribution of the medicinal materials. It is also highly susceptible to fluctuations in raw material quality due to subjective judgment bias, which in turn affects the stability of the clinical efficacy of the medicinal wine.

[0003] With the advancement of the modern TCM industrialization, the preparation of medicinal wines is gradually developing towards large-scale and standardized production, highlighting the limitations of traditional screening methods. On the one hand, the effective components of natural medicinal materials are affected by multiple factors such as growth environment, harvesting time, and storage conditions, exhibiting significant individual differences. Single-dimensional parameter detection alone cannot fully reflect their efficacy potential. For example, the distribution density of astragaloside A in the same batch of astragalus may fluctuate by more than 20% due to differences in heavy metal content in the soil of the producing area, and traditional methods struggle to capture such microscopic differences. On the other hand, the efficacy of medicinal wines often depends on the synergistic effect of multiple medicinal components. Existing screening technologies lack the ability to dynamically monitor the interactions between multiple components, often resulting in products failing to meet expectations due to individual components meeting standards but insufficient synergistic effects.

[0004] In industrialized production, the efficiency and stability of medicinal herb screening directly affect production costs and market competitiveness. Current mainstream screening processes involve multiple stages, including raw material cleaning, pre-analysis of components, and small-scale compatibility verification. The entire process typically takes over 72 hours, and the screening standards for different batches are difficult to standardize, resulting in a low raw material utilization rate of only 60%-70%. Furthermore, existing technologies cannot track the release patterns of components during the preparation process in real time. When the component release curve deviates from the preset benchmark, the problem is often only discovered during the finished product testing stage, by which time a significant amount of raw materials and energy have already been wasted.

[0005] From the perspective of industry technology development, although component detection methods based on technologies such as near-infrared spectroscopy and high-performance liquid chromatography have emerged in recent years, these methods mostly focus on the quantitative analysis of single components and lack the ability to integrate multimodal characteristic data. For example, near-infrared spectroscopy can quickly identify the macroscopic component distribution of medicinal materials, but it cannot correlate it with the peak efficacy time point in the preparation of medicinal wine; while chromatographic analysis can accurately determine the concentration of a single component, it is difficult to assess the synergistic reaction initiation time of multiple components. This technological gap has kept the screening of medicinal materials at the "static detection" level, unable to meet the complex needs of "dynamic fusion" in the preparation of medicinal wine.

[0006] As consumers increasingly demand higher efficacy and safety standards for medicinal wines, and with the stringent regulations on product traceability in the internationalization of traditional Chinese medicine, traditional screening methods can no longer meet the technical standards of modern industry. Industry statistics show that the incidence of efficacy fluctuations in medicinal wines due to improper raw material selection is as high as 35%, and related quality complaints account for over 40%, severely hindering the high-quality development of the traditional Chinese medicine industry. Therefore, developing a technical method that integrates multimodal feature analysis, dynamic monitoring of component synergistic effects, and achieves efficient and accurate screening has become a critical issue that urgently needs to be addressed in the field of medicinal wine preparation. Summary of the Invention

[0007] The purpose of this invention is to provide an efficient screening method for medicinal wine raw materials to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides a method for efficient screening of medicinal wine raw materials, the method comprising:

[0009] The basic parameters of raw materials of various candidate medicinal materials are collected, multimodal feature data in the basic parameters of raw materials are extracted, the distribution status of effective components of medicinal materials is identified based on the multimodal feature data, and a medicinal material efficacy feature map is generated.

[0010] By calling the active ingredient identification area in the medicinal herb efficacy feature map, the concentration range of active ingredients pointing to the target efficacy and the threshold of the duration of medicinal effect are obtained, the matching difference between the actual component concentration value and the theoretical duration of effect is detected, and an efficacy deviation feature set is generated.

[0011] Based on the herb numbers associated with the efficacy deviation feature set, extract the component release curves and efficacy peak time points of the herb in continuous preparation batches, calculate the stability gap between the component release curve and the preset efficacy benchmark, determine whether the component release has not met the standard, screen the stability gaps corresponding to the non-compliant herbs as the basis for optimization priority, and generate a dynamic adjustment table of herb ratio.

[0012] The system calls upon the baseline ratio values ​​in the dynamic adjustment table of medicinal material ratios, monitors the fusion status of effective components during real-time preparation, marks the start time nodes of the synergistic reaction of three adjacent components, and records the synergistic activation status and generates raw material combination optimization instructions if all synergistic reaction start time nodes are earlier than the preset fusion stage switching point.

[0013] Preferably, the basic parameters of the raw materials include microstructure image sequences, chemical composition spectral data, and bioactive marker values; the efficacy characteristic map of the medicinal materials includes a thermal map of the spatial distribution of active ingredients, a time axis of efficacy action, and a component interaction correlation matrix; the efficacy deviation feature set includes a target efficacy deviation code, actual duration of action deviation, and component concentration fluctuation anomaly identifier; the dynamic adjustment table of the medicinal material ratio includes the priority weight of the medicinal material compatibility, the adjustment range of the benchmark ratio, and the dynamic adjustment effective period; and the raw material combination optimization instruction includes a synergistic reaction trigger identifier, the component fusion start time, and the reaction process compression detection status.

[0014] Preferably, the steps for generating the medicinal herb efficacy feature map are as follows: acquiring multimodal feature data of multiple candidate medicinal herbs during the preprocessing stage; extracting the distribution density change and action time series data of the effective components of each medicinal herb; calculating the spatial offset of component distribution and the action time delay period based on the correspondence between the distribution density change and the action time series; generating a set of component distribution and time-effect parameters; based on the component distribution spatial offset and action time delay period in the set of component distribution and time-effect parameters, combined with the temperature change information of the medicinal herbs at continuous time nodes during the extraction process, counting the number of activity decays of each medicinal herb in the distribution offset and time delay stages, identifying the correlation distribution pattern between the number of activity decays and the time delay period, and generating medicinal herb activity retention rate features; based on the medicinal herb activity retention rate features, determining medicinal herb samples with an activity retention rate lower than a preset efficacy maintenance threshold, binding the identification code of the corresponding sample with the preparation batch number, and generating a medicinal herb efficacy feature map.

[0015] Preferably, the steps for generating the efficacy deviation feature set are as follows: First, the active ingredient identification regions marked in the medicinal herb efficacy feature map are called, and three parameters are obtained for the corresponding identification regions under the target efficacy direction: the theoretical active ingredient concentration range, the actual duration of action threshold, and the measured peak efficacy time. The difference between the actual duration of action and the theoretical duration of action, as well as the offset between the measured concentration value and the theoretical concentration range, are calculated to generate an efficacy deviation core parameter set. Second, based on the actual duration of action difference and concentration offset values ​​in the efficacy deviation core parameter set, the total amount and distribution density data of the active ingredients in the corresponding identification regions within the medicinal effect cycle are called, and the aggregation intensity and dispersion level of the effective ingredients are identified to obtain component distribution status information. Third, based on the component distribution status information and the time interval distribution of different spatial locations within the efficacy deviation core parameter set, the component distribution offset feature value is calculated, and the location range of the abnormal distribution area within the medicinal effect cycle is identified to generate a component distribution abnormality period. Fourth, based on the associated preparation batches within the component distribution abnormality period, the correlation between the time segment and component release behavior within the medicinal effect cycle is evaluated, and continuous preparation segments with abnormal component release behavior are screened and marked as efficacy deviation areas to generate an efficacy deviation feature set.

[0016] Preferably, the steps for generating the dynamic adjustment table of medicinal material ratios are as follows: Based on the medicinal material numbers associated with the efficacy deviation feature set, extract the lag duration of the peak efficacy time point and the time point of complete component release in two consecutive preparation batches of medicinal materials. Combine the time difference between the time point of complete component release and the theoretical end time point of the action cycle to obtain the lag information of the tail-end release of medicinal materials. Based on the lag information of the tail-end release of medicinal materials, determine whether the release lag amount exceeds the preset release completion benchmark value, screen the medicinal material numbers that have not met the release standard, extract the waiting time before the complete release of components in the corresponding medicinal materials, sort them according to the waiting time length of the non-compliant medicinal materials, and generate a medicinal material release priority sequence. Call the sorting results in the medicinal material release priority sequence, configure incremental fusion time periods for the medicinal materials in turn, adjust the medicinal material action duration within the total preparation cycle, record the medicinal material number and the adjusted action duration, and generate a dynamic adjustment table of medicinal material ratios.

[0017] Preferably, the specific steps for generating the raw material combination optimization instruction are as follows: calling the baseline ratio value recorded in the dynamic adjustment table of medicinal material ratio, detecting the fusion distance distribution of associated medicinal material groups during real-time preparation, marking the synergistic reaction start time nodes of the three adjacent active ingredients in the fusion stage, and generating a component synergistic start time set; based on the order of the synergistic reaction start time nodes of the three adjacent components in the component synergistic start time set, determining whether all time nodes are earlier than the fusion stage switching time point in the dynamic adjustment table of medicinal material ratio; if the determination condition is met, marking it as a synergistic activation state, associating the medicinal material group number and state parameters, and generating the raw material combination optimization instruction.

[0018] Preferably, the method further includes: calling the preparation cycle marked by the raw material combination optimization instruction, screening the release trajectory of the tail component in the fusion stage, comparing the trend of component activity change with the tail duration, and if the activity value continues to rise without entering the stable range, calculating the time required to reach the preset release endpoint and updating the tail control cycle to obtain the tail continuous effect regulation result; the tail continuous effect regulation result includes the remaining release path duration, the predicted period for complete release of tail components, and the suggested time window for fusion maintenance.

[0019] Preferably, the steps for generating the tail-end continuous action regulation result are as follows: The preparation cycle number marked in the raw material combination optimization instruction is called; the trajectory data of the tail-end component in the fusion stage under the corresponding cycle is screened; the continuous activity value sequence and time stamp data from the tail-end start time to the component reaching the release endpoint are extracted to generate the tail-end release trajectory sequence; based on the activity value sequence in the tail-end release trajectory sequence, the activity change trend of the tail-end component within the continuous action duration is analyzed; the activity increase at the end of the sequence is extracted and compared with the corresponding duration of the tail-end; if the activity value continues to rise and has not reached a stable threshold, the supplementary time required for the component to reach the preset release endpoint is calculated to generate the remaining tail-end release time interval; the supplementary time value required for the associated medicinal material group in the remaining tail-end release time interval is called; the real-time tail-end action control cycle is updated; the original tail-end action end time point is corrected; the control interval of the medicinal material group tail-end is reset to obtain the tail-end continuous action regulation result.

[0020] Preferably, the method further includes a dynamic optimization ratio feature generation step: based on the optimized ratio features output from the dynamic adjustment table of medicinal material ratios, extracting the light intensity fluctuation sequence and humidity change from the environmental factor data, and calculating the influence coefficient of environmental factors on the component release rate; performing environmental compensation calculation on the benchmark ratio value based on the influence coefficient to generate environmentally suitable ratio features; and fusing the environmentally suitable ratio features with the distribution data of active ingredients in the medicinal material efficacy feature map to generate environmentally compensated dynamic optimization ratio features.

[0021] Preferably, the application steps of the dynamic optimization ratio feature are as follows: calling the medicinal material compatibility parameters in the dynamic optimization ratio feature to construct a multi-batch preparation verification queue; collecting the matching degree data between the actual efficacy peak time point and the component release curve in the verification queue to generate a ratio verification result set; adjusting the priority weight parameters of the medicinal material ratio dynamic adjustment table according to the ratio verification result set to generate an optimized ratio dynamic adjustment table; and applying the optimized ratio dynamic adjustment table to the raw material screening process of subsequent batches.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] This efficient screening method for medicinal wine raw materials collects basic parameters of various candidate medicinal materials and extracts multimodal feature data. It can analyze the distribution of effective components from multiple dimensions, generating efficacy feature maps that provide a more comprehensive and accurate presentation of the efficacy characteristics of the medicinal materials. This multimodal feature-based analysis method overcomes the limitations of traditional single-parameter detection, allowing for a deeper exploration of the inherent efficacy potential of medicinal materials and providing richer reference information for raw material screening.

[0024] By utilizing the active ingredient identification regions in the efficacy characteristic map of medicinal materials, and combining them with the concentration range of active ingredients indicated by the target efficacy and the threshold of the duration of medicinal effect, the difference between actual and theoretical values ​​is detected, and an efficacy deviation feature set is generated. This allows for the precise identification of differences in efficacy among medicinal materials. In this way, medicinal materials that deviate from the target efficacy can be clearly identified, providing a clear direction for subsequent screening and adjustment, avoiding the misuse of ineffective raw materials, and reducing unnecessary cost investment.

[0025] Based on the herbal material numbers associated with the efficacy deviation feature set, the component release curves and efficacy peak time points of consecutive preparation batches are extracted. The stability gap between these curves and a preset baseline is calculated, and substandard herbs are screened. A dynamic adjustment table for herbal material ratios is generated, enabling dynamic tracking of herbal material performance during preparation. By analyzing the stability gaps to determine optimization priorities, adjustments to raw material ratios can be more targeted, resulting in more stable component release across different batches and reducing the impact of raw material fluctuations on the final product.

[0026] By calling upon the baseline ratio values ​​in the dynamic adjustment table of medicinal material proportions, monitoring the fusion state of effective components during real-time preparation, marking the synergistic reaction start time nodes of three adjacent components, determining whether all are earlier than the preset fusion stage switching point and recording the synergistic activation state, and generating raw material combination optimization instructions, this method can effectively capture the synergistic effect patterns between components. Through dynamic monitoring of the synergistic reaction start time, problems in the raw material combination can be identified in a timely manner, and the raw material combination method can be optimized to allow multiple medicinal material components to better exert synergistic effects during preparation, thereby improving the rationality and compatibility of the raw material combination. Attached Figure Description

[0027] Figure 1 This is a schematic diagram illustrating the working principle of the efficient screening method for medicinal wine raw materials described in this invention.

[0028] Figure 2 A flowchart of basic parameters and characteristic spectra of raw materials;

[0029] Figure 3 A flowchart for generating the efficacy bias feature set;

[0030] Figure 4 The flowchart is generated for the control results of the tail segment's continuous action. Detailed Implementation

[0031] 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.

[0032] Please see Figure 1 This invention provides a highly efficient screening method for medicinal wine raw materials. This method achieves precise screening and optimized formulation of effective components from medicinal materials through multimodal feature data acquisition and analysis. Specifically, it includes the following steps:

[0033] The system collects basic raw material parameters of various candidate medicinal materials, extracts multimodal feature data, identifies the distribution status of effective components, and generates a medicinal material efficacy feature map. Based on the active ingredient identification region, it detects the matching difference between the actual component concentration and the theoretical duration of action under the target efficacy, and generates an efficacy deviation feature set. According to the medicinal material number associated with the efficacy deviation feature set, it analyzes the component release curve and efficacy peak time point in consecutive preparation batches, calculates the stability gap with the preset efficacy benchmark, screens out non-compliant medicinal materials, and generates a dynamic adjustment table of medicinal material ratios. It calls the benchmark ratio value in the dynamic adjustment table, monitors the fusion status of effective components during real-time preparation, marks the synergistic reaction start time node, and generates a raw material combination optimization instruction if all nodes are earlier than the preset fusion stage switching point.

[0034] Example 1: See Figure 2 The acquisition of basic raw material parameters involves the collaborative application of multiple instruments and technologies. Microstructural image sequences were acquired using a combination of high-resolution optical and electron microscopy, performing layered scanning of the cellular and tissue structures of different medicinal materials to obtain multi-scale image data from macroscopic to microscopic levels. Uniform focal length and illumination conditions were set during image acquisition to avoid data deviations caused by environmental differences. Chemical composition spectral data were obtained using Fourier transform infrared spectroscopy coupled with liquid chromatography-mass spectrometry (LC-MS). The former was used for rapid screening of functional group distribution in medicinal materials, while the latter was used for precise determination of the molecular weight and structural characteristics of specific compounds. The determination of bioactivity marker values ​​relied on standardized in vitro experimental models. For example, enzyme-linked immunosorbent assay (ELISA) was used to determine the inhibitory ability of medicinal material extracts on specific inflammatory factors, or cell proliferation experiments were used to assess their regulatory effects on specific signaling pathways.

[0035] The construction of the medicinal herb efficacy characteristic map begins with the preprocessing of multimodal feature data. After noise reduction and enhancement, the microscopic image sequences are segmented using deep learning algorithms to identify the distribution regions of active ingredients, and the pixel density and spatial coordinates of each region are calculated. The spectral data of chemical components are converted into a standardized intensity matrix after baseline correction and peak area normalization. Bioactivity marker values ​​are statistically processed based on the number of experimental repetitions, and the mean value is taken after removing outliers as the final activity index. The calculation of distribution density changes is based on comparative analysis of time-series images, using a difference algorithm to identify the migration trajectory of active ingredients at different treatment stages. The action time-series data are obtained through dynamic monitoring using high-performance liquid chromatography, recording the release rate curves of the target components in the solvent.

[0036] The spatial offset of component distribution was quantified using spatial statistical methods, comparing the changes in the centroid position of the active ingredient in its initial state and after treatment. The reaction time delay period was determined through phase difference analysis of the release curves, calculating the lag time of the actual peak release based on the theoretical peak release time. Temperature change information was acquired using an embedded temperature sensor, recording the temperature gradient in different regions within the extraction container in real time. The statistical analysis of the number of activity decays, combined with chemical component stability data, identified the types of compounds that degraded under high temperature or prolonged exposure, along with their degradation rates.

[0037] The generation of medicinal herb activity retention rate characteristics employs a multi-index fusion algorithm, weighting and integrating spatial offset, time delay period, and activity decay count to output a comprehensive score reflecting the overall stability of the medicinal herb. A preset efficacy maintenance threshold is set based on the best performance of historical batches; samples below this threshold are identified as requiring optimization. The binding of identification codes to preparation batches is achieved through blockchain technology, ensuring the authenticity and immutability of data traceability. The final generated medicinal herb efficacy characteristic map includes a three-dimensional heatmap, a time-activity curve, and a component interaction network diagram. The heatmap displays the spatial aggregation state of active ingredients, the time-activity curve reflects the trend of efficacy changes over time, and the interaction network diagram reveals the synergistic or antagonistic relationships between different compounds.

[0038] The rendering of the thermal map of the spatial distribution of active ingredients employs a Gaussian kernel density estimation algorithm, transforming discrete component detection points into a continuous probability distribution surface. The construction of the pharmacodynamic action timeline integrates in vitro pharmacodynamic experiments and in vivo metabolic kinetic data, marking the time points at which key pharmacodynamic indicators reach effective concentrations. The component interaction correlation matrix is ​​filled using a combination of molecular docking simulations and experimental verification; the matrix element values ​​represent the binding free energy between pairs of components and the experimentally observed effect intensity.

[0039] The acquisition of microscopic structural image sequences follows a standardized sample preparation process, including unified parameter control for fixation, dehydration, embedding, and slicing. The acquisition of chemical composition spectral data adheres to the solvent system and detection wavelength specified in the International Pharmacopoeia, ensuring the comparability of data across batches. The determination of bioactive marker values ​​includes positive and negative controls, using relative activity values ​​to eliminate systematic errors. Synchronous acquisition of multimodal characteristic data relies on a laboratory information management system, with all instrument output data automatically stamped with a unified timestamp and associated with sample numbers.

[0040] Dynamic monitoring of distribution density changes employs computer vision technology to automatically track and quantify target regions in time-series images. The acquisition interval of the time-series data is dynamically adjusted based on the component release characteristics; high-frequency sampling is used during the rapid release phase, while the sampling rate is reduced during the stable phase to save storage space. Spatial offset calculation introduces a reference coordinate system, establishing a three-dimensional spatial grid with the geometric center of the medicinal material as the origin to quantify the displacement vector of the active ingredient clusters. A tolerance range is set for determining the time delay period to avoid misjudgments caused by instrument sampling errors.

[0041] Temperature change information is collected using a distributed sensor network, with multiple temperature measurement points deployed within the extraction container and uploaded to the central processing unit in real time. The recording of activity decay counts is correlated with specific temperature ranges and durations, establishing a three-dimensional relationship model of temperature-time-degradation. The calculation of activity retention rate characteristics incorporates a time decay factor, assigning different weights to early and late decay events. The efficacy maintenance threshold is set considering clinical application scenarios, differentiating between the requirements of acute treatment and chronic management.

[0042] The visualization of medicinal herb efficacy characteristics uses an interactive and dynamic presentation method, supporting multi-layer overlay display and parameter filtering. The heatmap's color gradation mapping employs a non-linear transformation to highlight the distribution differences of key active regions. The time-activity curve is plotted using a dual Y-axis design, simultaneously displaying absolute concentration values ​​and relative activity percentages. The component interaction network diagram is laid out using a force-directed algorithm, with node size representing compound content and edge thickness reflecting interaction intensity. The exported format of the graphs is compatible with mainstream data analysis software, supporting further data mining and model building.

[0043] Quality control in the data acquisition phase includes a comprehensive verification system encompassing instrument calibration, operator training, and standard sample testing. The feature extraction algorithm was selected after comparing multiple schemes, ultimately employing the model that best balances recall and precision. Manual verification nodes are set up at intermediate results during the spectrum generation process to perform secondary verification of outliers generated by automatic analysis. The final output spectrum of medicinal herb efficacy characteristics includes an integrity check code; any data tampering will result in verification failure. The spectrum update mechanism supports incremental revisions, ensuring that the addition of new batches of data does not affect the analytical conclusions of previous batches.

[0044] Example 2: See Figure 3 The generation process of the efficacy deviation feature set is based on the analysis of the efficacy feature map of medicinal materials. The retrieval of active ingredient identification regions is accomplished through spatial coordinate indexing in the map, with each identification region associated with a unique code that corresponds one-to-one with the original records in the medicinal material sample database. The theoretical active ingredient concentration range under the target efficacy is derived from pharmacopoeia standards, clinical drug guidelines, or dose-response curves established in previous studies, and the range is dynamically adjusted according to the type of medicinal material and preparation. The actual duration of action threshold is determined through in vitro release experiments simulating the human metabolic environment. The release dynamics of the active ingredient are continuously monitored using dialysis membranes or flow-through cells, recording the time span from the start of release to the concentration dropping below the effective threshold. The measured peak efficacy time is collected through a real-time monitoring system, with monitoring points set at key stages of the extraction process, including extraction, concentration, and alcohol precipitation. The data acquisition frequency is set according to the process characteristics, using second-level sampling during rapid changes and extending to minute-level sampling during stable stages.

[0045] The difference between the actual duration of action and the theoretical duration is calculated using a time series alignment algorithm. After synchronizing the measured and theoretical curves at their starting points, the time difference at the intersection of their falling edges is calculated. The concentration offset is quantified using statistical methods. The normality of the distribution of measured concentration values ​​within the theoretical range is tested, and the dispersion exceeding the upper and lower limits of the range is calculated. The generation of the core parameter set for efficacy bias uses a structured data storage format. Each parameter is associated with a timestamp, spatial location, and confidence score. The confidence score is dynamically adjusted based on the accuracy of the data acquisition equipment and the number of repeated experiments.

[0046] The data on total active ingredient content and distribution density are retrieved from a medicinal herb characteristic database, which integrates multi-source data from microscopic image analysis, spectral detection, and chromatographic separation. The calculation of the aggregation intensity of active ingredients employs image processing technology, performing cluster analysis on the pixel grayscale values ​​within the identified area to identify the contours and geometric centers of high-density aggregation zones. The dispersion level is assessed using the nearest neighbor index in spatial statistics, comparing the deviation of the actual distribution pattern from a completely random distribution. The output of component distribution status information includes two parts: a vector map and quantitative indicators. The vector map displays the spatial arrangement pattern of active ingredients, while the quantitative indicators include the aggregation coefficient, dispersion index, and anisotropy score.

[0047] The calculation of component distribution offset eigenvalues ​​introduces a reference coordinate system, establishing a three-dimensional grid with the geometric center of the medicinal material as the origin, mapping spatial location points to standardized volumetric units. Time interval distribution acquisition relies on a high-precision timing system, recording the detection time difference at different spatial locations within the medicinal effect cycle. Anomaly detection algorithms are used to identify distribution anomalies, performing principal component analysis on the spatiotemporal data matrix to mark data clusters exceeding the normal fluctuation range. Determining the location range combines process parameters and instrument logs, analyzing the operating conditions and environmental states at the time anomalies occurred.

[0048] The screening of continuously prepared fragments employed a sliding window algorithm. A fixed-length analysis window was set on the time axis, and the window was moved stepwise while calculating the component release behavior indicators within the window. Criteria for identifying abnormal behavior included abrupt changes in the slope of the release curve, missing peaks, or abnormally prolonged plateau periods. Regions exhibiting efficacy deviations were marked using a hierarchical coloring method, with different colors indicating mild, moderate, and severe deviations in a three-dimensional spatiotemporal model. The final efficacy deviation feature set was stored in a multidimensional array structure, with each data point containing a deviation type code, spatial coordinates, a time stamp, and associated process parameters.

[0049] The theoretical active ingredient concentration range was set considering batch-to-batch variations in medicinal materials and fluctuations in the preparation process. A dynamic boundary algorithm was used to automatically adjust the range width based on the standard deviation of historical data. An environmental control group was set up to determine the actual duration of action threshold, eliminating the influence of interfering factors such as temperature, pH, and stirring speed. Environmental sensor data, including parameters such as temperature, pressure, and solvent flow rate, were simultaneously recorded during the acquisition of the measured peak efficacy time for subsequent deviation attribution analysis. A smoothing process was incorporated into the difference calculation to eliminate random noise interference from the monitoring equipment.

[0050] The statistical processing of concentration offset employs a robust estimation method, with Winsorization applied to extreme values ​​to avoid excessive influence of individual outliers on the overall assessment. The data structure design of the efficacy deviation core parameter set supports rapid retrieval and batch processing; index fields include medicinal material number, preparation batch, and deviation type. The calculation of the total active ingredient content uses the standard curve method, converting the detection signal intensity into absolute content and correcting for the recovery rate of the extraction solvent. Spatial interpolation of distribution density data employs the Kriging algorithm, predicting the density distribution of unsampled areas based on measurements from known points.

[0051] Clustering analysis for aggregation intensity uses an adaptive threshold to dynamically adjust classification boundaries based on image contrast. Dispersion level assessment incorporates Monte Carlo simulation, establishing confidence intervals through random sampling. Visualization of component distribution state information employs volume rendering techniques, using transparency adjustments to achieve a hierarchical display of 3D data. Offset eigenvalue calculation considers spatial anisotropy, assigning differential weights to positional deviations in different directions. Analysis of time interval distribution uses a Poisson process model to examine the randomness of detected events along the time axis.

[0052] The anomaly detection algorithm's parameters were optimized through cross-validation, adjusting sensitivity thresholds by dividing historical data into training and testing sets. Location range determination was combined with process video recordings, comparing anomaly periods with the timeline of operational actions. An overlapping window mechanism was implemented for screening continuously prepared fragments to avoid analytical blind spots. The criteria for anomaly judgment were established based on an expert knowledge base, integrating empirical rules from pharmacology and formulation. The exported format of the efficacy deviation feature set supports integration with the quality management system, enabling automatic archiving and trend analysis of deviation data.

[0053] Environmental control during data acquisition employs a constant temperature and humidity system to minimize the impact of external condition fluctuations on monitoring results. Instrument calibration strictly adheres to metrological standards, and standard substances are used regularly to verify detection accuracy. The selection of analytical algorithms involved comparing multiple options to achieve a balance between computational efficiency and accuracy. Intermediate results are stored using a version control mechanism, preserving data snapshots from critical steps. The final output efficacy deviation feature set includes metadata documentation detailing the parameter settings and algorithm configurations used during its generation.

[0054] The application of efficacy deviation feature sets is not limited to single-batch analysis; it supports horizontal comparison and vertical trend prediction of multiple batches of data. The data structure design considers scalability, with reserved fields for storing newly added detection indicators and analysis dimensions. The visual interface provides interactive filtering functions, allowing users to dynamically filter data of interest by deviation level, spatial region, or time segment. Real-time data exchange is achieved through the interface with the production process control system, and an alarm mechanism is automatically triggered when a serious deviation is detected. The historical data backtracking analysis function supports statistical analysis of deviation occurrence patterns by dimensions such as medicinal material type, preparation process, or operator.

[0055] Example 3: The generation process of the dynamic adjustment table for medicinal material proportions uses an efficacy deviation feature set as input and achieves precise optimization by analyzing the differences in release behavior in consecutive preparation batches. The medicinal material numbers associated with the efficacy deviation feature set are mapped to the raw material database using an encrypted hash algorithm, extracting the complete release curves of the corresponding medicinal materials in two consecutive preparation batches. The calculation of the lag duration of the efficacy peak time point uses a dynamic time warping algorithm to non-linearly align the release curve of the current batch with the reference batch, eliminating the influence of time-series distortion caused by process fluctuations. The determination of the complete release time point of components is based on the second derivative analysis of the release curve; when the rate of change of the curve slope is lower than a threshold... The time marker is the release endpoint, where The value is dynamically adjusted within the range of 0.001-0.005% / min² based on the type of medicinal material. The theoretical end time of action is derived from the pharmacodynamic model prediction, and the difference between the actual complete release time and the predicted time is used to obtain the tail-end release lag information.

[0056] The release completion baseline is set using quantile statistics, analyzing the distribution of release lag in historical qualified batches of corresponding medicinal materials, and taking the 95th percentile as the critical threshold. Screening of substandard medicinal materials is achieved through comparison operators, comparing the current lag with the threshold in real time and triggering a flag. Waiting time is extracted using timestamp difference calculations, recording the time span from the start of the process to complete component release. The generation of the medicinal material release priority sequence relies on a min-heap data structure, arranged in ascending order of waiting time and dynamically maintained in a queue. The configuration of incremental fusion time periods uses a linear allocation algorithm, allocating longer action periods to high-priority medicinal materials while keeping the total preparation cycle unchanged; the adjustment magnitude is proportional to the priority score.

[0057] The dynamic ratio adjustment table is stored using a relational database design, with a primary key that is a composite index (medicinal herb number + preparation batch number), and fields containing compatibility weights. Adjustment amount of the standard ratio and effective period .in The calculation formula is:

[0058]

[0059] In the formula: Indicates the waiting time for the current batch. This is the historical average waiting time. This is the adjustment coefficient (default value 0.5). This function converts the waiting time difference into a weight value in the range of 0-1, realizing non-linear priority mapping.

[0060] The generation process of raw material combination optimization instructions is monitored in real time to track the fusion status in the preparation system. The reference ratio value is invoked via the OPC-UA protocol, interacting with the process control system and sampling the ratio parameters once per second. Near-infrared spectroscopy is used to detect the fusion distance distribution, scanning the spectral characteristics of different sites in the reaction vessel and calculating the spectral angular distance between adjacent herbal groups through principal component analysis. The identification of the synergistic reaction initiation node of three adjacent active ingredients relies on a mutation detection algorithm; when the cross-correlation coefficient of the spectral characteristics of any two components exceeds a threshold... When the value is typically 0.85, it is marked as a collaborative initiation event. The time node set is stored using a circular buffer structure, retaining the detection results within the most recent 30 seconds for time series analysis.

[0061] The determination of the fusion stage switching time point is based on the stage transition conditions set in the process specification, usually synchronized with changes in temperature or pH value. The synergistic activation state is marked using a finite state machine model; a state transition is triggered if and only if all synergistic start times are earlier than the switching point. The association between the herb group number and state parameters is encapsulated in XML format, including metadata such as timestamps, equipment numbers, and process segment identifiers. The evaluation of the reaction progress compression detection state is based on changes in the synergistic reaction rate, calculating the percentage of reaction progress completed per unit time; when this value exceeds 20% of the historical average, it is marked as a compression state.

[0062] The real-time preparation process monitoring system employs a distributed architecture, with data acquisition nodes deployed at key locations such as extraction tanks, reaction vessels, and delivery pipelines. Each node is equipped with an independent signal conditioning module to filter and amplify the raw sensor data. The quantization of the fusion distance incorporates standard reference calibration, and the detection model is periodically corrected using mixed samples with known component proportions. The calculation of spectral angular distance uses a vector space model, treating the characteristic spectrum of each component as an n-dimensional vector, and measuring similarity using the cosine of the included angle. The sliding window width of the mutation detection algorithm adaptively adjusts according to the material flow rate, shortening the window during rapid flow phases to improve temporal resolution.

[0063] The digital representation of the process specifications adopts the BPMN 2.0 standard, transforming stage switching conditions into executable business rules. Temperature monitoring utilizes platinum resistance temperature sensors with a measurement accuracy of ±0.1℃. pH changes are detected using a composite glass electrode, maintained within the set range by an automatic titration device. The finite state machine's state transition conditions are equipped with an anti-jitter mechanism to avoid false triggering caused by instantaneous fluctuations. XML message transmission uses the MQTT protocol, achieving cross-system data distribution through a topic subscription mechanism. Reaction rate calculation employs a numerical differentiation method, performing a five-point cubic smoothing process on the reaction progress curve before differentiation.

[0064] Historical data benchmarks are built upon large-sample statistical analysis, encompassing complete process parameters from at least 50 qualified batches. The preparation of standard reference materials strictly adheres to the operational procedures of weighing, mixing, homogenizing, and dispensing to ensure batch-to-batch consistency. Dimensional optimization of the vector space model is achieved through a feature selection algorithm, retaining feature bands with a signal-to-noise ratio higher than 3:1. The execution engine for business rules employs the Rete algorithm, supporting parallel matching and rapid response under complex conditions. The anti-jitter mechanism features an adjustable time constant, configurable within the 5-60 second range based on process stability requirements. Parameter selection for smoothing is based on spectral analysis to determine the cutoff frequency, preserving the true reaction dynamics.

[0065] The system's fault-tolerant design includes data verification and recovery mechanisms. Each data packet is appended with a CRC32 checksum, triggering a retransmission request in case of a transmission error. The real-time database employs a dual-buffer structure, with primary and backup storage areas periodically synchronized and automatically switching in case of failure. The signal conditioning module's gain setting retains a 20% margin to prevent signal saturation distortion. A variance threshold is set during feature selection to eliminate invalid bands with fluctuations less than 1%. The Rete algorithm's rule network is pre-compiled into binary code to improve pattern matching efficiency. The historical database undergoes periodic defragmentation and index rebuilding to maintain query performance.

[0066] The human-computer interface design prioritizes ease of operation. The proportioning adjustment table is visualized using a heatmap, with color depth indicating the adjustment range. The display of the synergistic reaction status incorporates animation effects, rendering the spatial distribution of the reaction process in real time. Alarm information is presented with tiered management; emergency events are indicated by flashing icons and audible alerts. The parameter setting interface provides preset templates, supporting quick access to typical process schemes. Data export formats are compatible with Excel and MATLAB, meeting diverse analytical needs. Operation logs include user IDs and modification details, enabling complete audit trails.

[0067] A comprehensive maintenance system ensures long-term stable operation. A full calibration is performed monthly to verify the measurement accuracy of all sensors. The spectrometer's light source and detector window are replaced quarterly to maintain optical performance. During annual overhauls, the contact resistance of all electrical connections is checked to prevent intermittent failures. The software system undergoes regular security updates to patch known vulnerabilities. Spare parts are managed using a first-in, first-out (FIFO) principle to ensure critical components are readily available. Maintenance records are electronically archived and linked to equipment QR codes; scanning the code allows access to the complete maintenance history.

[0068] The compilation and updating of technical documentation forms a complete knowledge system. The system architecture specification describes the hardware components and communication protocols. The data dictionary clearly defines all variable names and value ranges. The algorithm white paper explains the core mathematical model and parameter settings in detail. The operation manual provides step-by-step daily usage guidance. The fault code manual lists all possible alarm messages and handling methods. The API reference documentation describes the interface specifications for external system integration. The version update log records the content and scope of each modification.

[0069] The training system ensures that personnel capabilities match system requirements. New employee training includes theoretical courses and practical assessments, with a focus on handling abnormal situations. Regular refresher training updates skills and knowledge, and introduces new system features. Specialized training provides in-depth explanations of specific modules, such as advanced techniques for spectral data analysis. External certification training invites equipment manufacturers to deliver courses, obtaining official accreditation. Training effectiveness is evaluated through a combination of written tests and simulated operations; those who do not meet the standards must retake the training. Training materials include video tutorials that can be viewed on mobile devices for convenient review anytime.

[0070] Example 4: See Figure 4 The process of generating the tail-end effect regulation results starts with the raw material combination optimization command and achieves fine-tuning of the process by analyzing the release trajectory of the fused components. In the preparation of a certain batch of Astragalus-Angelica medicinal wine, the system calls the preparation cycle number P-202307-085 marked by the command to screen out the fusion trajectory data of the tail-end components ferulic acid and ligustilide. The trajectory acquisition adopts an online ultraviolet spectroscopy monitoring system, recording the absorbance values ​​at wavelengths of 254nm and 280nm every 15 seconds, and continuously tracking the dynamic changes from the tail-end start time (8 hours after alcohol precipitation) to the preset release endpoint (12 hours after alcohol precipitation).

[0071] Analysis of the activity value sequence showed that ferulic acid exhibited a continuous linear growth trend in the later stages, with absorbance at 12 hours increasing by 42.9% compared to 8 hours, without the expected plateau phase. The growth curve of ligustilide showed a significant decrease in slope after 11 hours, indicating a stable state. A systematic comparison with historical qualified batches of this medicinal material revealed that the delayed release of ferulic acid was related to the particle size distribution of the Angelica sinensis raw material. The current batch of Angelica sinensis had an 80-mesh sieve pass rate of 92%, while the historical best-performing batch had a pass rate of 95%, indicating that subtle particle size differences led to variations in the dissolution rate of the active ingredient.

[0072] The remaining release time was calculated using a trend extrapolation method, based on the change rate of ferulic acid absorbance in the last 30 minutes (0.015 / 30min), to predict the replenishment time required to reach the stable threshold of 0.520. The system determined that the tail-end control cycle needed to be extended by 85 minutes, adjusting the original 12-hour endpoint to 13 hours and 25 minutes. Real-time control parameter updates were performed through a distributed control system, specifically adjusting the following: the alcohol precipitation temperature maintenance module setpoint was increased from 60℃ to 61℃, the stirrer speed was increased from 105 rpm to 115 rpm, and simultaneously triggering the replenishment pump to add 65% ethanol solution at a flow rate of 0.8 L / min to maintain the liquid phase volume.

[0073] The recommended time window for maintaining the blend was generated based on the performance of this formulation in three historically best-performing batches. The first batch used a fixed 12-hour tail-end duration, achieving an average ferulic acid release rate of 89.2% of the nominal value. The second batch implemented dynamic adjustments, extending the average to 13 hours and 18 minutes, increasing the release rate to 94.7%. The third batch experimentally used a 14-hour tail-end duration, achieving a release rate of 96.3% but increasing energy consumption by 22%. After comprehensive system evaluation, a compromise of 13 hours and 25 minutes is recommended for the current batch, striking a balance between release efficiency and energy consumption.

[0074] The implementation of the control results is accompanied by multiple verification mechanisms. The temperature sensor adopts a triple-redundant design, and the measurement values ​​of the three probes are output as the final control signal through a median filtering algorithm. Torque detection is performed before adjusting the stirring speed to confirm that the motor load is within a safe range. Online density detection is performed before the replenishment operation is started to avoid the dissolution balance being affected by the ethanol concentration deviation. All adjustment commands are digitally signed and authenticated, and the operator ID, timestamp, and parameter snapshot before modification are recorded to form a complete audit trail chain.

[0075] The real-time monitoring interface for the tail-end release trajectory integrates multi-dimensional visualization tools. The main view displays a superimposed comparison of two release curves of the active ingredient, with a gray background area indicating the confidence interval of the historical best batch. The auxiliary panel displays environmental parameter trend graphs, using color gradients to indicate deviations between set and actual values ​​for temperature and stirring speed. The warning area dynamically lists abnormal indicators requiring attention, currently displaying the message "Ferulic acid release rate exceeds expected range +15%". Operators can manually intervene in the parameter weights of the prediction model by dragging markers on the timeline, and the system automatically saves all manual correction records for subsequent model optimization.

[0076] The data backtracking function supports in-depth analysis of the tail-end regulation effect. Comparing batch P-202307-085 with the historical reference batch R-202303-042, the system generates a parallel coordinate graph displaying five key indicators: total tail-end duration, ferulic acid release rate, ligustilide stabilization time, energy consumption index, and solvent consumption. The predicted ferulic acid release rate for the current batch is 94.2%, slightly higher than the reference batch's 93.8%, but the energy consumption index increases by 7.3%. This subtle difference triggers the system to automatically generate a suggestion: in subsequent batches, controlling the angelica powder particle size within a narrow range of 93-95% pass rate at 80 mesh may achieve a better energy efficiency ratio.

[0077] The quality control module correlates the tail-end regulation results with the final product testing indicators. After completing the 13-hour and 25-minute extended tail-end, sampling and testing showed that the ferulic acid content reached 4.78 mg / mL, an increase of 0.82 mg / mL compared to the control sample terminated at 12 hours. High-performance liquid chromatography (HPLC) chromatograms showed a 12.6% reduction in impurity peak area, indicating that extending the reaction time promoted more complete release of components and precipitation of impurities. These test data are automatically populated into the medicinal material processing parameter knowledge base to update the recommended range for the optimal particle size of Angelica sinensis.

[0078] The iterative update mechanism of the process knowledge graph continuously optimizes the control logic. Two new empirical rules have been added to the processing instructions for Angelica sinensis: when the 80-mesh sieve pass rate is below 94%, it is recommended to increase the base processing time for the final stage by 30 minutes; when the ambient humidity exceeds 65%, the powder flowability index needs to be monitored additionally. These rules are transformed into executable business rules using natural language processing technology and embedded into the automatic control process for the next batch. Simultaneously, the system establishes a direct correlation between the Angelica sinensis pulverization process parameters and the final stage control results in the material traceability module, supporting the direct pre-loading of optimization parameters by scanning the raw material batch QR code.

[0079] The anomaly handling protocol specifies emergency measures during the tail-end control process. When a sudden drop in the ferulic acid release rate exceeding 20% ​​is detected, the system automatically initiates a three-level response: the primary response increases the stirring speed by 10% and maintains it for 15 minutes; the intermediate response increases the temperature setpoint by 2°C; and the advanced response triggers a production halt and notifies the process engineer. In the current batch, a brief rate drop was detected at 11 hours and 45 minutes. After the system executed the primary response, the rate returned to normal. The event report indicated that the possible cause was temporary precipitation due to localized temperature fluctuations.

[0080] The personnel operating procedures emphasize human-machine collaboration during the control process. Although the system automates most adjustments, key decision points still require manual confirmation. For example, when the predicted make-up time exceeds 2 hours, two-factor authentication by the quality manager is required before execution; any parameter modifications that deviate from the standard operating procedure must be accompanied by a change order. These operation records, together with process data, constitute a complete electronic batch record, meeting GMP data integrity requirements.

[0081] Technical documentation is managed using a version control strategy. Each update to the tail-end control plan generates a new version number, and historical versions are retained for auditing purposes. The currently effective control plan is V3.2, with major improvements including: the introduction of an environmental humidity compensation coefficient, optimization of the weight parameters in the trend extrapolation algorithm, and the addition of an energy efficiency balance assessment module. The version release notes detail the technical basis and verification data for each modification, including test analysis reports from six comparative batches.

[0082] Example 5: The generation and application of dynamically optimized formulation characteristics achieve adaptive adjustment through the synergistic analysis of environmental factors and medicinal material characteristics. Environmental factor data is collected using a distributed sensor network. Light intensity fluctuation sequences are collected via a silicon photocell array installed on the top of the extraction workshop, recording the radiant flux density within the visible spectrum (400-700nm) every five minutes. During data preprocessing, outliers caused by equipment obstruction or transient interference are removed. Humidity changes are monitored using a capacitive polymer thin-film sensor. Three monitoring points are set up in the medicinal material pretreatment area, extraction area, and storage area, respectively, and the moving average value is taken as the representative value of the current ambient humidity.

[0083] The influence coefficients of environmental factors on the release rate of components were calculated using multivariate time series analysis. Light intensity data was decomposed into trend, periodic, and random terms, with the periodic term matching the switching patterns of workshop lighting equipment, and the trend term reflecting the diurnal variation of natural light. A dynamic correlation model was established between humidity data and the moisture content of medicinal materials, and key influencing periods were identified through partial least squares regression. The quantification of influence coefficients was based on historical batch data mining. Dynamic time-normalization and alignment were performed on the release curves of components with the same formulation under different environmental conditions, and the correlation coefficient matrix between environmental changes and changes in release rate was calculated.

[0084] The environmental compensation calculation employs an incremental update strategy. The baseline ratio is extracted from the dynamic adjustment table of medicinal herb ratios, containing three core parameters: herb type, weight percentage, and duration of action. The compensation calculation introduces an environmental adaptation factor, which is a weighted composite of the light and humidity influence coefficients. The weights are dynamically adjusted based on the photosensitivity and hygroscopic properties of the medicinal herbs. For photosensitizing herbs such as Scutellaria baicalensis, the light weight is set to 0.7; for hygroscopic herbs such as Rehmannia glutinosa, the humidity weight is set to 0.6. The generation of the environmentally adapted ratio characteristics is achieved through parameter space mapping. The baseline ratio vector is multiplied by the environmental adaptation factor matrix, outputting the compensated ratio parameter set.

[0085] The final integration of dynamically optimized formulation features requires the fusion of multi-source data. The distribution data of active ingredients in the medicinal herb efficacy feature map is transformed through spatial encoding, converting the pixel matrix of the heatmap into a component density distribution vector. The fusion process employs a feature concatenation method, weighting the environmentally adapted formulation features and the component density vector in the hidden layer space using an attention mechanism to highlight the formulation adjustment needs of key active regions. The output layer generates a three-dimensional formulation optimization surface, where the horizontal axis represents the type of medicinal herb, the vertical axis represents the combination of environmental conditions, and the height represents the recommended formulation adjustment range. The surface smoothness is controlled by regularization parameters to prevent overfitting.

[0086] The multi-batch preparation validation cohort was constructed using a stratified sampling strategy. Six validation groups were established based on typical combinations of environmental conditions: high light and high humidity, high light and low humidity, and moderate light and moderate humidity. Each validation group contained three parallel batches, and the preparation order was balanced using a Latin square design to account for learning effects. Monitoring of the actual efficacy peak time point employed an embedded fiber optic sensor, forming a three-dimensional monitoring grid within the extraction container. Transmission spectra at characteristic wavelengths were acquired every two minutes, and principal component analysis was used to determine the release peak positions of each herbal component.

[0087] The matching degree assessment of component release curves introduces a dynamic similarity index. Real-time acquired release curves are compared with theoretical curves at multiple scales. The relative difference in the area under the curve is calculated at the macro scale, and the temporal shift of the peak position is analyzed at the micro scale. The matching degree data is transformed into standardized scores using a fuzzy logic system, considering the contribution weight of different medicinal components to the overall efficacy. The storage of the formulation verification result set adopts a time-series database structure, with each data point associated with three-dimensional information: environmental condition labels, formulation parameters, and verification scores.

[0088] The priority weight parameters are adjusted using a reinforcement learning framework. The dynamic adjustment table of medicinal material proportions is treated as a policy network, and the validation score is used as a reward signal. The weight parameters are updated through a policy gradient method. A conservative learning rate is set during the adjustment process, with each update not exceeding 15% of the original value to avoid drastic fluctuations that could affect production stability. After the optimized dynamic adjustment table is generated, it must undergo a completeness check through a simulation validation module, including parameter range verification, logical consistency verification, and environmental boundary condition testing.

[0089] The implementation of subsequent batch raw material screening processes reflects dynamic adaptability. The initial formulation of new batches uses the moving average of the three most recent optimization results to reduce the impact of random fluctuations. Real-time environmental monitoring data is input into the prediction model every ten minutes, triggering formulation pre-adjustment suggestions when changes in light or humidity exceed set thresholds. The user interface displays the recommended formulation parameters under current environmental conditions, along with actual performance data from three of the most similar historical batches for reference. Final execution parameters require dual confirmation: one person is responsible for verifying the formulation calculations, and the other for verifying the authenticity of the environmental data.

[0090] The calibration and maintenance of the environmental sensor network have been standardized. Light sensors undergo monthly calibration using standard light sources, covering all operating ranges. Humidity sensors are calibrated weekly with saturated salt solutions, verifying measurement accuracy at 33%, 75%, and 97% humidity levels. Sensor fault diagnosis employs cross-validation; an alarm is automatically triggered when the difference in measurements between adjacent nodes exceeds 15%. Calibration records are stored on the blockchain, ensuring immutability and traceability to national metrological standards.

[0091] Strict version management is implemented for the ratio optimization algorithm. Each algorithm update requires five sets of backtesting reports from historical batches, demonstrating that the new version outperforms the old version in at least 80% of cases. The algorithm is containerized and deployed on edge computing nodes, supporting rapid rollback to any historical version. Detailed operation logs record the input parameters, calculation process, and output results for each prediction; log files are encrypted and retained for at least twice the product's validity period.

[0092] The personnel training system includes specialized courses on the relationship between environment and formulation. Basic training covers the mechanisms by which environmental factors affect the components of medicinal materials, including photodegradation kinetics and hygroscopic swelling effects. Intermediate training teaches manual correction methods for formulation compensation calculations, such as how to estimate environmental parameters based on weather forecast data when sensors malfunction. Advanced training cultivates the ability to diagnose abnormal situations, using twenty typical fault case studies to demonstrate how to distinguish between environmental interference and actual process anomalies. Training assessment utilizes a virtual simulation system, requiring trainees to complete six sets of formulation optimization tasks under different climatic conditions within a simulated environment.

[0093] The knowledge accumulation mechanism continuously enriches and optimizes the basis for improvement. Complete environmental data, formulation parameters, and product test results for each production batch are automatically archived into the knowledge base, and reusable experience is extracted through semantic analysis technology. After accumulating ten successful cases under similar environmental conditions, the system automatically generates new experience rules and submits them to process engineers for review. Adopted rules are transformed into executable logic in the business rule engine, such as specific guidance like "increase the proportion of Angelica sinensis by 3% under continuous rainy weather".

[0094] Cross-system integration enables seamless data flow throughout the entire process. The environmental monitoring system interfaces with the building automation system to obtain air conditioning unit operation plans in advance to predict humidity change trends. The proportioning optimization system integrates with the MES system to obtain production scheduling information in real time and avoid equipment conflicts. The quality management system automatically receives the fingerprint spectrum of the final product and performs closed-loop verification with the component release curves in the prediction model. All integration interfaces are isolated using middleware, ensuring that a failure in any system will not lead to a cascading shutdown.

[0095] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0096] 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 method for efficiently screening raw materials for medicinal wine, characterized in that, include: The basic parameters of raw materials from various candidate medicinal materials are collected, multimodal feature data from the basic parameters are extracted, the distribution of effective components of medicinal materials is identified based on the multimodal feature data, and a medicinal material efficacy feature map is generated. By calling the active ingredient identification area in the medicinal herb efficacy feature map, the concentration range of active ingredients pointing to the target efficacy and the threshold of the duration of medicinal effect are obtained, the matching difference between the actual component concentration value and the theoretical duration of effect is detected, and an efficacy deviation feature set is generated. Based on the medicinal material numbers associated with the efficacy deviation feature set, the component release curves and efficacy peak time points of the medicinal materials in continuous preparation batches are extracted. The stability gap between the component release curve and the preset efficacy benchmark is calculated to determine whether the component release has not met the standard. The stability gaps corresponding to the non-compliant medicinal materials are selected as the basis for optimization priority and a dynamic adjustment table of medicinal material ratio is generated. Call the baseline ratio value in the dynamic adjustment table of medicinal material ratio, monitor the fusion status of effective components during real-time preparation, mark the start time node of the synergistic reaction of three adjacent components, if all synergistic reaction start time nodes are earlier than the preset fusion stage switching point, record the synergistic activation status and generate raw material combination optimization instructions. The specific steps for generating the medicinal herb efficacy feature map are as follows: First, acquire multimodal feature data of multiple candidate medicinal herbs during the preprocessing stage. Second, extract the distribution density change and action time series data of the effective components of each medicinal herb. Third, based on the correspondence between the distribution density change and the action time series, calculate the spatial offset of component distribution and the action time delay period to generate a set of component distribution and time-effect parameters. Fourth, based on the component distribution spatial offset and action time delay period in the set of component distribution and time-effect parameters, and combined with the temperature change information of the medicinal herbs at continuous time points during extraction, count the number of activity decays of each medicinal herb in the distribution offset and time delay stages, identify the correlation distribution pattern between the number of activity decays and the time delay period, and generate the medicinal herb activity retention rate feature. Fifth, based on the medicinal herb activity retention rate feature, determine medicinal herb samples with an activity retention rate lower than a preset efficacy maintenance threshold, bind the corresponding sample's identification code to the preparation batch number, and generate the medicinal herb efficacy feature map. The specific steps for generating the efficacy deviation feature set are as follows: First, the active ingredient identification regions marked in the medicinal herb efficacy feature map are called. Then, three parameters are obtained for each identified region under the target efficacy: the theoretical active ingredient concentration range, the actual duration of action threshold, and the measured peak efficacy time. The difference between the actual duration of action and the theoretical duration of action, as well as the offset between the measured concentration value and the theoretical concentration range, are calculated to generate the efficacy deviation core parameter set. Based on the actual duration of action difference and concentration offset values ​​in the efficacy deviation core parameter set, the total amount and distribution density data of active ingredients in the corresponding identified region within the medicinal effect cycle are called. The aggregation intensity and dispersion level of the effective ingredients are identified to obtain the component distribution status information. Based on the component distribution status information and the time interval distribution of different spatial locations within the efficacy deviation core parameter set, the component distribution offset feature value is calculated. The location range of the abnormal distribution area within the medicinal effect cycle is identified, generating the component distribution abnormality period. Based on the associated preparation batches within the component distribution abnormality period, the correlation between the time segment and component release behavior within the medicinal effect cycle is evaluated. Continuous preparation segments with abnormal component release behavior are screened and marked as efficacy deviation areas, generating the efficacy deviation feature set. The specific steps for generating the dynamic adjustment table of medicinal material ratios are as follows: Based on the medicinal material numbers associated with the efficacy deviation feature set, extract the lag duration of the peak efficacy time point and the time point of complete component release in two consecutive preparation batches of medicinal materials. Combine the time difference between the time point of complete component release and the theoretical end time point of the action cycle to obtain the lag information of the tail segment release of medicinal materials. Based on the lag information of the tail segment release of medicinal materials, determine whether the release lag amount exceeds the preset release completion benchmark value, screen the medicinal material numbers that have not met the release standard, extract the waiting time before the complete release of components in the corresponding medicinal materials, sort them according to the waiting time length of the non-compliant medicinal materials, and generate a medicinal material release priority sequence. Call the sorting results in the medicinal material release priority sequence, configure incremental fusion time periods for the medicinal materials in turn, adjust the medicinal material action duration within the total preparation cycle, record the medicinal material number and the adjusted action duration, and generate a dynamic adjustment table of medicinal material ratios. The specific steps for generating the raw material combination optimization instruction are as follows: calling the baseline ratio value recorded in the dynamic adjustment table of medicinal material ratio, detecting the fusion distance distribution of related medicinal material groups during real-time preparation, marking the synergistic reaction start time nodes of the three adjacent active ingredients in the fusion stage, and generating a component synergistic start time set; based on the order of the synergistic reaction start time nodes of the three adjacent components in the component synergistic start time set, determining whether all time nodes are earlier than the fusion stage switching time point in the dynamic adjustment table of medicinal material ratio; if the determination condition is met, marking it as a synergistic activation state, associating the medicinal material group number and state parameters, and generating the raw material combination optimization instruction.

2. The method for efficient screening of medicinal wine raw materials according to claim 1, characterized in that, The raw material basic parameters include microstructure image sequences, chemical component spectral data, and bioactivity marker values. The medicinal material efficacy feature map includes a thermal map of the spatial distribution of active ingredients, a time axis of efficacy action, and a component interaction correlation matrix. The efficacy deviation feature set includes target efficacy deviation encoding, actual duration of action deviation, and component concentration fluctuation anomaly markers. The medicinal material ratio dynamic adjustment table includes medicinal material compatibility priority weights, baseline ratio adjustment range, and dynamic adjustment effective period. The raw material combination optimization instructions include synergistic reaction trigger markers, component fusion start time, and reaction process compression detection status.

3. The method for efficient screening of medicinal wine raw materials according to claim 1, characterized in that, Also includes: The preparation cycle marked by the raw material combination optimization instruction is called to screen the release trajectory of the tail component in the fusion stage. The change trend of component activity is compared with the tail duration. If the activity value continues to rise without entering the stable range, the time required to reach the preset release endpoint is calculated and the tail control cycle is updated to obtain the tail continuous effect regulation result. The tail continuous effect regulation result includes the remaining release path duration, the predicted time period for complete release of tail components, and the suggested time window for fusion maintenance.

4. The method for efficient screening of medicinal wine raw materials according to claim 3, characterized in that, The specific steps for generating the tail-end continuous action regulation result are as follows: The preparation cycle number marked in the raw material combination optimization instruction is called, and the trajectory data of the tail-end component in the fusion stage under the corresponding cycle is screened. A continuous activity value sequence and time stamp data from the tail-end start time to the component reaching the release endpoint are extracted to generate the tail-end release trajectory sequence. Based on the activity value sequence in the tail-end release trajectory sequence, the activity change trend of the tail-end component within the continuous action duration is analyzed. The activity increase at the end of the sequence is extracted and compared with the corresponding tail-end duration. If the activity value continues to rise and has not reached a stable threshold, the supplementary time required for the component to reach the preset release endpoint is calculated, generating the remaining tail-end release time interval. The supplementary time value required for the associated medicinal material group in the remaining tail-end release time interval is called, the real-time tail-end action control cycle is updated, the original tail-end action end time point is corrected, and the control interval of the medicinal material group's tail-end is reset to obtain the tail-end continuous action regulation result.

5. The method for efficient screening of medicinal wine raw materials according to claim 1, characterized in that, It also includes a dynamic optimization ratio feature generation step: based on the optimized ratio features output from the dynamic adjustment table of medicinal material ratios, extract the light intensity fluctuation sequence and humidity change from the environmental factor data, calculate the influence coefficient of environmental factors on the component release rate; perform environmental compensation calculation on the benchmark ratio value based on the influence coefficient, and generate environmentally suitable ratio features. By integrating the environmental adaptation ratio characteristics with the distribution data of active ingredients in the efficacy characteristic spectrum of medicinal materials, dynamic optimized ratio characteristics after environmental compensation are generated.

6. The method for efficient screening of medicinal wine raw materials according to claim 5, characterized in that, The specific steps for applying the dynamic optimization ratio feature are as follows: calling the medicinal material compatibility parameters in the dynamic optimization ratio feature to construct a multi-batch preparation verification queue; collecting the matching degree data between the actual efficacy peak time point and the component release curve in the verification queue to generate a ratio verification result set; adjusting the priority weight parameters of the medicinal material ratio dynamic adjustment table according to the ratio verification result set to generate an optimized ratio dynamic adjustment table; and applying the optimized ratio dynamic adjustment table to the raw material screening process of subsequent batches.

Images