A method for preparing veterinary ceftiofur hydrochloride injection
By establishing a database of dissolution behavior characteristics through machine learning algorithms, identifying contact hindrance moments, and optimizing temperature control, the problems of dissolution hindrance and temperature runaway in the preparation of high-concentration ceftiofur hydrochloride injection were solved, achieving product quality stability and consistency.
Patent Information
- Application Number
- CN202611386773.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-08
- Publication Date
- 2026-11-17
AI Technical Summary
In the preparation of high-concentration ceftiofur hydrochloride injection, the active pharmaceutical ingredient tends to form a high-viscosity coating layer in complex solvent systems, which hinders dissolution. Furthermore, local temperature runaway can lead to chemical degradation, resulting in an unstable production process.
By processing critical saturation solubility data and historical batch operation records using machine learning algorithms, a database of dissolution behavior characteristics is established. The timing of contact hindrance is identified, the safe temperature control range is analyzed, the timing of stirring intervention and the order of material addition are matched, the solvent ratio parameters are dynamically corrected in real time, and the cooling device is activated when the temperature exceeds the safe range to optimize the liquid preparation operation.
Precise control of high-concentration ceftiofur hydrochloride injection has been achieved, solving the problems of dissolution inhibition and temperature runaway, and ensuring the stability and consistency of product quality.
Smart Images

Figure CN123120109A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of information technology, and in particular to a method for preparing veterinary ceftiofur hydrochloride injection. Background Technology
[0002] Veterinary antibiotic injections play a crucial role in disease prevention and control in modern animal husbandry. To reduce injection volume and stress response in animals, developing high-concentration ceftiofur hydrochloride injections is a core direction for current pharmaceutical process upgrades. Currently, conventional preparation processes rely on fixed formulation ratios and conventional constant-temperature stirring methods. While this approach works adequately for low-concentration solutions, it often overlooks the significant fluctuations in the dissolution behavior of the active pharmaceutical ingredient in complex solvent systems when dealing with high-concentration formulations. This leads to crystallization or batch-to-batch concentration instability during production and storage. High-concentration drug administration requires extremely high solubility of the active pharmaceutical ingredient (API) in a polyethylene glycol-water mixture, which approaches the drug's saturation threshold. When the drug concentration nears this limit, its solubility becomes exceptionally sensitive to temperature fluctuations and solvent ratios. This sensitivity directly leads to contradictions between feed addition and dissolution during production. For example, in actual dispensing tank operations, if a large amount of API powder is added at once using conventional methods, the outer layer of the powder quickly binds with the solvent, forming a viscous coating that completely blocks contact between the internal powder and the solvent. Even blindly increasing the temperature or extending the stirring time indefinitely cannot break this coating, failing to achieve the expected solubility limit and instead causing localized overheating and drug degradation.
[0003] Therefore, accurately grasping the solubility variation of the active pharmaceutical ingredient at different temperatures and in proportions of compound solvents, and then scientifically planning the order of material addition and stirring timing in the mixing tank, has become a key issue in overcoming the bottleneck of high-concentration injection preparation. Summary of the Invention
[0004] This invention provides a method for preparing veterinary ceftiofur hydrochloride injection, mainly comprising: A set of critical saturation solubility data for the active pharmaceutical ingredient under different environmental temperature fluctuations and solvent ratio parameter combinations is obtained. This data set, along with historical batch feeding operation records, is processed using a machine learning regression algorithm to extract key parameters reflecting the dissolution characteristics of the active pharmaceutical ingredient, resulting in a dissolution behavior feature database. Based on the dissolution rate change curves in the database, the time points at which the outer surface layer of the active pharmaceutical ingredient powder particles reacts with solvent molecules to form a high-viscosity coating are identified. The contact retardation time series corresponding to these time points is extracted. Using this contact retardation time series data, the triggering conditions for the degradation reaction of the active pharmaceutical ingredient's chemical structure, where the temperature in a localized area within the mixing container exceeds the safe range, are analyzed. These triggering conditions include the local temperature value, duration, and temperature gradient distribution range. The upper and lower safe temperature thresholds corresponding to these triggering conditions are calculated to determine the appropriate temperature control safety range for mixing operations. For this temperature control safety range, target stirring intervention times are matched for different target concentration levels of the pharmaceutical solution. A corresponding correlation table between the material batch feeding order rules and the target stirring intervention times is established to obtain the initial mixing operation execution plan. The system acquires the current dissolution progress status parameter when executing the initial solution preparation operation plan, determines whether the current dissolution progress status parameter deviates from the pre-set normal fluctuation range threshold in the initial solution preparation operation plan, and if the current dissolution progress status parameter deviates from the normal fluctuation range threshold, a real-time dynamic correction command for the solvent ratio parameter is triggered. The initial solution preparation operation plan is updated based on the corrected solvent ratio parameter to obtain an adjusted solution preparation operation plan. Through the adjusted solution preparation operation plan, a multi-batch material delivery control signal sequence divided by time period is generated in the solution preparation container. The temperature fluctuation amplitude in the container during the execution of the multi-batch material delivery control signal sequence is acquired. If the temperature fluctuation amplitude in the container exceeds the upper or lower threshold of the temperature control safety range, a cooling intervention operation command to start the cooling device is output. The adjusted solution preparation operation plan is updated based on the actual temperature distribution after the cooling intervention to obtain a temperature-optimized solution preparation operation plan. Based on the optimized temperature control solution preparation operation plan, the timing of material batch addition and stirring equipment operation parameters in the solution preparation container are controlled, and temperature curve data, dissolution rate data and final product quality test data are recorded during the preparation operation to obtain a full-cycle dataset for the preparation of high-concentration ceftiofur hydrochloride injection.
[0005] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for preparing veterinary ceftiofur hydrochloride injection. Addressing the technical challenges of high-concentration drug solution preparation where the active pharmaceutical ingredient (API) easily forms a high-viscosity coating layer during dissolution, hindering dissolution and causing chemical degradation due to localized temperature runaway, the method utilizes machine learning algorithms to process critical saturation solubility data and historical batch operation records. This establishes a database of dissolution behavior characteristics and identifies the timing of contact hindrance. Based on this, the method analyzes the conditions for localized temperature exceeding limits and triggering degradation reactions, determines the safe temperature control range, and matches the optimal stirring intervention time and material batching sequence rules for different concentration levels. This forms an initial solution preparation operation execution plan. During execution, the dissolution progress status parameters are monitored in real time. When deviations from the normal fluctuation range occur, the solvent ratio parameters are dynamically corrected and the feeding control signal sequence is adjusted. If the temperature fluctuation within the container exceeds the safe range, a cooling device is activated to intervene in cooling. Finally, the optimized solution preparation operation execution plan controls the batching sequence of materials and the operating parameters of the stirring equipment, effectively solving the problems of dissolution hindrance and temperature runaway in high-concentration solution preparation, achieving precise control of the preparation process and stable improvement in product quality. Attached Figure Description
[0006] Figure 1 This is a flowchart of a method for preparing veterinary ceftiofur hydrochloride injection according to the present invention.
[0007] Figure 2 This is a schematic diagram illustrating a method for preparing a veterinary ceftiofur hydrochloride injection according to the present invention.
[0008] Figure 3 This is another schematic diagram of a method for preparing veterinary ceftiofur hydrochloride injection according to the present invention. Detailed Implementation
[0009] The technical solutions of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.
[0010] like Figures 1-3 The preparation method of veterinary ceftiofur hydrochloride injection in this embodiment may specifically include: Step S101: Obtain the critical data set of saturated solubility of the active pharmaceutical ingredient under different environmental temperature fluctuation conditions and solvent ratio parameter combinations. Process the critical data set of saturated solubility and historical batch feeding operation records through machine learning regression algorithm to extract key parameter indicators reflecting the dissolution characteristics of the active pharmaceutical ingredient and obtain a dissolution behavior feature database.
[0011] The solubility of the primary drug source at a preset temperature and solvent ratio is obtained, and the critical value under saturation is determined to obtain a critical solubility dataset. From this dataset, the target temperature and target solvent ratio are extracted, and the feed amounts corresponding to historical batches are obtained to generate a feed operation record table. A machine learning regression algorithm is used to process the critical solubility dataset and the feed operation record table to extract the solubility of the primary drug source, obtaining a solubility parameter. For each solubility parameter, it is determined whether it exceeds a preset threshold. If it does, a target solubility parameter is determined. Based on the target solubility parameter, the batch number of the primary drug source is matched to obtain a target solubility parameter with the batch number. The target solubility parameter with the batch number is processed using a structured storage method to establish a dissolution behavior feature library.
[0012] In one possible implementation, the solubility of the main drug source under different environments is obtained by controlling the heating jacket and solvent injection valve, under preset conditions such as 45 degrees Celsius and a solvent ratio of 1:5. When a trace amount of undissolved solid appears at the bottom of the solution and does not decrease for 30 minutes, the concentration at this point is determined to be the critical value under saturation, thereby compiling a critical dataset of saturated solubility.
[0013] For example, the target temperature value of 60 degrees Celsius and the target solvent ratio of 1:8 are extracted from the dataset, and historical batch data from the manufacturing execution system are retrieved to obtain the corresponding actual feed amount, such as 500 kg. This generates a feed operation record table containing temperature, solvent ratio, and feed amount. This provides accurate data support for subsequent solubility prediction and effectively avoids raw material waste caused by blind feeding.
[0014] It should be noted that after obtaining the above data, the random forest regression algorithm in machine learning regression algorithms was used to process the critical dataset of saturated solubility and the feeding operation record table.
[0015] Specifically, temperature, solvent ratio, and historical feed rates are used as input features, and the actual measured solubility rate is used as the output label for model training. Through ensemble learning of decision trees and feature weight allocation, the solubility rate of the main drug source under specific process conditions is extracted, obtaining specific solubility rate parameters such as 95%. This process can deeply explore the nonlinear mapping relationship between multidimensional process parameters and dissolution behavior, significantly improving the accuracy of solubility rate prediction.
[0016] In one possible implementation, a preset threshold of 90% is set for the obtained solubility parameter. It is then determined whether 95% of the extracted solubility parameter exceeds this threshold; if so, it is identified as the target solubility parameter.
[0017] For example, based on the target solubility parameter, the batch number corresponding to the main drug source, such as 20231001, is matched in the production database to obtain the target solubility parameter with the batch number. Subsequently, using a structured storage method of a relational database, the batch number, temperature, solvent ratio, and target solubility parameter are mapped to a data table and stored as fields, establishing a solubility behavior feature library. Establishing this feature library enables the accumulation and traceability of historically excellent solubility processes, providing a reliable reference benchmark for optimizing process parameters for future new batches of drugs, and significantly improving batch-to-batch consistency in drug production.
[0018] Step S102: Based on the dissolution rate change curve in the dissolution behavior feature database, identify the time node at which a high-viscosity coating layer is formed after the initial reaction between the outer surface layer of the active pharmaceutical ingredient powder particles and the solvent molecules, extract the contact retardation time sequence corresponding to the time node, and obtain contact retardation time sequence data.
[0019] The contact hindrance time series set is obtained, and the hindrance occurrence time is extracted from the set to obtain the initial hindrance time node. Based on the initial hindrance time node, the corresponding dissolution rate change curve is matched to obtain the instantaneous dissolution rate value corresponding to the initial hindrance time node. The instantaneous dissolution rate value is processed using a K-means clustering algorithm to divide different rate decrease intervals, obtaining rate decay clusters. For each rate decay cluster, the minimum rate value within the cluster is extracted, and it is determined whether the minimum rate value is less than a preset rate threshold. If the minimum rate value is less than the preset rate threshold, the rate decay cluster is determined to be a severely hindrance cluster. By analyzing the severely hindrance cluster, the high-viscosity coating thickness value corresponding to the cluster is extracted, and a correlation data table between coating thickness and hindrance state is established.
[0020] In one possible implementation, to address the dissolution stagnation phenomenon caused by the formation of a gel layer due to surface hydration of the active pharmaceutical ingredient in the solvent, a set of contact retardation time series during the dissolution process is obtained using an online turbidity monitoring device.
[0021] For example, the moment when the turbidity change rate approaches zero is extracted from the time series set as the moment of stagnation, thus obtaining the initial stagnation time node, such as the 15th minute. Based on this initial stagnation time node, the corresponding dissolution rate change curve is matched in a pre-built dissolution kinetics database to obtain the instantaneous dissolution rate value corresponding to the initial stagnation time node, such as 2 grams per minute.
[0022] It should be noted that after obtaining a large number of instantaneous dissolution rate values, the mean clustering algorithm was used to process these values.
[0023] Specifically, the instantaneous dissolution rate values extracted from different batches are used as input features. The number of cluster centers is set to 3. By iteratively calculating the Euclidean distance from each data point to the cluster center, different rate decrease intervals are divided, and finally, rate decay clusters containing different degrees of decay are output.
[0024] In one possible implementation, for each output rate decay cluster, the minimum rate value within that cluster is extracted, such as 1 gram per minute. Then, a preset rate threshold of 3 grams per minute is set, and it is determined whether the extracted minimum rate value is less than this preset threshold. Since 1 gram per minute is less than 3 grams per minute, the currently analyzed rate decay cluster is determined to be a heavily blocked cluster.
[0025] For example, by analyzing the identified heavily stagnant clusters and combining them with in-situ microscopic imaging records, the thickness value of the high-viscosity coating layer corresponding to each heavily stagnant cluster, such as 5 mm, is extracted. Finally, the extracted coating layer thickness value of 5 mm is structurally mapped to the corresponding heavily stagnant state label to establish a correlation data table between coating layer thickness and stagnant state.
[0026] Step S103: Using the contact retardation time series data, analyze the triggering conditions for the degradation reaction of the main drug's chemical structure when the temperature in a local area of the solution preparation container exceeds the safe range. The triggering conditions include the local temperature value, duration, and temperature gradient distribution range. Calculate the upper and lower safe temperature thresholds corresponding to the triggering conditions to determine the safe temperature control range suitable for solution preparation operations.
[0027] Contact retardation time series data is acquired, and the local temperature values and durations within the solution preparation container are extracted to obtain an initial temperature state sequence. The isolated forest algorithm is used to process this initial temperature state sequence, identifying abnormal temperature nodes that exceed normal fluctuations and obtaining candidate nodes for degradation reactions. For each candidate node, the corresponding temperature gradient distribution range is extracted, and the product of the temperature gradient distribution range and the local temperature value is calculated to obtain a structural bond breakage risk index. It is determined whether the structural bond breakage risk index is greater than a preset risk threshold. If the index is greater than the threshold, the candidate node is identified as a true degradation point. Based on the true degradation point, the corresponding local temperature values and durations are extracted to construct a degradation reaction triggering condition set. By analyzing this set, the highest and lowest temperatures where degradation has not occurred are calculated to obtain a safe upper temperature threshold and a safe lower temperature threshold. Based on these thresholds, the temperature operating range of the solution preparation container is defined, determining a safe temperature control zone suitable for solution preparation operations.
[0028] In one possible implementation, the local temperature value and duration during the contact retardation period of the active pharmaceutical ingredient powder within the solution preparation container are obtained. A local temperature of 45 degrees Celsius and a duration of 120 seconds are extracted to form an initial temperature state sequence. This sequence is then input into an isolated forest algorithm model. This algorithm constructs multiple isolated trees to continuously and randomly partition the data space. Because abnormal temperature data is relatively sparse in the feature space, it is easier to isolate it near the root node of the tree. The output is an anomaly score. When the score exceeds a set threshold, the temperature node is marked as a candidate node for degradation reaction.
[0029] For example, for the identified candidate degradation reaction nodes, the temperature gradient distribution range around the node is further extracted. Assume the extracted temperature gradient distribution range value is 5, and the local temperature value is 45 degrees Celsius. Multiplying the temperature gradient distribution range by the local temperature value yields a structural bond breakage risk index of 225. Setting a preset risk threshold of 200, since 225 is greater than 200, it is determined that the localized high temperature and drastic temperature change are sufficient to break the chemical bonds of the active pharmaceutical ingredient, thus identifying the candidate degradation reaction node as a true degradation point.
[0030] It should be noted that, based on the identified multiple actual degradation points, the corresponding local temperature values and durations were extracted to construct a set of degradation reaction triggering conditions. This set was analyzed, and all temperature records that did not trigger degradation reactions were examined. The highest temperature at which degradation did not occur was calculated to be 38 degrees Celsius, and the lowest temperature was 15 degrees Celsius. These were used as the upper and lower safe temperature thresholds, respectively. Based on these two thresholds, the temperature operating range of the dispensing container was defined, and the final safe temperature control range for dispensing this active pharmaceutical ingredient powder was determined to be 15 to 38 degrees Celsius. This served as the basis for setting subsequent dissolution temperature control parameters.
[0031] Step S104: For the temperature control safety range, match the target stirring intervention time under different drug solution target concentration levels, establish a corresponding correlation table between the material batch addition order rule and the target stirring intervention time, and obtain the initial solution preparation operation execution plan.
[0032] The process involves acquiring the temperature control safety range and historical solution preparation records, extracting the target concentration values of the drug solutions from the historical preparation records, and classifying these values using a K-means clustering algorithm to obtain different target concentration levels. Based on these different target concentration levels, corresponding historical stirring start-up time nodes are extracted, and the distribution density of these time nodes within the temperature control safety range is calculated to obtain the target stirring intervention time. For the target stirring intervention time, the dissolution rate of the material to be prepared is acquired, and the product of the dissolution rate and the target stirring intervention time is calculated to obtain a material accumulation risk index. It is then determined whether the material accumulation risk index is greater than a preset accumulation threshold. If the index is greater than the threshold, the material to be prepared is identified as a material to be added in batches. The mass values of the materials to be added in batches are acquired, and a decision tree algorithm is used to process these mass values and the target stirring intervention time to generate a material batching order rule. Based on the material batching order rule and the target stirring intervention time, a two-dimensional matching matrix is constructed, and non-zero elements are extracted to establish a corresponding association table. By parsing the corresponding association table, the time nodes and action instructions in the corresponding association table are extracted, and the time nodes and action instructions are combined to obtain the initial solution preparation operation execution plan.
[0033] In one possible implementation, a pre-determined temperature control safety range of 15°C to 38°C and 500 historical drug preparation records are obtained. The target concentration values of the drug solutions are extracted from these records and used as input feature vectors to a clustering algorithm model. With a cluster size of 3, the algorithm iteratively calculates the distance from each concentration value to the initial cluster center and continuously updates the center point, outputting three target concentration levels: low, medium, and high. For the high concentration level, the corresponding set of historical stirring start-up time nodes is extracted. A kernel density estimation algorithm is used to analyze the distribution of this set on the time axis corresponding to the 15°C to 38°C range, finding the peak point of the probability density curve. The time corresponding to this peak, i.e., 45 seconds after the start of drug preparation, is taken as the target stirring intervention time.
[0034] For example, the dissolution rate of the active pharmaceutical ingredient powder to be prepared is obtained as 2 grams per second. This dissolution rate value of 2 is multiplied by the target stirring intervention time of 45 to calculate the material accumulation risk index of 90. A preset accumulation threshold of 80 is set. Since 90 is greater than 80, it is determined that if the material is added all at once, it will form clumps at the bottom of the container, thus determining that the material to be prepared is to be added in batches.
[0035] It should be noted that the total mass of the materials to be added in batches is 500 grams. The mass value of 500 and the target stirring intervention time of 45 seconds are used as input features and input into a pre-trained decision tree algorithm model. This model splits the features into nodes using the information gain criterion, outputting a rule for the order of material batch additions: an initial addition of 200 grams and a subsequent addition of 300 grams. Based on this rule and the target stirring intervention time, a two-dimensional matching matrix is constructed with time as the horizontal axis and addition quantity as the vertical axis. Elements with non-zero values are extracted from the matrix, establishing a corresponding association table. This table is parsed to extract the instruction for the first batch addition at second 0 and the instruction for the second batch addition and stirring start at second 45. Combining these time nodes and action instructions, an initial liquid preparation operation execution plan containing a timeline and specific operational actions is finally generated.
[0036] Step S105: Obtain the current dissolution progress status parameter when executing the initial solution preparation operation plan, determine whether the current dissolution progress status parameter deviates from the normal fluctuation range threshold preset in the initial solution preparation operation plan, if the current dissolution progress status parameter deviates from the normal fluctuation range threshold, trigger a real-time dynamic correction command for the solvent ratio parameter, update the initial solution preparation operation plan based on the corrected solvent ratio parameter, and obtain the adjusted solution preparation operation plan.
[0037] The adjusted solution preparation operation plan is obtained, and the corrected solvent ratio parameters are extracted from the adjusted solution preparation operation plan. The corresponding target solvent injection flow rate is obtained based on the corrected solvent ratio parameters. For the target solvent injection flow rate, the real-time fluid pressure value in the current solution preparation pipeline is collected. The target solvent injection flow rate and the real-time fluid pressure value are processed using a random forest algorithm to obtain the flow rate resistance coefficient. A preset pipeline pressure threshold is obtained, and it is determined whether the flow rate resistance coefficient is greater than the pipeline pressure threshold. If the flow rate resistance coefficient is greater than the pipeline pressure threshold, a pipeline pressure reduction adjustment command is generated. Based on the pipeline pressure reduction adjustment command, the liquid level height value in the current solution preparation container is obtained, and the product of the liquid level height value and the flow rate resistance coefficient is calculated to obtain the overflow risk index. It is determined whether the overflow risk index exceeds a preset safety warning value. If the overflow risk index exceeds the safety warning value, a flow throttling action signal for the target solvent injection flow rate is triggered. The duration of the flow throttling action signal is obtained, and the final liquid preparation flow control strategy is generated by combining the duration of the flow throttling action with the corrected solvent ratio parameter.
[0038] For example, after obtaining the adjusted solution preparation operation plan, the corrected solvent ratio parameter is extracted. Assuming the current solution preparation task is to prepare a high-concentration drug solution, the corrected solvent ratio parameter indicates a volume ratio of purified water to active solvent of 3:1. Based on this, the corresponding target solvent injection flow rate is obtained as 50 liters per minute. For this target solvent injection flow rate, the real-time fluid pressure value in the current solution preparation pipeline is collected, for example, a real-time fluid pressure value of 200 kPa. The target solvent injection flow rate of 50 and the real-time fluid pressure value of 200 are used as input features and input into a pre-trained random forest algorithm model. This random forest algorithm constructs multiple decision trees and performs ensemble voting, outputting a current flow rate resistance coefficient of 8.
[0039] Specifically, a preset pipeline pressure threshold is obtained and set to 5. If the flow resistance coefficient 8 is greater than the pipeline pressure threshold 5, a pipeline pressure reduction adjustment command is generated. Based on this command, the current liquid level in the dispensing container is obtained, for example, 150 cm. The liquid level 150 is multiplied by the flow resistance coefficient 8 to obtain an overflow risk index of 1200. A preset safety warning value of 1000 is obtained. If the overflow risk index 1200 exceeds the safety warning value 1000, a flow control signal is triggered to halt further solvent injection.
[0040] It should be noted that the duration of the flow throttling action signal is obtained, for example, by setting the duration to 30 seconds. This 30-second throttling duration is then combined with the corrected solvent ratio of purified water to active solvent (3:1) to generate the final solution preparation flow control strategy. This final strategy includes control logic that restarts solvent injection at a new segmented flow rate after 30 seconds of flow throttling, thus completing the closed-loop dynamic fluid control of the entire solution preparation pipeline.
[0041] Step S106: Through the adjusted liquid preparation operation execution plan, generate a sequence of multiple batches of material release control signals divided into time periods in the liquid preparation container, obtain the temperature fluctuation range in the container during the execution of the multiple batches of material release control signal sequences, and if the temperature fluctuation range in the container exceeds the upper or lower threshold of the temperature control safety range, output a cooling intervention operation command to start the cooling device, and update the adjusted liquid preparation operation execution plan based on the actual temperature distribution after the cooling intervention to obtain the temperature-controlled optimized liquid preparation operation execution plan.
[0042] The optimized liquid preparation operation plan is obtained, and the material delivery time period and corresponding target material quantity are extracted from the optimized liquid preparation operation plan. A mixer speed adjustment command is generated based on the material delivery time period and the target material quantity. For the mixer speed adjustment command, the real-time torque value of the stirring shaft in the liquid preparation container is collected. A support vector machine algorithm is used to process the target material quantity and the real-time torque value to obtain the material viscosity index. A preset viscosity warning threshold is obtained, and it is determined whether the material viscosity index is greater than the viscosity warning threshold. If the material viscosity index is greater than the viscosity warning threshold, a diluent replenishment trigger signal is generated. Based on the diluent replenishment trigger signal, the remaining space capacity in the current liquid preparation container is obtained, and the ratio of the remaining space capacity to the material viscosity index is calculated to obtain the target diluent injection volume. For the target diluent injection volume, the injection time window corresponding to the target diluent injection volume is obtained. Combining the injection time window with the optimized liquid preparation operation plan, a final material mixing control strategy is generated.
[0043] For example, after obtaining the temperature-controlled optimized liquid preparation operation plan, the material delivery time period and the corresponding target material quantity are extracted. Assuming the current liquid preparation task is in the polymer excipient addition stage, the extracted material delivery time period is from the 10th to the 15th minute, and the corresponding target material quantity is 50 kg. Based on this, a mixer speed adjustment command is generated, instructing the mixer's base speed to be set to 120 rpm during the above time period.
[0044] Specifically, in response to the mixer speed adjustment command, the real-time torque value of the stirring shaft in the liquid preparation container is collected during the material feeding process. For example, the collected real-time torque value is 85 Nm. The target material quantity of 50 and the real-time torque value of 85 are used as input feature vectors and fed into a pre-trained support vector machine (SVM) algorithm model. This SVM algorithm maps the input features to a high-dimensional space through a kernel function and finds the optimal classification hyperplane, outputting the current material viscosity index as 75.
[0045] It should be noted that a preset viscosity warning threshold is obtained, and this threshold is set to 60. If the viscosity index of the material is greater than the viscosity warning threshold of 60, it indicates that there is local agglomeration in the current mixture, and a diluent replenishment trigger signal is generated.
[0046] For example, based on the diluent replenishment trigger signal, the remaining space capacity in the current dispensing container is obtained, for instance, the remaining space capacity is calculated to be 300 liters through space scanning. The remaining space capacity of 300 is divided by the material viscosity index of 75 to obtain the target diluent injection volume of 4 liters.
[0047] Specifically, for a target diluent injection volume of 4 liters, the corresponding injection time window is obtained, for example, set to 2 minutes. Combining this 2-minute injection time window with the aforementioned temperature-optimized liquid preparation operation plan, a final material mixing control strategy is generated. This final material mixing control strategy includes control logic that, within the 2-minute time window, 4 liters of diluent are injected according to a specific flow rate curve, and a variable frequency agitator is simultaneously activated to perform dynamic adjustment of the material state during the liquid preparation process.
[0048] Step S107: According to the optimized temperature control liquid preparation operation plan, control the batch addition sequence of materials in the liquid preparation container and the operating parameters of the stirring equipment, record the temperature curve data, dissolution rate data and final product quality test data during the preparation operation, and obtain the full-cycle dataset of high-concentration ceftiofur hydrochloride injection preparation.
[0049] A complete dataset of the preparation cycle of the high-concentration ceftiofur hydrochloride injection was obtained. Dissolution rate data and temperature curve data were extracted from this dataset. A random forest algorithm was used to process the dissolution rate data and temperature curve data to obtain dissolution lag time nodes. For each dissolution lag time node, real-time transmittance and real-time pH values were extracted from the complete dataset. The product of the real-time transmittance and pH values was calculated to obtain the solution turbidity index. A preset turbidity warning threshold was obtained, and it was determined whether the solution turbidity index exceeded the threshold. If the turbidity index exceeded the threshold, a crystallization risk warning signal was generated. Based on the crystallization risk warning signal, the mass of undissolved solute in the current dispensing container was obtained, and the ratio of the undissolved solute mass to the solution turbidity index was calculated to determine the target co-solvent addition volume. For the target co-solvent addition volume, a flow rate control command corresponding to the target co-solvent addition volume is obtained. Combining the flow rate control command with the temperature-optimized solution preparation operation plan, an anti-crystallization solution preparation operation plan is generated. The anti-crystallization solution preparation operation plan is obtained, and the target co-solvent addition time period is extracted from the plan. A circulation pump start / stop control signal is generated based on the target co-solvent addition time period.
[0050] In one possible implementation, a complete dataset of the preparation process of high-concentration ceftiofur hydrochloride injection is obtained, containing multidimensional parameters of different batches during the dispensing process. Dissolution rate data and temperature curve data are extracted as input features. The amount of solute loss per minute is used as the dissolution rate feature, and the corresponding container temperature is used as the temperature curve feature. A random forest algorithm is used to construct a prediction model, and the above input features are integrated and processed through multiple decision trees to output the dissolution lag time point. When the dissolution rate and temperature data of the first 30 minutes of a certain batch are input, the dissolution lag time point output by the random forest algorithm is the 42nd minute, indicating that the dissolution process of the solute significantly stalls at this time point.
[0051] For example, for the obtained dissolution lag time node of 42 minutes, the real-time transmittance and real-time pH values at that moment are extracted from the dataset. If the extracted real-time transmittance value is 85.5 and the real-time pH value is 6.2, the two are multiplied to obtain a solution turbidity index of 530.1. A preset turbidity warning threshold is obtained, assuming that the threshold is set to 500.0. If the current solution turbidity index of 530.1 is determined to be greater than the turbidity warning threshold of 500.0, a crystallization risk warning signal is generated, indicating that there is a tendency for solute to precipitate and crystallize in the current solution preparation environment.
[0052] It should be noted that, based on the generated crystallization risk warning signal, the mass of undissolved solute in the current solution preparation container is further obtained. If the current mass of undissolved solute is 1500 grams, the ratio of this mass to the solution turbidity index of 530.1 is calculated, and the target co-solvent addition volume is determined to be 2.8 liters. For this target co-solvent addition volume, the corresponding flow rate control command is obtained, such as setting the flow rate to 0.5 liters per minute. Combined with the aforementioned temperature-controlled optimized solution preparation operation execution plan, an anti-crystallization solution preparation operation execution plan is generated. The target co-solvent addition time period in this anti-crystallization solution preparation operation execution plan is extracted, and a circulation pump start / stop control signal is generated based on this time period to control the circulation pump to start within the specified co-solvent addition time period, performing the co-solvent injection and mixing operation.
[0053] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for preparing veterinary ceftiofur hydrochloride injection, characterized in that, The method includes: A set of critical saturation solubility data for the active pharmaceutical ingredient under different environmental temperature fluctuations and solvent ratio parameter combinations is obtained. This data set, along with historical batch feeding operation records, is processed using a machine learning regression algorithm to extract key parameters reflecting the dissolution characteristics of the active pharmaceutical ingredient, resulting in a dissolution behavior feature database. Based on the dissolution rate change curves in the database, the time points at which the outer surface layer of the active pharmaceutical ingredient powder particles reacts with solvent molecules to form a high-viscosity coating are identified. The contact retardation time series corresponding to these time points is extracted. Using this contact retardation time series data, the triggering conditions for the degradation reaction of the active pharmaceutical ingredient's chemical structure, where the temperature in a localized area within the mixing container exceeds the safe range, are analyzed. These triggering conditions include the local temperature value, duration, and temperature gradient distribution range. The upper and lower safe temperature thresholds corresponding to these triggering conditions are calculated to determine the appropriate temperature control safety range for mixing operations. For this temperature control safety range, target stirring intervention times are matched for different target concentration levels of the pharmaceutical solution. A corresponding correlation table between the material batch feeding order rules and the target stirring intervention times is established to obtain the initial mixing operation execution plan. The system acquires the current dissolution progress status parameter when executing the initial solution preparation operation plan, determines whether the current dissolution progress status parameter deviates from the pre-set normal fluctuation range threshold in the initial solution preparation operation plan, and if the current dissolution progress status parameter deviates from the normal fluctuation range threshold, a real-time dynamic correction command for the solvent ratio parameter is triggered. The initial solution preparation operation plan is updated based on the corrected solvent ratio parameter to obtain an adjusted solution preparation operation plan. Through the adjusted solution preparation operation plan, a multi-batch material delivery control signal sequence divided by time period is generated in the solution preparation container. The temperature fluctuation amplitude in the container during the execution of the multi-batch material delivery control signal sequence is acquired. If the temperature fluctuation amplitude in the container exceeds the upper or lower threshold of the temperature control safety range, a cooling intervention operation command to start the cooling device is output. The adjusted solution preparation operation plan is updated based on the actual temperature distribution after the cooling intervention to obtain a temperature-optimized solution preparation operation plan. Based on the optimized temperature control solution preparation operation plan, the timing of material batch addition and stirring equipment operation parameters in the solution preparation container are controlled, and temperature curve data, dissolution rate data and final product quality test data are recorded during the preparation operation to obtain a full-cycle dataset for the preparation of high-concentration ceftiofur hydrochloride injection.
2. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The process involves acquiring a set of critical saturation solubility data for the active pharmaceutical ingredient under different environmental temperature fluctuations and solvent ratio parameter combinations. This data, along with historical batch feeding operation records, is processed using a machine learning regression algorithm to extract key parameter indicators reflecting the dissolution characteristics of the active pharmaceutical ingredient, resulting in a dissolution behavior feature database, including: The solubility of the main drug source at a preset temperature and solvent ratio is obtained, the critical value under saturation is determined, and the critical solubility dataset is obtained. Extract the target temperature value and target solvent ratio from the saturated solubility critical dataset, obtain the feed amount corresponding to the historical batch, and generate a feed operation record table; The saturation solubility critical dataset and the feeding operation record table are processed using a machine learning regression algorithm to extract the solubility of the main drug source and obtain the solubility parameter. For the solubility parameter, it is determined whether the solubility parameter is greater than a preset threshold. If the solubility parameter is greater than the preset threshold, the target solubility parameter is determined. Based on the target solubility parameter, the batch number of the main drug source is matched to obtain the target solubility parameter with the batch number; The target solubility parameter with batch number is processed using a structured storage method to establish a solubility behavior feature library.
3. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The step involves identifying the time points where a high-viscosity coating layer forms after the initial reaction between the outer surface layer of the active pharmaceutical ingredient powder particles and solvent molecules, based on the dissolution rate change curves in the dissolution behavior feature database. The step also involves extracting the contact retardation time series data corresponding to these time points, including: Obtain the contact retardation time series set, extract the retardation occurrence time from the contact retardation time series set, and obtain the initial retardation time node; Based on the initial stagnation time node, match the corresponding dissolution rate change curve to obtain the instantaneous dissolution rate value corresponding to the initial stagnation time node; The instantaneous dissolution rate value was processed using the K-means clustering algorithm to divide different rate decrease intervals and obtain rate decay clusters. For the rate decay cluster, extract the minimum rate value within the rate decay cluster, and determine whether the minimum rate value is less than a preset rate threshold. If the minimum rate value is less than the preset rate threshold, then the rate decay cluster is determined to be a severely blocked cluster. By analyzing the heavily blocked clusters, the thickness values of the high-viscosity coating layer corresponding to the heavily blocked clusters are extracted, and a correlation data table between the coating layer thickness and the blocking state is established.
4. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The method utilizes the contact retardation time series data to analyze the triggering conditions for degradation reactions of the active pharmaceutical ingredient's chemical structure in localized areas of the solution preparation container where the temperature exceeds the safe range. These triggering conditions include the local temperature value, duration, and temperature gradient distribution range. The method calculates the upper and lower safe temperature thresholds corresponding to these triggering conditions, determining a suitable safe temperature control range for solution preparation operations, including: Obtain contact retardation time series data, extract the local temperature value and duration in the solution preparation container, and obtain the initial temperature state sequence; The initial temperature state sequence is processed using the isolated forest algorithm to identify abnormal temperature nodes that exceed normal fluctuations and obtain candidate nodes for degradation reactions. For the candidate nodes of the degradation reaction, the corresponding temperature gradient distribution range is extracted, and the product of the temperature gradient distribution range and the local temperature value is calculated to obtain the structural bond breakage risk index. Determine whether the structural bond breakage risk index is greater than a preset risk threshold. If the structural bond breakage risk index is greater than the preset risk threshold, then determine the candidate degradation reaction node as the actual degradation point. Based on the actual degradation point, extract the local temperature value and duration corresponding to the actual degradation point to construct a set of degradation reaction triggering conditions; By analyzing the set of degradation reaction triggering conditions, the highest and lowest temperatures at which no degradation occurred are calculated to obtain the upper and lower threshold values of the safe temperature. Based on the upper and lower safe temperature thresholds, the temperature operating range of the dispensing container is defined, and a safe temperature control zone suitable for dispensing operations is determined.
5. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, For the aforementioned temperature control safety range, the target stirring intervention time is matched under different target concentration levels of the drug solution, and a corresponding correlation table is established between the material batching order rules and the target stirring intervention time to obtain the initial solution preparation operation execution plan, including: The temperature control safety range and historical solution preparation records are obtained. The target concentration values of the drug solution in the historical solution preparation records are extracted. The K-means clustering algorithm is used to classify the target concentration values of the drug solution to obtain different target concentration levels of the drug solution. Based on the different target concentration levels of the drug solution, the corresponding historical stirring start time nodes are extracted, and the distribution density of the historical stirring start time nodes within the temperature control safety range is calculated to obtain the target stirring intervention time. For the target stirring intervention time, obtain the dissolution rate value of the material to be prepared, calculate the product of the dissolution rate value and the target stirring intervention time, and obtain the material accumulation risk index. Determine whether the material accumulation risk index is greater than a preset accumulation threshold. If the material accumulation risk index is greater than the preset accumulation threshold, then determine that the material to be prepared is a material to be released in batches. The mass values of the materials added in batches are obtained, and the decision tree algorithm is used to process the mass values and the target stirring intervention time to generate the material batch addition order rules. Based on the material batching order rules and the target stirring intervention time, a two-dimensional matching matrix is constructed, and the non-zero elements in the two-dimensional matching matrix are extracted to establish a corresponding association table. By parsing the corresponding association table, the time nodes and action instructions in the corresponding association table are extracted, and the time nodes and action instructions are combined to obtain the initial solution preparation operation execution plan.
6. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The process involves obtaining the current dissolution progress status parameter when executing the initial solution preparation operation plan, determining whether the current dissolution progress status parameter deviates from a pre-set normal fluctuation range threshold in the initial solution preparation operation plan, and if the current dissolution progress status parameter deviates from the normal fluctuation range threshold, triggering a real-time dynamic correction command for the solvent ratio parameter, updating the initial solution preparation operation plan based on the corrected solvent ratio parameter, and obtaining an adjusted solution preparation operation plan, including: Obtain the adjusted solution preparation operation execution plan, extract the corrected solvent ratio parameters from the adjusted solution preparation operation execution plan, and obtain the corresponding target solvent injection flow rate based on the corrected solvent ratio parameters; For the target solvent injection flow rate, the real-time fluid pressure value in the current liquid preparation pipeline is collected, and the random forest algorithm is used to process the target solvent injection flow rate and the real-time fluid pressure value to obtain the flow rate resistance coefficient; Obtain a preset pipeline pressure threshold, determine whether the flow velocity resistance coefficient is greater than the pipeline pressure threshold, and if the flow velocity resistance coefficient is greater than the pipeline pressure threshold, generate a pipeline pressure reduction adjustment command. According to the pipeline pressure reduction adjustment command, the liquid level height value in the current liquid preparation container is obtained, and the product of the liquid level height value and the flow resistance coefficient is calculated to obtain the overflow risk index. Determine whether the overflow risk index exceeds the preset safety warning value. If the overflow risk index exceeds the safety warning value, trigger a flow throttling action signal for the target solvent injection flow rate. The duration of the flow throttling action signal is obtained, and the final liquid preparation flow control strategy is generated by combining the duration of the flow throttling action with the corrected solvent ratio parameter.
7. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The adjusted liquid preparation operation execution plan generates a sequence of control signals for multiple batches of materials added to the liquid preparation container, divided into time periods. It then acquires the temperature fluctuation range within the container during the execution of these control signal sequences. If the temperature fluctuation range exceeds the upper or lower threshold of the temperature control safety range, a cooling intervention command is output to activate the cooling device. The adjusted liquid preparation operation execution plan is updated based on the actual temperature distribution after the cooling intervention, resulting in a temperature-optimized liquid preparation operation execution plan, including: Obtain the optimized liquid preparation operation plan after temperature control, extract the material delivery time period and the corresponding target material quantity in the optimized liquid preparation operation plan, and generate a mixer speed adjustment command based on the material delivery time period and the target material quantity. In response to the mixer speed adjustment command, the real-time torque value of the stirring shaft in the liquid preparation container is collected, and the target material quantity and the real-time torque value are processed by the support vector machine algorithm to obtain the material viscosity index. Obtain a preset viscosity warning threshold, determine whether the viscosity index of the material is greater than the viscosity warning threshold, and if the viscosity index of the material is greater than the viscosity warning threshold, generate a diluent replenishment trigger signal. Based on the diluent replenishment trigger signal, the remaining space capacity in the current liquid preparation container is obtained, the ratio of the remaining space capacity to the material viscosity index is calculated, and the target diluent injection volume is obtained. For the target diluent injection volume, the injection time window corresponding to the target diluent injection volume is obtained. Combining the injection time window with the temperature-controlled optimized liquid preparation operation execution plan, a final material mixing control strategy is generated.
8. The method for preparing veterinary ceftiofur hydrochloride injection according to claim 1, characterized in that, The process involves controlling the batch addition sequence of materials in the preparation container and the operating parameters of the stirring equipment according to the optimized temperature control solution preparation plan. Temperature curve data, dissolution rate data, and final product quality testing data are recorded during the preparation operation to obtain a complete dataset for the preparation of high-concentration ceftiofur hydrochloride injection, including: The complete preparation cycle dataset of the high-concentration ceftiofur hydrochloride injection was obtained. The dissolution rate data and temperature curve data in the complete preparation cycle dataset of the high-concentration ceftiofur hydrochloride injection were extracted. The dissolution rate data and temperature curve data were processed by the random forest algorithm to obtain the dissolution lag time nodes. For the dissolution lag time point, the real-time transmittance and real-time pH values corresponding to the dissolution lag time point are extracted from the full-cycle dataset of the preparation of the high-concentration ceftiofur hydrochloride injection. The product of the real-time transmittance and real-time pH values is calculated to obtain the solution turbidity index. Obtain a preset turbidity warning threshold, determine whether the turbidity index of the solution is greater than the turbidity warning threshold, and if the turbidity index of the solution is greater than the turbidity warning threshold, generate a crystallization risk warning signal. Based on the crystallization risk warning signal, obtain the mass of undissolved solute in the current solution preparation container, calculate the ratio of the mass of undissolved solute to the turbidity index of the solution, and determine the target co-solvent addition volume; For the target cosolvent addition volume, obtain the flow rate control command corresponding to the target cosolvent addition volume, and combine the flow rate control command with the temperature-controlled optimized solution preparation operation execution plan to generate an anti-crystallization solution preparation operation execution plan. Obtain the anti-crystallization solution preparation operation plan, extract the target co-solvent addition time period in the anti-crystallization solution preparation operation plan, and generate a circulation pump start / stop control signal based on the target co-solvent addition time period.
